[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

Allen added inline comments. Comment at: llvm/lib/Target/NVPTX/NVPTXInstrInfo.td:186 + !eq(name, "v2f16"): Float16x2Regs, + !eq(name, "bf16"): Float16Regs, + !eq(name, "v2bf16"): Float16x2Regs, tra wrote: > tra wrote: > > Allen wrote: > > > sorry for

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG0e8a414ab3d3: [CUDA, NVPTX] Added basic __bf16 support for NVPTX. (authored by tra). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTIO

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra updated this revision to Diff 470544. tra added a comment. whitespace fix Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D136311/new/ https://reviews.llvm.org/D136311 Files: clang/lib/Basic/Targets/NVPTX.cpp clang/lib/Basic/Targets/NVPTX.h

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra added inline comments. Comment at: llvm/lib/Target/NVPTX/NVPTXInstrInfo.td:186 + !eq(name, "v2f16"): Float16x2Regs, + !eq(name, "bf16"): Float16Regs, + !eq(name, "v2bf16"): Float16x2Regs, tra wrote: > Allen wrote: > > sorry for a basic question: w

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra added a comment. In D136311#3882748 , @yaxunl wrote: > LGTM. Thanks. > > Do you plan to support arithmetic operators for bf16 or implement the FMA > instruction support? Yes. sm_90 has introduced a handful of new bf16 operations that will be eventu

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

Allen added inline comments. Comment at: llvm/lib/Target/NVPTX/NVPTXInstrInfo.td:186 + !eq(name, "v2f16"): Float16x2Regs, + !eq(name, "bf16"): Float16Regs, + !eq(name, "v2bf16"): Float16x2Regs, sorry for a basic question: what's the different between

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

yaxunl accepted this revision. yaxunl added a comment. LGTM. Thanks. Do you plan to support arithmetic operators for bf16 or implement the FMA instruction support? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D136311/new/ https://reviews.llvm.org

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

jchlanda accepted this revision. jchlanda added a comment. This revision is now accepted and ready to land. Looks good. Comment at: llvm/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp:838 } } +static int getLdStRegType(EVT VT) { New line here. Co

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra added a comment. @yaxunl It appears that AMDGPU also does not support `__bf16`, but for some reason it does not error out in clang headers: https://godbolt.org/z/GrTGMn49f Any ideas why that may be the case? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra added a comment. @jchlanda PTAL. You probably have the most context for NVPTX and bf16 instructions there. We need this change to unbreak CUDA compilation after D132329 exposed __bf16 to GPU-side compilation. https://godbolt.org/z/Kz8PYfPj5 Repository:

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra updated this revision to Diff 469683. tra added a comment. Added LLVM test for bfloat load/stores. Fixed asm output for bf16 constants. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D136311/new/ https://reviews.llvm.org/D136311 Files: clang/l

[PATCH] D136311: [CUDA,NVPTX] Implement __bf16 support for NVPTX.

tra updated this revision to Diff 469676. tra added a comment. Cosmetic refectoring. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D136311/new/ https://reviews.llvm.org/D136311 Files: clang/lib/Basic/Targets/NVPTX.cpp clang/lib/Basic/Targets/NV