[all-commits] [llvm/llvm-project] 1ee463: AMDGPU: Add cost model tests for minimumnum/maximu...

[Matt Arsenault via All-commits]

  Branch: refs/heads/users/arsenm/amdgpu/add-baseline-test-special-argument-intrinsics-costs
  Home:   https://github.com/llvm/llvm-project
  Commit: 1ee463bf5768bc41b25b4e888327fff16e6297c8
      https://github.com/llvm/llvm-project/commit/1ee463bf5768bc41b25b4e888327fff16e6297c8
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-06-17 (Tue, 17 Jun 2025)

  Changed paths:
    A llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll
    A llvm/test/Analysis/CostModel/AMDGPU/minimumnum.ll

  Log Message:
  -----------
  AMDGPU: Add cost model tests for minimumnum/maximumnum

The f16 cases in particular look broken since every vector size
has the same reported cost.


  Commit: e4fcaabbdfa400d015a6bb4b80d5831b3ee9b92f
      https://github.com/llvm/llvm-project/commit/e4fcaabbdfa400d015a6bb4b80d5831b3ee9b92f
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-06-17 (Tue, 17 Jun 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/test/Analysis/CostModel/AMDGPU/canonicalize.ll
    M llvm/test/Analysis/CostModel/AMDGPU/copysign.ll
    M llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll
    M llvm/test/Analysis/CostModel/AMDGPU/minimumnum.ll
    M llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll

  Log Message:
  -----------
  AMDGPU: Fix cost model for 16-bit operations on gfx8

We should only divide the number of pieces to fit the packed instructions
if we actually have pk instructions. This increases the cost of copysign,
but is closer to the current codegen output. It could be much cheaper
than it is now.


  Commit: 0ddc81d117497e6caea3334f7e62ff1aa62f0e3a
      https://github.com/llvm/llvm-project/commit/0ddc81d117497e6caea3334f7e62ff1aa62f0e3a
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-06-17 (Tue, 17 Jun 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/test/Analysis/CostModel/AMDGPU/copysign.ll

  Log Message:
  -----------
  AMDGPU: Reduce cost of f64 copysign

The real implementation is 1 real instruction plus a constant
materialize. Call that a 1, it's not a real f64 operation.


  Commit: 63d221ae57f3fb5f2e41bc29ff93338c209ab0fe
      https://github.com/llvm/llvm-project/commit/63d221ae57f3fb5f2e41bc29ff93338c209ab0fe
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-06-17 (Tue, 17 Jun 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h

  Log Message:
  -----------
  AMDGPU: Move fpenvIEEEMode into TTI


  Commit: 06cc9b7a57026c71069f38f91a30cc8a9532c48b
      https://github.com/llvm/llvm-project/commit/06cc9b7a57026c71069f38f91a30cc8a9532c48b
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-06-17 (Tue, 17 Jun 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll
    M llvm/test/Analysis/CostModel/AMDGPU/minimumnum.ll
    M llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll

  Log Message:
  -----------
  AMDGPU: Cost model for minimumnum/maximumnum


  Commit: 4497b8d17e9aa65d0eccb13a179a6dfb544d808d
      https://github.com/llvm/llvm-project/commit/4497b8d17e9aa65d0eccb13a179a6dfb544d808d
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2025-06-17 (Tue, 17 Jun 2025)

  Changed paths:
    A llvm/test/Analysis/CostModel/AMDGPU/special-argument-intrinsics.ll

  Log Message:
  -----------
  AMDGPU: Add baseline cost model tests for special argument intrinsics


Compare: https://github.com/llvm/llvm-project/compare/2eb6a6ed25e2...4497b8d17e9a

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