3 ms·

> Intel keeps getting it right with numerical libraries. They're open. They work well. They work on AMD. What Intel numerical libraries are you thinking of? Wh

> Intel keeps getting it right with numerical libraries. They're open. They work well. They work on AMD. What Intel numerical libraries are you thinking of? When I think of Intel numerical libraries, the first that comes to mind is MKL. MKL is neither open-source nor does it work well on AMD without some fragile hacks [0]. [0] https://www.pugetsystems.com/labs/hpc/How-To-Use-MKL-with-AMD-Ryzen-and-Threadripper-CPU-s-Effectively-for-Python-Numpy-And-Other-Applications-1637/ https://www.pugetsystems.com/labs/hpc/How-To-Use-MKL-with-AM...

hyperbovine 6y ago
Well, OP didn't say MKL works well on AMD. But you can at least run it on a non-Intel CPU. Compare CUDA.
- Klinky 6y ago
  CUDA appears to have come out well before even OpenCL. I don't see why there would be expectation that nVidia would design their framework to work on a competitors product.
  - andromeduck 6y ago
    Cuda was also a response to ATI's own efforts at a proprietary effort which they eventually have up on.
    - TomVDB 6y ago
      ATI came out with CTM, which was just an assembler. CUDA was released a month or so after that. It was a full C compiler and already had a pretty large set of examples and library functions. I downloaded CUDA about the day it was released, and used it for real some months later when I bought a 8600 GT GPU. To call CUDA a response to CTM is too much praise for Nvidia, because it suggests that their response included cobbling a compiler and SDK in just a month. :-)
- fluffything 6y ago
  The nvidia pgi compiler compiles CUDA to multi-core x86-64. There are other third-party compilers for CUDA->x86-64 (one LLVM-based one from Intel). There is a "library replacement" for CUDA from AMD called HIP, that you can use to map CUDA programs to ROCm. But... it doesn't work very well. NVIDIA also open-sourced CUDA support for Clang and LLVM. So anybody can extend clang to map CUDA to any hardware supported by LLVM, including SPIRV. The only company that would benefit from doing this would be AMD, but AMD doesn't have many LLVM contributors. Intel drives clang and LLVM development for x86_64, paying a lot of people to work on that.
  - xigency 6y ago
    It sounds like people want nvidia to write drivers for AMD. This criticism makes even less sense when any bystander could implement CUDA suppport on AMD by connecting open source software.
    - Reelin 6y ago
      > any bystander You aren't seriously implying than any bystander is capable of extending LLVM to map CUDA to SPIR-V? What percentage of present day gainfully employed software engineers do you suppose even has the background knowledge? How many hours do you suppose the work would require?
      - fluffything 6y ago
        If LLVM has a SPIRV backend, probably very little. For a proof of concept, a bachelor CS thesis would probably do. Clang already has a CUDA parser, and all the code to lower CUDA specific constructs to LLVM-IR, some of which are specific for the PTX backend. If you try to compile CUDA code for a different target, like SPIRV, you'll probably get some errors saying that some of the LLVM-IR instructions generated by clang are not available in that backend, and you'll need to generate the proper SPIRV calls in clang instead. Its probably a great beginner task to get started with clang and LLVM. You don't need to worry about the C++ frontend side of things because that's already done, and can focus on understanding the LLVM-IR and how to emit it from clang when you already have a proper AST.
- gnufx 6y ago
  Not on ARM, or POWER, you can't. Why you'd want to run it on AMD, I don't understand. I don't know what fraction of peak BLIS and OpenBLAS get, but it will be high.

hyperbovine 6y ago

Well, OP didn't say MKL works well on AMD. But you can at least run it on a non-Intel CPU. Compare CUDA.

Klinky 6y ago
CUDA appears to have come out well before even OpenCL. I don't see why there would be expectation that nVidia would design their framework to work on a competitors product.
- andromeduck 6y ago
  Cuda was also a response to ATI's own efforts at a proprietary effort which they eventually have up on.
  - TomVDB 6y ago
    ATI came out with CTM, which was just an assembler. CUDA was released a month or so after that. It was a full C compiler and already had a pretty large set of examples and library functions. I downloaded CUDA about the day it was released, and used it for real some months later when I bought a 8600 GT GPU. To call CUDA a response to CTM is too much praise for Nvidia, because it suggests that their response included cobbling a compiler and SDK in just a month. :-)
fluffything 6y ago
The nvidia pgi compiler compiles CUDA to multi-core x86-64. There are other third-party compilers for CUDA->x86-64 (one LLVM-based one from Intel). There is a "library replacement" for CUDA from AMD called HIP, that you can use to map CUDA programs to ROCm. But... it doesn't work very well. NVIDIA also open-sourced CUDA support for Clang and LLVM. So anybody can extend clang to map CUDA to any hardware supported by LLVM, including SPIRV. The only company that would benefit from doing this would be AMD, but AMD doesn't have many LLVM contributors. Intel drives clang and LLVM development for x86_64, paying a lot of people to work on that.
- xigency 6y ago
  It sounds like people want nvidia to write drivers for AMD. This criticism makes even less sense when any bystander could implement CUDA suppport on AMD by connecting open source software.
  - Reelin 6y ago
    > any bystander You aren't seriously implying than any bystander is capable of extending LLVM to map CUDA to SPIR-V? What percentage of present day gainfully employed software engineers do you suppose even has the background knowledge? How many hours do you suppose the work would require?
    - fluffything 6y ago
      If LLVM has a SPIRV backend, probably very little. For a proof of concept, a bachelor CS thesis would probably do. Clang already has a CUDA parser, and all the code to lower CUDA specific constructs to LLVM-IR, some of which are specific for the PTX backend. If you try to compile CUDA code for a different target, like SPIRV, you'll probably get some errors saying that some of the LLVM-IR instructions generated by clang are not available in that backend, and you'll need to generate the proper SPIRV calls in clang instead. Its probably a great beginner task to get started with clang and LLVM. You don't need to worry about the C++ frontend side of things because that's already done, and can focus on understanding the LLVM-IR and how to emit it from clang when you already have a proper AST.
gnufx 6y ago
Not on ARM, or POWER, you can't. Why you'd want to run it on AMD, I don't understand. I don't know what fraction of peak BLIS and OpenBLAS get, but it will be high.