I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM,and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
Isn't it time we had something more like these designs upgraded? 16
bit or more versions?
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth forARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
Wayne morellini schrieb am Dienstag, 7. Juni 2022 um 11:52:48 UTC+2:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
A Micropython chip will be more attractive, I guess.A more fancy fashion accessory. Why waste the time on a dead end performance implementation that could never be as good?
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
I wasn't expecting such negative answers, but more community spirit. If I wanted to do my own design, I would do it under agreement, this is more about community design. I remember when I wanted to implement my own uniquely improved design, and peoplewanted to seize control of it. In real life money flows one way, in business money tend to flow one opposite way (that's not a Chinese proverb, that's just me writing). :)
I had thought of colorforth as a sort of model of a stripped down x86 mode, like Arm has Thumb, where the instructions can use existing instruction circuitry. But it could also be an alternative to Thumb on the Arm too. A future basis They could make asuper light weight computing chip (SLWC) out of. Not as light as the recent misc's, but also able to be put in arrays, and used in a similar way that Arm is used in custom chips. So, that companies could incorporate into many forms of chips.
So. a cell, isa and interfacing spec (basically what they use for Arm chips), and pursue a manufacturer to kick off with a io logic processor version (also suitable in arrays), a minimal microcontroller (also suitable as an array controller) and a fullsystem on a chip (also useful to have an array it controls) in 16 and 32 bit versions. All the same two cells with different attachments. To kick things off, and attract other supporters instead of RiscV and Arm. Basically, largely coming in the low end,
As far as my own. It's unaffordable, but maybe in the future if success is already obtained to afford it. Isn't Google non commercial open source or something. Maybe that helps here, but not for myself. Control of design and use of resources to enhancedesign and market is needed to maximise public good here. But, 2 micron chips? I'm talking about modern process sizes, certainly not less than the 180nm which is currently used for misc.
Commercially myself, I am more interested in simple printed circuits at 2.5 micro to 1 micron. I think I finally have a clean room solution to do this at home. That's more doable on an individual's level. 1 billion dollars to $100, it doesn't matter somuch about making chips so small if the cost of making is cheap enough. I tried to contact Jeff a long time ago about a seperate technology, which might even be suitable for 180nm, where they could role stamp a circuit. I was envisioning printing on
Anyway, I remember Jeff telling me he couldn't get any further help with money for chips from his family, unless it was the kind you get with fish. Considering what else happened to him with his prototype chip run, it was a real loss to the computerindustry that he couldn't do the F series of processors. He had good intent of ideas, and his wafer arrays. Since my father unexpectedly died recently, before I could get back in to spend time with him, I am also coming to a similar dilemma, even if the
Anyway, this is not about what I can do and my work. It's about the community. There is a guy with a channel called the Unknown Cat, that I've seen. And, he was saying he's getting cranky and slowing down because he's getting older. Well, it seems tobe a lot of that going around here, instead of community.
Everybody loosing their edge or mind, is getting negative. I've seen
Why bother with b16, I understand that is the one which was cut down for FPGA? Instead of say Dr Ting's 16 or 32 bit, or Jeff's VLSI design? What's wrong with using that compared to modern misc instruction set format?
FPGA makes more sense when the product is FPGA based, or you are prototyping towards final product. There were misc chip simulators out there, after you go through the extending forth to simulate the logicality of the instruction set architecture.
Anyway, some thoughts.
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM,and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
Wayne morellini schrieb am Mittwoch, 8. Juni 2022 um 13:59:50 UTC+2:
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
people wanted to seize control of it. In real life money flows one way, in business money tend to flow one opposite way (that's not a Chinese proverb, that's just me writing). :)I wasn't expecting such negative answers, but more community spirit. If I wanted to do my own design, I would do it under agreement, this is more about community design. I remember when I wanted to implement my own uniquely improved design, and
a super light weight computing chip (SLWC) out of. Not as light as the recent misc's, but also able to be put in arrays, and used in a similar way that Arm is used in custom chips. So, that companies could incorporate into many forms of chips.I had thought of colorforth as a sort of model of a stripped down x86 mode, like Arm has Thumb, where the instructions can use existing instruction circuitry. But it could also be an alternative to Thumb on the Arm too. A future basis They could make
full system on a chip (also useful to have an array it controls) in 16 and 32 bit versions. All the same two cells with different attachments. To kick things off, and attract other supporters instead of RiscV and Arm. Basically, largely coming in the lowSo. a cell, isa and interfacing spec (basically what they use for Arm chips), and pursue a manufacturer to kick off with a io logic processor version (also suitable in arrays), a minimal microcontroller (also suitable as an array controller) and a
enhance design and market is needed to maximise public good here. But, 2 micron chips? I'm talking about modern process sizes, certainly not less than the 180nm which is currently used for misc.As far as my own. It's unaffordable, but maybe in the future if success is already obtained to afford it. Isn't Google non commercial open source or something. Maybe that helps here, but not for myself. Control of design and use of resources to
so much about making chips so small if the cost of making is cheap enough. I tried to contact Jeff a long time ago about a seperate technology, which might even be suitable for 180nm, where they could role stamp a circuit. I was envisioning printing onCommercially myself, I am more interested in simple printed circuits at 2.5 micro to 1 micron. I think I finally have a clean room solution to do this at home. That's more doable on an individual's level. 1 billion dollars to $100, it doesn't matter
industry that he couldn't do the F series of processors. He had good intent of ideas, and his wafer arrays. Since my father unexpectedly died recently, before I could get back in to spend time with him, I am also coming to a similar dilemma, even if theAnyway, I remember Jeff telling me he couldn't get any further help with money for chips from his family, unless it was the kind you get with fish. Considering what else happened to him with his prototype chip run, it was a real loss to the computer
be a lot of that going around here, instead of community.Anyway, this is not about what I can do and my work. It's about the community. There is a guy with a channel called the Unknown Cat, that I've seen. And, he was saying he's getting cranky and slowing down because he's getting older. Well, it seems to
Everybody loosing their edge or mind, is getting negative. I've seen
Why bother with b16, I understand that is the one which was cut down for FPGA? Instead of say Dr Ting's 16 or 32 bit, or Jeff's VLSI design? What's wrong with using that compared to modern misc instruction set format?
FPGA makes more sense when the product is FPGA based, or you are prototyping towards final product. There were misc chip simulators out there, after you go through the extending forth to simulate the logicality of the instruction set architecture.
Anyway, some thoughts.Just because you don't agree with what others wrote, you call them negative?
Isn't it positive at all that they "warned" before entering a possible dead end?
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
I saw your last post, but let me respond to this one. There have been many stack chips, even if virtually no Forth chips (a pedantic distinction, I know). You seem to want to fix a problem in some previous chip without indicating what needs to beaddressed. If the Mup21 was a good CPU chip, why is it no longer in production? Without an understanding of that, why do you think a redo would be successful?
I also don't understand the interest in a 16 bit chip. In the microprocessor world, nearly all devices are 4/8, 16 or 32 bits. Over time the trends which have emerged are that 4/8 bit MCUs are still dominant in low end applications where cost is thedeterminant. 32 bit processors have come down in price enough to dominate in nearly every other application space. 16 bit processors are actually becoming less popular, both in percentages of design wins and numbers of devices shipped. There is no reason
The next stack processor that should be marketed is a 32 bit design... IF there is a genuine interest in selling the chip to a general market. It also needs a number of features lacking in most stack processors.included in an external interface, but if no such interface exists, then a high speed clock is needed in the chip. A low speed clock is useful for wake up calls to poll the environment for activity. If the activity has a signal that can wake up the CPU,
1) A crystal clock oscillator, if not two. 32.768 kHz and some high frequency such as 12.0 MHz. While an asynch processor is a great idea, applications are seldom time invariant, so need clocking at some point in the design. Possibly this clock can be
2) 3.3V I/Os, if not 5V tolerant. Having to add parts to a design to provide level shifting is an absurdity in today's MCU world. With the vast majority of MCUs having 3.3V I/O capability, if not 5V tolerance, it is hard for a device to compete whichrequires level shifting.
3) There are others, but this is the real issue with a stack processor... C based development tools. Yes, this may be blasphemy in this group, but if an MCU is going to succeed in the world today, it must support the language that the vast majority ofembedded programmers use. There must be the debugging tools the world is used to. Such a chip must allow the world to use it, the way *they* are accustomed to working. The conventional Forth programming paradigm is too radical, and too crude (compared to
Just my opinion with a few facts.
On Wednesday, June 8, 2022 at 10:24:51 PM UTC+10, minf...@arcor.de wrote:
Wayne morellini schrieb am Mittwoch, 8. Juni 2022 um 13:59:50 UTC+2:
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
people wanted to seize control of it. In real life money flows one way, in business money tend to flow one opposite way (that's not a Chinese proverb, that's just me writing). :)I wasn't expecting such negative answers, but more community spirit. If I wanted to do my own design, I would do it under agreement, this is more about community design. I remember when I wanted to implement my own uniquely improved design, and
make a super light weight computing chip (SLWC) out of. Not as light as the recent misc's, but also able to be put in arrays, and used in a similar way that Arm is used in custom chips. So, that companies could incorporate into many forms of chips.I had thought of colorforth as a sort of model of a stripped down x86 mode, like Arm has Thumb, where the instructions can use existing instruction circuitry. But it could also be an alternative to Thumb on the Arm too. A future basis They could
full system on a chip (also useful to have an array it controls) in 16 and 32 bit versions. All the same two cells with different attachments. To kick things off, and attract other supporters instead of RiscV and Arm. Basically, largely coming in the lowSo. a cell, isa and interfacing spec (basically what they use for Arm chips), and pursue a manufacturer to kick off with a io logic processor version (also suitable in arrays), a minimal microcontroller (also suitable as an array controller) and a
enhance design and market is needed to maximise public good here. But, 2 micron chips? I'm talking about modern process sizes, certainly not less than the 180nm which is currently used for misc.As far as my own. It's unaffordable, but maybe in the future if success is already obtained to afford it. Isn't Google non commercial open source or something. Maybe that helps here, but not for myself. Control of design and use of resources to
matter so much about making chips so small if the cost of making is cheap enough. I tried to contact Jeff a long time ago about a seperate technology, which might even be suitable for 180nm, where they could role stamp a circuit. I was envisioningCommercially myself, I am more interested in simple printed circuits at 2.5 micro to 1 micron. I think I finally have a clean room solution to do this at home. That's more doable on an individual's level. 1 billion dollars to $100, it doesn't
computer industry that he couldn't do the F series of processors. He had good intent of ideas, and his wafer arrays. Since my father unexpectedly died recently, before I could get back in to spend time with him, I am also coming to a similar dilemma,Anyway, I remember Jeff telling me he couldn't get any further help with money for chips from his family, unless it was the kind you get with fish. Considering what else happened to him with his prototype chip run, it was a real loss to the
to be a lot of that going around here, instead of community.Anyway, this is not about what I can do and my work. It's about the community. There is a guy with a channel called the Unknown Cat, that I've seen. And, he was saying he's getting cranky and slowing down because he's getting older. Well, it seems
better is harder and more limited. Community becomes non community, and non community becomes community. It's extremely short sighted and in some places, reverse.Everybody loosing their edge or mind, is getting negative. I've seen
Why bother with b16, I understand that is the one which was cut down for FPGA? Instead of say Dr Ting's 16 or 32 bit, or Jeff's VLSI design? What's wrong with using that compared to modern misc instruction set format?
FPGA makes more sense when the product is FPGA based, or you are prototyping towards final product. There were misc chip simulators out there, after you go through the extending forth to simulate the logicality of the instruction set architecture.
Anyway, some thoughts.Just because you don't agree with what others wrote, you call them negative?
Isn't it positive at all that they "warned" before entering a possible dead end?Don't you think that's a bit of a twist? They cut accross the conversation negatively towards incompatible directions. This is not a fantasy where down is dysfunctionally up and up is dysfunctionally down. Harder and more limited is better, and actual
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
I saw your last post, but let me respond to this one. There have been many stack chips, even if virtually no Forth chips (a pedantic distinction, I know). You seem to want to fix a problem in some previous chip without indicating what needs to beaddressed. If the Mup21 was a good CPU chip, why is it no longer in production? Without an understanding of that, why do you think a redo would be successful?
I also don't understand the interest in a 16 bit chip. In the microprocessor world, nearly all devices are 4/8, 16 or 32 bits. Over time the trends which have emerged are that 4/8 bit MCUs are still dominant in low end applications where cost is thedeterminant. 32 bit processors have come down in price enough to dominate in nearly every other application space. 16 bit processors are actually becoming less popular, both in percentages of design wins and numbers of devices shipped. There is no reason
The next stack processor that should be marketed is a 32 bit design... IF there is a genuine interest in selling the chip to a general market. It also needs a number of features lacking in most stack processors.included in an external interface, but if no such interface exists, then a high speed clock is needed in the chip. A low speed clock is useful for wake up calls to poll the environment for activity. If the activity has a signal that can wake up the CPU,
1) A crystal clock oscillator, if not two. 32.768 kHz and some high frequency such as 12.0 MHz. While an asynch processor is a great idea, applications are seldom time invariant, so need clocking at some point in the design. Possibly this clock can be
2) 3.3V I/Os, if not 5V tolerant. Having to add parts to a design to provide level shifting is an absurdity in today's MCU world. With the vast majority of MCUs having 3.3V I/O capability, if not 5V tolerance, it is hard for a device to compete whichrequires level shifting.
3) There are others, but this is the real issue with a stack processor... C based development tools. Yes, this may be blasphemy in this group, but if an MCU is going to succeed in the world today, it must support the language that the vast majority ofembedded programmers use. There must be the debugging tools the world is used to. Such a chip must allow the world to use it, the way *they* are accustomed to working. The conventional Forth programming paradigm is too radical, and too crude (compared to
Just my opinion with a few facts.
--
Rick C.
- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209
gnuarm.del...@gmail.com schrieb am Mittwoch, 8. Juni 2022 um 14:52:08 UTC+2:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
addressed. If the Mup21 was a good CPU chip, why is it no longer in production? Without an understanding of that, why do you think a redo would be successful?I saw your last post, but let me respond to this one. There have been many stack chips, even if virtually no Forth chips (a pedantic distinction, I know). You seem to want to fix a problem in some previous chip without indicating what needs to be
determinant. 32 bit processors have come down in price enough to dominate in nearly every other application space. 16 bit processors are actually becoming less popular, both in percentages of design wins and numbers of devices shipped. There is no reasonI also don't understand the interest in a 16 bit chip. In the microprocessor world, nearly all devices are 4/8, 16 or 32 bits. Over time the trends which have emerged are that 4/8 bit MCUs are still dominant in low end applications where cost is the
be included in an external interface, but if no such interface exists, then a high speed clock is needed in the chip. A low speed clock is useful for wake up calls to poll the environment for activity. If the activity has a signal that can wake up theThe next stack processor that should be marketed is a 32 bit design... IF there is a genuine interest in selling the chip to a general market. It also needs a number of features lacking in most stack processors.
1) A crystal clock oscillator, if not two. 32.768 kHz and some high frequency such as 12.0 MHz. While an asynch processor is a great idea, applications are seldom time invariant, so need clocking at some point in the design. Possibly this clock can
requires level shifting.2) 3.3V I/Os, if not 5V tolerant. Having to add parts to a design to provide level shifting is an absurdity in today's MCU world. With the vast majority of MCUs having 3.3V I/O capability, if not 5V tolerance, it is hard for a device to compete which
of embedded programmers use. There must be the debugging tools the world is used to. Such a chip must allow the world to use it, the way *they* are accustomed to working. The conventional Forth programming paradigm is too radical, and too crude (compared3) There are others, but this is the real issue with a stack processor... C based development tools. Yes, this may be blasphemy in this group, but if an MCU is going to succeed in the world today, it must support the language that the vast majority
Just my opinion with a few facts.
Second this. In addition, minimal power consumption, idle states, and flexible interrupt
handling capabilities could be some more selling points. Since Forth code can be extremely
compact, memories can be held relatively small
.
Still one would have to be able to beat eg Arduino Pico.
Leave out such "annoying edge computing facilities", one would have to race against atiny4.
Guess who'll win.
Thanks for that. I'll have to have a look at the AVR stuff. The
thing we offer is better performance for less transistors on the
ground, equalling less cost. [2 typos fixed, I think]
I would be interested in finding out about anything that can clearly
best that, or match it.
Commercially myself, I am more interested in simple printed circuits
at 2.5 micro to 1 micron. I think I finally have a clean room
solution to do this at home. That's more doable on an individual's
level.
Facts, hardly anything you said is not accommodated in what I'm
saying. You can make it level tolerant. The starting point is not
the finish. How many of those 4 bit CPU's are C processors.
I do like AVR8 but really it's a legacy design.
gnuarm.del...@gmail.com schrieb am Mittwoch, 8. Juni 2022 um 14:52:08 UTC+2:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
addressed. If the Mup21 was a good CPU chip, why is it no longer in production? Without an understanding of that, why do you think a redo would be successful?I saw your last post, but let me respond to this one. There have been many stack chips, even if virtually no Forth chips (a pedantic distinction, I know). You seem to want to fix a problem in some previous chip without indicating what needs to be
determinant. 32 bit processors have come down in price enough to dominate in nearly every other application space. 16 bit processors are actually becoming less popular, both in percentages of design wins and numbers of devices shipped. There is no reasonI also don't understand the interest in a 16 bit chip. In the microprocessor world, nearly all devices are 4/8, 16 or 32 bits. Over time the trends which have emerged are that 4/8 bit MCUs are still dominant in low end applications where cost is the
be included in an external interface, but if no such interface exists, then a high speed clock is needed in the chip. A low speed clock is useful for wake up calls to poll the environment for activity. If the activity has a signal that can wake up theThe next stack processor that should be marketed is a 32 bit design... IF there is a genuine interest in selling the chip to a general market. It also needs a number of features lacking in most stack processors.
1) A crystal clock oscillator, if not two. 32.768 kHz and some high frequency such as 12.0 MHz. While an asynch processor is a great idea, applications are seldom time invariant, so need clocking at some point in the design. Possibly this clock can
requires level shifting.2) 3.3V I/Os, if not 5V tolerant. Having to add parts to a design to provide level shifting is an absurdity in today's MCU world. With the vast majority of MCUs having 3.3V I/O capability, if not 5V tolerance, it is hard for a device to compete which
of embedded programmers use. There must be the debugging tools the world is used to. Such a chip must allow the world to use it, the way *they* are accustomed to working. The conventional Forth programming paradigm is too radical, and too crude (compared3) There are others, but this is the real issue with a stack processor... C based development tools. Yes, this may be blasphemy in this group, but if an MCU is going to succeed in the world today, it must support the language that the vast majority
Just my opinion with a few facts.
Second this. In addition, minimal power consumption, idle states, and flexible interrupt
handling capabilities could be some more selling points. Since Forth code can be extremely
compact, memories can be held relatively small
Still one would have to be able to beat eg Arduino Pico.
Leave out such "annoying edge computing facilities", one would have to race against atiny4.
Guess who'll win.
Wayne morellini <waynemo...@gmail.com> writes:
Facts, hardly anything you said is not accommodated in what I'mI couldn't read that very long post either (could you try shorter
saying. You can make it level tolerant. The starting point is not
the finish. How many of those 4 bit CPU's are C processors.
ones?), but there really are no 4 bit processors these days, except
maybe as sequencer cells inside ASIC's. The Padauk PA150 is a super
cheap 8 bit processor, costing 3 cents (0.03 USD) retail in small quantities. It can run C code. I was interested in it, but I figure I
can afford to splurge and get the more powerful 10 cent version.
The AVR is a step up from the Padauk and is also 8 bits and was designed
to run C. It has 32 8-bit registers and is a convenient compiler
target. THe Padauk is a minimal load-store design with a single
accumulator, so not that great for HLL's, but people do use it.
On Wednesday, June 8, 2022 at 10:12:52 AM UTC-4, minf...@arcor.de wrote:for ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
gnuarm.del...@gmail.com schrieb am Mittwoch, 8. Juni 2022 um 14:52:08 UTC+2:
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth
addressed. If the Mup21 was a good CPU chip, why is it no longer in production? Without an understanding of that, why do you think a redo would be successful?I saw your last post, but let me respond to this one. There have been many stack chips, even if virtually no Forth chips (a pedantic distinction, I know). You seem to want to fix a problem in some previous chip without indicating what needs to be
the determinant. 32 bit processors have come down in price enough to dominate in nearly every other application space. 16 bit processors are actually becoming less popular, both in percentages of design wins and numbers of devices shipped. There is noI also don't understand the interest in a 16 bit chip. In the microprocessor world, nearly all devices are 4/8, 16 or 32 bits. Over time the trends which have emerged are that 4/8 bit MCUs are still dominant in low end applications where cost is
be included in an external interface, but if no such interface exists, then a high speed clock is needed in the chip. A low speed clock is useful for wake up calls to poll the environment for activity. If the activity has a signal that can wake up theThe next stack processor that should be marketed is a 32 bit design... IF there is a genuine interest in selling the chip to a general market. It also needs a number of features lacking in most stack processors.
1) A crystal clock oscillator, if not two. 32.768 kHz and some high frequency such as 12.0 MHz. While an asynch processor is a great idea, applications are seldom time invariant, so need clocking at some point in the design. Possibly this clock can
which requires level shifting.2) 3.3V I/Os, if not 5V tolerant. Having to add parts to a design to provide level shifting is an absurdity in today's MCU world. With the vast majority of MCUs having 3.3V I/O capability, if not 5V tolerance, it is hard for a device to compete
of embedded programmers use. There must be the debugging tools the world is used to. Such a chip must allow the world to use it, the way *they* are accustomed to working. The conventional Forth programming paradigm is too radical, and too crude (compared3) There are others, but this is the real issue with a stack processor... C based development tools. Yes, this may be blasphemy in this group, but if an MCU is going to succeed in the world today, it must support the language that the vast majority
Just my opinion with a few facts.
Second this. In addition, minimal power consumption, idle states, and flexible interruptThe "idle states" are probably not required, IF a design technique is used like the F18, where CPUs are automatically stopped when there is no further data for them to process. The "idle state" is basically OFF with no penalty for restarting.
handling capabilities could be some more selling points. Since Forth code can be extremely
compact, memories can be held relatively small
Forth code can be smaller, but "extremely" is a bit of an overstatement I think.
Still one would have to be able to beat eg Arduino Pico.Beat it in what regard exactly? Do you mean power consumption?
Leave out such "annoying edge computing facilities", one would have to race against atiny4.Not sure what you are saying with this.
Guess who'll win.
gnuarm.del...@gmail.com schrieb am Donnerstag, 9. Juni 2022 um 09:15:10 UTC+2:for ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Wednesday, June 8, 2022 at 10:12:52 AM UTC-4, minf...@arcor.de wrote:
gnuarm.del...@gmail.com schrieb am Mittwoch, 8. Juni 2022 um 14:52:08 UTC+2:
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth
addressed. If the Mup21 was a good CPU chip, why is it no longer in production? Without an understanding of that, why do you think a redo would be successful?I saw your last post, but let me respond to this one. There have been many stack chips, even if virtually no Forth chips (a pedantic distinction, I know). You seem to want to fix a problem in some previous chip without indicating what needs to be
the determinant. 32 bit processors have come down in price enough to dominate in nearly every other application space. 16 bit processors are actually becoming less popular, both in percentages of design wins and numbers of devices shipped. There is noI also don't understand the interest in a 16 bit chip. In the microprocessor world, nearly all devices are 4/8, 16 or 32 bits. Over time the trends which have emerged are that 4/8 bit MCUs are still dominant in low end applications where cost is
can be included in an external interface, but if no such interface exists, then a high speed clock is needed in the chip. A low speed clock is useful for wake up calls to poll the environment for activity. If the activity has a signal that can wake upThe next stack processor that should be marketed is a 32 bit design... IF there is a genuine interest in selling the chip to a general market. It also needs a number of features lacking in most stack processors.
1) A crystal clock oscillator, if not two. 32.768 kHz and some high frequency such as 12.0 MHz. While an asynch processor is a great idea, applications are seldom time invariant, so need clocking at some point in the design. Possibly this clock
which requires level shifting.2) 3.3V I/Os, if not 5V tolerant. Having to add parts to a design to provide level shifting is an absurdity in today's MCU world. With the vast majority of MCUs having 3.3V I/O capability, if not 5V tolerance, it is hard for a device to compete
majority of embedded programmers use. There must be the debugging tools the world is used to. Such a chip must allow the world to use it, the way *they* are accustomed to working. The conventional Forth programming paradigm is too radical, and too crude (3) There are others, but this is the real issue with a stack processor... C based development tools. Yes, this may be blasphemy in this group, but if an MCU is going to succeed in the world today, it must support the language that the vast
Just my opinion with a few facts.
Second this. In addition, minimal power consumption, idle states, and flexible interruptThe "idle states" are probably not required, IF a design technique is used like the F18, where CPUs are automatically stopped when there is no further data for them to process. The "idle state" is basically OFF with no penalty for restarting.
handling capabilities could be some more selling points. Since Forth code can be extremely
compact, memories can be held relatively small
Forth code can be smaller, but "extremely" is a bit of an overstatement I think.A hypothetical Forth chip could IMO only be successful in the embedded or microcontroller domain.
Still one would have to be able to beat eg Arduino Pico.Beat it in what regard exactly? Do you mean power consumption?
Leave out such "annoying edge computing facilities", one would have to race against atiny4.Not sure what you are saying with this.
Guess who'll win.
There flexible IOs to communicate with the outer world (edges) are essential. This has to be reflected
in the chip design.
For instance the Arduino Pico offers https://arduino-pico.readthedocs.io/en/latest/pins.html
Or of the atiny family the small atiny4 https://www.microchip.com/en-us/product/ATTINY4
Yeah, the GA144 could emulate pretty much any common peripheral with the
I/O CPUs. They seem to have munged the dedicated memory interface so it >would not mate well with DRAM, so instead they developed apps with very >expensive and power hungry static RAM (also very low density and on the
way out). I tried to see if I could figure out how to use it with DRAM
(you don't actually have to clock DRAM as fast as it can possibly run,
you just need to meet all the timing specs). But at one point I needed
to understand the timing between the three CPUs that were handling the >interface, including the timing of the internal comms between them.
They would not release any internal timing info. I was told to "play"
with the chips to see if the design would work.
waste my time
Rick C.--
Virtually every coffee maker, microwave oven, and remote control has a 4 bi= >t processor in it.
There's only one reason why anyone would use a 4 bit MC=
U, because it costs less.
Rick C <gnuarm.del...@gmail.com> writes:
Virtually every coffee maker, microwave oven, and remote control has a 4 bi= >t processor in it.
Citation needed
When googling for "4-bit microcontroller market", I only find
references that don't give 4-bit market share. E.g., <https://www.verifiedmarketresearch.com/product/microcontroller-market/>
says
|[A microcontroller] is capable of processing a word length that ranges |between 4-bit up to 64-bit.
But when it comes to market segmentation, one of the segmentations
they provide is into 8-bit, 16-bit, and 32-bit microcontrollers. It
seems that 4-bit microcontrollers are not longer relevant.
There's only one reason why anyone would use a 4 bit MC=
U, because it costs less.
These days, the main reason for 8-bit probably is that you don't need
to redesign the thing. And for 4-bit the prime is sufficiently far in
the past that even that is not a good reason for most uses: the
designs with 4-bit controllers from maybe the 1980s have been
redesigned with 8-bit or bigger microcontrollers in the meantime.
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/about
Wayne morellini <waynemo...@gmail.com> writes:
Commercially myself, I am more interested in simple printed circuitsI couldn't read that whole long post, but tons of very interesting chips
at 2.5 micro to 1 micron. I think I finally have a clean room
solution to do this at home. That's more doable on an individual's
level.
were done in 2.5 micron and larger sizes. The Mead-Conway revolution
happened in the 3 to 5 micron era. There was a #homecmos channel on
freenode for a while that was trying to do stuff at, iirc, something
like 12 microns. If you have an affordable way to make 2.5 micron chips
at home, that is really quite revolutionary, especially if you can do
mixed signal.
As for a MISC-like chip though, yes of course there is skepticism: who
would want a thing like that, at least as a separate chip rather than a
macro cell? DIY satisfaction (and I'm all in favor of that) seems like
the main reason to make it, unless you've got some pretty clear and quantitative claims about how it would outperform conventional chips at
some meaningful application.
This morning I was thinking it would be interesting to have a chip with
wide but single core SMT, to allow super fast coroutine switching
without having to save and restore registers, sort of like a Forth multitasker (that only has a couple of registers to save and restore, so
the task switcher is fast), but allowing something more like
conventional OS's and compilers, which do use lots of registers. That
might be more interesting than a MISC chip. It could be a RISC-V with
a special instruction added to call another context, and there might be
8 or so contexts available in a core, sort of like the Parallax
Propeller.
Wayne morellini <waynemo...@gmail.com> writes:Your following post wasn't that much shorter Paul. Time is passing on.
Facts, hardly anything you said is not accommodated in what I'mI couldn't read that very long post either (could you try shorter
saying. You can make it level tolerant. The starting point is not
the finish. How many of those 4 bit CPU's are C processors.
ones?), but there really are no 4 bit processors these days, except
maybe as sequencer cells inside ASIC's. The Padauk PA150 is a super
cheap 8 bit processor, costing 3 cents (0.03 USD) retail in small
quantities. It can run C code. I was interested in it, but I figure I
can afford to splurge and get the more powerful 10 cent version.
The AVR is a step up from the Padauk and is also 8 bits and was designed
to run C. It has 32 8-bit registers and is a convenient compiler
target. THe Padauk is a minimal load-store design with a single
accumulator, so not that great for HLL's, but people do use it.
Wayne morellini <waynemo...@gmail.com> writes:
Commercially myself, I am more interested in simple printed circuitsI couldn't read that whole long post, but tons of very interesting chips were done in 2.5 micron and larger sizes. The Mead-Conway revolution happened in the 3 to 5 micron era. There was a #homecmos channel on
at 2.5 micro to 1 micron. I think I finally have a clean room
solution to do this at home. That's more doable on an individual's
level.
freenode for a while that was trying to do stuff at, iirc, something
like 12 microns. If you have an affordable way to make 2.5 micron chips
at home, that is really quite revolutionary, especially if you can do
mixed signal.
As for a MISC-like chip though, yes of course there is skepticism: who
would want a thing like that, at least as a separate chip rather than a macro cell? DIY satisfaction (and I'm all in favor of that) seems like
the main reason to make it, unless you've got some pretty clear and quantitative claims about how it would outperform conventional chips at
some meaningful application.
This morning I was thinking it would be interesting to have a chip with
wide but single core SMT, to allow super fast coroutine switching
without having to save and restore registers, sort of like a Forth multitasker (that only has a couple of registers to save and restore, so
the task switcher is fast), but allowing something more like
conventional OS's and compilers, which do use lots of registers. That
might be more interesting than a MISC chip. It could be a RISC-V with
a special instruction added to call another context, and there might be
8 or so contexts available in a core, sort of like the Parallax
Propeller.
Wayne morellini <waynemo...@gmail.com> writes:Paul the transistors are consuming the energy doing the processing. That's how the GA gets its advantage
Thanks for that. I'll have to have a look at the AVR stuff. The
thing we offer is better performance for less transistors on the
ground, equalling less cost. [2 typos fixed, I think]
ALU transistor costs in MCU's today are almost irrelevant compared to transistors driving i/o pins, transistors in memory arrays, and mixed
signal peripherals. Raspberry Pi Foundation decided to make its own MCU
a few years ago (the RP2040). It sells for $1 retail, it has two 32-bit
ARM cores, 264KB of ram, its own weird programmable digital i/o
peripheral, A-D converters, timers, etc. The ARM cores are tiny
compared to the ram array and the other stuff. If you decreased the transistor count in the ARM cores to zero by replacing ARM with a stack architecture, the chip cost wouldn't decrease by enough for anyone to
care.
I wouldn't pay attention to AVR when making comparisons to your MISC
part. I do like AVR8 but really it's a legacy design. You have to
compare to ARM, RISC-V, and the like.
I would be interested in finding out about anything that can clearlyBefore anyone can suggesting things better than X, you first have to
best that, or match it.
tell us what X is. X is your MISC design and V is the conventional
thing (RISC-V, say). Does X use fewer transistors? Don't care (see
above), ALU transistors are free. If your rationale involves transistor count then you have already declared irrelevance. Does X provide more operations per unit of energy? That is much more interesting. GA at
least focused on that. But you have to give us some numbers for X,
since you are the one who knows anything about it.
We do have the GA chips, which were done by smart people. Do they have
the advantage that GA claimed? I am not sure, but can provisionally
accept "yes". Is the advantage so overwhelming as to get people to use GA144's in favor of the stuff they are used to? Apparently not.
Jurgen Pitaske <jpit...@gmail.com> writes:
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit")
on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit")
on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
Features
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
Rick C <gnuarm.del...@gmail.com> writes:
Virtually every coffee maker, microwave oven, and remote control has a 4 bi=
t processor in it.
Citation needed
When googling for "4-bit microcontroller market", I only find
references that don't give 4-bit market share. E.g., <https://www.verifiedmarketresearch.com/product/microcontroller-market/> says
|[A microcontroller] is capable of processing a word length that ranges |between 4-bit up to 64-bit.
But when it comes to market segmentation, one of the segmentations
they provide is into 8-bit, 16-bit, and 32-bit microcontrollers. It
seems that 4-bit microcontrollers are not longer relevant.
There's only one reason why anyone would use a 4 bit MC=
U, because it costs less.
These days, the main reason for 8-bit probably is that you don't need
to redesign the thing. And for 4-bit the prime is sufficiently far in
the past that even that is not a good reason for most uses: the
designs with 4-bit controllers from maybe the 1980s have been
redesigned with 8-bit or bigger microcontrollers in the meantime.
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit")
on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
There is literally no reason on earth to use an 8-bit processor in a
$20 coffee maker.
The software running on the 4 bit processor is probably not more than
200 instructions. Zero reason to use anything other than the cheapest
4-bit MCU they can find.
They also use 4-bit MCUs in remote controls, microwave ovens and many,
many other relatively low cost appliances and devices.
Rick C <gnuarm.del...@gmail.com> writes:
There is literally no reason on earth to use an 8-bit processor in aI have a $20 coffee maker and afaict it has no processor at all. There
$20 coffee maker.
is a tank of water, a heating element, and a thermostat. You close the
cover over the water tank, which seals it with a gasket, and then turn
on the heat. Steam builds up in the tank which pushes the water up
through a siphon-like tube where it drops through the coffee grounds
into the carafe. When all the water is pushed out that way, the heating element gets above 100C, and a thermostat shuts off the power. Rice
cookers work roughly the same way.
The software running on the 4 bit processor is probably not more thanGiven that a flash programmable 8 bit cpu is 3 cents retail (i.e. you
200 instructions. Zero reason to use anything other than the cheapest
4-bit MCU they can find.
can order 100 of them for 3 bucks from lcsc.com), I don't see how much
they can save by using 4 bitters. Stuff made in really huge quantity
likely uses ASIC's rather than MCU's.
They also use 4-bit MCUs in remote controls, microwave ovens and many,I'd want to see a teardown before being convinced of this. It has to be
many other relatively low cost appliances and devices.
a model introduced in the past 10 years, not something from the 1980's,
since we're talking about new designs.
The TV remote my mom uses has voice recognition. It wouldn't surprise
me if it has a 32 bit cpu.
The TV remote my mom uses has voice recognition. It wouldn't surprise
me if it has a 32 bit cpu.
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit")
on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
I like to see somebody try to make a Linux desktop computer out of these, like people say that problems can be broken down into smaller bit sizes. But, hie would they fix the resulting bias worth of computer on the desk alongside an equivalent top endPC configuration. Imagine the pipelining you would need. You might be able to run alongside the data as it propagates down to the other end of the bus, as it goes through a million deep pipeline. I can actually do this with normal chips and beat the
On Thursday, June 9, 2022 at 4:27:44 AM UTC-4, Anton Ertl wrote:
Rick C <gnuarm.del...@gmail.com> writes:=20bi=3D=20
Virtually every coffee maker, microwave oven, and remote control has a 4=
t processor in it.=20=20
Citation needed=20
Yeah, wait here while I get that for you.=20
When googling for "4-bit microcontroller market", I only find=20
references that don't give 4-bit market share. E.g.,=20
<https://www.verifiedmarketresearch.com/product/microcontroller-market/>= >=20
says=20
=20
|[A microcontroller] is capable of processing a word length that ranges= >=20
|between 4-bit up to 64-bit.=20
=20
But when it comes to market segmentation, one of the segmentations=20
they provide is into 8-bit, 16-bit, and 32-bit microcontrollers. It=20
seems that 4-bit microcontrollers are not longer relevant.=20
Not sure who "they" is.
I don't put all my faith in any individual report.=
These reports have a target audience. If you understand the market you w=
ould understand why 4-bit processors are not very well reported. =20
These days, the main reason for 8-bit probably is that you don't need=20
to redesign the thing. And for 4-bit the prime is sufficiently far in=20
the past that even that is not a good reason for most uses: the=20
designs with 4-bit controllers from maybe the 1980s have been=20
redesigned with 8-bit or bigger microcontrollers in the meantime.=20
You don't need to redesign what "thing"?
There are NEW designs that use 4-bit MCUs...
You don't have to belie=
ve it, but it is reality.
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
In article <9b86fdb3-8c42-4954...@googlegroups.com>,
Jurgen Pitaske <jpit...@gmail.com> wrote:
<SNIP>
DescriptionThis is certainly a 3 eurocent processor.
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
Groetjes Albert
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst
There is literally no reason on earth to use an 8-bit processor in a $20 coffee maker. Every > > >piece of this appliance is cost optimized. The software running on the 4 bit processor is >probably not more than 200 instructions. Zero reason to useanything other than the cheapest >4-bit MCU they can find.
On Friday, 10 June 2022 at 03:20:13 UTC+1, Wayne morellini wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit") on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
end PC configuration. Imagine the pipelining you would need. You might be able to run alongside the data as it propagates down to the other end of the bus, as it goes through a million deep pipeline. I can actually do this with normal chips and beat theI like to see somebody try to make a Linux desktop computer out of these, like people say that problems can be broken down into smaller bit sizes. But, hie would they fix the resulting bias worth of computer on the desk alongside an equivalent top
No, this was for Anton to find this 4 bit documentation at EM, and for others who want to see products out there now.
4 bit is useful where needed, and mostly for cost and volume.
Any Forth chip should have at least 16 bit
Anyway, I'm going have to leave you guys. My father passed away unexpectedly last month, after not been able to see him more than a handful.of times I'm the last 6 months due to covid restrictions, and not been able to talk with him on the phone due tohis hearing. Now, there are unnecessary petty will games, and the stress is getting too much, after a number of other overlapping stressful things trying to take advantage of my disability, which has partly lifted a bit for now, to address this. Anybody
But keep up the conversation, it almost seems like people favour a 4 bit misc version do far. :)
Thanks.
Anyway, I'm going have to leave you guys. My father passed away unexpectedly last month, after not been able to see him more than a handful.of times I'm the last 6 months due to covid restrictions, and not been able to talk with him on the phone due tohis hearing. Now, there are unnecessary petty will games, and the stress is getting too much, after a number of other overlapping stressful things trying to take advantage of my disability, which has partly lifted a bit for now, to address this. Anybody
But keep up the conversation, it almost seems like people favour a 4 bit misc version do far. :)
Thanks.
On Friday, June 10, 2022 at 2:24:30 AM UTC-4, jpit...@gmail.com wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Friday, 10 June 2022 at 03:20:13 UTC+1, Wayne morellini wrote:
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit") on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
end PC configuration. Imagine the pipelining you would need. You might be able to run alongside the data as it propagates down to the other end of the bus, as it goes through a million deep pipeline. I can actually do this with normal chips and beat theI like to see somebody try to make a Linux desktop computer out of these, like people say that problems can be broken down into smaller bit sizes. But, hie would they fix the resulting bias worth of computer on the desk alongside an equivalent top
agnostic. It can have any size instruction (if you recode them to suit), data path, address path. There is no direct dependence between the three.No, this was for Anton to find this 4 bit documentation at EM, and for others who want to see products out there now.
4 bit is useful where needed, and mostly for cost and volume.That is simply an opinion. There's no reason why a 8 or even 4 bit CPU can't be a stack processor and be useful to those who build things with 4 and 8 bit CPUs. Heck, it might turn out that a 5 bit CPU is preferable. My stack processor is data path
Any Forth chip should have at least 16 bit
--
Rick C.
++- Get 1,000 miles of free Supercharging
++- Tesla referral code - https://ts.la/richard11209
On Friday, 10 June 2022 at 03:20:13 UTC+1, Wayne morellini wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit") on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
end PC configuration. Imagine the pipelining you would need. You might be able to run alongside the data as it propagates down to the other end of the bus, as it goes through a million deep pipeline. I can actually do this with normal chips and beat theI like to see somebody try to make a Linux desktop computer out of these, like people say that problems can be broken down into smaller bit sizes. But, hie would they fix the resulting bias worth of computer on the desk alongside an equivalent top
No, this was for Anton to find this 4 bit documentation at EM, and for others who want to see products out there now.
4 bit is useful where needed, and mostly for cost and volume.
Any Forth chip should have at least 16 bit
Rick C <gnuarm.del...@gmail.com> writes:
On Thursday, June 9, 2022 at 4:27:44 AM UTC-4, Anton Ertl wrote:
Rick C <gnuarm.del...@gmail.com> writes:=20bi=3D=20
Virtually every coffee maker, microwave oven, and remote control has a 4=
t processor in it.=20=20
Citation needed=20
Yeah, wait here while I get that for you.=20
When googling for "4-bit microcontroller market", I only find=20
references that don't give 4-bit market share. E.g.,=20
<https://www.verifiedmarketresearch.com/product/microcontroller-market/>= >=20
says=20
=20
|[A microcontroller] is capable of processing a word length that ranges= >=20
|between 4-bit up to 64-bit.=20
=20
But when it comes to market segmentation, one of the segmentations=20
they provide is into 8-bit, 16-bit, and 32-bit microcontrollers. It=20
seems that 4-bit microcontrollers are not longer relevant.=20
Not sure who "they" is.In this case https://www.verifiedmarketresearch.com
I don't put all my faith in any individual report.=
These reports have a target audience. If you understand the market you w=
ould understand why 4-bit processors are not very well reported. =20
So you claim that 4-bit processors are relevant,
don't want to provide
any evidence for that,
dismiss evidence against it out of hand,
then say that one has to be in the know to understand why there is no evidence for your claims. Maybe you should found the church of 4-bit processing.
These days, the main reason for 8-bit probably is that you don't need=20 >> to redesign the thing. And for 4-bit the prime is sufficiently far in=20 >> the past that even that is not a good reason for most uses: the=20
designs with 4-bit controllers from maybe the 1980s have been=20
redesigned with 8-bit or bigger microcontrollers in the meantime.=20
You don't need to redesign what "thing"?Whatever things still use 8-bit MCUs. Maybe "every coffee maker,
microwave oven, and remote control".
There are NEW designs that use 4-bit MCUs...That's what you claim.
You don't have to belie=You don't provide any evidence, so the only choice is to believe or
ve it, but it is reality.
not to believe.
On Friday, 10 June 2022 at 09:41:01 UTC+1, none albert wrote:
In article <9b86fdb3-8c42-4954...@googlegroups.com>,
Jurgen Pitaske <jpit...@gmail.com> wrote:
<SNIP>
DescriptionThis is certainly a 3 eurocent processor.
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
Groetjes AlbertNot many know EM in Switzerland or have been there.
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst
Wikipedia gives a nice overview of the 4 bit history with more known names. How many of these are still available is a question I am not really interested in.
I have my own MISC processor 16 bit in FPGA
A similar version had been done about 20 years ago as ASIC as a student project.
It uses minimum ressources, so not optimised for speed, but flexible
List of 4-bit processors at wikipedia, copy and paste from there
Intel C4004
NEC D63GS 4-bit microcontroller
NEC D63GS: a 4-bit microcontroller for infrared remote control transmission card-edge PCB
Olympia CD700 Desktop Calculator using the National Semiconductor MAPS MM570X bit-serial 4-bit microcontroller
16-pin DIP
National Semiconductor MM5700CA/D bit-serial 4-bit microcontroller
Intel 4004
Intel 4040
TMS 1000
Atmel MARC4 core[26][27] – (discontinued: "Last ship date: 7 March 2015"[28])
Samsung S3C7 (KS57 Series) 4-bit microcontrollers (RAM: 512 to 5264 nibbles, 6 MHz clock)
Toshiba TLCS-47 series
HP Saturn
NEC μPD75X
NEC μCOM-4
NEC (now Renesas) μPD612xA (discontinued), μPD613x, μPD6x[23][29] and μPD1724x[30] infrared remote control transmitter microcontrollers[31][32]
EM Microelectronic-Marin EM6600 family,[33] EM6580,[34][35] EM6682,[36] etc. Epson S1C63 family
National Semiconductor "COPS I" and "COPS II" ("COP400") 4-bit microcontroller families[37]
National Semiconductor MAPS MM570X
Sharp SM590/SM591/SM595[38]: 26–34
Sharp SM550/SM551/SM552[38]: 36–48
Sharp SM578/SM579[38]: 49–64
Sharp SM5E4[38]: 65–74
Sharp LU5E4POP[38]: 75–82
Sharp SM5J5/SM5J6[38]: 83–99
Sharp SM530[38]: 100–109
Sharp SM531[38]: 110–118
Sharp SM500[38]: 119–127 (ROM 1197×8 bit, RAM 40×4 bit, a divider and 56-segment LCD driver circuit)
Sharp SM5K1[38]: 128–140
Sharp SM4A[38]: 141–148
Sharp SM510[38]: 149–158 (ROM 2772×8 bit, RAM 128×4 bit, a divider and 132-segment LCD driver circuit)
Sharp SM511/SM512[38]: 159–171 (ROM 4032×8 bit, RAM 128/142×4 bit, a divider and 136/200-segment LCD driver circuit)
Sharp SM563[38]: 172–186
On Friday, 10 June 2022 at 21:35:59 UTC+1, gnuarm.del...@gmail.com wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Friday, June 10, 2022 at 2:24:30 AM UTC-4, jpit...@gmail.com wrote:
On Friday, 10 June 2022 at 03:20:13 UTC+1, Wayne morellini wrote:
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit")
on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
top end PC configuration. Imagine the pipelining you would need. You might be able to run alongside the data as it propagates down to the other end of the bus, as it goes through a million deep pipeline. I can actually do this with normal chips and beatI like to see somebody try to make a Linux desktop computer out of these, like people say that problems can be broken down into smaller bit sizes. But, hie would they fix the resulting bias worth of computer on the desk alongside an equivalent
agnostic. It can have any size instruction (if you recode them to suit), data path, address path. There is no direct dependence between the three.No, this was for Anton to find this 4 bit documentation at EM, and for others who want to see products out there now.
4 bit is useful where needed, and mostly for cost and volume.That is simply an opinion. There's no reason why a 8 or even 4 bit CPU can't be a stack processor and be useful to those who build things with 4 and 8 bit CPUs. Heck, it might turn out that a 5 bit CPU is preferable. My stack processor is data path
Any Forth chip should have at least 16 bit
--
Rick C.
++- Get 1,000 miles of free SuperchargingYou have an opinion, I have an opinion.
++- Tesla referral code - https://ts.la/richard11209
My opinion about 4 bit very special and not for the general public seems to be supported by reality and market overviews.
Apart from this, anybody can have any processor.
Let's have as many Forth processors as there are Forths out there.
And none is successful. As it is now.
Why do you think a stack processor has to have a 16 bit data path?
On Friday, June 10, 2022 at 2:24:30 AM UTC-4, jpit...@gmail.com wrote:potentially so confusing! Sorry, hard to resist giving a misconstrued reply that mistakes everything, like I get around here. :)
On Friday, 10 June 2022 at 03:20:13 UTC+1, Wayne morellini wrote:
On Friday, June 10, 2022 at 3:14:53 AM UTC+10, jpit...@gmail.com wrote:
On Thursday, 9 June 2022 at 16:06:03 UTC+1, Anton Ertl wrote:
Jurgen Pitaske <jpit...@gmail.com> writes:see the data sheet: https://www.emmicroelectronic.com/sites/default/files/products/datasheets/em6607_ds.pdf
I at least found one with a quick google: >https://www.emmicroelectronic.com/product >https://www.emmicroelectronic.com/aboutNo mention of "4 bit" or "4-bit" (except in the context of "64 bit") on these pages (I did not follow any links).
- anton
--
M. Anton Ertl http://www.complang.tuwien.ac.at/anton/home.html comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
New standard: https://forth-standard.org/
EuroForth 2022: http://www.euroforth.org/ef22/cfp.html
FeaturesSo, are you saying we should use this chip instead, or 24 pin 4 bit misc chip, or piggyback this ochip on a misc chip to give a high energy mode, or get EM's advanced low powered CMOS process, or EM should take over misc and buy out GA. It's
‰ Low Power typical 1.8µA active mode
typical 0.5µA standby mode
typical 0.1µA sleep mode
@ 1.5V, 32kHz, 25 °C
‰ Low Voltage 1.2 to 3.3 V
‰ ROM 2k × 16 (Mask Programmed)
‰ RAM 96 × 4 (User Read/Write)
‰ 2 clocks per instruction cycle
‰ RISC architecture
‰ 5 software configurable 4-bit ports
‰ 1 High drive output port
‰ Up to 20 inputs (5 ports)
‰ Up to 16 outputs (4 ports)
‰ buzzer three tone
‰ Serial Write buffer – SWB
‰ Supply Voltage level detection (SVLD).
‰ Analogue and timer watchdog
‰ 8 bit timer / event counter
‰ Internal interrupt sources (timer, event counter,
prescaler)
‰ External interrupt sources (portA + portC)
Description
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
end PC configuration. Imagine the pipelining you would need. You might be able to run alongside the data as it propagates down to the other end of the bus, as it goes through a million deep pipeline. I can actually do this with normal chips and beat theI like to see somebody try to make a Linux desktop computer out of these, like people say that problems can be broken down into smaller bit sizes. But, hie would they fix the resulting bias worth of computer on the desk alongside an equivalent top
agnostic. It can have any size instruction (if you recode them to suit), data path, address path. There is no direct dependence between the three.No, this was for Anton to find this 4 bit documentation at EM, and for others who want to see products out there now.
4 bit is useful where needed, and mostly for cost and volume.That is simply an opinion. There's no reason why a 8 or even 4 bit CPU can't be a stack processor and be useful to those who build things with 4 and 8 bit CPUs. Heck, it might turn out that a 5 bit CPU is preferable. My stack processor is data path
Any Forth chip should have at least 16 bit
--
Rick C.
++- Get 1,000 miles of free Supercharging
++- Tesla referral code - https://ts.la/richard11209
The EM6607 is a single chip ... 4-bit microcontroller.This is certainly a 3 eurocent processor.
Indeed. I simply have been paying attention to that world long enough
to know. You seem to be a total newbie.
So, are you claiming that 4-bit processors are no longer designed into
new products?
Show me a simple product, that could make use of a 4-bit processor, is
made in qty over a million and has an 8-bit processor.
Then again very few true Forth or stack designs under 200 LUTs?
Anyway, I'm going have to leave you guys. My father passed away
unexpectedly last month, after not been able to see him more than a handful.of times I'm the last 6 months due to covid restrictions, and
not been able to talk with him on the phone due to his hearing.
albert@cherry.(none) (albert) writes:a
I like how it comes with 2kx16 bits (4k bytes) of code space (maskThe EM6607 is a single chip ... 4-bit microcontroller.This is certainly a 3 eurocent processor.
rom). Somehow GA thought that 64 18-bit words was enough.
It has 96 nibbles of ram, no mention of a hardware stack. In a Forth
chip with 4-bit data, how is the return stack supposed to be stored?
Rick C <gnuarm.del...@gmail.com> writes:
Why do you think a stack processor has to have a 16 bit data path?If we're talking about Forth, we usually expect data cells to be able to hold addresses. 4 bits is awfully small for that. 8 bits, maybe.
Also, I had thought we were talking about MCU's, which are packaged, multipurpose programmable devices that control the rest of the product through i/o pins. Stuff inside an ASIC or FPGA doesn't count as that.
Rick C <gnuarm.del...@gmail.com> writes:
Indeed. I simply have been paying attention to that world long enoughI would say paying attention to the world for a long time means you witnessed some stuff that happened in the distant past, which can be the basis of much wisdom. But it doesn't make you an authority about what
to know. You seem to be a total newbie.
is or isn't happening in the present.
So, are you claiming that 4-bit processors are no longer designed intoI don't think such a claim was made. Only that there hasn't been
new products?
convincing evidence shown against it.
Show me a simple product, that could make use of a 4-bit processor, is made in qty over a million and has an 8-bit processor.Every PC keyboard including in the pre-USB era had an 8035-style
processor even though it could have used a 4 bitter. In USB keyboards,
using 4 bitters may be unfeasible so I won't count them.
On Friday, June 10, 2022 at 8:42:34 AM UTC-4, jpit...@gmail.com wrote:
On Friday, 10 June 2022 at 09:41:01 UTC+1, none albert wrote:
In article <9b86fdb3-8c42-4954...@googlegroups.com>,
Jurgen Pitaske <jpit...@gmail.com> wrote:
<SNIP>
DescriptionThis is certainly a 3 eurocent processor.
The EM6607 is a single chip low power, mask
programmed CMOS 4-bit microcontroller. It contains
ROM, RAM, watchdog timer, oscillation detection circuit,
combined timer / event counter, prescaler, voltage level
detector and a number of clock functions. Its low voltage
and low power operation make it the most suitable
controller for battery, stand alone and mobile equipment.
The EM6607 microcontroller is manufactured using EM’s
Advanced Low Power CMOS Process.
In 24 Pin package it is direct replacement for EM6603.
Groetjes AlbertNot many know EM in Switzerland or have been there.
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst
Wikipedia gives a nice overview of the 4 bit history with more known names.
How many of these are still available is a question I am not really interested in.
I have my own MISC processor 16 bit in FPGA
A similar version had been done about 20 years ago as ASIC as a student project.
It uses minimum ressources, so not optimised for speed, but flexible
List of 4-bit processors at wikipedia, copy and paste from there
The Intel 4004 was the first single chip CPU ever made, no?Intel C4004
NEC D63GS 4-bit microcontroller
NEC D63GS: a 4-bit microcontroller for infrared remote control transmissionWikipedia is not very useful on this issue. This is largely an historical list of devices made by mainstream companies. I see a mention of a remote control MCU. That might be current and any of the Sharp devices may be current.
card-edge PCB
Olympia CD700 Desktop Calculator using the National Semiconductor MAPS MM570X bit-serial 4-bit microcontroller
16-pin DIP
National Semiconductor MM5700CA/D bit-serial 4-bit microcontroller
Intel 4004
Intel 4040
TMS 1000
Atmel MARC4 core[26][27] – (discontinued: "Last ship date: 7 March 2015"[28])
Samsung S3C7 (KS57 Series) 4-bit microcontrollers (RAM: 512 to 5264 nibbles, 6 MHz clock)
Toshiba TLCS-47 series
HP Saturn
NEC μPD75X
NEC μCOM-4
NEC (now Renesas) μPD612xA (discontinued), μPD613x, μPD6x[23][29] and μPD1724x[30] infrared remote control transmitter microcontrollers[31][32]
EM Microelectronic-Marin EM6600 family,[33] EM6580,[34][35] EM6682,[36] etc.
Epson S1C63 family
National Semiconductor "COPS I" and "COPS II" ("COP400") 4-bit microcontroller families[37]
National Semiconductor MAPS MM570X
Sharp SM590/SM591/SM595[38]: 26–34
Sharp SM550/SM551/SM552[38]: 36–48
Sharp SM578/SM579[38]: 49–64
Sharp SM5E4[38]: 65–74
Sharp LU5E4POP[38]: 75–82
Sharp SM5J5/SM5J6[38]: 83–99
Sharp SM530[38]: 100–109
Sharp SM531[38]: 110–118
Sharp SM500[38]: 119–127 (ROM 1197×8 bit, RAM 40×4 bit, a divider and 56-segment LCD driver circuit)
Sharp SM5K1[38]: 128–140
Sharp SM4A[38]: 141–148
Sharp SM510[38]: 149–158 (ROM 2772×8 bit, RAM 128×4 bit, a divider and 132-segment LCD driver circuit)
Sharp SM511/SM512[38]: 159–171 (ROM 4032×8 bit, RAM 128/142×4 bit, a divider and 136/200-segment LCD driver circuit)
Sharp SM563[38]: 172–186
--
Rick C.
---- Get 1,000 miles of free Supercharging
---- Tesla referral code - https://ts.la/richard11209
??? What does the data path have to do with the return stack?
People like TI don't want =
to bother with 4-bit MCUs because the profit margins are too small for them=
to make a profit. =20
There are NEW designs that use 4-bit MCUs...That's what you claim.=20
So, are you claiming that 4-bit processors are no longer designed into new = >products?
Show me a simple product, that could make use of a 4-bit process=
or, is made in qty over a million and has an 8-bit processor. The coffee m= >aker is a perfect example. Can you show me an under $25 coffee maker that = >uses an 8-bit processor? They don't exist because the chip may only cost a=
penny more, but that's $10,000 in lost profits.
I'm not your search engine.
I'm also a bit busy.
The coffee maker is a perfect example. Can you show me an under $25
coffee maker that uses an 8-bit processor? They don't exist because
the chip may only cost a penny more,
Not sure how you can compare the two. The GA144 has an unlimited code
space since the storage is all external to the chip.
??? What does the data path have to do with the return stack? ....
An 8080 has 8 bit data paths, and 16 bit addresses. Why do you keep
getting confused about this?
And what if it costs 0 pennies more? In <https://bernd-paysan.de/b16-presentation.pdf> (from 2005), Bernd
Paysan gives the size of the (small) b16 in the XC035 process (a
0.35um (350nm) process) as 0.16mm^2. Intel and AMD were at 90nm by
that time. These days, my guess is that it takes 100x less area in a
process used now for embedded controllers. How much does 0.0016mm^2
in such a process cost? How much do you save by using, say 0.0000mm^2
(a 4-bit CPU will certainly need more area than that)? Will it save a
penny? I doubt it.
albert@cherry.(none) (albert) writes:The stack pointer is two bits, so likely the stack is 4 x 12b.
The EM6607 is a single chip ... 4-bit microcontroller.This is certainly a 3 eurocent processor.
I like how it comes with 2kx16 bits (4k bytes) of code space (mask
rom). Somehow GA thought that 64 18-bit words was enough.
It has 96 nibbles of ram, no mention of a hardware stack.
In a Forth
chip with 4-bit data, how is the return stack supposed to be stored?
Rick C <gnuarm.del...@gmail.com> writes:
??? What does the data path have to do with the return stack?In Forth, the words >R, R>, and R@ connect the two.
Rick C <gnuarm.del...@gmail.com> writes:
The coffee maker is a perfect example. Can you show me an under $25Anton made a similar point, but "a penny more" is another claim I'm skeptical of. Would the 4-bit processor cost less by some amount x>0?
coffee maker that uses an 8-bit processor? They don't exist because
the chip may only cost a penny more,
Let's assume yes. Is x > $1? Obviously not. Is x > $0.0000001? I'm
ok assuming yes. You are claiming x is around $0.01 but I would like to
see evidence that it is that large. I'll toss out $0.001 as an
alternative number. Most of the cost of that part is likely to be in
the package, pin drivers, on-chip peripherals, etc. Not the ALU. The
memory arrays are similar in size to those on small 8 bitters.
One data point: the EM6607 mentioned earlier has 4k bytes (2k 16-bit
words) of mask rom, and 96 nibbles of ram. Do they bother making
versions with less ram and rom, since a coffee maker shouldn't need that much code? If not, it is because the cost of those extra transistors is insignificant compared to the other stuff in the chip.
I couldn't easily find a teardown of the Walmart coffee maker you
linked, but here is a Mr. Coffee:
https://www.eetimes.com/mr-coffee-teardown-simple-effective-design/
The board in that thing isn't wasteful but you can see they were ok with passing costs along to the consumer.
There was BOLTR video where AvE took apart an Instant Pot pressure
cooker, which is more than $25 but is sort of a glorified coffee maker.
IIRC it had an 8 bit cpu inside.
Cheap FM broadcast receivers of 30 years ago had analog tuners and discriminators, but that stuff is all DSP now. Cost is low enough that
basic 8 bit MCU's must be close to free.
I made an error about the 3-cent PA150 earlier. Its program memory is
OTP rather than reprogrammable flash. The more expensive versions (5
cents etc.) are reprogrammable. For huge quantities you'd use mask rom,
of course.
https://jaycarlson.net/2019/09/06/whats-up-with-these-3-cent-microcontrollers/
Rick C <gnuarm.del...@gmail.com> writes:
Not sure how you can compare the two. The GA144 has an unlimited code space since the storage is all external to the chip.It has 64 words of program rom available at each node.
??? What does the data path have to do with the return stack? ....Idk about "data paths" which I thought was a feature of the hardware implementation, not the cpu architecture. E.g. the 8088 also had 8 bit
An 8080 has 8 bit data paths, and 16 bit addresses. Why do you keep getting confused about this?
data paths iirc, but it was a 16 bit architecture (size of the registers
and operands for most instructions).
The PDP 8/S was a 12-bit
architecture with 1-bit data path (S stood for serial). It was very
slow.
We're talking about stack cpu's in a Forth context, one would hope that
the idea is to program that cpu in Forth pretty much directly. So the
word size (accumulator size, register size, whatever) would be 4 bits,
in a 4 bit arch. What is the return stack going to look like in the
Forth for that cpu? Will there be instructions like >R R> for the stack juggling of Forth tradition? Will the return stack hold return
addresses?
On Saturday, June 11, 2022 at 9:34:13 AM UTC+2, Anton Ertl wrote:
[..]
And what if it costs 0 pennies more? In
<https://bernd-paysan.de/b16-presentation.pdf> (from 2005), Bernd
Paysan gives the size of the (small) b16 in the XC035 process (a
0.35um (350nm) process) as 0.16mm^2. Intel and AMD were at 90nm by
that time. These days, my guess is that it takes 100x less area in a
process used now for embedded controllers. How much does 0.0016mm^2
in such a process cost? How much do you save by using, say 0.0000mm^2
(a 4-bit CPU will certainly need more area than that)? Will it save a
penny? I doubt it.
I think the (possible) reason is not so much the number of bits in an >instruction, but the the number of pins on the package (or the effort
needed to check all pins). At least, that is the way it works for analog >discrete stuff: all small-signal bipolar transistors have the same
minimum cost. That's why I was surprised to see a 4-bit processor with
24 pins in the list -- that should make absolutely no sense.
-marcel--
AFAIK some 4-bitters are produced for extreme environmental conditions
(250 deg C ambient)
for which my argument does not hold (and evidently nothing holds for
the current 4-bit superconducting processors).
[GA144] It has 64 words of program rom available at each node.It has 64 words of program RAM at each node.
It can also execute instructions arriving through a pipe
Your example of a 0.0016mm^2 chip is not realistic. Find an 8 bit
device that is made on a 35nm process node you are talking about. You
won't.
https://bbs.21ic.com/icview-2816080-1-1.html
if you give them your program for an 8051, for example, they will
translate it into their chip's instruction set.
Rick C <gnuarm.del...@gmail.com> writes:
Your example of a 0.0016mm^2 chip is not realistic. Find an 8 bitThat means that the cost of fabbing transistors is not a driving factor
device that is made on a 35nm process node you are talking about. You won't.
in making such parts, compared with packaging etc. That also says not
much cost advantage for 4 bits over 8 bits. It's just transistors after
all.
IIRC there are AVR 8-bit processors with 384KB of program flash,
hardware multipliers and crypto instructions, etc. These can't be using
super ancient fab processes. 4
I believe the RP2040 is made in 40nm. It is quite a powerful chip, with
two ARM cores and 264KB of ram. My impression was that its purpose was
to provide Arduino-like control capabilities to the Raspberry Pi
ecosystem which was previously made of small Linux computers that were computationally powerful but without much control capability. In other
words they could have used an 8 bit design and probably thought about
it, but the RP2040 turned out to be sufficiently cheap for their
purposes. The RP2040 is about $1 retail, similar to an AVR.
Rick C <gnuarm.del...@gmail.com> writes:
Yes, it can run code from ram, but the rom is also there. RAM used for program code is of course unavailable for data.[GA144] It has 64 words of program rom available at each node.It has 64 words of program RAM at each node.
It can also execute instructions arriving through a pipeYes, that is mostly a feature of the implementation. Nothing stops
other architectures from getting such an ability if there was a use for
it.
Rick C <gnuarm.del...@gmail.com> writes:
https://bbs.21ic.com/icview-2816080-1-1.html
Thanks, this is interesting,
one of the embedded charts says it has 1K
instruction words (12 bit words) and 128 nibbles of ram. In fact it
looks like the program memory is OTP PROM rather than mask rom. I don't
see a description of the instruction set, but it says "4 bit risc",
and runs at up to 4 MIPS.
As for peripherals, the SOP8 version has five i/o pins (1 I), 3 timers,
2 PWM, WDT, and a 32khz xtal connection with an RTC. There is an
SOT23-6 version with less stuff connected to the pins. I don't see a
whole lot of difference from what you find in cheap 8-bit processors.
Actually here is another page about that part, with maybe more info:
http://www.upt-ic.com/en/products.aspx?id=0
It mentions that the "4-bit MCU can evade the patent of sop-8 package, because it adopts 4-bit bus architecture wire." No idea what that is
about.
There are quite a few configurations listed there. But, 1) I'd
like to know how the cost compares with the Padauk stuff, 2) how old the design is. Certainly the market for super low cost stuff will never go
away. So we should continue to see new designs once in a while. What
do they look like?
if you give them your program for an 8051, for example, they willTheir chip runs at 16 mhz while the original 8051 was a heck of a lot
translate it into their chip's instruction set.
slower, besides using more cycles per instruction. That suggests a more modern fab process than was used in the early 8-bit era.
Those 4-bit parts all seem to have program OTP, some of them up to 2K*16 bits. That is enough code space that if I expected to ship multiple generations or variants of a product, recurring software development
would become a significant cost component, so a C environment (or Forth, given where we are) would start looking more attractive than assembler despite possibly being slightly less hardware-efficient.
Thanks, this is interesting,Interesting for what? Do you have an application that needs 100,000 devices?
So we should continue to see new designs once in a while. WhatThey look like coffee makers and remote controls.
do they look like?
I have no idea what you are talking about. Software development costs
have to do with the software development. What does the memory on the
chip have to do with it?
It says nothing of the sort. It says the more modern processes are
not appropriate for making MCU devices that interface to 3.3 and 5
volt circuits. Nobody wants an MCU with 1.2V interfaces.
I like the way you invent a narrative based on virtually no facts. If
this part actually contains 264 kB of RAM, that is what drives the die
size. That's a lot of RAM for an MCU.
Rick C <gnuarm.del...@gmail.com> writes:I'll drop in here. It is strangely calming to catch up on others debates (except when protagonists are getting to long winded, resulting in everybody else matching them).
It says nothing of the sort. It says the more modern processes areThe RP2040 and ESP32 are both made in 40nm. I don't think they use 1.2V interfaces. RP2040 might have been doable in a larger process:
not appropriate for making MCU devices that interface to 3.3 and 5
volt circuits. Nobody wants an MCU with 1.2V interfaces.
https://semiwiki.com/forum/index.php?threads/cost-tradeoffs-at-28nm-vs-40nm-arm-m0.13887/
I like the way you invent a narrative based on virtually no facts. IfISTM that they had a budget and that told them the die size, and then
this part actually contains 264 kB of RAM, that is what drives the die size. That's a lot of RAM for an MCU.
they put on whatever amount of ram would fit. They certainly didn't
need two cores or 264kB. I earlier said they wanted control
capabilities similar to an AVR, but it occurs to me, they likely wanted
to program it in MicroPython rather than just in C, so they needed 32
bits. But, we know from the BBC micro:bit v1 that MicroPython can limp
along with 16KB of ram, and from the Adafruit SAMD21 boards that it is pleasant to use with 32KB. The RP2040 would have been quite comfortable
with 64KB. They put in 4x that much because they could afford to, not because they had to.
Regarding coffee pots: here is a Cuisinart DCC 1200 teardown. No MCU
visible but it looks expensive inside. It is around $100 retail. Yet I
don't see significant features that the super cheap pots don't also
have.
https://fullychargd.blogspot.com/2015/06/cuisinart-dcc-1200-teardown-and-power_7.html
If I had a product feature list that I thought would take 2Kw of code to implement, the software cost difference between C and assembler might
easily be in the 10k dollar range over the program's lifecycle. So with
1e6 units, that is 1 cent per unit all by itself. Therefore if the 4
bit cpu has to be programmed in asm while the 8 bitter can use C, then
the 4 bit hardware has to be at least 1 cent cheaper to be worth
thinking about. Is it that much cheaper? Only actual data can tell us
that, not philosophy.
Did Commodore charge a premium because somebody had to program the
software in asm? It was essentially free compared to the millions of
units pumped out.
dxforth <dxforth@gmail.com> writes:
Did Commodore charge a premium because somebody had to program the
software in asm? It was essentially free compared to the millions of
units pumped out.
This I don't know. Hardware back then (cpu and memory) was stupendously expensive by today's standards.
Being able to stretch it a little
further made a real difference in your product's profitability. And compilers at the time were not very good. That era was a bit before my
time, but I think asm language programming then was at least basically
sane.
However, here is one of my other schemes for cost savings compared to
4 bits, a misc processor in a single pin that attaches to the main
board, using a broadcast protocols like ethernet does.
Component count was expensive. Why Tramiel pushed at every step to reduce it. It's the reason VIC and 64 had dedicated graphics ships and software UART.
The 80's needed asm programmers. It may be less so today. Programmer
costs (whatever it may be) remain.
dxforth <dxforth@gmail.com> writes:
Component count was expensive. Why Tramiel pushed at every step to reduce >> it. It's the reason VIC and 64 had dedicated graphics ships and software
UART.
I'd be interested to know if the C64 factory software did much more than
the VIC and CBM software from earlier on. I thought the C64 was a
response to the then-amazing 64 kilobit DRAMs, giving vast amounts of
unused memory space to its users, i.e. coming with a similar BASIC to
what the VIC had. If yes, Commodore may not have used the C64's
inherently greater software flexibility.
The 80's needed asm programmers. It may be less so today. Programmer
costs (whatever it may be) remain.
Asm is super niche now. Even C is niche, outside of the embedded
sector, which itself is niche. I've certainly worked around some very
good fulltime web programmers who have never seen C code.
good fulltime web programmers who have never seen C code.Do they do less or cost less?
Rick C <gnuarm.del...@gmail.com> writes:
It says nothing of the sort. It says the more modern processes areThe RP2040 and ESP32 are both made in 40nm. I don't think they use 1.2V interfaces. RP2040 might have been doable in a larger process:
not appropriate for making MCU devices that interface to 3.3 and 5
volt circuits. Nobody wants an MCU with 1.2V interfaces.
https://semiwiki.com/forum/index.php?threads/cost-tradeoffs-at-28nm-vs-40nm-arm-m0.13887/
I like the way you invent a narrative based on virtually no facts. IfISTM that they had a budget and that told them the die size, and then
this part actually contains 264 kB of RAM, that is what drives the die size. That's a lot of RAM for an MCU.
they put on whatever amount of ram would fit. They certainly didn't
need two cores or 264kB. I earlier said they wanted control
capabilities similar to an AVR, but it occurs to me, they likely wanted
to program it in MicroPython rather than just in C, so they needed 32
bits. But, we know from the BBC micro:bit v1 that MicroPython can limp
along with 16KB of ram, and from the Adafruit SAMD21 boards that it is pleasant to use with 32KB. The RP2040 would have been quite comfortable
with 64KB. They put in 4x that much because they could afford to, not
because they had to.
Regarding coffee pots: here is a Cuisinart DCC 1200 teardown. No MCU
visible but it looks expensive inside. It is around $100 retail. Yet I
don't see significant features that the super cheap pots don't also
have.
https://fullychargd.blogspot.com/2015/06/cuisinart-dcc-1200-teardown-and-power_7.html
I think the (possible) reason is not so much the number of bits in an >instruction, but the the number of pins on the package (or the effort
needed to check all pins). At least, that is the way it works for analog >discrete stuff: all small-signal bipolar transistors have the same
minimum cost. That's why I was surprised to see a 4-bit processor with
24 pins in the list -- that should make absolutely no sense.
Rick C <gnuarm.del...@gmail.com> writes:
Interesting as info about the state of current hardware, which is partThanks, this is interesting,Interesting for what? Do you have an application that needs 100,000 devices?
of the state of the universe. I just read something in the news today
about the discovery of the first known black hole without a known
companion star (not counting the SMBH in galactic nuclei). Other people other thought that was interesting, or else it wouldn't be news. Do any
of us have a use for it? Probably not. Things can be interesting
without being useful.
Yes I've worked on products made by the millions (a family of payment terminals built around an ASIC). The ASIC had an ARM core and various
other stuff specific to the product family. Yes they were cost driven.
My boss told me they would sometimes launch significant engineering
projects with the sole purpose of getting 50 cents out of the unit cost
of those things. That 50 cents per unit figure suggested that they
would not do the same for 1 cent per unit.
The largest quantity of a real product I remember a regular here being involved with is Bernd Paysan's b16-based battery controller, which supposedly lives in a corner of an ASIC inside of around 1e8 Apple
phones. That is a 16 bit processor, programmed in Forth. Do you think
he could have saved by using 4 bits? His 16 bit cpu actually replaced
an 8051-based one.
No I mean what do the microprocessors look like. 4 bits? 8 bits? New architecture? Existing? Part of a remote control SOC? Many of thoseSo we should continue to see new designs once in a while. WhatThey look like coffee makers and remote controls.
do they look like?
Phaeton chips had seemingly special features like 150 volt switches.
I have no idea what you are talking about. Software development costsIf an MCU has 64 words of code space, it can only run an extremely
have to do with the software development. What does the memory on the
chip have to do with it?
simple program. You can write such a program in assembly language and maintain it across multiple versions without going crazy. If it has a megabyte of code space, it is made to run complex software. Using
assembler would be nuts. Languages and dev tools to manage the
complexity will be indispensible.
That particular chip had 2K words of code space which is around the
point where complexity becomes significant enough that you benefit from managing it using HLL's etc. If I'm going to use the whole 2Kw, I'd certainly prefer using C or Forth instead of assembler. There are tons
of C and Forth implementations that work nicely on 8 bitters of that
size. I don't know if there are any for 4 bitters.
If I had a product feature list that I thought would take 2Kw of code to implement, the software cost difference between C and assembler might
easily be in the 10k dollar range over the program's lifecycle. So with
1e6 units, that is 1 cent per unit all by itself. Therefore if the 4
bit cpu has to be programmed in asm while the 8 bitter can use C, then
the 4 bit hardware has to be at least 1 cent cheaper to be worth
thinking about. Is it that much cheaper? Only actual data can tell us
that, not philosophy.
In the old days, good sailors knew how to navigate by the stars. Now
they use GPS. It's like that. One less thing to deal with, so you can
direct your energy elsewhere.
But recently, I saw a document on Colorforth for ARM, and comparisons to Swift
Forth etc.
Rick C <gnuarm.del...@gmail.com> writes:
You said, "launch significant engineering projects". That's notYeah fair enough. The processor was already on an ASIC and had been
remotely the same as picking a processor as a part of development.
since long before I got there. I don't know if earlier versions of the product used commodity processors or how they were picked.
His CPU was in an ASIC. We are talking about discrete 4 bit MCUsHe'd know better, but I don't think they were designing a new chip, as opposed to iterating an existing design. The old chip had an 8051 core
vs. 8 bit MCUs. I hope you understand the difference. You also know nothing of the requirements. Obviously they were designing a new chip
for a reason, such as asking it to do something new or additional.
doing the battery stuff, and the new chip had a b16 (maybe -small).
But if you don't have an application, you probably don't understandI think you are obfuscating this. I'm using your picture of a coffee
the requirements that drive such a selection. I believe that you
don't do much hardware design, correct?
pot engineer choosing an MCU to deploy in 1e6 coffee pots. She or he
chooses between an 8-bit chip that costs $X and a 4-bit chip that costs
$Y. The main issue here would be the cost difference $(X-Y). I can
accept the idea that if $(X-Y) > $0.01 and there aren't consequential additional costs from the choice, then the 4 bit chip wins. I would
like to see evidence that the difference, today, in 2022, not 30 years
ago, is that large. If it's $0.0001 then I don't know what happens. If
it's negative, which AFAIK it might be, then 8 bits win.
I'm not even convinced that a coffee pot with a timer would have an MCU,
as opposed to e.g. some kind of dedicated power timer chip containing a
high voltage switch. I'm not convinced that a designer in 2022 would
face such a choice.
You are right that I don't design hardware, but I studied a lot of mathematics and I can look at two numbers and figure out which one is bigger. All the insinuation about hardware design is misdirection. I
need to see the actual numbers in order to be convinced.
I still don't get your point. Forth is traditionally a self writtenI have never heard of a Forth (or C) cross compiler targeted to a 4 bit
tool. If it doesn't exist, it is not so hard to write. It would need
to be a cross compiler, which is even easier in many respects
MCU. Do you know of any? For 8 bit MCU's, both are plentiful.
In either case, porting or writing a Forth compiler for a small MCU and
then writing an application with it has to be more work than programming
the application in assembler directly. "Small" here = 2k code words.
If you're going to write a lot of such applications, developing the
compiler as a one-time project might be worthwhile, but you'd budget it differently.
the software cost difference between C and assembler might easily be
in the 10k dollar range over the program's lifecycle.
That's your thinking. Most of these devices have zero software maintenance.The device itself has no sw maintenance once shipped, but the company
will keep making new variants of the product, each with slightly
different features. White label coffee pot = customers keep coming up
with new requests. The software has to be modified for each of those.
I was hoping to hear from people who actually use theseComp.arch.embedded maybe.
devices. Whatever. This is not the thread to discuss it. Too much
drift.
You said, "launch significant engineering projects". That's not
remotely the same as picking a processor as a part of development.
His CPU was in an ASIC. We are talking about discrete 4 bit MCUs
vs. 8 bit MCUs. I hope you understand the difference. You also know
nothing of the requirements. Obviously they were designing a new chip
for a reason, such as asking it to do something new or additional.
But if you don't have an application, you probably don't understand
the requirements that drive such a selection. I believe that you
don't do much hardware design, correct?
I still don't get your point. Forth is traditionally a self written
tool. If it doesn't exist, it is not so hard to write. It would need
to be a cross compiler, which is even easier in many respects
the software cost difference between C and assembler might easily be
in the 10k dollar range over the program's lifecycle.
That's your thinking. Most of these devices have zero software
maintenance.
I was hoping to hear from people who actually use these
devices. Whatever. This is not the thread to discuss it. Too much
drift.
I thought the MARK4 was mentioned in this thread, no? It was intended for the key fob market and was programmed in Forth.
dxforth <dxforth@gmail.com> writes:
good fulltime web programmers who have never seen C code.Do they do less or cost less?
In the old days, good sailors knew how to navigate by the stars. Now
they use GPS. It's like that. One less thing to deal with, so you can direct your energy elsewhere.
On 13/06/2022 11:38, Rick C wrote:
I thought the MARK4 was mentioned in this thread, no? It was intended for the key fob market and was programmed in Forth.
I wasn't aware of any of that. Does this mean I have to change my opinion
of Atmel and its designs? :)
https://en.wikichip.org/w/images/4/44/MARC4_4-bit_Microcontrollers_Programmer%27s_Guide.pdf
The compiler https://en.wikichip.org/w/images/2/25/MARC4_User%27s_Guide_qFORTH_Compiler.pdf
Sometimes we are in different worlds. "Iterating" a chip means making
a new chip in my world. You don't "iterate" a chip unless the
requirements are different, or there are bugs that must be fixed.
Ok, so are you looking for the evidence?
Coffee makers in my world have a digital clock with a 4 digit display
and a handful of buttons that make it very hard to set the timer and
use the damn thing. Why so few buttons? Because they cost something
like $0.01 each and if they aren't essential, they get thrown away.
You have to know what numbers to calculate. Any boob can use a
calculator.
I thought the MARK4 was mentioned in this thread, no? It was intended
for the key fob market and was programmed in Forth.
mentioned earlier in the conversation, like the digital time display.
If the company makes a new product, that's a new product.
dxforth <dxforth@gmail.com> writes:
The compiler
https://en.wikichip.org/w/images/2/25/MARC4_User%27s_Guide_qFORTH_Compiler.pdf
Wow, that is neat, and the programmers' guide also talks about qFORTH a
lot. The return stack lives in ram is it looks like its slots are 4
nibbles, a 12-bit code address plus 4 data bits. The manual advises
against too many levels of subroutines. Saving temporary data on the R
stack with >R etc. also sounds bad. But qFORTH does have those words.
I wonder if any actual application code is around that we can look at.
Talk to someone who designs toys.
=20=20There are NEW designs that use 4-bit MCUs...That's what you claim.=3D20=20
So, are you claiming that 4-bit processors are no longer designed into n= >ew =3D=20
products?=20
I wrote that you made a claim without providing evidence.=20
You mean the claim that 4-bit processors are in use?=20
m=3D=20Show me a simple product, that could make use of a 4-bit process=3D=20
or, is made in qty over a million and has an 8-bit processor. The coffee=
a=3Daker is a perfect example. Can you show me an under $25 coffee maker tha= >t =3D=20
uses an 8-bit processor? They don't exist because the chip may only cost=
penny more, but that's $10,000 in lost profits.And what if it costs 0 pennies more? In=20
<https://bernd-paysan.de/b16-presentation.pdf> (from 2005), Bernd=20
Paysan gives the size of the (small) b16 in the XC035 process (a=20
0.35um (350nm) process) as 0.16mm^2. Intel and AMD were at 90nm by=20
that time. These days, my guess is that it takes 100x less area in a=20
process used now for embedded controllers. How much does 0.0016mm^2=20
in such a process cost? How much do you save by using, say 0.0000mm^2=20
(a 4-bit CPU will certainly need more area than that)? Will it save a=20
penny? I doubt it.=20
You are suggesting that the chips are small enough that die size is no long= >er a cost factor?
I can't follow your logic above. I don't believe anyone=
uses even 90nm technology for 4-bit MCUs.
The cost is related to the equi=
pment used, the capital cost. By using fully amortized equipment they get = >the lowest cost.
Unfortunately, I have not found anything on cost per area, but I have=20
found=20
<https://www.fabricatedknowledge.com/p/the-rising-tide-of-semiconductor>,= >=20
which states that the cost per 100M gates is <$2 for processes more=20
recent than 45/50nm, and $1.3 for 28nm; so for new designs MCU=20
manufacturers are likely to use 28nm.=20
None of these processes are used for low end MCUs that work with 5 volt tol= >erant I/Os.=20
dxforth <dxforth@gmail.com> writes:
The compiler
https://en.wikichip.org/w/images/2/25/MARC4_User%27s_Guide_qFORTH_Compiler.pdf
Wow, that is neat, and the programmers' guide also talks about qFORTH a
lot. The return stack lives in ram is it looks like its slots are 4
nibbles, a 12-bit code address plus 4 data bits. The manual advises
against too many levels of subroutines. Saving temporary data on the R
stack with >R etc. also sounds bad.
But qFORTH does have those words.
I wonder if any actual application code is around that we can look at.
I have never heard of a Forth (or C) cross compiler targeted to a 4 bit
MCU. Do you know of any? For 8 bit MCU's, both are plentiful.
On 7 Jun 2022 at 11:52:46 CEST, "Wayne morellini" <waynemo...@gmail.com> wrote:
But recently, I saw a document on Colorforth for ARM, and comparisons to Swift
Forth etc.
Reference? Link?
Stephen
--
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM,and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM,and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
It's always time for another Forth chip. I've been experimenting with FPGA implementations of stack computers for 20+ years using inexpensive evaluation boards. Each project revealed design strengths and weaknesses in the hardware and software domains.Experience using the current 18-bit axe is giving me ideas for a 32-bit version.
The 18-bit axe manual is at https://drive.google.com/file/d/1PN6igK_-DLcrx2OWT1H5mXyfrT4OAAq3/view?usp=sharinglangauge.
I use Icarus Verilog to design and simulate hardware modules before attempting an FPGA fit.
An assembler is required to test the CPU at simulation time, and to launch the final SoC on powerup. I've been using Python3 for that job lately. The 18-bit axe uses the Python assembler to generate a run-time Forth native machine assembler scripting
All the tools I've used are free. I'd like to see more opinions based on original R&D experience rather than arguing the theoretical merits of historic silicon.
- Myron Plichota
"Jimbo is not James Bond" - an ancient koan
On Tuesday, June 14, 2022 at 3:41:51 PM UTC+10, Myron Plichota wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
domains. Experience using the current 18-bit axe is giving me ideas for a 32-bit version.It's always time for another Forth chip. I've been experimenting with FPGA implementations of stack computers for 20+ years using inexpensive evaluation boards. Each project revealed design strengths and weaknesses in the hardware and software
langauge.The 18-bit axe manual is at https://drive.google.com/file/d/1PN6igK_-DLcrx2OWT1H5mXyfrT4OAAq3/view?usp=sharing
I use Icarus Verilog to design and simulate hardware modules before attempting an FPGA fit.
An assembler is required to test the CPU at simulation time, and to launch the final SoC on powerup. I've been using Python3 for that job lately. The 18-bit axe uses the Python assembler to generate a run-time Forth native machine assembler scripting
All the tools I've used are free. I'd like to see more opinions based on original R&D experience rather than arguing the theoretical merits of historic silicon.
- Myron PlichotaHave people got worn out arguing with people and suddenly vanished?
"Jimbo is not James Bond" - an ancient koan
Your effort, I appreciate.
Is Kopin still alive, to do another book that examines all the new forth processors? Would be a great book, and you could compare how design features enhance things.
Anyway, really worn out tonight, had a few good days. Apparently, if I don't eat when well, I can be better while well. So, I'm here to mention something people don't know.
On Tuesday, June 14, 2022 at 3:41:51 PM UTC+10, Myron Plichota wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Tuesday, June 7, 2022 at 5:52:48 AM UTC-4, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
domains. Experience using the current 18-bit axe is giving me ideas for a 32-bit version.It's always time for another Forth chip. I've been experimenting with FPGA implementations of stack computers for 20+ years using inexpensive evaluation boards. Each project revealed design strengths and weaknesses in the hardware and software
langauge.The 18-bit axe manual is at https://drive.google.com/file/d/1PN6igK_-DLcrx2OWT1H5mXyfrT4OAAq3/view?usp=sharing
I use Icarus Verilog to design and simulate hardware modules before attempting an FPGA fit.
An assembler is required to test the CPU at simulation time, and to launch the final SoC on powerup. I've been using Python3 for that job lately. The 18-bit axe uses the Python assembler to generate a run-time Forth native machine assembler scripting
All the tools I've used are free. I'd like to see more opinions based on original R&D experience rather than arguing the theoretical merits of historic silicon.
- Myron PlichotaHave people got worn out arguing with people and suddenly vanished?
"Jimbo is not James Bond" - an ancient koan
Your effort, I appreciate.
Is Kopin still alive, to do another book that examines all the new forth processors? Would be a great book, and you could compare how design features enhance things.
Anyway, really worn out tonight, had a few good days. Apparently, if I don't eat when well, I can be better while well. So, I'm here to mention something people don't know.
The news from a Forth Processor Manufacturer as a hint:platform development using polyFORTH/GA144 running on the Host chip of an evaluation board (or equivalent) with much higher speed capability for communicating with microcode running in F18A nodes on Host and/or Target chips.
What's New at GreenArrays
arrayForth 3 Released 21 May 2019:
The new development platform for the GA144 is arrayForth 3, which is released today for general use. This package supports cross-development from saneFORTH/x86 running on a PC or win32 equivalent using high speed serial communication, and/or within-
New Evaluation Board EVB002 Available for Purchase 20 May 2019:
The new evaluation board is manufactured and may be purchased here. New documentation is also posted: DB014, EVB002 Eval Board Reference,
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on .. for ARM, andcomparisons to .. Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these designs
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:comparisons to .. Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these designs
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on .. for ARM, and
I have been interested in working on a ring computer. Something Misc would be very good at. I did many draft designs, and new techniques.
It has to be defined
what this Forth processor does and
how it is implemented:
Wow! Nothing new in three years. That's not promising. We keep hearing rumors of new chip designs, but no announcements of new products. I wonder when some of these rumors might pan out to be true? Isn't Stephen involved in one?
On 15/06/2022 02:39, Rick C wrote:
Wow! Nothing new in three years. That's not promising. We keep hearing rumors of new chip designs, but no announcements of new products. I wonder when some of these rumors might pan out to be true? Isn't Stephen involved in one?
"Good things come to those who wait" - an English proverb
On Tuesday, June 14, 2022 at 7:30:14 PM UTC-4, dxforth wrote:
On 15/06/2022 02:39, Rick C wrote:
Wow! Nothing new in three years. That's not promising. We keep hearing rumors of new chip designs, but no announcements of new products. I wonder when some of these rumors might pan out to be true? Isn't Stephen involved in one?"Good things come to those who wait" - an English proverb
That's not the "proverb". It's a misquote from the French which literally translated means, "All things come to those who know how to wait".
I don't think it applies to wishful thinking.
I wonder when some of these rumors might pan out to be true? Isn't Stephen involved in one?
On 14 Jun 2022 at 18:39:23 CEST, "Rick C" <gnuarm.del...@gmail.com>
wrote:
I wonder when some of these rumors might pan out to be true? Isn't Stephen involved in one?I have nothing to do with GreenArrays except to wish them well.
I cannot release information about the chip MPE is involved with without permission,
except to say that the current application is very specialised.
In general, the answer to the question "Is it time for another Forth chip?" is
no.
The reason is that it's much cheaper to write a good Forth compiler for an existing CPU than to bring a new chip to market. We learned this way back when Digital brought StrongARM to market in 18 months, and a VFX compiler took about three months. Note that Digital's team was exceptional.
In rare special cases, usually where deterministic execution is required, silicon
stack machines can be used to advantage.
Stephen
--
Stephen Pelc, ste...@vfxforth.com
MicroProcessor Engineering, Ltd. - More Real, Less Time
133 Hill Lane, Southampton SO15 5AF, England
tel: +44 (0)23 8063 1441, +44 (0)78 0390 3612, +34 649 662 974 http://www.mpeforth.com - free VFX Forth downloads
On 14 Jun 2022 at 18:39:23 CEST, "Rick C" <gnuarm.del...@gmail.com>
wrote:
I wonder when some of these rumors might pan out to be true? Isn't Stephen involved in one?I have nothing to do with GreenArrays except to wish them well.
I cannot release information about the chip MPE is involved with without permission,
except to say that the current application is very specialised.
In general, the answer to the question "Is it time for another Forth chip?" is
no.
The reason is that it's much cheaper to write a good Forth compiler for an existing CPU than to bring a new chip to market. We learned this way back when Digital brought StrongARM to market in 18 months, and a VFX compiler took about three months. Note that Digital's team was exceptional.
In rare special cases, usually where deterministic execution is required, silicon
stack machines can be used to advantage.
Stephen
--
Stephen Pelc, ste...@vfxforth.com
MicroProcessor Engineering, Ltd. - More Real, Less Time
133 Hill Lane, Southampton SO15 5AF, England
tel: +44 (0)23 8063 1441, +44 (0)78 0390 3612, +34 649 662 974 http://www.mpeforth.com - free VFX Forth downloads
On Tuesday, June 14, 2022 at 2:31:25 PM UTC-4, jpit...@gmail.com wrote:
It has to be definedBingo. And if you can't define it, you'll never achieve it.
what this Forth processor does and
how it is implemented:
On Tuesday, 14 June 2022 at 19:09:52 UTC+1, Wayne morellini wrote:and comparisons to .. Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on .. for ARM,
I have been interested in working on a ring computer. Something Misc would be very good at. I did many draft designs, and new techniques.It has to be defined
what this Forth processor does and
how it is implemented:
in CMOS chips
as FPFA
FPGA transferred into Gates - if you have the money for the masks
Gate Array - If you find an opportunity - if you have the money for the masks
Standard Cell - if you have the money for the masks
Full Custom Design - if you have the money for the masks
But who is interested in a Forth chip anyway; the world now lives without it and who defines the requirements.
Where is the comparison table of
-- Gates
-- Cost
-- Speed
-- Power consumption
-- Availability?
On Wednesday, 15 June 2022 at 10:39:53 UTC+1, Stephen Pelc wrote:
On 14 Jun 2022 at 18:39:23 CEST, "Rick C" <gnuarm.del...@gmail.com>
wrote:
I wonder when some of these rumors might pan out to be true? Isn't StephenI have nothing to do with GreenArrays except to wish them well.
involved in one?
I cannot release information about the chip MPE is involved with without permission,
except to say that the current application is very specialised.
In general, the answer to the question "Is it time for another Forth chip?" is
no.
The reason is that it's much cheaper to write a good Forth compiler for an existing CPU than to bring a new chip to market. We learned this way back when Digital brought StrongARM to market in 18 months, and a VFX compiler took about three months. Note that Digital's team was exceptional.
In rare special cases, usually where deterministic execution is required, siliconI would kindly disagree.
stack machines can be used to advantage.
Stephen
--
Stephen Pelc, ste...@vfxforth.com
MicroProcessor Engineering, Ltd. - More Real, Less Time
133 Hill Lane, Southampton SO15 5AF, England
tel: +44 (0)23 8063 1441, +44 (0)78 0390 3612, +34 649 662 974 http://www.mpeforth.com - free VFX Forth downloads
There are probably 20 to 200 or more Forth implementations that people use in applications , or just did for fun.
Plus the commercial versions.
The same should actually apply to Forth chips.
There are just not enough people here
who know Forth well AND HDLs
to have fun implementing Forth processors in VHDL / Verilog.
The only 2 I am aware of as "products" recently
is Bernd's b16
and the J1
who made it into silicon.
Now. This modern EU free microchip fab design. I suppose all we would need to know of an EU citizen to work with to get to use it to design and make chips. Is that right?
That problem is solved? It wouldn't worry me. The sort of designs I would look at wouldn't be available in a cheaper plant elsewhere.
Misc?
On Wednesday, 15 June 2022 at 10:39:53 UTC+1, Stephen Pelc wrote:
On 14 Jun 2022 at 18:39:23 CEST, "Rick C" <gnuarm.del...@gmail.com>
wrote:
I wonder when some of these rumors might pan out to be true? Isn't StephenI have nothing to do with GreenArrays except to wish them well.
involved in one?
I cannot release information about the chip MPE is involved with without permission,
except to say that the current application is very specialised.
In general, the answer to the question "Is it time for another Forth chip?" is
no.
The reason is that it's much cheaper to write a good Forth compiler for an existing CPU than to bring a new chip to market. We learned this way back when Digital brought StrongARM to market in 18 months, and a VFX compiler took about three months. Note that Digital's team was exceptional.
In rare special cases, usually where deterministic execution is required, siliconI would kindly disagree.
stack machines can be used to advantage.
Stephen
--
Stephen Pelc, ste...@vfxforth.com
MicroProcessor Engineering, Ltd. - More Real, Less Time
133 Hill Lane, Southampton SO15 5AF, England
tel: +44 (0)23 8063 1441, +44 (0)78 0390 3612, +34 649 662 974 http://www.mpeforth.com - free VFX Forth downloads
There are probably 20 to 200 or more Forth implementations that people use in applications , or just did for fun.
Plus the commercial versions.
The same should actually apply to Forth chips.
There are just not enough people here
who know Forth well AND HDLs
to have fun implementing Forth processors in VHDL / Verilog.
The only 2 I am aware of as "products" recently
is Bernd's b16
and the J1
who made it into silicon.
Ok, let's get back to reality.some useful devices.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more compact, and for
Now. This modern EU free microchip fab design. I suppose all we would need to know of an EU citizen to work with to get to use it to design and make chips. Is that right?
That problem is solved? It wouldn't worry me. The sort of designs I would look at wouldn't be available in a cheaper plant elsewhere.
On Wednesday, June 15, 2022 at 8:23:22 AM UTC-4, Wayne morellini wrote:for some useful devices.
Ok, let's get back to reality.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more compact, and
The journey of a thousand miles, begins with a single step.
--
Rick C.
--+ Get 1,000 miles of free Supercharging
--+ Tesla referral code - https://ts.la/richard11209
On Wednesday, June 15, 2022 at 8:28:41 AM UTC-4, Wayne morellini wrote:with a clear, requirements definition. Something that has enough detail to actually be able to start the design process.
Now. This modern EU free microchip fab design. I suppose all we would need to know of an EU citizen to work with to get to use it to design and make chips. Is that right?
That problem is solved? It wouldn't worry me. The sort of designs I would look at wouldn't be available in a cheaper plant elsewhere.I don't think I would worry with the later stages of implementation, until I had a design I was interested in implementing. So far, this discussion has been extremely vague with no indication of what is needed from the device. Perhaps it should start
--
Rick C.
-+- Get 1,000 miles of free Supercharging
-+- Tesla referral code - https://ts.la/richard11209
On Wednesday, June 15, 2022 at 11:39:31 PM UTC+10, gnuarm.del...@gmail.com wrote:for some useful devices.
On Wednesday, June 15, 2022 at 8:23:22 AM UTC-4, Wayne morellini wrote:
Ok, let's get back to reality.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more compact, and
The journey of a thousand miles, begins with a single step.And a million dollars.
I have been interested in working on a ring computer.
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:
Ok, I forgot. Chick made okcad free didn't he? Is that opensourced? Maybe we could use that? Anybody still got it?
My last feedback from Greenarrays was, that okcad is a secret
weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a
new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not
available for the world,
I have not seen anything comparable.
Wayne morellini <waynemo...@gmail.com> writes:Thanks for that Paul. I'm thinking of something useful.
I have been interested in working on a ring computer.https://www.nngroup.com/articles/javaring-wearable-computer/
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:Oh well, somebody could maybe a similar freeware design
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
On Wednesday, June 15, 2022 at 12:06:43 PM UTC-4, Wayne morellini wrote:and for some useful devices.
On Wednesday, June 15, 2022 at 11:39:31 PM UTC+10, gnuarm.del...@gmail.com wrote:
On Wednesday, June 15, 2022 at 8:23:22 AM UTC-4, Wayne morellini wrote:
Ok, let's get back to reality.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more compact,
You completely fail to understand what I'm saying. Your options are not going to diminish if you start work on the CPU design now. You don't even know what you want to design as far as I can tell.The journey of a thousand miles, begins with a single step.And a million dollars.
Once you've designed the CPU, and figured out what else needs to go with it, maybe then you can look into how to get it built.
Presently, you are just tilting at windmills. Narrow your task by focusing on the first step rather than trying to plan a trip when you don't even know where you are going.
--
Rick C.
-++ Get 1,000 miles of free Supercharging
-++ Tesla referral code - https://ts.la/richard11209
On Thursday, June 16, 2022 at 1:55:15 AM UTC-4, jpit...@gmail.com wrote:ago, but far from being useful with more modern process technologies without massive updates which means, building chips and adapting the tools to suit what doesn't work.
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:I'm sure it would also be very hard for anyone other than Chuck to use without tons of documentation, which I expect doesn't exist.
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
I find it amusing they feel the need to protect it from being used by "competitors". I seriously doubt GA has any competitors, they are too far behind anyone else in the real world. The tools may be good enough for process technologies in use 20 years
I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth group where there are few hardware oriented people.don't have a commercially viable chip. But since there are no concrete goals for this chip, it's hard to imagine it will ever happen, much less be a success.
But as I've said several times, trying to plan the chip, is far ahead of the game at this point. The chip design is just an implementation of your logic design... unless you plan to do a "Chuck", and make successive iterations a decade apart and still
Many in the Forth community feel the way to design is to play and develop interesting technical features, even if they don't fit any particular model or goal. When you do that, you end up with a GA144 and very little sales.
--
Rick C.
+-- Get 1,000 miles of free Supercharging
+-- Tesla referral code - https://ts.la/richard11209
On Thursday, June 16, 2022 at 2:20:28 AM UTC+10, gnuarm.del...@gmail.com wrote:and for some useful devices.
On Wednesday, June 15, 2022 at 12:06:43 PM UTC-4, Wayne morellini wrote:
On Wednesday, June 15, 2022 at 11:39:31 PM UTC+10, gnuarm.del...@gmail.com wrote:
On Wednesday, June 15, 2022 at 8:23:22 AM UTC-4, Wayne morellini wrote:
Ok, let's get back to reality.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more compact,
You completely fail to understand what I'm saying. Your options are not going to diminish if you start work on the CPU design now. You don't even know what you want to design as far as I can tell.The journey of a thousand miles, begins with a single step.And a million dollars.
Once you've designed the CPU, and figured out what else needs to go with it, maybe then you can look into how to get it built.
Presently, you are just tilting at windmills. Narrow your task by focusing on the first step rather than trying to plan a trip when you don't even know where you are going.
--
Rick C.
-++ Get 1,000 miles of free SuperchargingRick you don't listen too well!
-++ Tesla referral code - https://ts.la/richard11209
On Thursday, June 16, 2022 at 3:55:15 PM UTC+10, jpit...@gmail.com wrote:
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:Oh well, somebody could maybe a similar freeware design
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
On Thursday, June 16, 2022 at 4:41:15 PM UTC+10, gnuarm.del...@gmail.com wrote:years ago, but far from being useful with more modern process technologies without massive updates which means, building chips and adapting the tools to suit what doesn't work.
On Thursday, June 16, 2022 at 1:55:15 AM UTC-4, jpit...@gmail.com wrote:
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:I'm sure it would also be very hard for anyone other than Chuck to use without tons of documentation, which I expect doesn't exist.
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
I find it amusing they feel the need to protect it from being used by "competitors". I seriously doubt GA has any competitors, they are too far behind anyone else in the real world. The tools may be good enough for process technologies in use 20
still don't have a commercially viable chip. But since there are no concrete goals for this chip, it's hard to imagine it will ever happen, much less be a success.I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth group where there are few hardware oriented people.
But as I've said several times, trying to plan the chip, is far ahead of the game at this point. The chip design is just an implementation of your logic design... unless you plan to do a "Chuck", and make successive iterations a decade apart and
Many in the Forth community feel the way to design is to play and develop interesting technical features, even if they don't fit any particular model or goal. When you do that, you end up with a GA144 and very little sales.
--
Rick C.
+-- Get 1,000 miles of free SuperchargingI think you lost the point Rick.
+-- Tesla referral code - https://ts.la/richard11209
I thought there were open source IC design tools? A post in a more
relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth group
where there are few hardware oriented people.
Rick C <gnuarm.del...@gmail.com> writes:
I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth groupThe Berkeley stuff from the 1980s was sort of comparable to OKAD in the sense of being rectangle editors with macros etc. HDL's weren't really
where there are few hardware oriented people.
a thing back then. Today in principle most of the code exists to go
from HDL to rectangles, but a few steps are missing. I don't know if
they are difficult.
Certainly even without FOSS tools, it is possible to make a chip without spending money, by sending a HDL file to the fab service that runs it through closed source stuff.
I don't understand the thought process that starts with the idea of
another Forth chip and jumps immediately to fabrication or even design tools. I'd start with figuring out the architecture, then simulating it
in software, then on an FPGA, before pursuing silicon fab. Otherwise it
is a solution in search of a problem.
Fwiw I don't know if it is a stack arch, but Forth on the Parallax P2 is supposed to be pretty nice, and it is already in the rom.
On Thursday, June 16, 2022 at 8:59:26 AM UTC-4, Wayne morellini wrote:and for some useful devices.
On Thursday, June 16, 2022 at 2:20:28 AM UTC+10, gnuarm.del...@gmail.com wrote:
On Wednesday, June 15, 2022 at 12:06:43 PM UTC-4, Wayne morellini wrote:
On Wednesday, June 15, 2022 at 11:39:31 PM UTC+10, gnuarm.del...@gmail.com wrote:
On Wednesday, June 15, 2022 at 8:23:22 AM UTC-4, Wayne morellini wrote:
Ok, let's get back to reality.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more compact,
You completely fail to understand what I'm saying. Your options are not going to diminish if you start work on the CPU design now. You don't even know what you want to design as far as I can tell.The journey of a thousand miles, begins with a single step.And a million dollars.
Once you've designed the CPU, and figured out what else needs to go with it, maybe then you can look into how to get it built.
Presently, you are just tilting at windmills. Narrow your task by focusing on the first step rather than trying to plan a trip when you don't even know where you are going.
--
Rick C.
Quite the illuminating comment. Do you wish to have a discussion, or just be insulting?-++ Get 1,000 miles of free SuperchargingRick you don't listen too well!
-++ Tesla referral code - https://ts.la/richard11209
--
Rick C.
+-+ Get 1,000 miles of free Supercharging
+-+ Tesla referral code - https://ts.la/richard11209
On Thursday, June 16, 2022 at 9:05:31 AM UTC-4, Wayne morellini wrote:years ago, but far from being useful with more modern process technologies without massive updates which means, building chips and adapting the tools to suit what doesn't work.
On Thursday, June 16, 2022 at 4:41:15 PM UTC+10, gnuarm.del...@gmail.com wrote:
On Thursday, June 16, 2022 at 1:55:15 AM UTC-4, jpit...@gmail.com wrote:
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:I'm sure it would also be very hard for anyone other than Chuck to use without tons of documentation, which I expect doesn't exist.
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
I find it amusing they feel the need to protect it from being used by "competitors". I seriously doubt GA has any competitors, they are too far behind anyone else in the real world. The tools may be good enough for process technologies in use 20
still don't have a commercially viable chip. But since there are no concrete goals for this chip, it's hard to imagine it will ever happen, much less be a success.I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth group where there are few hardware oriented people.
But as I've said several times, trying to plan the chip, is far ahead of the game at this point. The chip design is just an implementation of your logic design... unless you plan to do a "Chuck", and make successive iterations a decade apart and
..Many in the Forth community feel the way to design is to play and develop interesting technical features, even if they don't fit any particular model or goal. When you do that, you end up with a GA144 and very little sales.
think he has ever said if it is for a particular product, or if it will be sold as a chip. Otherwise, there has never really been a successful stack processor. I suppose you can count the RTX2000 as "successful". It did get used in various spacecraft.I think you lost the point Rick.Which is what? That the Forth community is more interested in playing than producing a useful and profitable product? GA has not shown anything viable in the market. The device Stephen is working on may turn out to be useful... to someone. I don't
I'm not knocking stack processors. I like them and use them in FPGAs. I'm just pointing out the facts of the matter. They are simply not a significant success, and are unlikely to ever become one.
I don't understand the thought process that starts with the idea of
another Forth chip and jumps immediately to fabrication or even design tools. I'd start with figuring out the architecture, then simulating it
in software, then on an FPGA, before pursuing silicon fab. Otherwise it
is a solution in search of a problem.
On Thursday, June 16, 2022 at 9:03:12 AM UTC-4, Wayne morellini wrote:duplicates the many tools others have developed over the decades.
On Thursday, June 16, 2022 at 3:55:15 PM UTC+10, jpit...@gmail.com wrote:While Okad is not a universal tool, it does a job and does it well I would assume. But it was a *lot* of work and involved calibration from test chip fabrications. It is very unlikely to ever be reinvented. Then there is the fact that is largely
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:Oh well, somebody could maybe a similar freeware design
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
--
Rick C.
++- Get 1,000 miles of free Supercharging
++- Tesla referral code - https://ts.la/richard11209
On Thursday, June 16, 2022 at 12:12:53 PM UTC-4, Paul Rubin wrote:the story presents his first draft to his boss saying the purpose was to demonstrate something and the goal was something else. His boss replied back, no, the demonstration is the goal. So what is your purpose?
Rick C <gnuarm.del...@gmail.com> writes:
I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth groupThe Berkeley stuff from the 1980s was sort of comparable to OKAD in the sense of being rectangle editors with macros etc. HDL's weren't really
where there are few hardware oriented people.
a thing back then. Today in principle most of the code exists to go
from HDL to rectangles, but a few steps are missing. I don't know if
they are difficult.
Certainly even without FOSS tools, it is possible to make a chip without spending money, by sending a HDL file to the fab service that runs it through closed source stuff.
I don't understand the thought process that starts with the idea of another Forth chip and jumps immediately to fabrication or even design tools. I'd start with figuring out the architecture, then simulating itEven before figuring out the architecture, I'd start with the purpose. I recall a humorous story about writing government IRAD reports. They had both a purpose and a goal. It was hard for most people to understand the difference. So the character in
in software, then on an FPGA, before pursuing silicon fab. Otherwise it
is a solution in search of a problem.
He takes the report back and rewrites it for his department manager. The department manager says, this isn't right, what you have for the purpose, is really your goal! So he takes it back again.stack processor".
Now he presents the third draft to he division head who says, "No, this purpose is really your goal. What's your purpose?"
Finally, he is presenting the report to the government. He says, "My purpose is to get into heaven!"
Designing an MCU is something that needs to have it's purposes defined, well defined, which will guide all further decisions about the design, the goals. Maybe starting with something less supreme than getting into heaven, but higher than "making a
Fwiw I don't know if it is a stack arch, but Forth on the Parallax P2 is supposed to be pretty nice, and it is already in the rom.Ok, then, "Mission Complete".
On Friday, June 17, 2022 at 1:07:53 AM UTC+10, gnuarm.del...@gmail.com wrote:compact, and for some useful devices.
On Thursday, June 16, 2022 at 8:59:26 AM UTC-4, Wayne morellini wrote:
On Thursday, June 16, 2022 at 2:20:28 AM UTC+10, gnuarm.del...@gmail.com wrote:
On Wednesday, June 15, 2022 at 12:06:43 PM UTC-4, Wayne morellini wrote:
On Wednesday, June 15, 2022 at 11:39:31 PM UTC+10, gnuarm.del...@gmail.com wrote:
On Wednesday, June 15, 2022 at 8:23:22 AM UTC-4, Wayne morellini wrote:
Ok, let's get back to reality.
I'm thinking of using a version of colorforth on ARM, as a starting piont. Then go to misc chip to perform the colorforth code, if enough revenue is generated.
I'm thinking of making a retro gaming misc chip with a sort of super economical blitter., But not really a blitter. Having that as a basis for a series of chips to sell for embedded, and a system on the level of the Raspberry PI, more
You completely fail to understand what I'm saying. Your options are not going to diminish if you start work on the CPU design now. You don't even know what you want to design as far as I can tell.The journey of a thousand miles, begins with a single step.And a million dollars.
Once you've designed the CPU, and figured out what else needs to go with it, maybe then you can look into how to get it built.
Presently, you are just tilting at windmills. Narrow your task by focusing on the first step rather than trying to plan a trip when you don't even know where you are going.
--
Rick C.
Quite the illuminating comment. Do you wish to have a discussion, or just be insulting?-++ Get 1,000 miles of free SuperchargingRick you don't listen too well!
-++ Tesla referral code - https://ts.la/richard11209
--
Rick C.
+-+ Get 1,000 miles of free SuperchargingJust statement. Still got lots happening here.
+-+ Tesla referral code - https://ts.la/richard11209
On Friday, June 17, 2022 at 1:16:30 AM UTC+10, gnuarm.del...@gmail.com wrote:years ago, but far from being useful with more modern process technologies without massive updates which means, building chips and adapting the tools to suit what doesn't work.
On Thursday, June 16, 2022 at 9:05:31 AM UTC-4, Wayne morellini wrote:
On Thursday, June 16, 2022 at 4:41:15 PM UTC+10, gnuarm.del...@gmail.com wrote:
On Thursday, June 16, 2022 at 1:55:15 AM UTC-4, jpit...@gmail.com wrote:
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:I'm sure it would also be very hard for anyone other than Chuck to use without tons of documentation, which I expect doesn't exist.
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
I find it amusing they feel the need to protect it from being used by "competitors". I seriously doubt GA has any competitors, they are too far behind anyone else in the real world. The tools may be good enough for process technologies in use 20
people.I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth group where there are few hardware oriented
still don't have a commercially viable chip. But since there are no concrete goals for this chip, it's hard to imagine it will ever happen, much less be a success.But as I've said several times, trying to plan the chip, is far ahead of the game at this point. The chip design is just an implementation of your logic design... unless you plan to do a "Chuck", and make successive iterations a decade apart and
think he has ever said if it is for a particular product, or if it will be sold as a chip. Otherwise, there has never really been a successful stack processor. I suppose you can count the RTX2000 as "successful". It did get used in various spacecraft...Many in the Forth community feel the way to design is to play and develop interesting technical features, even if they don't fit any particular model or goal. When you do that, you end up with a GA144 and very little sales.
I think you lost the point Rick.Which is what? That the Forth community is more interested in playing than producing a useful and profitable product? GA has not shown anything viable in the market. The device Stephen is working on may turn out to be useful... to someone. I don't
Very successful.
You are mistaken, you don't know what confidential products greenarrays had done.
I don't know about a lot of stuff in products,
and if I rang up manufacturers about it, I'm sure I would be denied knowledge of it as a trade secret. Greenarrays existed for many years with a large staff before getting difficulty. They obviously we're generating income to pay for it.
The issue is, if they even had an F21 like design, I would have been using it. Look at the complete consumption of cores to simulate such a thing om the 144, in the recent thread. Sure it's a nice example, but the 144 is not designed towards thatapplication. It would have been better if one of the cores had been F21 like, to interface to the rest of the world, and stream process data with the rest of the array, bit by bit as a giant pipeline. That makes it more useful. You don't have to run the
I'm not knocking stack processors. I like them and use them in FPGAs. I'm just pointing out the facts of the matter. They are simply not a significant success, and are unlikely to ever become one.No you are not. There are some people in the world in general, who seem to take pleasure in knocking things and being negative towards people, and insulting as they don't get their way. Anybody around here know about that?
On Friday, June 17, 2022 at 1:11:05 AM UTC+10, gnuarm.del...@gmail.com wrote:duplicates the many tools others have developed over the decades.
On Thursday, June 16, 2022 at 9:03:12 AM UTC-4, Wayne morellini wrote:
On Thursday, June 16, 2022 at 3:55:15 PM UTC+10, jpit...@gmail.com wrote:While Okad is not a universal tool, it does a job and does it well I would assume. But it was a *lot* of work and involved calibration from test chip fabrications. It is very unlikely to ever be reinvented. Then there is the fact that is largely
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:Oh well, somebody could maybe a similar freeware design
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
But, you remember how Chuck had figured out the bugs, and could simply move from process to process? I forget if that was by changing parameters without constant test runs, or not. But Chuck did it, and now it's known, so somebody could do somethingsimilar
Most of the effort involved in coming up with a solution, is realising the solution. Now people realise how he did it, they can redo it.
On Friday, June 17, 2022 at 2:44:20 AM UTC+10, gnuarm.del...@gmail.com wrote:the story presents his first draft to his boss saying the purpose was to demonstrate something and the goal was something else. His boss replied back, no, the demonstration is the goal. So what is your purpose?
On Thursday, June 16, 2022 at 12:12:53 PM UTC-4, Paul Rubin wrote:
Rick C <gnuarm.del...@gmail.com> writes:
I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might be willing to assist someone getting started. This is a Forth groupThe Berkeley stuff from the 1980s was sort of comparable to OKAD in the sense of being rectangle editors with macros etc. HDL's weren't really
where there are few hardware oriented people.
a thing back then. Today in principle most of the code exists to go
from HDL to rectangles, but a few steps are missing. I don't know if they are difficult.
Certainly even without FOSS tools, it is possible to make a chip without spending money, by sending a HDL file to the fab service that runs it through closed source stuff.
I don't understand the thought process that starts with the idea of another Forth chip and jumps immediately to fabrication or even design tools. I'd start with figuring out the architecture, then simulating it in software, then on an FPGA, before pursuing silicon fab. Otherwise it is a solution in search of a problem.Even before figuring out the architecture, I'd start with the purpose. I recall a humorous story about writing government IRAD reports. They had both a purpose and a goal. It was hard for most people to understand the difference. So the character in
stack processor".He takes the report back and rewrites it for his department manager. The department manager says, this isn't right, what you have for the purpose, is really your goal! So he takes it back again.
Now he presents the third draft to he division head who says, "No, this purpose is really your goal. What's your purpose?"
Finally, he is presenting the report to the government. He says, "My purpose is to get into heaven!"
Designing an MCU is something that needs to have it's purposes defined, well defined, which will guide all further decisions about the design, the goals. Maybe starting with something less supreme than getting into heaven, but higher than "making a
The purpose is to reach a goal of a general purpose microcontroller into various markets.Fwiw I don't know if it is a stack arch, but Forth on the Parallax P2 is supposed to be pretty nice, and it is already in the rom.Ok, then, "Mission Complete".
Specific purpose designs are the goals of GA, which you have problems with :).
A specific purpose chip, may result in not a single chip being sold as part of a volume. A very bad way of doing business, unless you have something that is likely to sell, or contract to do it lined up. A general purpose chip is something which can besold to anybody and for a wide range of applications. If somebody wants, a specific purpose design can be made out of it, the general core of the chip remains the same often, which saves money. So, Rick, is it your goal or purpose, to try to kick in
BTW, brain injury, I forgot to mention, I would like to know how ok cad goes on a modern process fab node, as not only are structures more complex, but as you get smaller things become more fuzzy and less definable. Just looking at some mode processes,you see density of interference on the chip, with the lines being the main density, but with interfering abbarations around it. How closely can you tile the tolerances of a process like that.
On Thursday, June 16, 2022 at 6:13:23 PM UTC-4, Wayne morellini wrote:
Just statement. Still got lots happening here.Ok, just insulting. Got it. Let me know when you wish to have an actual discussion. This is a discussion group after all. :)
On Thursday, June 16, 2022 at 6:22:32 PM UTC-4, Wayne morellini wrote:duplicates the many tools others have developed over the decades.
On Friday, June 17, 2022 at 1:11:05 AM UTC+10, gnuarm.del...@gmail.com wrote:
On Thursday, June 16, 2022 at 9:03:12 AM UTC-4, Wayne morellini wrote:
On Thursday, June 16, 2022 at 3:55:15 PM UTC+10, jpit...@gmail.com wrote:While Okad is not a universal tool, it does a job and does it well I would assume. But it was a *lot* of work and involved calibration from test chip fabrications. It is very unlikely to ever be reinvented. Then there is the fact that is largely
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:Oh well, somebody could maybe a similar freeware design
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
similarBut, you remember how Chuck had figured out the bugs, and could simply move from process to process? I forget if that was by changing parameters without constant test runs, or not. But Chuck did it, and now it's known, so somebody could do something
Most of the effort involved in coming up with a solution, is realising the solution. Now people realise how he did it, they can redo it.I'm not sure what you are talking about. Chuck did an iterative design on the tool. Each time he designed a tested a chip, he honed the assumptions the tool makes. This process is only valid for the process geometry nodes it has been honed to work with.
Of course, this is true for pretty much every semiconductor design tool. The difference is, most tools are used by many, many designers. Okad is used by a relative handful. As someone pointed out, it is really a rectangle drawing tool. So you need tocreate your own libraries of common elements... much like Forth.
--
Rick C.
+-+ Get 1,000 miles of free Supercharging
+-+ Tesla referral code - https://ts.la/richard11209
On Friday, June 17, 2022 at 10:57:46 AM UTC+10, gnuarm.del...@gmail.com wrote:duplicates the many tools others have developed over the decades.
On Thursday, June 16, 2022 at 6:22:32 PM UTC-4, Wayne morellini wrote:
On Friday, June 17, 2022 at 1:11:05 AM UTC+10, gnuarm.del...@gmail.com wrote:
On Thursday, June 16, 2022 at 9:03:12 AM UTC-4, Wayne morellini wrote:
On Thursday, June 16, 2022 at 3:55:15 PM UTC+10, jpit...@gmail.com wrote:While Okad is not a universal tool, it does a job and does it well I would assume. But it was a *lot* of work and involved calibration from test chip fabrications. It is very unlikely to ever be reinvented. Then there is the fact that is largely
On Wednesday, 15 June 2022 at 17:19:54 UTC+1, Wayne morellini wrote:Oh well, somebody could maybe a similar freeware design
Ok, I forgot. Chick made okcad free didn't he? Is that open sourced? Maybe we could use that? Anybody still got it?My last feedback from Greenarrays was, that okcad is a secret weapon of theirs, so not freely available.
Parts are on the internet, but not to the level of using it for a new design as I understand.
Another great Minimum Design of Chuck, but unfortunately not available for the world,
I have not seen anything comparable.
something similarBut, you remember how Chuck had figured out the bugs, and could simply move from process to process? I forget if that was by changing parameters without constant test runs, or not. But Chuck did it, and now it's known, so somebody could do
with.Most of the effort involved in coming up with a solution, is realising the solution. Now people realise how he did it, they can redo it.I'm not sure what you are talking about. Chuck did an iterative design on the tool. Each time he designed a tested a chip, he honed the assumptions the tool makes. This process is only valid for the process geometry nodes it has been honed to work
create your own libraries of common elements... much like Forth.Of course, this is true for pretty much every semiconductor design tool. The difference is, most tools are used by many, many designers. Okad is used by a relative handful. As someone pointed out, it is really a rectangle drawing tool. So you need to
--
Rick C.
+-+ Get 1,000 miles of free SuperchargingIt suites simple designs, but thats what we are looking at. You must have missed the updates where he can change parameters to suit much better than originally.
+-+ Tesla referral code - https://ts.la/richard11209
On Thursday, June 16, 2022 at 6:42:20 PM UTC-4, Wayne morellini wrote:
similar, without violating any NDAs.You are mistaken, you don't know what confidential products greenarrays had done.We've had this discussion before. If GA had a new product they would be blabing about the fact that they had money flowing through the company, even if they could not divulge any details. Stephen is able to acknowledge that he is working on something
I don't know about a lot of stuff in products,Yes, I get that.
and if I rang up manufacturers about it, I'm sure I would be denied knowledge of it as a trade secret. Greenarrays existed for many years with a large staff before getting difficulty. They obviously we're generating income to pay for it.That is patently not true. GA brags about getting by on a shoestring budget. I'm not sure any of them get paid. After the GA144 chip was released, they advertised for people to come, work in their offices to develop applications... for free.
application. It would have been better if one of the cores had been F21 like, to interface to the rest of the world, and stream process data with the rest of the array, bit by bit as a giant pipeline. That makes it more useful. You don't have to run theThe issue is, if they even had an F21 like design, I would have been using it. Look at the complete consumption of cores to simulate such a thing om the 144, in the recent thread. Sure it's a nice example, but the 144 is not designed towards that
Sorry, I don't get what you are talking about, but it's not important. You have once again gone off topic and have started to wistfully re-engineer the GA144 to no end. The GA144 is not going to be re-engineered. You are talking about designing a stackprocessor, but you can't stay on topic long enough to get through a single post.
I've not insulted anyone. I point out shortcomings in your approach to designing a "Forth CPU". That's not insulting, it factual. You can disagree with my thoughts, but that doesn't make them insults.I'm not knocking stack processors. I like them and use them in FPGAs. I'm just pointing out the facts of the matter. They are simply not a significant success, and are unlikely to ever become one.No you are not. There are some people in the world in general, who seem to take pleasure in knocking things and being negative towards people, and insulting as they don't get their way. Anybody around here know about that?
I'm not knocking stack processors. I'm discussing their limitations, which are mostly in acceptance in the market place. You seem to take every comment as an overly negative criticism.
If you know other facts, please present them, but if they aren't based in reality, don't expect me to ignore the errors.
On Thursday, June 16, 2022 at 7:22:56 PM UTC-4, Wayne morellini wrote:in the story presents his first draft to his boss saying the purpose was to demonstrate something and the goal was something else. His boss replied back, no, the demonstration is the goal. So what is your purpose?
On Friday, June 17, 2022 at 2:44:20 AM UTC+10, gnuarm.del...@gmail.com wrote:
On Thursday, June 16, 2022 at 12:12:53 PM UTC-4, Paul Rubin wrote:
Rick C <gnuarm.del...@gmail.com> writes:
I thought there were open source IC design tools? A post in a more relevant group might help find good tools and some people who might beThe Berkeley stuff from the 1980s was sort of comparable to OKAD in the
willing to assist someone getting started. This is a Forth group where there are few hardware oriented people.
sense of being rectangle editors with macros etc. HDL's weren't really a thing back then. Today in principle most of the code exists to go from HDL to rectangles, but a few steps are missing. I don't know if they are difficult.
Certainly even without FOSS tools, it is possible to make a chip without
spending money, by sending a HDL file to the fab service that runs it through closed source stuff.
I don't understand the thought process that starts with the idea of another Forth chip and jumps immediately to fabrication or even design tools. I'd start with figuring out the architecture, then simulating itEven before figuring out the architecture, I'd start with the purpose. I recall a humorous story about writing government IRAD reports. They had both a purpose and a goal. It was hard for most people to understand the difference. So the character
in software, then on an FPGA, before pursuing silicon fab. Otherwise it
is a solution in search of a problem.
stack processor".He takes the report back and rewrites it for his department manager. The department manager says, this isn't right, what you have for the purpose, is really your goal! So he takes it back again.
Now he presents the third draft to he division head who says, "No, this purpose is really your goal. What's your purpose?"
Finally, he is presenting the report to the government. He says, "My purpose is to get into heaven!"
Designing an MCU is something that needs to have it's purposes defined, well defined, which will guide all further decisions about the design, the goals. Maybe starting with something less supreme than getting into heaven, but higher than "making a
ones can be shipped without a year wait.I don't know of any markets without MCUs, very capable MCUs. There are so many of them you can walk without tripping over them. The hard part is picking just one! Well, maybe not in today's market where MCU selection is often done according to whichThe purpose is to reach a goal of a general purpose microcontroller into various markets.Fwiw I don't know if it is a stack arch, but Forth on the Parallax P2 isOk, then, "Mission Complete".
supposed to be pretty nice, and it is already in the rom.
Specific purpose designs are the goals of GA, which you have problems with :).That's not correct, and I don't have problems with GA.
of interference on the chip, with the lines being the main density, but with interfering abbarations around it. How closely can you tile the tolerances of a process like that.I forgot to mention, I would like to know how ok cad goes on a modern process fab node, as not only are structures more complex, but as you get smaller things become more fuzzy and less definable. Just looking at some mode processes, you see density
Okad will handle "modern" process nodes when it is completely rewritten to handle them. Going from 180 nm to 12 nm, is a *huge* difference. It's also not remotely feasible to attempt to use any process beyond what you can get in the low cost fabsbecause of the need to hand tune the software to match the process. This is both the feature size, but more importantly, the process provider. Every chip maker is a little different if not a *lot* different.
But you will be entertaining to watch. I suggest you not spend so much time responding to me, but focus on the work.
On Friday, June 17, 2022 at 9:04:36 AM UTC-4, Wayne morellini wrote:..
It suites simple designs, but thats what we are looking at. You must have missed the updates where he can change parameters to suit much better than originally.I think you would benefit from learning about semiconductor design. Between 180 nm and modern processes, the changes are not just a matter of tweaking a few parameters.
But none of this changes the basic fact that there is pretty much nothing significant about Okad as a chip design tool. There's nothing to protect really. It's like keeping secret, the manufacturing details on the original Ford V8.
That's not a putdown. It's a simple fact. Like all of Chuck's tools, it was designed to be used by him, to do the job at hand, at that time. Those are a lot of limitations.
--
Rick C.
++- Get 1,000 miles of free Supercharging
++- Tesla referral code - https://ts.la/richard11209
But, if you emulated arm mobile chipset on an GA, I don't think
you will get using less power. Maybe emulating the instruction set or JavaScript.
Wayne morellini <waynemo...@gmail.com> writes:
But, if you emulated arm mobile chipset on an GA, I don't thinkWait, is this the vision? Now we're in backyard moon rocket territory.
you will get using less power. Maybe emulating the instruction set or JavaScript.
Wayne morellini <waynemo...@gmail.com> writes:
But, if you emulated arm mobile chipset on an GA, I don't thinkWait, is this the vision? Now we're in backyard moon rocket territory.
you will get using less power. Maybe emulating the instruction set or JavaScript.
I'm sorry but I just can't read those long walls of text. Could you
give a description of the chip you're proposing, the kind of description
one would see on a data sheet? Not a 500 page manual or anything like
that. Just a few lines with the basics. Examples:
GREENARRAYS GA144
- 144 nodes of F18A MISC processors, about 750 mhz each
- 4 interconnect paths from each node, connecting to the adjacent
nodes in rectangular layout
- 18 bit word size, 64 words RAM and 64 words ROM per node, plus
8 level control stack and 10 level data stack at each node
ATMEGA ATTINY1616
- 8 bit word size, 32 registers, RISC-like architecture, 16 mhz
- 2KB ram, 16KB program flash
- GPIO, ADC, programmable timers yada yada
Just basic outline. Otherwise the picture is way too vague.
Next question: who would want to use it? That is, why wouldn't they use
chip XYZ instead, where XYZ is some popular existing chip?
A data sheet is not a chip. A chip is more like 500 pages.
On Saturday, June 18, 2022 at 1:07:07 PM UTC+10, Paul Rubin wrote:great deal of people dieing from 3 years and onwards.
I'm sorry but I just can't read those long walls of text. Could you
give a description of the chip you're proposing, the kind of description one would see on a data sheet? Not a 500 page manual or anything like that. Just a few lines with the basics. Examples:
GREENARRAYS GA144
- 144 nodes of F18A MISC processors, about 750 mhz each
- 4 interconnect paths from each node, connecting to the adjacent
nodes in rectangular layout
- 18 bit word size, 64 words RAM and 64 words ROM per node, plus
8 level control stack and 10 level data stack at each node
ATMEGA ATTINY1616
- 8 bit word size, 32 registers, RISC-like architecture, 16 mhz
- 2KB ram, 16KB program flash
- GPIO, ADC, programmable timers yada yada
Just basic outline. Otherwise the picture is way too vague.
Next question: who would want to use it? That is, why wouldn't they use chip XYZ instead, where XYZ is some popular existing chip?A data sheet is not a chip. A chip is more like 500 pages. Chuck's a lot less.
Paul, you should be able to read that. Ever since the vaccines, a lot of people can't understand so well. Are you feeling ok?
Ok, some info I shared before. Vaccines escape the muscle into lymph system and blood stream, around the body organs and brain
That's what I heard.
Spike proteins themselves are very harmful, and go in and damage the cells. It's of bad problems are possible
A vaccine can inject you with more spike proteins than the virus might.
Your cells build up a coating of debris and immune system components, your immune system sees this and thinks that is where the action is and attacks that. Producing a chronic inflammatory disease. This may happen.
They suspect that the chronic inflammatory condition is similar to the one after the Spanish flu out break. Where recovered people started getting sick say 3 months after recivery, and git sicker and sicker sitting in a semi vegetative state with a
It's bada clearly gut. Most are treatable, but racoon what it's name, parasite. Is one of the most deadly, with large death rate even with treatment. The normal form of it is also highly deadly without treatment and even with treatment it is supposed to be bad.
I used the Russian or Siberian pine needle tea treatment to nock the spikes down. I used vitamin C 6 to 20 grams distributed accross the day. I used other recommended things
Look up artery health and vitamin C, and cayenne chilli pepper. Maybe lookup Black tea, maybe bicarbonate soda (not belong powder, different) see what it says.
I used cryptopoles? for Lyme, and boy that's working, another thing that will crush your thinking. Fungus infection through body will, toxoplasmosis (in schizophrenia), parasites causing semi processed food leakage in gut will, causing irratable bowel
You should be able to read that. Reply to message and cut up the paragraphs into double lines with a blank carriage return line inbetwen. That should make it easier to read.
Wayne morellini <waynemo...@gmail.com> writes:
A data sheet is not a chip. A chip is more like 500 pages.The topic line talks about another Forth chip. You have already said
the cpu arch is MISC. Ok fine. I'm just asking for the basic features
of the chip. How many cpu cores, how much memory, etc? If you don't
have that figured out even approximately, it sounds way too early to
talk abot making a chip.
Wayne morellini <waynemo...@gmail.com> writes:
A data sheet is not a chip. A chip is more like 500 pages.The topic line talks about another Forth chip. You have already said
the cpu arch is MISC. Ok fine. I'm just asking for the basic features
of the chip. How many cpu cores, how much memory, etc? If you don't
have that figured out even approximately, it sounds way too early to
talk abot making a chip.
On Friday, June 17, 2022 at 11:57:31 PM UTC-4, Wayne morellini wrote:great deal of people dieing from 3 years and onwards.
On Saturday, June 18, 2022 at 1:07:07 PM UTC+10, Paul Rubin wrote:
I'm sorry but I just can't read those long walls of text. Could you
give a description of the chip you're proposing, the kind of description one would see on a data sheet? Not a 500 page manual or anything like that. Just a few lines with the basics. Examples:
GREENARRAYS GA144
- 144 nodes of F18A MISC processors, about 750 mhz each
- 4 interconnect paths from each node, connecting to the adjacent
nodes in rectangular layout
- 18 bit word size, 64 words RAM and 64 words ROM per node, plus
8 level control stack and 10 level data stack at each node
ATMEGA ATTINY1616
- 8 bit word size, 32 registers, RISC-like architecture, 16 mhz
- 2KB ram, 16KB program flash
- GPIO, ADC, programmable timers yada yada
Just basic outline. Otherwise the picture is way too vague.
Next question: who would want to use it? That is, why wouldn't they use chip XYZ instead, where XYZ is some popular existing chip?A data sheet is not a chip. A chip is more like 500 pages. Chuck's a lot less.
Paul, you should be able to read that. Ever since the vaccines, a lot of people can't understand so well. Are you feeling ok?
Ok, some info I shared before. Vaccines escape the muscle into lymph system and blood stream, around the body organs and brain
That's what I heard.
Spike proteins themselves are very harmful, and go in and damage the cells. It's of bad problems are possible
A vaccine can inject you with more spike proteins than the virus might.
Your cells build up a coating of debris and immune system components, your immune system sees this and thinks that is where the action is and attacks that. Producing a chronic inflammatory disease. This may happen.
They suspect that the chronic inflammatory condition is similar to the one after the Spanish flu out break. Where recovered people started getting sick say 3 months after recivery, and git sicker and sicker sitting in a semi vegetative state with a
bowel a clearly gut. Most are treatable, but racoon what it's name, parasite. Is one of the most deadly, with large death rate even with treatment. The normal form of it is also highly deadly without treatment and even with treatment it is supposed to beIt's bad
I used the Russian or Siberian pine needle tea treatment to nock the spikes down. I used vitamin C 6 to 20 grams distributed accross the day. I used other recommended things
Look up artery health and vitamin C, and cayenne chilli pepper. Maybe lookup Black tea, maybe bicarbonate soda (not belong powder, different) see what it says.
I used cryptopoles? for Lyme, and boy that's working, another thing that will crush your thinking. Fungus infection through body will, toxoplasmosis (in schizophrenia), parasites causing semi processed food leakage in gut will, causing irratable
You should be able to read that. Reply to message and cut up the paragraphs into double lines with a blank carriage return line inbetwen. That should make it easier to read.One thing I'm sure of. This conversation will never lead to any chips being designed. The discussion is far too amorphous, starting with vagueness and leading nowhere useful.
Whatever. What else is new?
--
Rick C.
+++ Get 1,000 miles of free Supercharging
+++ Tesla referral code - https://ts.la/richard11209
A data sheet is not a chip. A chip is more like 500 pages.
In article <2e774fb8-5186-402d...@googlegroups.com>,
Wayne morellini <waynemo...@gmail.com> wrote:
A data sheet is not a chip. A chip is more like 500 pages.If you can't do a data sheet, I will ignore you.
Groetjes Albert
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst
It is asking you guys a question. It's not saying I'm working out one
for you. I thought it might have been a good community effort spirit
excerize for us.
Those are not even the right questions to ask, when starting out. The
first questions are about the use of the chip. What does the chip
need to do to be useful in an application.
You don't even know what goes into a data sheet, a finale chip.
This is only a proposal stage.
Wayne morellini <waynemo...@gmail.com> writes:
You don't even know what goes into a data sheet, a finale chip.It isn't even that though. There's lots of MCU's out there. There is
This is only a proposal stage.
an implicit idea that this Forth chip can possibly designed in place of
some existing MCU. So I'm wondering what specific existing MCUs the
proposed Forth chip is comparable to. No answer so far.
Wayne morellini <waynemo...@gmail.com> writes:
It is asking you guys a question. It's not saying I'm working out oneI think it doesn't come up here because most people don't see a use for
for you. I thought it might have been a good community effort spirit excerize for us.
the idea. There have been Forth chips in the past, they had limited usefulness, there doesn't seem to be much use for them now, and that is
the current state of knowledge.
Re-opening the topic is fine if you have something new to add. So you
get a bunch of responses asking what about your idea we didn't know
already. It starts to sound like there wasn't anything.
Rick C <gnuarm.del...@gmail.com> writes:
Those are not even the right questions to ask, when starting out. The first questions are about the use of the chip. What does the chipWell, MCU's these days mostly fall into a few categories and the basics specs I mentioned are enough to identify the category. I know what the RP2040 does (two Cortex-M0 cores, 264KB ram, etc) and how it is used.
need to do to be useful in an application.
If someone said "I want to make a chip like the RP2040 except with
RISC-V cores instead of Cortex M0", I wouldn't have to ask about applications. It would work in mostly the same applications where the
RP2040 works, and be better in a few ways, by not needing workarounds
for the M0's limitations.
So I was hoping for something like that about Wayne's chip. E.g. "like
the Atmega328 but MISC", or "like the RP2040" etc.
I just don't know of any conventional MCU design that could benefit at
all from MISC. Even something like an RP2040 with MISC peripheral
processors seems a bit dubious, though maybe it's a good place to start.
That leaves unconventional designs (the GA144 is unconventional), but
those normally wouldn't be called general purpose.
On Saturday, June 18, 2022 at 2:03:52 PM UTC+10, gnuarm.del...@gmail.com wrote:
One thing I'm sure of. This conversation will never lead to any chips being designed. The discussion is far too amorphous, starting with vagueness and leading nowhere useful.
Whatever. What else is new?
Rick, the only thing specific here is a general purpose controller. The subject is not even that, it's about something to play with for you guys and use to gain work.
You keep putting vague stuff in here, like some target, but not specifying a target or description, just disrupting. A general purpose CPU/controller, is a target.
You say vague sounding things about GA got it wrong because they didn't have a target market.
Oh, they did, and I said that target market was going nearly disappear.
The problem is they didn't design it to be general purpose enough to suit more markets.
They simply didn't design it appealing enough, simple enough to develop for, or high enough functonality and performance.
Given those things, you could fit it many places. My ideas make it suitable for a wide range of market per chip. I'm wanting to do my own retro chip. And that is a lot narrower market. Except for my own platform, the chip itself would have limitedappeal, but the core is to be reusable in silicon products with a range of uses. So, while the processor development costs might look like it only has a certain return, it usefully could go into many things with good sales. Of you have 100 places trying
I'm actually thinking up a easy you might do custom rom on a chip after manufacture at the moment. That would be useful. Load chips into carrier, mass program them at once 1000 to one million at a time, in X seconds.
I forgot another original one. Testing the chip. Execution test code, activation pin flip or gate to remove faulty item. As the pins are under the stacking direction it tricky. But manipulation of the package can align a contact test device with codeto do durability testing on range of conditions, levels, timing, clocks etc, and custom in placement environment emulation tests to be sure. On the chip line, this could be fine as part of production.
But, as you guys weren't going do anything useful, I've put up a basic proposal above.
Let's just hope the the tools are as easy and logic related as they can be. I din't want to learn material physics or other analogue stuff (simulators to verify before prototyping).
Whatever happened to the greenarrays silicon design school thing? Really need to do that. "What! You mean that green thing is a transistor?". :)
Rick C <gnuarm.del...@gmail.com> writes:
Those are not even the right questions to ask, when starting out. The first questions are about the use of the chip. What does the chipWell, MCU's these days mostly fall into a few categories and the basics specs I mentioned are enough to identify the category. I know what the RP2040 does (two Cortex-M0 cores, 264KB ram, etc) and how it is used.
need to do to be useful in an application.
If someone said "I want to make a chip like the RP2040 except with
RISC-V cores instead of Cortex M0", I wouldn't have to ask about applications. It would work in mostly the same applications where the
RP2040 works, and be better in a few ways, by not needing workarounds
for the M0's limitations.
So I was hoping for something like that about Wayne's chip. E.g. "like
the Atmega328 but MISC", or "like the RP2040" etc.
I just don't know of any conventional MCU design that could benefit at
all from MISC. Even something like an RP2040 with MISC peripheral
processors seems a bit dubious, though maybe it's a good place to start.
That leaves unconventional designs (the GA144 is unconventional), but
those normally wouldn't be called general purpose.
On Sunday, June 19, 2022 at 7:36:24 AM UTC+10, Paul Rubin wrote:
Wayne morellini <waynemo...@gmail.com> writes:
It is asking you guys a question. It's not saying I'm working out oneI think it doesn't come up here because most people don't see a use for
for you. I thought it might have been a good community effort spirit excerize for us.
the idea. There have been Forth chips in the past, they had limited usefulness, there doesn't seem to be much use for them now, and that is
the current state of knowledge.
Re-opening the topic is fine if you have something new to add. So youThere has definitely been success, that makes sense. It's due to poor design decisions that their has been lack of appeal, that makes sense.
get a bunch of responses asking what about your idea we didn't know already. It starts to sound like there wasn't anything.
It wasn't about me design, that's commercial on confidence, trade secret, it was about you guys getting together. That makes sense.
I put forwards a proposal which is far ahead of what you guys are asking of 16 bits, that makes sense. Bit, not looking at that, and pretending it isnt anything, or not reading it, does not make sense.
On Sunday, June 19, 2022 at 7:33:11 AM UTC+10, Paul Rubin wrote:those, then it's an interesting hobby chip. What I describes probably blows the 16 bit markets away, and can go after 8 and 4 bits in many places. The thing you don't get is all the custom circuit features, like LTE modem , GPU (well on this design) etc.
Rick C <gnuarm.del...@gmail.com> writes:
Those are not even the right questions to ask, when starting out. The first questions are about the use of the chip. What does the chipWell, MCU's these days mostly fall into a few categories and the basics specs I mentioned are enough to identify the category. I know what the RP2040 does (two Cortex-M0 cores, 264KB ram, etc) and how it is used.
need to do to be useful in an application.
If someone said "I want to make a chip like the RP2040 except with
RISC-V cores instead of Cortex M0", I wouldn't have to ask about applications. It would work in mostly the same applications where the RP2040 works, and be better in a few ways, by not needing workarounds
for the M0's limitations.
So I was hoping for something like that about Wayne's chip. E.g. "like
the Atmega328 but MISC", or "like the RP2040" etc.
I just don't know of any conventional MCU design that could benefit at
all from MISC. Even something like an RP2040 with MISC peripheral processors seems a bit dubious, though maybe it's a good place to start.
That leaves unconventional designs (the GA144 is unconventional), but those normally wouldn't be called general purpose.Speed, energy, timeliness, and price. Because you don't design things like that, you don't realise that you can spec out the performance to meet a range of applications. It's all about performance and price metrics. If I designed a misc chip without
But. Once you go to high-end, or even mid range, the processing costs get prohibitive . Also, there is only certain applications that need top end pricesding. So, you can play with the layout and technologies to present a good offering , or out perform,the high end. On the low end. You. An emulate the chipset likely, and be close enough developers won't be bothered about the energy difference.
Don't bother mentioning the GA144 it's in a different market.
On Sunday, June 19, 2022 at 5:04:00 AM UTC-4, Wayne morellini wrote:
On Sunday, June 19, 2022 at 7:36:24 AM UTC+10, Paul Rubin wrote:
Wayne morellini <waynemo...@gmail.com> writes:
It is asking you guys a question. It's not saying I'm working out one for you. I thought it might have been a good community effort spirit excerize for us.I think it doesn't come up here because most people don't see a use for the idea. There have been Forth chips in the past, they had limited usefulness, there doesn't seem to be much use for them now, and that is the current state of knowledge.
Re-opening the topic is fine if you have something new to add. So you get a bunch of responses asking what about your idea we didn't know already. It starts to sound like there wasn't anything.There has definitely been success, that makes sense. It's due to poor design decisions that their has been lack of appeal, that makes sense.
It wasn't about me design, that's commercial on confidence, trade secret, it was about you guys getting together. That makes sense.
I put forwards a proposal which is far ahead of what you guys are asking of 16 bits, that makes sense. Bit, not looking at that, and pretending it isn't anything, or not reading it, does not make sense.So you are saying you are the "idea guy" and we all need to get behind your ideas and make them happen. Great!
The first problem is, what are you talking about???
On Monday, June 20, 2022 at 2:02:11 AM UTC+10, gnuarm.del...@gmail.com wrote:maintainers, might see you as reasonable, but I'm awake to you. Your logic is often flawed and not going far enough. Yet. You think you waste time.
On Sunday, June 19, 2022 at 5:04:00 AM UTC-4, Wayne morellini wrote:
On Sunday, June 19, 2022 at 7:36:24 AM UTC+10, Paul Rubin wrote:
Wayne morellini <waynemo...@gmail.com> writes:
It is asking you guys a question. It's not saying I'm working out oneI think it doesn't come up here because most people don't see a use for
for you. I thought it might have been a good community effort spirit excerize for us.
the idea. There have been Forth chips in the past, they had limited usefulness, there doesn't seem to be much use for them now, and that is
the current state of knowledge.
Re-opening the topic is fine if you have something new to add. So you get a bunch of responses asking what about your idea we didn't know already. It starts to sound like there wasn't anything.There has definitely been success, that makes sense. It's due to poor design decisions that their has been lack of appeal, that makes sense.
It wasn't about me design, that's commercial on confidence, trade secret, it was about you guys getting together. That makes sense.
I've got no time, so I'll answer this one for now.I put forwards a proposal which is far ahead of what you guys are asking of 16 bits, that makes sense. Bit, not looking at that, and pretending it isn't anything, or not reading it, does not make sense.So you are saying you are the "idea guy" and we all need to get behind your ideas and make them happen. Great!
Don't twist things again. You have mostly constantly been a no ideas man, just moaning and misunderstanding things. So you ask, I do, and instead of saying anything about it, you divert and subvert, uselessly. The ignorant masses, of contract
You seem to fail to understand that the reconfigurable proposal is capable of taking on the form of many MCU's competitively. But, I know when you avoid talking about something technical like this you have been proudly going on about for a whileerrantly, I've got it pretty spot on, but you don't want to admit it and look foolish. It's unreal, that you have been coming here for years doing this unobjective unreal stuff. Don't you have a real life Rick, to follow people around doing this? You
Now, you were asked to come up with something yourself. But, being in denial, you won't, but instead, just vaguely go on about objections to do with a limited mistaken view of design and marketing. When I am asked, and eventually come up with somethingfor you, you then accuse me of being the ideas man you have to follow, this trying to steal the attention away for people to follow yourself. You don't discuss the stuff requested instead. Stuff like that is not worth following. In this situation, you
You don't seem ti realise that customisable rom, is a descriptive way of saying PROM. You seem to get what I'm taking about mixed up with a particular Prom technology. This is not that.
The first problem is, what are you talking about???Yep. Effectively saying you lack sufficient intelligence and reasoning skills. When there is something sophisticated that is rational, logical and true, you say you don't understand. How is that going turn out in front of customers? Put in more effort.
I'll say what a lot of people are feeling for a while, who don't usually come here anymore. The place had to many losers. This has been increasing over time here.
If someone said "I want to make a chip like the RP2040 except with
RISC-V cores instead of Cortex M0"...
No one will give you money to make a chip with such a justification.
Why would anyone think it would sell?
It's a terrible MCU for low power apps, so there's a huge mobile
market gone. I don't know much else about the RP2040. What else is
it unsuited for?
Rick C <gnuarm.del...@gmail.com> writes:
It is what the RPI foundation should have made instead of the RP2040If someone said "I want to make a chip like the RP2040 except withNo one will give you money to make a chip with such a justification.
RISC-V cores instead of Cortex M0"...
Why would anyone think it would sell?
imho. Maybe the next iteration will be something like that.
It's a terrible MCU for low power apps, so there's a huge mobileIt's a midrange MCU with midrange power consumption, it seems to me.
market gone. I don't know much else about the RP2040. What else is
it unsuited for?
Comparable to the ESP32 series if you don't want wireless. Yes it uses
more power than some tiny 8 bit cpu. It wouldn't be good for a
wristwatch that has to run for years on a coin cell. On the other hand
it uses a lot less power than a mobile phone processor that runs Android
or whatever.
Phones are getting more and more power hungry rather than less. I think
web bloat is driving that, but don't know.
It's a bloody pair of M0+ cores! Hardly "mid-range".
I can't think of a comparable processor that uses as much idle power
as the RP2040. I also can't think of an application that it is more
suited to than other MCUs, except maybe on price. I think you said it
is only $1, or do I remember that wrong?
Rick C <gnuarm.del...@gmail.com> writes:
It's a bloody pair of M0+ cores! Hardly "mid-range".It runs at 125 mhz, has a pair of 32 bit cores and 264K of on-chip ram,
plus the PIO's, that memory crossbar system, etc. Low end would be an
Atmega or one of those Cortex M0's with 4k of ram. High end would be a Cortex M7 like on the Teensy 4, running at 600+ MHz with DSP stuff and double precision FPU, or maybe even a Cortex A series Linux chip. I
don't see how the RP2040 can be called anything but midrange.
I can't think of a comparable processor that uses as much idle powerYes, it's $1, comparable to the ESP32-C3 which has a RISC-V core plus
as the RP2040. I also can't think of an application that it is more
suited to than other MCUs, except maybe on price. I think you said it
is only $1, or do I remember that wrong?
400k ram and wifi. I haven't compared the idle power. I think actual
low powered stuff is almost a lost art. The Casio G-shock wristwatch
was introduced 40 years ago and it ran 10 years on a CR2025 coin cell. Today's Apple wristwatch has a non-removable lipo cell that has to be recharged every few days.
Why do we care about idle power anyway? It is important some of the
time. But we were also just talking about 4 bit MCU's in coffee pots.
Those are powered by AC wall current and they use 1000+ watts when the heater is running. So if the heater run for 5 minutes a day and the
idle power is 1 milliwatt, idling is less than 3% of the total energy consumption.
Ok, fine, call it mid-range. Now what are the apps it's superior in
compared to what other devices?
BTW, high end stuff is what shows up in cell phones, running at 1 GHz
The RP2040 M0 MCU runs about half as fast as an M3 at the same clock.
The reality is the RP2040 is still a low end processor.
That's exactly my point. You have no idea what applications the
RP2040 is useful in.
Rick C <gnuarm.del...@gmail.com> writes:
Ok, fine, call it mid-range. Now what are the apps it's superior in compared to what other devices?I dunno about superior, but it is reasonably adequate for tons of MCU
stuff. I think you said it is fine to make a chip that is ok at
everything instead of the best at one thing.
Fwiw I got an RP2040 board to use with an NDIR (nondispersive infrared)
CO2 detector module. That is a gas sensor with a heating element that
iirc uses a few hundred milliwatts. If you want to you run it
continuously, you power it with a wall wart. If you run it from a
battery, you power it up now and then to take a reading. Either way,
the MCU power consumption doesn't matter much.
BTW, high end stuff is what shows up in cell phones, running at 1 GHzHmm ok, I had not thought of those as MCU's, but fine.
The RP2040 M0 MCU runs about half as fast as an M3 at the same clock.
The reality is the RP2040 is still a low end processor.It's midrange in that it can run stuff that the once-popular
STM32F103CBT6 (Cortex M3, 72 mhz with 20k of ram) can't without a lot of pain. The extra ram helps a lot for MicroPython. Also, the 125mhz
clock makes up for a lot of the 2x ipc disadvantage, even before
bringing in the second M0 core or the PIO's.
I'm told that the GDVF103xxx (very similar chip to the STM32F103 but
with a RISC-V Bumblebee core) beats the M3 and maybe M4 in both compute speed and power consumption, so that's another reason to use RISC-V in
the RP2040.
My understanding is that the RPI Foundation chose the M0
because they didn't have to pay ARM license fees that they would have
had to if they wanted to use the M4. They put in various kludges like
an outboard hardware divider to get around the M0's lack of one. But
RISC-V would have sidestepped all of that.
That's exactly my point. You have no idea what applications theTake any list of 1000 microprocessor applications. Cross off the ones
RP2040 is useful in.
where the RP2040 won't work. There will still be plenty left. That's
all that matters. The RP2040 is cheap, readily available, well
supported, well documented, and easy to program. You can buy the Pi
Pico board for $4, plug it into USB, drag and drop a software file to
it, press boot, and it is running MicroPython interactively. Most
anything else is a world of pain by comparison.
So why does the RP2040 do a better job than other devices?
with a RISC-V Bumblebee core) beats the M3 and maybe M4When you say beats, is this by 20%, 50%, 1%?
Is Giga taking over the market? No.
I'd like to see something that confirms that. So you are saying the
RP2040 is a kludge? I wasn't aware of that. I suppose the divide is
very slow then? Ok, good to know.
I don't typically have a use for an MCU with a mA range idle current.
That takes the RP2040 off my list of candidates for virtually
everything I do.
You can buy the Pi
Pico board for $4, plug it into USB, drag and drop a software file to
it, press boot, and it is running MicroPython interactively. Most
anything else is a world of pain by comparison.
[Raspberry Pi Pico] Until you discover that you have to power cycle
the board and the only way is to disconnect the USB cable.
Compare that with Willem Ouwerkerk MSP430 board. You can flash the
basic Forth, flash an application, remove the application, and
flash an improved one. Or flash the basic Forth afresh.
Now that is painless.
Wayne morellini <waynemo...@gmail.com> writes:
Commercially myself, I am more interested in simple printed circuitsI couldn't read that whole long post, but tons of very interesting chips were done in 2.5 micron and larger sizes. The Mead-Conway revolution happened in the 3 to 5 micron era. There was a #homecmos channel on
at 2.5 micro to 1 micron. I think I finally have a clean room
solution to do this at home. That's more doable on an individual's
level.
freenode for a while that was trying to do stuff at, iirc, something
like 12 microns. If you have an affordable way to make 2.5 micron chips
at home, that is really quite revolutionary, especially if you can do
mixed signal.
As for a MISC-like chip though, yes of course there is skepticism: who
would want a thing like that, at least as a separate chip rather than a macro cell? DIY satisfaction (and I'm all in favor of that) seems like
the main reason to make it, unless you've got some pretty clear and quantitative claims about how it would outperform conventional chips at
some meaningful application.
This morning I was thinking it would be interesting to have a chip with
wide but single core SMT, to allow super fast coroutine switching
without having to save and restore registers, sort of like a Forth multitasker (that only has a couple of registers to save and restore, so
the task switcher is fast), but allowing something more like
conventional OS's and compilers, which do use lots of registers. That
might be more interesting than a MISC chip. It could be a RISC-V with
a special instruction added to call another context, and there might be
8 or so contexts available in a core, sort of like the Parallax
Propeller.
wide but single core SMT, to allow super fast coroutine switching ...If you have lots of registers Inna chop to address, you expand your
average instruction size, increase complexity, increase the amount of
energy per instruction ...
Wayne morellini <waynemo...@gmail.com> writes:
wide but single core SMT, to allow super fast coroutine switching ...If you have lots of registers Inna chop to address, you expand your
average instruction size, increase complexity, increase the amount of energy per instruction ...
Nono the idea is not to have lots of named registers in an execution
context, but rather to have lots of execution contexts (like maybe 8 of them), like the Propeller does. The Propeller itself is pretty neat,
though quite niche. But the current version (Propeller 2) has Forth
built in...
So Parallax has gone Forth, is the propellor Misc?
In article <87pmj3n...@nightsong.com>,
Paul Rubin <no.e...@nospam.invalid> wrote:
You can buy the PiUntil you discover that you have to power cycle the board and the only
Pico board for $4, plug it into USB, drag and drop a software file to
it, press boot, and it is running MicroPython interactively. Most
anything else is a world of pain by comparison.
way is to disconnect the USB cable.
Compare that with Willem Ouwerkerk MSP430 board. You can flash the
basic Forth, flash an application, remove the application, and
flash an improved one. Or flash the basic Forth afresh.
Now that is painless.
I see nobody is really interested in Forth or Forth chips. Such a shame, such a loss. Not even GA seems that interested.
A word on what Stephen said. Highly specialised chip, can be code for a specialised chip lacking general purpose features, not available for general sale. Let's hope that it is made of macro components, from which the core can be licensed, otherwise weare back at square one again.
Rick C <gnuarm.del...@gmail.com> writes:
So why does the RP2040 do a better job than other devices?1. I got the board for free
2. Easy to program as I mentioned below. Plenty of cpu, ram, and
software tooling, particularly MicroPython ready to go.
3. Board has convenient JST peripheral connector that talks to an
ecosystem of peripherals (such as the CO2 sensor) that use this
connector. RP2040 software knows how to communicate with these
peripherals.
4. Very thorough documentation and well informed dev community.
What other device do you think would have done this better than the
RP2040? ESP32 is a possibility, but is in the same general class.
Supposedly by 3x, but I haven't measured.with a RISC-V Bumblebee core) beats the M3 and maybe M4When you say beats, is this by 20%, 50%, 1%?
Is Giga taking over the market? No.Espressif has done a lot of taking over despite weird instruction sets
and other issues. The Giga part is kind of niche and hard to find.
Maybe more interesting stuff will appear.
I'd like to see something that confirms that. So you are saying theIdk if the divide is slow compared to an M3 or whatever. The kludge is
RP2040 is a kludge? I wasn't aware of that. I suppose the divide is
very slow then? Ok, good to know.
that it's an add-on peripheral instead of a normal machine instruction.
From page 31 of the data sheet ( https://datasheets.raspberrypi.com/pico/pico-datasheet.pdf ):
2.3.1.5. Integer Divider
The SIO provides one 8-cycle signed/unsigned divide/modulo module to
each of the cores. Calculation is started by writing a dividend and
divisor to the two argument registers, DIVIDEND and DIVISOR . The
divider calculates the quotient / and remainder % of this division over
the next 8 cycles, and on the 9th cycle the results can be read from the
two result registers DIV_QUOTIENT and DIV_REMAINDER. A 'ready' bit in register DIV_CSR can be polled to wait for the calculation to complete,
or software can insert a fixed 8-cycle delay.
I don't typically have a use for an MCU with a mA range idle current.It's off of my list for some things too, for the same reason, but not
That takes the RP2040 off my list of candidates for virtually
everything I do.
for anywhere near "almost everything". And the idle current can be
perhaps fixed by having a MISC ULP that keeps an eye on things while the main processors are shut down.
I might not be reading the data sheet (p. 619) properly but it looks
like the "dormant" (whatever that means) current is 0.18 mA and the
sleep current is 0.39 mA. Neither of those is so terrible. They
correspond to around a year of on-time with an 18650 cell, or months
with a small lipo pouch cell. That is about what I see from tons of
small gadgets.
More constructively, what device would you suggest instead of the
RP2040? Requirement #1 is cheap available well-supported dev boards.
They can be more expensive than RP2040 boards but not by too much. Requirement #2 is running MicroPython out of the box, or as close to
that as possible.
What other device do you think would have done this better than theWould have done what, fit on your desk?
RP2040? ESP32 is a possibility, but is in the same general class.
I select components for designs. I can never remember what you do for
a living, but it sure does not look like you design electronics using
MCUs.
Supposedly by 3x, but I haven't measured."Supposedly"? Where did you read this?
Did someone actually produce a benchmark? There are a million other parameters when selecting a part. I get that you judge MCUs the same
way you pick your favorite sports car.
It is not important that you call it a kludge.
sleep current is 0.39 mA.Sorry, they are orders of magnitude higher than most parts.
Provide a set of requirements, based on use cases for a product, and
I'll be happy to select your part for you... for my usual hourly rate,
I'm already rejecting the "cheap" development board requirement,
because that was never relevant to any project I've ever worked on.
That sounds like a hobby project.
You find a new "toy" product that sounds "neat" and then try to find
uses for it.
if I return, I will simply use an rPi I/O port pin to control a power
switch, constructed myself. Yeah, "simply". That's the ticket!
Wayne morellini <waynemo...@gmail.com> writes:
So Parallax has gone Forth, is the propellor Misc?No, it's closer to a conventional register-oriented design, except for
the multitasking ("propeller") feature. There have always been software Forths for the Propeller chip (original version, Propeller 1) going way back. The new version (Propeller 2) has it in rom so it is possible to
boot the chip directly to a Forth prompt. But it is still a software
Forth.
The guy who did the port, Peter Jakacki, is or was a regular here. See:
https://forums.parallax.com/discussion/167868/taqoz-tachyon-forth-for-the-p2-boot-rom
and: https://forums.parallax.com/categories/forthspace
MISC doesn't have much traction these days because, imho, there are big microprocessors, medium sized ones, small ones, and itsy-bitsy-teeny
ones. MISC is of value mostly in the itsy-bitsy-teeny region, and with advancing technology, that region itself (at least for packaged chips)
has become less interesting. Thus, the idea that MISC cores might be
more useful as peripheral processors on medium and bigger chips, than as main processors on small or tiny chips.
On Tuesday, June 21, 2022 at 8:16:27 AM UTC-4, Wayne morellini wrote:
I see nobody is really interested in Forth or Forth chips. Such a shame, such a loss. Not even GA seems that interested.
we are back at square one again.A word on what Stephen said. Highly specialised chip, can be code for a specialised chip lacking general purpose features, not available for general sale. Let's hope that it is made of macro components, from which the core can be licensed, otherwise
Specialized for what??? Ok, let's talk about it. What do you want to specialize it for?
--
Rick C.
++-- Get 1,000 miles of free Supercharging
++-- Tesla referral code - https://ts.la/richard11209
Ok, who would like to go into the Forth chip business with me?
On Wednesday, June 22, 2022 at 7:57:23 PM UTC+10, Wayne morellini wrote:one chip in a large company.
Ok, who would like to go into the Forth chip business with me?- I'm interested in an angle of modular electronics you can use professionally, and sell into the Pi and Arduino markets too.
- able to act as drop in replacements for parts. The architecture is very light physycally, so can fit on board made for purpose.
- To use as much of the open architecture as desirable.
- Something I have been thinking about for years, much better than the clumsy form factors used with them. Better than the Google modular concept, which though smaller than the PI or Arduino, was still bulky fit phones.
- old and new interfaces.
- optical interconnects.
- old and new products.
While I'm not actually against doing the products Rick sort, it's to finance general chips aswell. Coffee maker chips are probably saturated and out. But, the chip I proposed is capable of doing most things in software. It is a major boost to only need
I had planned a language decades ago without need to do the layout in a stack load fashion )the compiler handles) so it is close to normal languages. But you may do it the normal Forth way also.
Sell macro cell licensing.
Sell products based on.
With a number of irons in the fire, one or more may stick. But their combined number might produce enough revenue to keep going to do better in revenue.
Hi Wayne,Assembly, as well as FPGA using Verilog. And most important, my new passion on Forth and its HW implementation. Recently I got my GA144 EVB002 and currently learning arrayForth using the colorForth version. But later will also learn the saneForth version
I'm new here as well as new to Forth programming. I am interested to join your venture on Forth chip business, but I'm afraid I cannot offer any support related to financial. What I can offer though is my experience on embedded systems using C and
On Thursday, June 23, 2022 at 1:01:19 PM UTC+10, Geo wrote:Assembly, as well as FPGA using Verilog. And most important, my new passion on Forth and its HW implementation. Recently I got my GA144 EVB002 and currently learning arrayForth using the colorForth version. But later will also learn the saneForth version
Hi Wayne,
I'm new here as well as new to Forth programming. I am interested to join your venture on Forth chip business, but I'm afraid I cannot offer any support related to financial. What I can offer though is my experience on embedded systems using C and
Great.bits. With that a 4 bit versions possible. By using a simple DMA circuit and a few other things. GA didn't get it right, there are simple ways to do cache, interprocessor high speed comms, global memory access and comms, and other things I can't remember
Yes, a new functional based chip, rather than a more purely forth based. But not like the current one. It's designed to be orientated to real world tasks.
There is actually a lot happening, and it's going on. It's going take a while to speed up.
Can we go slow for the time being? If everything is good, look at preliminary steps and see who turns up along the way. Mostly pre steps before funded production.
I'm actually interested in pursuing a thumbish like architecture strategy, that can be used on other chips for low power mode. Jeff's work on machine forth identified that. I'm planning on dumping some instructions to get the instruction set into 4
On Thursday, June 23, 2022 at 11:08:15 PM UTC+10, Wayne morellini wrote:Assembly, as well as FPGA using Verilog. And most important, my new passion on Forth and its HW implementation. Recently I got my GA144 EVB002 and currently learning arrayForth using the colorForth version. But later will also learn the saneForth version
On Thursday, June 23, 2022 at 1:01:19 PM UTC+10, Geo wrote:
Hi Wayne,
I'm new here as well as new to Forth programming. I am interested to join your venture on Forth chip business, but I'm afraid I cannot offer any support related to financial. What I can offer though is my experience on embedded systems using C and
bits. With that a 4 bit versions possible. By using a simple DMA circuit and a few other things. GA didn't get it right, there are simple ways to do cache, interprocessor high speed comms, global memory access and comms, and other things I can't rememberGreat.
Yes, a new functional based chip, rather than a more purely forth based. But not like the current one. It's designed to be orientated to real world tasks.
There is actually a lot happening, and it's going on. It's going take a while to speed up.
Can we go slow for the time being? If everything is good, look at preliminary steps and see who turns up along the way. Mostly pre steps before funded production.
I'm actually interested in pursuing a thumbish like architecture strategy, that can be used on other chips for low power mode. Jeff's work on machine forth identified that. I'm planning on dumping some instructions to get the instruction set into 4
I'm going to contact Green Arrays about using OKCAD and their silicon school. I'm more a hands in that type of way. I've thought about a way to do low energy at small feature size, and might have come up with an answer.
On Friday, 24 June 2022 at 11:46:09 UTC+1, Wayne morellini wrote:and Assembly, as well as FPGA using Verilog. And most important, my new passion on Forth and its HW implementation. Recently I got my GA144 EVB002 and currently learning arrayForth using the colorForth version. But later will also learn the saneForth
On Thursday, June 23, 2022 at 11:08:15 PM UTC+10, Wayne morellini wrote:
On Thursday, June 23, 2022 at 1:01:19 PM UTC+10, Geo wrote:
Hi Wayne,
I'm new here as well as new to Forth programming. I am interested to join your venture on Forth chip business, but I'm afraid I cannot offer any support related to financial. What I can offer though is my experience on embedded systems using C
bits. With that a 4 bit versions possible. By using a simple DMA circuit and a few other things. GA didn't get it right, there are simple ways to do cache, interprocessor high speed comms, global memory access and comms, and other things I can't rememberGreat.
Yes, a new functional based chip, rather than a more purely forth based. But not like the current one. It's designed to be orientated to real world tasks.
There is actually a lot happening, and it's going on. It's going take a while to speed up.
Can we go slow for the time being? If everything is good, look at preliminary steps and see who turns up along the way. Mostly pre steps before funded production.
I'm actually interested in pursuing a thumbish like architecture strategy, that can be used on other chips for low power mode. Jeff's work on machine forth identified that. I'm planning on dumping some instructions to get the instruction set into 4
I'm going to contact Green Arrays about using OKCAD and their silicon school. I'm more a hands in that type of way. I've thought about a way to do low energy at small feature size, and might have come up with an answer.Good luck with Greg.
When I tried to get more details a few years ago,
his answer was,
this OKCAD very valuable
and there is no additional info available than what is on the Internet already.
I wonder, how many people on this planet have actually worked with this software,
and more importantly:
which current processes can it be adapted to,
it was on 180nm MOSIS if I remember correctly.
I received an inquiry from Wayne this morning and thought I'd do a little searching to learn where he was coming from. That led me to this group and now that I've made it through this thread a reply seems appropriate. So I shall write one post hopingto address a few things but will not be back to the group per se. gr...@greenarraychips.com if you wish to discuss anything further. Will try to stick to facts and avoid opinions.
First, we have not yet devoted full time to seeking funding for our 501(c)(3) nonprofit called POLYSANCE, founded in Wyoming, whose charter is to teach young people to program and to design semiconductors in our fashion. By "young" we are principallythinking of bright kids not yet out of high school, because (1) they can certainly understand what is required, and (2) they have not yet been told that FORTH sucks and that asynchronous computers are infeasible. Kids will use a form of GLOW, the
Second, OKAD has produced chips in 180 and 130nm CMOS, the latter with 5V I/O. GLOW has produced 180nm CMOS and 28nm designs that passed all design rule checks and simulated successfully. The time to adjust our tiled design methods to a new geometrywas minimal; the effort is in addressing increasingly bizarre design rules that are visible in the higher level design, which thanks to foundry nondisclosures can't be discussed.
Third, modern processes cost fabulous prices while, until FinFETs emerge (literally :), leakage increases fabulously as well. Most of us cannot individually afford the $80k (2010 prices) for a mask set at 180nm, let alone the $6e6 at 28nm. When I askedGlobal Foundries for the design rules pertinent to 14nm FinFETs they told me they'd allow me to see them if I paid them $25 million for a *shuttle run*. POLYSANCE will be looking among other things for a good complementary process that can be done with
Fourth, a big reason SWATCH group owns a 180nm fab is the same reason we like that process: Decent performance, low leakage, low cost. We met a design group of SWATCH in Colorado Springs who designed chips for Logitech and others who also appreciatelow leakage.
Fifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch and therefore, unlike most others, we do not have to pay on the order of $100k per year per guy to rent the software forthose guys to use. For asynchronous designs like ours analog simulation is also required. We paid $12k for a commercial license for TSpice which took four MONTHS to simulate sweeping a handful of transistors to get IV curves. In the past I watched
Sixth, our chips have driven SDRAM and John Rible's code to do so is in fact in ROM in nodes 7, 8, 9, 107, 108. The source code for those ROMs is included in the free distribution of arrayForth-3 starting in block 2241; to know that one need only look.made aware of any such inquiry. I don't know who provided that answer; they must at minimum have been distracted at the time. At any rate, you have GreenArrays' abject apology for such stupid advice and if you ask again by email to hot...@greenarraychips.
Seventh, we routinely run external timing devices (crystals) for exactly the reasons cited earlier in this thread when the application requires a time base meeting stringent specifications. See for example AN002 and AN012.
Eighth, someone above said he'd asked "them" for timing details on the memory interface nodes in the context of wanting to drive DRAMs, and had been advised to just "tinker". I've been here for all the time we were called GreenArrays, Inc. and was not
FINALLY, GreenArrays invites constructive dialog via direct email. We decided early on that we did not have the lifespan available to spend engaging chat groups, which requires reading them forever after. I read comp.lang.forth for long enough to graspthe problem.
With that, then, I wish you all well, and greetings to Anton and other names I recognize over a period of many years.
- Greg Bailey, President, GreenArrays, Inc.
Fifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW
*are* our crown jewels. We wrote these tools from scratch and
therefore, unlike most others, we do not have to pay on the order of
$100k per year per guy to rent the software for those guys to use.
I received an inquiry from Wayne this morning and thought I'd do a little searching to learn where he was coming from. That led me to this group and now that I've made it through this thread a reply seems appropriate. So I shall write one post hopingto address a few things but will not be back to the group per se. gr...@greenarraychips.com if you wish to discuss anything further. Will try to stick to facts and avoid opinions.
First, we have not yet devoted full time to seeking funding for our 501(c)(3) nonprofit called POLYSANCE, founded in Wyoming, whose charter is to teach young people to program and to design semiconductors in our fashion. By "young" we are principallythinking of bright kids not yet out of high school, because (1) they can certainly understand what is required, and (2) they have not yet been told that FORTH sucks and that asynchronous computers are infeasible. Kids will use a form of GLOW, the
Second, OKAD has produced chips in 180 and 130nm CMOS, the latter with 5V I/O. GLOW has produced 180nm CMOS and 28nm designs that passed all design rule checks and simulated successfully. The time to adjust our tiled design methods to a new geometrywas minimal; the effort is in addressing increasingly bizarre design rules that are visible in the higher level design, which thanks to foundry nondisclosures can't be discussed.
Third, modern processes cost fabulous prices while, until FinFETs emerge (literally :), leakage increases fabulously as well. Most of us cannot individually afford the $80k (2010 prices) for a mask set at 180nm, let alone the $6e6 at 28nm. When I askedGlobal Foundries for the design rules pertinent to 14nm FinFETs they told me they'd allow me to see them if I paid them $25 million for a *shuttle run*. POLYSANCE will be looking among other things for a good complementary process that can be done with
Fourth, a big reason SWATCH group owns a 180nm fab is the same reason we like that process: Decent performance, low leakage, low cost. We met a design group of SWATCH in Colorado Springs who designed chips for Logitech and others who also appreciatelow leakage.
Fifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch and therefore, unlike most others, we do not have to pay on the order of $100k per year per guy to rent the software forthose guys to use. For asynchronous designs like ours analog simulation is also required. We paid $12k for a commercial license for TSpice which took four MONTHS to simulate sweeping a handful of transistors to get IV curves. In the past I watched
Sixth, our chips have driven SDRAM and John Rible's code to do so is in fact in ROM in nodes 7, 8, 9, 107, 108. The source code for those ROMs is included in the free distribution of arrayForth-3 starting in block 2241; to know that one need only look.made aware of any such inquiry. I don't know who provided that answer; they must at minimum have been distracted at the time. At any rate, you have GreenArrays' abject apology for such stupid advice and if you ask again by email to hot...@greenarraychips.
Seventh, we routinely run external timing devices (crystals) for exactly the reasons cited earlier in this thread when the application requires a time base meeting stringent specifications. See for example AN002 and AN012.
Eighth, someone above said he'd asked "them" for timing details on the memory interface nodes in the context of wanting to drive DRAMs, and had been advised to just "tinker". I've been here for all the time we were called GreenArrays, Inc. and was not
FINALLY, GreenArrays invites constructive dialog via direct email. We decided early on that we did not have the lifespan available to spend engaging chat groups, which requires reading them forever after. I read comp.lang.forth for long enough to graspthe problem.
With that, then, I wish you all well, and greetings to Anton and other names I recognize over a period of many years.
- Greg Bailey, President, GreenArrays, Inc.
On Thursday, June 23, 2022 at 11:08:15 PM UTC+10, Wayne morellini wrote:Assembly, as well as FPGA using Verilog. And most important, my new passion on Forth and its HW implementation. Recently I got my GA144 EVB002 and currently learning arrayForth using the colorForth version. But later will also learn the saneForth version
On Thursday, June 23, 2022 at 1:01:19 PM UTC+10, Geo wrote:
Hi Wayne,
I'm new here as well as new to Forth programming. I am interested to join your venture on Forth chip business, but I'm afraid I cannot offer any support related to financial. What I can offer though is my experience on embedded systems using C and
bits. With that a 4 bit versions possible. By using a simple DMA circuit and a few other things. GA didn't get it right, there are simple ways to do cache, interprocessor high speed comms, global memory access and comms, and other things I can't rememberGreat.
Yes, a new functional based chip, rather than a more purely forth based. But not like the current one. It's designed to be orientated to real world tasks.
There is actually a lot happening, and it's going on. It's going take a while to speed up.
Can we go slow for the time being? If everything is good, look at preliminary steps and see who turns up along the way. Mostly pre steps before funded production.
I'm actually interested in pursuing a thumbish like architecture strategy, that can be used on other chips for low power mode. Jeff's work on machine forth identified that. I'm planning on dumping some instructions to get the instruction set into 4
I'm going to contact Green Arrays about using OKCAD and their silicon school. I'm more a hands in that type of way. I've thought about a way to do low energy at small feature size, and might have come up with an answer.
Greg Bailey <honky...@gmail.com> writes:
Fifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch andThis is the first I've heard of GLOW. I saw a video demo of OKAD
therefore, unlike most others, we do not have to pay on the order of
$100k per year per guy to rent the software for those guys to use.
though, and it was basically a rectangle/tile editor and timing
simulator from what I could tell, similar to the stuff written at
Berkeley and elsewhere in the 1980s. It's one thing to say you don't
want to release the code, but I wonder if there would be any issue with
just giving some description of its advantages over CAESAR, MAGIC, and
that other old timey software.
https://en.wikipedia.org/wiki/Magic_(software)
On Saturday, June 25, 2022 at 8:59:34 AM UTC+10, Paul Rubin wrote:this, the more I think a simpler ignored manufacturing process, which can't take a regular PC design, could take this design. I'm starting to wonder about silicon stamp frabrication technology, which should be cheap and maybe reach the 180nm design sizes,
Greg Bailey <honky...@gmail.com> writes:
Fifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch andThis is the first I've heard of GLOW. I saw a video demo of OKAD
therefore, unlike most others, we do not have to pay on the order of $100k per year per guy to rent the software for those guys to use.
though, and it was basically a rectangle/tile editor and timing
simulator from what I could tell, similar to the stuff written at
Berkeley and elsewhere in the 1980s. It's one thing to say you don't
want to release the code, but I wonder if there would be any issue with just giving some description of its advantages over CAESAR, MAGIC, and that other old timey software.
https://en.wikipedia.org/wiki/Magic_(software)
Thank for that Paul. Looks interesting. Is it less simple than OKCad? I can do the simple digital circuit, but dealing with analogue features might be behind my ability. The hatch patterns are definitely hard to look through. :). The more I consider
Is there anything else apart from magic and Ceasar?
Sorry for the delay in replying, under it here.
Thanks.
On Saturday, June 25, 2022 at 2:29:43 AM UTC+10, Greg Bailey wrote:to address a few things but will not be back to the group per se. gr...@greenarraychips.com if you wish to discuss anything further. Will try to stick to facts and avoid opinions.
I received an inquiry from Wayne this morning and thought I'd do a little searching to learn where he was coming from. That led me to this group and now that I've made it through this thread a reply seems appropriate. So I shall write one post hoping
thinking of bright kids not yet out of high school, because (1) they can certainly understand what is required, and (2) they have not yet been told that FORTH sucks and that asynchronous computers are infeasible. Kids will use a form of GLOW, theFirst, we have not yet devoted full time to seeking funding for our 501(c)(3) nonprofit called POLYSANCE, founded in Wyoming, whose charter is to teach young people to program and to design semiconductors in our fashion. By "young" we are principally
was minimal; the effort is in addressing increasingly bizarre design rules that are visible in the higher level design, which thanks to foundry nondisclosures can't be discussed.Second, OKAD has produced chips in 180 and 130nm CMOS, the latter with 5V I/O. GLOW has produced 180nm CMOS and 28nm designs that passed all design rule checks and simulated successfully. The time to adjust our tiled design methods to a new geometry
asked Global Foundries for the design rules pertinent to 14nm FinFETs they told me they'd allow me to see them if I paid them $25 million for a *shuttle run*. POLYSANCE will be looking among other things for a good complementary process that can be doneThird, modern processes cost fabulous prices while, until FinFETs emerge (literally :), leakage increases fabulously as well. Most of us cannot individually afford the $80k (2010 prices) for a mask set at 180nm, let alone the $6e6 at 28nm. When I
low leakage.Fourth, a big reason SWATCH group owns a 180nm fab is the same reason we like that process: Decent performance, low leakage, low cost. We met a design group of SWATCH in Colorado Springs who designed chips for Logitech and others who also appreciate
those guys to use. For asynchronous designs like ours analog simulation is also required. We paid $12k for a commercial license for TSpice which took four MONTHS to simulate sweeping a handful of transistors to get IV curves. In the past I watchedFifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch and therefore, unlike most others, we do not have to pay on the order of $100k per year per guy to rent the software for
look.Sixth, our chips have driven SDRAM and John Rible's code to do so is in fact in ROM in nodes 7, 8, 9, 107, 108. The source code for those ROMs is included in the free distribution of arrayForth-3 starting in block 2241; to know that one need only
not made aware of any such inquiry. I don't know who provided that answer; they must at minimum have been distracted at the time. At any rate, you have GreenArrays' abject apology for such stupid advice and if you ask again by email to hot...@Seventh, we routinely run external timing devices (crystals) for exactly the reasons cited earlier in this thread when the application requires a time base meeting stringent specifications. See for example AN002 and AN012.
Eighth, someone above said he'd asked "them" for timing details on the memory interface nodes in the context of wanting to drive DRAMs, and had been advised to just "tinker". I've been here for all the time we were called GreenArrays, Inc. and was
grasp the problem.FINALLY, GreenArrays invites constructive dialog via direct email. We decided early on that we did not have the lifespan available to spend engaging chat groups, which requires reading them forever after. I read comp.lang.forth for long enough to
With that, then, I wish you all well, and greetings to Anton and other names I recognize over a period of many years.
- Greg Bailey, President, GreenArrays, Inc.I thank you Greg. I understand. Rather than chat here, I propose a constructive FAQ like info base on your website, for marketing purposes, that can answer many questions, and occasionally publishing a link here?
I will suggest the same funding source I aim to approach .
On Saturday, 25 June 2022 at 02:57:52 UTC+1, Wayne morellini wrote:hoping to address a few things but will not be back to the group per se. gr...@greenarraychips.com if you wish to discuss anything further. Will try to stick to facts and avoid opinions.
On Saturday, June 25, 2022 at 2:29:43 AM UTC+10, Greg Bailey wrote:
I received an inquiry from Wayne this morning and thought I'd do a little searching to learn where he was coming from. That led me to this group and now that I've made it through this thread a reply seems appropriate. So I shall write one post
principally thinking of bright kids not yet out of high school, because (1) they can certainly understand what is required, and (2) they have not yet been told that FORTH sucks and that asynchronous computers are infeasible. Kids will use a form of GLOW,First, we have not yet devoted full time to seeking funding for our 501(c)(3) nonprofit called POLYSANCE, founded in Wyoming, whose charter is to teach young people to program and to design semiconductors in our fashion. By "young" we are
geometry was minimal; the effort is in addressing increasingly bizarre design rules that are visible in the higher level design, which thanks to foundry nondisclosures can't be discussed.Second, OKAD has produced chips in 180 and 130nm CMOS, the latter with 5V I/O. GLOW has produced 180nm CMOS and 28nm designs that passed all design rule checks and simulated successfully. The time to adjust our tiled design methods to a new
asked Global Foundries for the design rules pertinent to 14nm FinFETs they told me they'd allow me to see them if I paid them $25 million for a *shuttle run*. POLYSANCE will be looking among other things for a good complementary process that can be doneThird, modern processes cost fabulous prices while, until FinFETs emerge (literally :), leakage increases fabulously as well. Most of us cannot individually afford the $80k (2010 prices) for a mask set at 180nm, let alone the $6e6 at 28nm. When I
appreciate low leakage.Fourth, a big reason SWATCH group owns a 180nm fab is the same reason we like that process: Decent performance, low leakage, low cost. We met a design group of SWATCH in Colorado Springs who designed chips for Logitech and others who also
those guys to use. For asynchronous designs like ours analog simulation is also required. We paid $12k for a commercial license for TSpice which took four MONTHS to simulate sweeping a handful of transistors to get IV curves. In the past I watchedFifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch and therefore, unlike most others, we do not have to pay on the order of $100k per year per guy to rent the software for
look.Sixth, our chips have driven SDRAM and John Rible's code to do so is in fact in ROM in nodes 7, 8, 9, 107, 108. The source code for those ROMs is included in the free distribution of arrayForth-3 starting in block 2241; to know that one need only
not made aware of any such inquiry. I don't know who provided that answer; they must at minimum have been distracted at the time. At any rate, you have GreenArrays' abject apology for such stupid advice and if you ask again by email to hot...@Seventh, we routinely run external timing devices (crystals) for exactly the reasons cited earlier in this thread when the application requires a time base meeting stringent specifications. See for example AN002 and AN012.
Eighth, someone above said he'd asked "them" for timing details on the memory interface nodes in the context of wanting to drive DRAMs, and had been advised to just "tinker". I've been here for all the time we were called GreenArrays, Inc. and was
grasp the problem.FINALLY, GreenArrays invites constructive dialog via direct email. We decided early on that we did not have the lifespan available to spend engaging chat groups, which requires reading them forever after. I read comp.lang.forth for long enough to
With that, then, I wish you all well, and greetings to Anton and other names I recognize over a period of many years.
- Greg Bailey, President, GreenArrays, Inc.I thank you Greg. I understand. Rather than chat here, I propose a constructive FAQ like info base on your website, for marketing purposes, that can answer many questions, and occasionally publishing a link here?
I will suggest the same funding source I aim to approach .Just had a look at GA now, and it is not encouraging:
The new chip design that had been there for many years has been removed.
I wonder what this means.
Copied from the GA website just now:
Latest developments:
As of Summer 2022, ongoing activities continue
such as shipment of EVB002 evaluation kits and G144A12 chips,
improvements in chip design,
improvement of hardware and software development tools,
and production of application notes and other documentation.
https://en.wikipedia.org/wiki/Magic_(software)Thank for that Paul. Looks interesting. Is it less simple than OKCad?
Is there anything else apart from magic and Ceasar?
Wayne morellini <waynemo...@gmail.com> writes:
I've never used OKAD so can't really compare. Based on descriptionshttps://en.wikipedia.org/wiki/Magic_(software)Thank for that Paul. Looks interesting. Is it less simple than OKCad?
I've seen of OKAD though, they were in the same general class back in
the day. I worked on something unrelated to both that was also
comparable, but that was never released.
Is there anything else apart from magic and Ceasar?I'm sure there is, but I don't know what is out there these days. Did
you look on efabless? I think the tile and rectangle editing approach
is not all that interesting any more. You want something that starts
with an HDL, so you can use stuff from Opencores or whatever.
This might be interesting:
https://www.eejournal.com/article/meet-the-flex6502-a-flexible-6502-cpu/
On Saturday, 25 June 2022 at 02:57:52 UTC+1, Wayne morellini wrote:hoping to address a few things but will not be back to the group per se. gr...@greenarraychips.com if you wish to discuss anything further. Will try to stick to facts and avoid opinions.
On Saturday, June 25, 2022 at 2:29:43 AM UTC+10, Greg Bailey wrote:
I received an inquiry from Wayne this morning and thought I'd do a little searching to learn where he was coming from. That led me to this group and now that I've made it through this thread a reply seems appropriate. So I shall write one post
principally thinking of bright kids not yet out of high school, because (1) they can certainly understand what is required, and (2) they have not yet been told that FORTH sucks and that asynchronous computers are infeasible. Kids will use a form of GLOW,First, we have not yet devoted full time to seeking funding for our 501(c)(3) nonprofit called POLYSANCE, founded in Wyoming, whose charter is to teach young people to program and to design semiconductors in our fashion. By "young" we are
geometry was minimal; the effort is in addressing increasingly bizarre design rules that are visible in the higher level design, which thanks to foundry nondisclosures can't be discussed.Second, OKAD has produced chips in 180 and 130nm CMOS, the latter with 5V I/O. GLOW has produced 180nm CMOS and 28nm designs that passed all design rule checks and simulated successfully. The time to adjust our tiled design methods to a new
asked Global Foundries for the design rules pertinent to 14nm FinFETs they told me they'd allow me to see them if I paid them $25 million for a *shuttle run*. POLYSANCE will be looking among other things for a good complementary process that can be doneThird, modern processes cost fabulous prices while, until FinFETs emerge (literally :), leakage increases fabulously as well. Most of us cannot individually afford the $80k (2010 prices) for a mask set at 180nm, let alone the $6e6 at 28nm. When I
appreciate low leakage.Fourth, a big reason SWATCH group owns a 180nm fab is the same reason we like that process: Decent performance, low leakage, low cost. We met a design group of SWATCH in Colorado Springs who designed chips for Logitech and others who also
those guys to use. For asynchronous designs like ours analog simulation is also required. We paid $12k for a commercial license for TSpice which took four MONTHS to simulate sweeping a handful of transistors to get IV curves. In the past I watchedFifth, Jpit, I apologize for the reality that OKAD and, more so, GLOW *are* our crown jewels. We wrote these tools from scratch and therefore, unlike most others, we do not have to pay on the order of $100k per year per guy to rent the software for
look.Sixth, our chips have driven SDRAM and John Rible's code to do so is in fact in ROM in nodes 7, 8, 9, 107, 108. The source code for those ROMs is included in the free distribution of arrayForth-3 starting in block 2241; to know that one need only
not made aware of any such inquiry. I don't know who provided that answer; they must at minimum have been distracted at the time. At any rate, you have GreenArrays' abject apology for such stupid advice and if you ask again by email to hot...@Seventh, we routinely run external timing devices (crystals) for exactly the reasons cited earlier in this thread when the application requires a time base meeting stringent specifications. See for example AN002 and AN012.
Eighth, someone above said he'd asked "them" for timing details on the memory interface nodes in the context of wanting to drive DRAMs, and had been advised to just "tinker". I've been here for all the time we were called GreenArrays, Inc. and was
grasp the problem.FINALLY, GreenArrays invites constructive dialog via direct email. We decided early on that we did not have the lifespan available to spend engaging chat groups, which requires reading them forever after. I read comp.lang.forth for long enough to
With that, then, I wish you all well, and greetings to Anton and other names I recognize over a period of many years.
- Greg Bailey, President, GreenArrays, Inc.I thank you Greg. I understand. Rather than chat here, I propose a constructive FAQ like info base on your website, for marketing purposes, that can answer many questions, and occasionally publishing a link here?
I will suggest the same funding source I aim to approach .Just had a look at GA now, and it is not encouraging:
The new chip design that had been there for many years has been removed.
I wonder what this means.
Copied from the GA website just now:
Latest developments:
As of Summer 2022, ongoing activities continue
such as shipment of EVB002 evaluation kits and G144A12 chips,
improvements in chip design,
improvement of hardware and software development tools,
and production of application notes and other documentation.
I like the tile to get closer to the silicon and experiment with new designs in a way HDL doesn't normally. I thought OKCad outputs a vhdl file these days? Does magic?
I thought OKCad outputs a vhdl file these days? Does magic?
Efabless, they are the ones where you have to use a RiscV next to the
blank die space?
Wayne morellini <waynemo...@gmail.com> writes:
I thought OKCad outputs a vhdl file these days? Does magic?I never heard that about OKAD (I think that is the right spelling).
Magic predates HDL's and doesn't make HDL output either. Both basically output rectangles and maybe netlists.
Efabless, they are the ones where you have to use a RiscV next to theI don't think so, but you could look at their website.
blank die space?
Wayne morellini <waynemo...@gmail.com> writes:
I thought OKCad outputs a vhdl file these days? Does magic?I never heard that about OKAD (I think that is the right spelling).
Magic predates HDL's and doesn't make HDL output either. Both basically output rectangles and maybe netlists.
Efabless, they are the ones where you have to use a RiscV next to the blank die space?I don't think so, but you could look at their website.
On Sunday, July 3, 2022 at 12:50:26 AM UTC-4, Paul Rubin wrote:semiconductor behavior other than what can be expressed using simple math.
Wayne morellini <waynemo...@gmail.com> writes:The idea of a chip layout tool producing a VHDL file is pretty backwards. VHDL is more to the logical side of chip design and expresses none of the layout information. In particular, it isn't much good for analog simulation and particularly
I thought OKCad outputs a vhdl file these days? Does magic?I never heard that about OKAD (I think that is the right spelling).
Magic predates HDL's and doesn't make HDL output either. Both basically output rectangles and maybe netlists.
Efabless, they are the ones where you have to use a RiscV next to the blank die space?I don't think so, but you could look at their website.
An output of VHDL could be useful for logic simulation to verify the logical operation of the chip. But this is a bit like circuit design and layout. Normally, the circuit would be designed in a schematic source file, possibly simulated, converted to anetlist and input to the PCB layout tool. VHDL would serve as the logical source for a chip, simulated to eliminate logical bugs, then converted to a format an IC layout tool could use.
Of course, Chuck Moore is not a conventional thinker and could have done anything. However, if Okad produces VHDL, it isn't what the next step in the process uses, it is for going backwards in the process for validation/verification.
--
Rick C.
+++- Get 1,000 miles of free Supercharging
+++- Tesla referral code - https://ts.la/richard11209
On Sunday, July 3, 2022 at 11:53:17 PM UTC+10, gnuarm.del...@gmail.com wrote:semiconductor behavior other than what can be expressed using simple math.
On Sunday, July 3, 2022 at 12:50:26 AM UTC-4, Paul Rubin wrote:
Wayne morellini <waynemo...@gmail.com> writes:The idea of a chip layout tool producing a VHDL file is pretty backwards. VHDL is more to the logical side of chip design and expresses none of the layout information. In particular, it isn't much good for analog simulation and particularly
I thought OKCad outputs a vhdl file these days? Does magic?I never heard that about OKAD (I think that is the right spelling). Magic predates HDL's and doesn't make HDL output either. Both basically output rectangles and maybe netlists.
Efabless, they are the ones where you have to use a RiscV next to the blank die space?I don't think so, but you could look at their website.
a netlist and input to the PCB layout tool. VHDL would serve as the logical source for a chip, simulated to eliminate logical bugs, then converted to a format an IC layout tool could use.An output of VHDL could be useful for logic simulation to verify the logical operation of the chip. But this is a bit like circuit design and layout. Normally, the circuit would be designed in a schematic source file, possibly simulated, converted to
Of course, Chuck Moore is not a conventional thinker and could have done anything. However, if Okad produces VHDL, it isn't what the next step in the process uses, it is for going backwards in the process for validation/verification.
I see. Whatever it was , it was some compatible standard format. What would that be?
I see. Whatever it was , it was some compatible standard format.
What would that be?
Wayne morellini <waynemorellini@gmail.com> writes:
I see. Whatever it was , it was some compatible standard format.
What would that be?
Maybe this: https://en.wikipedia.org/wiki/GDSII
Or CIF:
Wayne morellini <waynemo...@gmail.com> writes:
I see. Whatever it was , it was some compatible standard format.Maybe this: https://en.wikipedia.org/wiki/GDSII
What would that be?
On Sunday, July 3, 2022 at 11:04:28 AM UTC-4, Wayne morellini
Of course, Chuck Moore is not a conventional thinker and could have done anything. However, if Okad produces VHDL, it isn't what the next step in the process uses, it is for going backwards in the process for validation/verification.
I see. Whatever it was , it was some compatible standard format. What would that be?Sorry, I'm lost as to what "it" is. If you are asking what the format for an IC layout tool would
Rick C.Arius.
++++ Get 1,000 miles of free Supercharging
++++ Tesla referral code - https://ts.la/richard11209
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like,and what Green Arrays will release for the glasses (which type of thing demands an advanced design).
I know we are waiting to hear what the 6Ghz chip Stephen has been workingwith will turn out like, and what Green Arrays will release for the glasses
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been workingwith will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM, and comparisons to Swift Forth etc. Which
got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work
latter. Isn't it time we had something more like these designs upgraded?
16 bit or more versions?
Ok. Rewritten to break up line length.
We are still waiting.
Looks le Rock has come a road that FPGA fabric in another thread, which I was trying to remember a few years ago. Efinix
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been workingwith will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM, and comparisons to Swift Forth etc. Which
got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work
latter. Isn't it time we had something more like these designs upgraded?
16 bit or more versions?
Ok. Rewritten to break up line length.
We are still waiting.
Looks le Rock has come a road that FPGA fabric in another thread, which I was trying to remember a few years ago. Efinix
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM,and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
But after 253 messages in this thread what is the answer to the main question?
Is it time for another Forth chip — or not yet?
On Saturday, July 16, 2022 at 5:02:02 PM UTC-4, Zbig wrote:
But after 253 messages in this thread what is the answer to the main question?
Is it time for another Forth chip — or not yet?
Not quite yet. Try back a week from next Tuesday.
On Saturday, July 16, 2022 at 11:58:03 PM UTC-4, dxforth wrote:
On 17/07/2022 13:18, Rick C wrote:
On Saturday, July 16, 2022 at 5:02:02 PM UTC-4, Zbig wrote:That looks like a deadline. We don't like deadlines.
But after 253 messages in this thread what is the answer to the main question?
Is it time for another Forth chip — or not yet?
Not quite yet. Try back a week from next Tuesday.
Who's "we"? Do you have a mouse in your pocket?
On 17/07/2022 13:18, Rick C wrote:
On Saturday, July 16, 2022 at 5:02:02 PM UTC-4, Zbig wrote:
But after 253 messages in this thread what is the answer to the main question?
Is it time for another Forth chip — or not yet?
Not quite yet. Try back a week from next Tuesday.That looks like a deadline. We don't like deadlines.
dxforth <dxf...@gmail.com> writes:
The compiler https://en.wikichip.org/w/images/2/25/MARC4_User%27s_Guide_qFORTH_Compiler.pdf
Wow, that is neat, and the programmers' guide also talks about qFORTH a
lot. The return stack lives in ram is it looks like its slots are 4
nibbles, a 12-bit code address plus 4 data bits. The manual advises
against too many levels of subroutines. Saving temporary data on the R
stack with >R etc. also sounds bad. But qFORTH does have those words.
I wonder if any actual application code is around that we can look at.
Ok, enjoy. I think it's pretty clear now that you are an armchair
designer.
Rick C <gnuarm.del...@gmail.com> writes:
Ok, enjoy. I think it's pretty clear now that you are an armchairI'm not designing anything and I've never claimed to be any sort of (hardware) designer. The comment you replied to was about the compiler software. It had nothing to do with the electronics.
designer.
On Sunday, July 17, 2022 at 8:11:42 PM UTC-4, Paul Rubin wrote:
Rick C <gnuarm.del...@gmail.com> writes:That post was not to you. I'm not sure what happened, but sometimes when I reply to a post, Google Groups actually brings up a different post. That was supposed to be a reply to a post by Wayne, in another thread even. I usually catch it.
Ok, enjoy. I think it's pretty clear now that you are an armchair designer.I'm not designing anything and I've never claimed to be any sort of (hardware) designer. The comment you replied to was about the compiler software. It had nothing to do with the electronics.
Sorry.
--
Rick C.
---+- Get 1,000 miles of free Supercharging
---+- Tesla referral code - https://ts.la/richard11209
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for ARM,and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like these
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
Syncing forth processor project threads.
On Sunday, September 4, 2022 at 2:59:31 PM UTC+10, Wayne morellini wrote:ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth for
Syncing forth processor project threads.Forth processor project
Is it time for another Forth chip?
https://groups.google.com/u/2/g/comp.lang.forth/c/6adve-Z1ppU
Designing a Forth Processor?
https://groups.google.com/u/2/g/comp.lang.forth/c/9lpG9yey_NQ
A low cost chip prototyping technique.
https://groups.google.com/u/2/g/comp.lang.forth/c/s27tSebmF-I
Comments: ColorForth binary in JavaScript!
https://groups.google.com/u/2/g/comp.lang.forth/c/3py7TwKu6b0
Looking for some advice on Offete p8, p16, p24, p32, p64. Ep16, ep24, ep32, and others.
https://groups.google.com/u/2/g/comp.lang.forth/c/EMgCYdV8NR8
On Sunday, September 4, 2022 at 11:26:12 PM UTC+8, Wayne morellini wrote:for ARM, and comparisons to Swift Forth etc. Which got me wondering about a lower end design. Now, with the passing of Doctor Ting, it reminds me of the Mup21 he had that kicked things off, and Jeff's work latter. Isn't it time we had something more like
On Sunday, September 4, 2022 at 2:59:31 PM UTC+10, Wayne morellini wrote:
On Tuesday, June 7, 2022 at 7:52:48 PM UTC+10, Wayne morellini wrote:
I know we are waiting to hear what the 6Ghz chip Stephen has been working with will turn out like, and what Green Arrays will release for the glasses (which type of thing demands an advanced design). But recently, I saw a document on Colorforth
status quo, especially if it has the ability to emulate any IO and still have all the atributes above, BUT it also must be able to access large memories or else it will not be able to do memory intensive product. still, as a super duper IO controller, itSyncing forth processor project threads.Forth processor project
Is it time for another Forth chip?
https://groups.google.com/u/2/g/comp.lang.forth/c/6adve-Z1ppU
Designing a Forth Processor?
https://groups.google.com/u/2/g/comp.lang.forth/c/9lpG9yey_NQ
A low cost chip prototyping technique.
https://groups.google.com/u/2/g/comp.lang.forth/c/s27tSebmF-I
Comments: ColorForth binary in JavaScript!
https://groups.google.com/u/2/g/comp.lang.forth/c/3py7TwKu6b0
Looking for some advice on Offete p8, p16, p24, p32, p64. Ep16, ep24, ep32, and others.
https://groups.google.com/u/2/g/comp.lang.forth/c/EMgCYdV8NR8after all that has been said, ultimately, what matters is 1) low power 2) high performance 3) easy to program 4) tools available 5) cheap. so if a 8bit cpu can do the job, why use 32bit, green array's version may be the best candidate to upset the
To conceive, design and introduce a new product, you should first ask,
"What problem am I trying to solve"? I think you will find there are no
more problems in the CPU world other than the tradeoffs of power,
performance and cost. I see no reason to think a Forth oriented CPU
design would be any better at this than what's available today.
Rick C.--
In article <270b46a1-fb7d-420c...@googlegroups.com>,
Lorem Ipsum <gnuarm.del...@gmail.com> wrote:
<SNIP>
To conceive, design and introduce a new product, you should first ask, >"What problem am I trying to solve"? I think you will find there are no >more problems in the CPU world other than the tradeoffs of power, >performance and cost. I see no reason to think a Forth oriented CPU+1
design would be any better at this than what's available today.
On Monday, February 27, 2023 at 1:31:47 PM UTC+1, none albert wrote:
In article <270b46a1-fb7d-420c...@googlegroups.com>,That invites a contrary opinion :--)
Lorem Ipsum <gnuarm.del...@gmail.com> wrote:
<SNIP>
To conceive, design and introduce a new product, you should first ask, >"What problem am I trying to solve"? I think you will find there are no >more problems in the CPU world other than the tradeoffs of power, >performance and cost. I see no reason to think a Forth oriented CPU >design would be any better at this than what's available today.+1
From personal experience, development of switch-mode power supplies
involves a joint effort of a group of highly specialized engineers. Developing
prototype is hell, because a tiny error in a board layout, the software, or in
doing measurements for debugging, can lead to catastrophic failure with
no option for repair. For some reason they always use the latest chips (not all bugs known), and for some reason the manuals become bigger and
bigger (1000 pages for a digital controller with programmable I/O?). The result is that for some types of bug, only a single person knows enough
to work on it, and nobody is able to help him because they have not read
the 1000 pages yet.
It would certainly help to have chips with an on-board RTOS and open
source drivers for all on-board I/O, plus ways to easily configure and
test stuff interactively.
The idea would be to make development less dependent on one or two
software engineers that know everything, and make it possible for
hardware people to at least intelligently step in for testing and
debugging.
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 300 |
Nodes: | 16 (3 / 13) |
Uptime: | 44:05:30 |
Calls: | 6,710 |
Calls today: | 3 |
Files: | 12,243 |
Messages: | 5,354,109 |