[all-commits] [llvm/llvm-project] 695944: [AMDGPU] Add LDS encoding granularity to TargetParser

[Chinmay Deshpande via All-commits]

  Branch: refs/heads/users/chinmaydd/lds-encoding-granule
  Home:   https://github.com/llvm/llvm-project
  Commit: 6959442887f9e795b4b745070a216759111ce17f
      https://github.com/llvm/llvm-project/commit/6959442887f9e795b4b745070a216759111ce17f
  Author: Chinmay Deshpande <chdeshpa at amd.com>
  Date:   2026-09-20 (Sun, 20 Sep 2026)

  Changed paths:
    M llvm/include/llvm/TargetParser/AMDGPUTargetParser.h
    M llvm/lib/Target/AMDGPU/AMDGPU.td
    M llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUFeatures.td
    M llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
    M llvm/lib/Target/AMDGPU/R600Processors.td
    M llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
    M llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
    M llvm/lib/TargetParser/AMDGPUTargetParser.cpp
    A llvm/test/CodeGen/AMDGPU/lds-size-gfx1030.ll
    A llvm/test/CodeGen/AMDGPU/lds-size-gfx9-4-generic.ll
    M llvm/test/TableGen/AMDGPUTargetDefLDSAllocGranularity.td
    M llvm/unittests/TargetParser/TargetParserTest.cpp

  Log Message:
  -----------
  [AMDGPU] Add LDS encoding granularity to TargetParser

Model LDS encoding granularity with dedicated features and expose the
byte-valued getLDSEncodingGranule query for GPUKind and subarch. Keep
encoding independent of the hardware allocation granularity used for
occupancy; GFX10.3, GFX11 and GFX12.0 encode in 512-byte units while
allocating 1024-byte blocks.

Migrate program resource register and PAL metadata encoding to the new
query and remove getLdsDwGranularity from AMDGPUBaseInfo. gfx9-4-generic
uses gfx950's 1280-byte encoding granule independently of LDS capacity.

Test encoding queries, feature membership, generic-target validation and
encoded LDS sizes, including the GFX10.3 allocation/encoding distinction.

Change-Id: I9d3c2c041605e9a45fa8fbda09fc3460a74953ea


  Commit: 147b3d9cf0f2fdbef8cc6ff0a5e09d8eea423cb7
      https://github.com/llvm/llvm-project/commit/147b3d9cf0f2fdbef8cc6ff0a5e09d8eea423cb7
  Author: Chinmay Deshpande <chdeshpa at amd.com>
  Date:   2026-09-20 (Sun, 20 Sep 2026)

  Changed paths:
    M llvm/include/llvm/TargetParser/AMDGPUTargetParser.h
    M llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
    M llvm/lib/TargetParser/AMDGPUTargetParser.cpp
    A llvm/test/CodeGen/AMDGPU/lds-size-default-device.ll
    M llvm/unittests/TargetParser/TargetParserTest.cpp

  Log Message:
  -----------
  [AMDGPU] Return zero LDS encoding granularity for dummy targets

Remove the redundant early return and assertion from getLDSEncodingGranule. Return zero when the target has no encoding granularity feature, and document and test the result for unknown and legacy generic targets.

Keep the existing 256-byte default in the assembly printer so compiling without a GPU still produces valid LDS sizes. Test default-target Mesa and PAL encodings and HSA metadata.

Change-Id: Ia0c69a0d7ce2858b31ee64100d37ae34bba6c42c


  Commit: 84291b02031ce730c82613be488dda1833f97757
      https://github.com/llvm/llvm-project/commit/84291b02031ce730c82613be488dda1833f97757
  Author: Chinmay Deshpande <chdeshpa at amd.com>
  Date:   2026-09-21 (Mon, 21 Sep 2026)

  Changed paths:
    M llvm/test/CodeGen/AMDGPU/lds-size-default-device.ll

  Log Message:
  -----------
  [AMDGPU] Use new triples in default LDS encoding test

Change-Id: I54360f63a4b7c64f968c4e8f81c61542a2856f69


Compare: https://github.com/llvm/llvm-project/compare/8cafdaaadf3a...84291b02031c

To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications