≡
[00:21:57] <bill-auger> frankly, most people over the age of 20 remember all of those things - it is barely an interesting conversation
[12:26:56] <clarity> I really hope we start seeing more powerful fpgas with fully open source toolchain and relevant backing (at least docs) from the manufacturer
[12:27:08] <clarity> Ideally contributions to the toolchain..
[12:27:50] <jonas> "start"? Olimex has been in this game for a looong time
[12:28:21] <clarity> Olimex doesn't manufacture fpgas
[13:38:22] <bill-auger> clarity: as i understand, the icestorm project is the best effort so far https://clifford.at/icestorm
[13:41:34] <mps> what is needed for this usb ethernet https://www.olimex.com/Products/USB-Modules/USB-GIGABIT/open-source-hardware to work properly with latest stable kernels (5.18.*)
[13:41:58] <mps> card is detected but driver can't read mac address from it and assigns random ones on every boot
[13:42:16] <mps> I enabled needed eeprom modules in kernel but still no luck
[14:57:21] <clarity> bill-auger: Yea, that's a reverse engineering effort though. And the ice40 are pretty slow as far as fpgas go
[14:57:25] <clarity> No support from the manufacturer
[14:57:31] <clarity> That's the frustrating part about fpgas
[14:58:12] <clarity> From what I gathered, quicklogic is the only manufacturer to officially support open source toolchain
[15:02:49] <bill-auger> its no different from any other popular hardware - without reverse engineering, most computer hardware would not work without the manufacturer's windows blobs
[15:03:52] <clarity> It's different from e.g. most microcontrollers, which are perfectly usable without reverse engineering
[15:06:30] <bill-auger> im no expert but as i understand it, FPGAs are "usable", but they are not programmable (yet) without the mnufacturer's blobs
[15:06:41] <clarity> Yes, I mean the same thing
[15:08:05] <bill-auger> its not though - only the person who programs it need to worry about that problem - it does not affect users of FPGA-cased gadgets , in the same way as say a wifi of graphics device does
[15:08:23] <clarity> Sigh
[15:08:50] <clarity> When you build a device, you use chips to do so
[15:09:02] <clarity> Programming those chips is a part of the deal
[15:10:42] <clarity> And arguably programming an fpga on the fly is as much "use" as is programming bits into RAM
[15:10:54] <bill-auger> as a hobby toy yes - but that implies nothing is being made which is generally useful to others
[15:11:27] <bill-auger> thats like a libre distro - only the distro devs need to touch the non-free upstream code - they do that so users dont need to
[15:12:43] <bill-auger> so it is only a serious problem if the developer is the only user
[15:12:56] <clarity> Well, I am a developer
[15:13:03] <clarity> For me to use things is to develop things
[15:13:15] <clarity> Just as I use gcc or kicad or whatever
[15:13:28] <bill-auger> ok so if you make a gadget for other people, they should never need to re-program the FPGA
[15:14:00] <clarity> Why not?
[15:14:09] <bill-auger> because you programmed it for them
[15:14:11] <clarity> FPGAs are interesting for their ability to be programmed on the fly
[15:14:28] <clarity> Just like RAM is interesting for its ability to have its contents changed on the fly, unlike ROM
[15:14:42] <bill-auger> they are most interesting because you can make stuff without an expensive silicon fab
[15:15:15] <bill-auger> the fact that they can be re-programmed is not nearly as interesting
[15:15:27] <clarity> Ok, whatever *shrug*
[15:15:55] <bill-auger> if they only cost 10 cents each. im sure you would not care if they can be re-programmed
[15:16:53] <clarity> I would, because asking a user to desolder a chip and replace it with a new one when they run a different program is terrible ux
[15:18:15] <bill-auger> so dont ask anyone to do that
[15:19:57] <bill-auger> i understand your concern; but it is not such dramatic - the typical gadget owner does not want to modify the gadget in any way - if they want it to do something differently, they buy a different gadget
[15:21:17] <bill-auger> intel does not expect anyone to desolder their MOBO to replace a silicon chip, or ever to replcae the CPU - they expect people to buy a new computer
[15:21:40] <bill-auger> im not saying thats great; but it is not a huge problem
[15:23:13] <bill-auger> youre basically assuming that because somethng is possible, that everyone should want to do it - FPGAs are entirely a nerd thing, a tool for experimentation - most people would never want to re-program them
[15:25:46] <clarity> I'm saying that something is possible and damn useful and therefore it would be damn nice if we could all do it with an open source toolchain
[15:25:55] <clarity> "most people" arguments are completely irrelevant
[15:26:12] <bill-auger> but we all do not want to do it - only very few people need to or want to
[15:26:20] <clarity> And that's fine
[15:26:50] <clarity> We all don't proram microcontrollres, I'm still glad they use openly documented ISAs and come with extensive datasheets
[15:27:45] <bill-auger> i totally agree with the goal - i am only putting it in context -in context, there is very low demand for it
[15:28:46] <bill-auger> if there was a high demand for it, the libre tools would already exist
[15:29:46] <clarity> I just want to see manufacturers behind it. Reverse engineering projects always lag behind and means commercial vendors will always choose prorietary tools
[15:30:07] <clarity> It just makes the whole ecosystem messy
[15:30:47] <bill-auger> yes - i am only making the point that it has always been that way
[15:30:56] <clarity> It hasn't been for a lot of things
[15:31:52] <clarity> I work for a company that makes custom hardware and 99% of the chips we use can be programmed and built into a product without NDAs and proprietary tools
[15:32:35] <clarity> Some chips have parts that need proprietary tools or blobs for a little thing, such as DRAM init on some sunxi platforms or bootloader signing on rockchip
[15:32:48] <clarity> Some have features that would require reverse engineering but these are rarely vital
[15:33:10] <bill-auger> ok but my point is that 99% of tech users do not use "chips" - they use complete "gadgets" - only the gadget maker needs to worry about the chips
[15:33:23] <clarity> Yea. Like I said, I'm not about 99% users
[15:34:05] <bill-auger> i understand - it seems urgent an important to you - im just explaining why there are not any libre options - it is the low demand
[15:34:12] <nedko> bill-auger: bandwagon is a logical fallacy
[15:35:41] <clarity> bill-auger: The demand for microcontrollers, switch ICs, ADC ICs, PMICs, USB HSIC hub ICs, etcetra by the "99%" is very low and those are still documented
[15:36:18] <bill-auger> i was not suggesting bandwagon ,entality - the implication is the opposite direction
[15:36:29] <clarity> It's mostly that some manufacturers realize opennes has benefits and/or that closing things up doesn't actually give them and edge
[15:36:33] <bill-auger> things that more people want are simply more likely to happen - if few people want something, it is probably not going to happen
[15:36:43] <clarity> And I'm just waiting for one of the FPGA manufacturers to come to the same realization
[15:37:05] <clarity> Sure, but this is an area where it's never been about a popularity vote
[15:37:18] <clarity> Manufacturers pretty much do what they want
[15:38:24] <bill-auger> manufacturers do the cheapest thing they can do while making a profit
[15:38:31] <bill-auger> in this specific case, manufacturers would all offer libre tools and drivers, if people stopped buying the models which dont
[15:39:02] <bill-auger> but people happily buy gadgets which are extremely difficult to hack - so that is what manufacturers make
[15:39:46] <clarity> And as far as FPGAs go, I could certainly envision a future where computers include similar hardware for on-the-fly reprogrammable hardware acceleration. Kinda like GPGPU, but fitting for things that aren't necessarily a great fit for GPUs. There's a world of possibilities out there
[15:40:12] <clarity> People certainly don't think that they want to reprogram GPUs, but they sure do want to run applications that run code on the GPU..
[15:40:15] <clarity> :P
[15:40:29] <clarity> (Once they realize how much of a game changer it its wrt performance)
[15:41:34] <bill-auger> thats why silicon os better - an FPGA CPU is begging for a virus infection, isnt it?
[15:42:14] <bill-auger> the main value of FPGAs, is to experiment and get the design correct, then bake that design into silicon
[15:43:15] <bill-auger> then the manufacturer can document that, then people can understand it and trust it
[15:47:09] <nedko> to me the main value of FGPAs is that you can update them
[15:47:18] <nedko> kind of x86_64 microcode, but better
[15:48:02] <bill-auger> from the hacker perspective sure
[15:49:38] <bill-auger> but for a real machine, that you want to use every day (and hope it actually works every day, and does not get hacked), you probably dont want a re-programmanble CPU
[15:50:51] <nedko> depends on what you do everyday
[15:51:18] <nedko> apple inc. is already good at producing computers in form of appliances
[15:51:39] <bill-auger> yea, anything important, which would be connected to the internet, or you may install third-party code onto
[15:52:45] <nedko> i for one would use it if powerful enough for DSP
[15:52:59] <bill-auger> a doorbell, light switch, or coffee pot, not so imprtant
[15:53:00] <nedko> if course i can get a DSP chip
[15:53:22] <nedko> i actually have a teapot with computer in it
[15:53:33] <nedko> if i was able to program it i'd pay more
[15:53:40] <nedko> but no such offers exist
[15:53:49] <bill-auger> but anyone who wants their doorbell, light switchs, or coffee pot, to be on the internet is kinda silly person if you ask me
[15:54:21] <nedko> i want it reprogramable, FPGA is good for this
[15:54:41] <nedko> in case of the teapot, not-vendor-locked in firmware would do
[15:55:11] <nedko> i bought that teapot only because it was the only available on the market back then
[15:55:17] <nedko> now it is discontinued product
[15:55:45] <nedko> to fix the non OSHW and software design i would have to reverse engineer
[15:55:49] <nedko> this is expensive
[15:56:30] <nedko> repairable electronics are better
[15:57:17] <nedko> for the reference, apple products are expensive as well
[15:58:52] <nedko> the bugs in the teapot are not that bad and it still works
[15:59:05] <nedko> also teapot will wear out for other reasons
[16:17:34] <bill-auger> i dont think anyone in this channel or the libre community at large would argue that
[16:17:34] <bill-auger> it just not so specific - it has been that way for 50 years or more
[16:17:59] <bill-auger> to fix _any_hardware_ you would have to reverse engineer
[16:18:52] <bill-auger> manufacturers have not offered schematics for their products since 50 years ago or more
[16:19:22] <bill-auger> before then, it mas most typical that every machine came with schematics glued inside the chassis
[16:20:35] <bill-auger> its not a new problem - what is new, is that people complain about it more today
[16:21:28] <bill-auger> probably because of things like FPGAs and the OSHW trend, gives people an inkling of hope that hardware freedom could be possible
[16:21:37] <nedko> well, in post-soviet space it is more like 30 years
[16:22:32] <nedko> to my knowledge the planned obsolescence was hyped 50 years ago in west
[16:23:08] <nedko> i see olimex products as preferable because they have schematics
[16:23:29] <nedko> the OSHW ones i mean
[16:23:56] <bill-auger> yep - back then, a TV repairman would be outraged if the manufacturer would not release the schematics - and he would probably reccommend to all clients not to buy from that company
[16:24:59] <nedko> electronic waste grows with population and with vendor lockin
[16:25:21] <nedko> IMO this makes the half century trend short sighted and doomed to fail in long term
[16:25:21] <bill-auger> then cheap foreign models became available, and people would just buy a enew one when it breaks, rather than repair anything
IRC #olimex 2022-07-03
[00:13:11] <mps> Guest62 is left, and I have to tell that I remember all things about talked above[00:21:57] <bill-auger> frankly, most people over the age of 20 remember all of those things - it is barely an interesting conversation
[12:26:56] <clarity> I really hope we start seeing more powerful fpgas with fully open source toolchain and relevant backing (at least docs) from the manufacturer
[12:27:08] <clarity> Ideally contributions to the toolchain..
[12:27:50] <jonas> "start"? Olimex has been in this game for a looong time
[12:28:21] <clarity> Olimex doesn't manufacture fpgas
[13:38:22] <bill-auger> clarity: as i understand, the icestorm project is the best effort so far https://clifford.at/icestorm
[13:41:34] <mps> what is needed for this usb ethernet https://www.olimex.com/Products/USB-Modules/USB-GIGABIT/open-source-hardware to work properly with latest stable kernels (5.18.*)
[13:41:58] <mps> card is detected but driver can't read mac address from it and assigns random ones on every boot
[13:42:16] <mps> I enabled needed eeprom modules in kernel but still no luck
[14:57:21] <clarity> bill-auger: Yea, that's a reverse engineering effort though. And the ice40 are pretty slow as far as fpgas go
[14:57:25] <clarity> No support from the manufacturer
[14:57:31] <clarity> That's the frustrating part about fpgas
[14:58:12] <clarity> From what I gathered, quicklogic is the only manufacturer to officially support open source toolchain
[15:02:49] <bill-auger> its no different from any other popular hardware - without reverse engineering, most computer hardware would not work without the manufacturer's windows blobs
[15:03:52] <clarity> It's different from e.g. most microcontrollers, which are perfectly usable without reverse engineering
[15:06:30] <bill-auger> im no expert but as i understand it, FPGAs are "usable", but they are not programmable (yet) without the mnufacturer's blobs
[15:06:41] <clarity> Yes, I mean the same thing
[15:08:05] <bill-auger> its not though - only the person who programs it need to worry about that problem - it does not affect users of FPGA-cased gadgets , in the same way as say a wifi of graphics device does
[15:08:23] <clarity> Sigh
[15:08:50] <clarity> When you build a device, you use chips to do so
[15:09:02] <clarity> Programming those chips is a part of the deal
[15:10:42] <clarity> And arguably programming an fpga on the fly is as much "use" as is programming bits into RAM
[15:10:54] <bill-auger> as a hobby toy yes - but that implies nothing is being made which is generally useful to others
[15:11:27] <bill-auger> thats like a libre distro - only the distro devs need to touch the non-free upstream code - they do that so users dont need to
[15:12:43] <bill-auger> so it is only a serious problem if the developer is the only user
[15:12:56] <clarity> Well, I am a developer
[15:13:03] <clarity> For me to use things is to develop things
[15:13:15] <clarity> Just as I use gcc or kicad or whatever
[15:13:28] <bill-auger> ok so if you make a gadget for other people, they should never need to re-program the FPGA
[15:14:00] <clarity> Why not?
[15:14:09] <bill-auger> because you programmed it for them
[15:14:11] <clarity> FPGAs are interesting for their ability to be programmed on the fly
[15:14:28] <clarity> Just like RAM is interesting for its ability to have its contents changed on the fly, unlike ROM
[15:14:42] <bill-auger> they are most interesting because you can make stuff without an expensive silicon fab
[15:15:15] <bill-auger> the fact that they can be re-programmed is not nearly as interesting
[15:15:27] <clarity> Ok, whatever *shrug*
[15:15:55] <bill-auger> if they only cost 10 cents each. im sure you would not care if they can be re-programmed
[15:16:53] <clarity> I would, because asking a user to desolder a chip and replace it with a new one when they run a different program is terrible ux
[15:18:15] <bill-auger> so dont ask anyone to do that
[15:19:57] <bill-auger> i understand your concern; but it is not such dramatic - the typical gadget owner does not want to modify the gadget in any way - if they want it to do something differently, they buy a different gadget
[15:21:17] <bill-auger> intel does not expect anyone to desolder their MOBO to replace a silicon chip, or ever to replcae the CPU - they expect people to buy a new computer
[15:21:40] <bill-auger> im not saying thats great; but it is not a huge problem
[15:23:13] <bill-auger> youre basically assuming that because somethng is possible, that everyone should want to do it - FPGAs are entirely a nerd thing, a tool for experimentation - most people would never want to re-program them
[15:25:46] <clarity> I'm saying that something is possible and damn useful and therefore it would be damn nice if we could all do it with an open source toolchain
[15:25:55] <clarity> "most people" arguments are completely irrelevant
[15:26:12] <bill-auger> but we all do not want to do it - only very few people need to or want to
[15:26:20] <clarity> And that's fine
[15:26:50] <clarity> We all don't proram microcontrollres, I'm still glad they use openly documented ISAs and come with extensive datasheets
[15:27:45] <bill-auger> i totally agree with the goal - i am only putting it in context -in context, there is very low demand for it
[15:28:46] <bill-auger> if there was a high demand for it, the libre tools would already exist
[15:29:46] <clarity> I just want to see manufacturers behind it. Reverse engineering projects always lag behind and means commercial vendors will always choose prorietary tools
[15:30:07] <clarity> It just makes the whole ecosystem messy
[15:30:47] <bill-auger> yes - i am only making the point that it has always been that way
[15:30:56] <clarity> It hasn't been for a lot of things
[15:31:52] <clarity> I work for a company that makes custom hardware and 99% of the chips we use can be programmed and built into a product without NDAs and proprietary tools
[15:32:35] <clarity> Some chips have parts that need proprietary tools or blobs for a little thing, such as DRAM init on some sunxi platforms or bootloader signing on rockchip
[15:32:48] <clarity> Some have features that would require reverse engineering but these are rarely vital
[15:33:10] <bill-auger> ok but my point is that 99% of tech users do not use "chips" - they use complete "gadgets" - only the gadget maker needs to worry about the chips
[15:33:23] <clarity> Yea. Like I said, I'm not about 99% users
[15:34:05] <bill-auger> i understand - it seems urgent an important to you - im just explaining why there are not any libre options - it is the low demand
[15:34:12] <nedko> bill-auger: bandwagon is a logical fallacy
[15:35:41] <clarity> bill-auger: The demand for microcontrollers, switch ICs, ADC ICs, PMICs, USB HSIC hub ICs, etcetra by the "99%" is very low and those are still documented
[15:36:18] <bill-auger> i was not suggesting bandwagon ,entality - the implication is the opposite direction
[15:36:29] <clarity> It's mostly that some manufacturers realize opennes has benefits and/or that closing things up doesn't actually give them and edge
[15:36:33] <bill-auger> things that more people want are simply more likely to happen - if few people want something, it is probably not going to happen
[15:36:43] <clarity> And I'm just waiting for one of the FPGA manufacturers to come to the same realization
[15:37:05] <clarity> Sure, but this is an area where it's never been about a popularity vote
[15:37:18] <clarity> Manufacturers pretty much do what they want
[15:38:24] <bill-auger> manufacturers do the cheapest thing they can do while making a profit
[15:38:31] <bill-auger> in this specific case, manufacturers would all offer libre tools and drivers, if people stopped buying the models which dont
[15:39:02] <bill-auger> but people happily buy gadgets which are extremely difficult to hack - so that is what manufacturers make
[15:39:46] <clarity> And as far as FPGAs go, I could certainly envision a future where computers include similar hardware for on-the-fly reprogrammable hardware acceleration. Kinda like GPGPU, but fitting for things that aren't necessarily a great fit for GPUs. There's a world of possibilities out there
[15:40:12] <clarity> People certainly don't think that they want to reprogram GPUs, but they sure do want to run applications that run code on the GPU..
[15:40:15] <clarity> :P
[15:40:29] <clarity> (Once they realize how much of a game changer it its wrt performance)
[15:41:34] <bill-auger> thats why silicon os better - an FPGA CPU is begging for a virus infection, isnt it?
[15:42:14] <bill-auger> the main value of FPGAs, is to experiment and get the design correct, then bake that design into silicon
[15:43:15] <bill-auger> then the manufacturer can document that, then people can understand it and trust it
[15:47:09] <nedko> to me the main value of FGPAs is that you can update them
[15:47:18] <nedko> kind of x86_64 microcode, but better
[15:48:02] <bill-auger> from the hacker perspective sure
[15:49:38] <bill-auger> but for a real machine, that you want to use every day (and hope it actually works every day, and does not get hacked), you probably dont want a re-programmanble CPU
[15:50:51] <nedko> depends on what you do everyday
[15:51:18] <nedko> apple inc. is already good at producing computers in form of appliances
[15:51:39] <bill-auger> yea, anything important, which would be connected to the internet, or you may install third-party code onto
[15:52:45] <nedko> i for one would use it if powerful enough for DSP
[15:52:59] <bill-auger> a doorbell, light switch, or coffee pot, not so imprtant
[15:53:00] <nedko> if course i can get a DSP chip
[15:53:22] <nedko> i actually have a teapot with computer in it
[15:53:33] <nedko> if i was able to program it i'd pay more
[15:53:40] <nedko> but no such offers exist
[15:53:49] <bill-auger> but anyone who wants their doorbell, light switchs, or coffee pot, to be on the internet is kinda silly person if you ask me
[15:54:21] <nedko> i want it reprogramable, FPGA is good for this
[15:54:41] <nedko> in case of the teapot, not-vendor-locked in firmware would do
[15:55:11] <nedko> i bought that teapot only because it was the only available on the market back then
[15:55:17] <nedko> now it is discontinued product
[15:55:45] <nedko> to fix the non OSHW and software design i would have to reverse engineer
[15:55:49] <nedko> this is expensive
[15:56:30] <nedko> repairable electronics are better
[15:57:17] <nedko> for the reference, apple products are expensive as well
[15:58:52] <nedko> the bugs in the teapot are not that bad and it still works
[15:59:05] <nedko> also teapot will wear out for other reasons
[16:17:34] <bill-auger> i dont think anyone in this channel or the libre community at large would argue that
[16:17:34] <bill-auger> it just not so specific - it has been that way for 50 years or more
[16:17:59] <bill-auger> to fix _any_hardware_ you would have to reverse engineer
[16:18:52] <bill-auger> manufacturers have not offered schematics for their products since 50 years ago or more
[16:19:22] <bill-auger> before then, it mas most typical that every machine came with schematics glued inside the chassis
[16:20:35] <bill-auger> its not a new problem - what is new, is that people complain about it more today
[16:21:28] <bill-auger> probably because of things like FPGAs and the OSHW trend, gives people an inkling of hope that hardware freedom could be possible
[16:21:37] <nedko> well, in post-soviet space it is more like 30 years
[16:22:32] <nedko> to my knowledge the planned obsolescence was hyped 50 years ago in west
[16:23:08] <nedko> i see olimex products as preferable because they have schematics
[16:23:29] <nedko> the OSHW ones i mean
[16:23:56] <bill-auger> yep - back then, a TV repairman would be outraged if the manufacturer would not release the schematics - and he would probably reccommend to all clients not to buy from that company
[16:24:59] <nedko> electronic waste grows with population and with vendor lockin
[16:25:21] <nedko> IMO this makes the half century trend short sighted and doomed to fail in long term
[16:25:21] <bill-auger> then cheap foreign models became available, and people would just buy a enew one when it breaks, rather than repair anything
