3 ms·

> This is likely a compiler issue where the v0 += acc * v0 sequence couldn’t be converted into a FMA instruction. Err, is the ISA undocumented/impossible to i

> This is likely a compiler issue where the v0 += acc * v0 sequence couldn’t be converted into a FMA instruction. Err, is the ISA undocumented/impossible to inspect in the execution pipeline? Seems like an important thing to verify/fix for a hardware benchmark...

tremon 3y ago
Yes, at least for as far as I know. The actual micro-ops resulting from the instruction stream are invisible. You can count the number of uops issued and partly deduce how the instructions were decoded, but not view the uops themselves.
wtallis 3y ago
From the preceding paragraph: > We weren’t able to get to the bottom of this because we don’t have the profiling tools necessary to get disassembly from the GPU.
- touisteur 3y ago
  And that's all I need to know about replacing all NVIDIA stuff. I know it's pretty hard to get there, but Intel should know that having a serious general purpose computing thing means solid compilers, toolchains, optimized libraries, and a whole lot of mindshare (as in 'a large number of people willing to throw their time to test your stuff').
  - ndneighbor 3y ago
    I am an Intel shill lately but I think it's more of a time thing rather than the desire to keep stuff a secret. They've been pretty good about open documentation on the stuff that matters (like this) such as OpenVINO.
    - touisteur 3y ago
      I was a bit annoyed about the OpenVINO reference, because I felt they closed most of the things about myriad-x and the SHAVE arch. And last time I tried OpenVINO on TigerLake I was left with a very thick pile of undebuggable, uninspectable opencl-y stuff, very bad taste in my mouth. I mean OpenVINO's perf is up there on Intel CPUs and it's a great optimising compiler, I've thrown a lot of weird stuff in there and it didn't crap out with complaints about unsupported layers or unsupported combination of layers. It also has an OK batching story (as opposed to TVM last time I checked...) if you're ready to perform some network surgery. I also feel it's very bad at reporting errors, and stepping through with gdb is one of the worst experiences... BUT but yeah most of the code is available now. Now if they could stop moving shit around, and renaming stuff, it'd be great. Hoping they settle on 'OneAPI' for some time.
      - bigbillheck 3y ago
        SHAVE was such a cool architecture, it's too bad about all the secrecy.
- colejohnson66 3y ago
  Is the Intel Xe ISA even publicly documented? I’ve searched before and I can’t find a PDF detailing the instruction set. AMD releases them,[0] but I can’t find anything from Intel (or Nvidia for that matter). [0]: RDNA2 ISA: https://www.amd.com/content/dam/amd/en/documents/radeon-tech-docs/instruction-set-architectures/rdna2-shader-instruction-set-architecture.pdf https://www.amd.com/content/dam/amd/en/documents/radeon-tech...
  - wmf 3y ago
    https://www.intel.com/content/www/us/en/docs/graphics-for-linux/developer-reference/1-0/overview.html https://www.intel.com/content/www/us/en/docs/graphics-for-li... (Alchemist is a variant of Xe)

tremon 3y ago

Yes, at least for as far as I know. The actual micro-ops resulting from the instruction stream are invisible. You can count the number of uops issued and partly deduce how the instructions were decoded, but not view the uops themselves.

wtallis 3y ago

From the preceding paragraph: > We weren’t able to get to the bottom of this because we don’t have the profiling tools necessary to get disassembly from the GPU.

touisteur 3y ago
And that's all I need to know about replacing all NVIDIA stuff. I know it's pretty hard to get there, but Intel should know that having a serious general purpose computing thing means solid compilers, toolchains, optimized libraries, and a whole lot of mindshare (as in 'a large number of people willing to throw their time to test your stuff').
- ndneighbor 3y ago
  I am an Intel shill lately but I think it's more of a time thing rather than the desire to keep stuff a secret. They've been pretty good about open documentation on the stuff that matters (like this) such as OpenVINO.
  - touisteur 3y ago
    I was a bit annoyed about the OpenVINO reference, because I felt they closed most of the things about myriad-x and the SHAVE arch. And last time I tried OpenVINO on TigerLake I was left with a very thick pile of undebuggable, uninspectable opencl-y stuff, very bad taste in my mouth. I mean OpenVINO's perf is up there on Intel CPUs and it's a great optimising compiler, I've thrown a lot of weird stuff in there and it didn't crap out with complaints about unsupported layers or unsupported combination of layers. It also has an OK batching story (as opposed to TVM last time I checked...) if you're ready to perform some network surgery. I also feel it's very bad at reporting errors, and stepping through with gdb is one of the worst experiences... BUT but yeah most of the code is available now. Now if they could stop moving shit around, and renaming stuff, it'd be great. Hoping they settle on 'OneAPI' for some time.
    - bigbillheck 3y ago
      SHAVE was such a cool architecture, it's too bad about all the secrecy.
colejohnson66 3y ago
Is the Intel Xe ISA even publicly documented? I’ve searched before and I can’t find a PDF detailing the instruction set. AMD releases them,[0] but I can’t find anything from Intel (or Nvidia for that matter). [0]: RDNA2 ISA: https://www.amd.com/content/dam/amd/en/documents/radeon-tech-docs/instruction-set-architectures/rdna2-shader-instruction-set-architecture.pdf https://www.amd.com/content/dam/amd/en/documents/radeon-tech...
- wmf 3y ago
  https://www.intel.com/content/www/us/en/docs/graphics-for-linux/developer-reference/1-0/overview.html https://www.intel.com/content/www/us/en/docs/graphics-for-li... (Alchemist is a variant of Xe)