[all-commits] [llvm/llvm-project] bab513: [AMDGPU] Use new CSR cost calculation (#219220)

[Lukas Sommer via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: bab513bedcf5c7582a24c97976042af4dba19c01
      https://github.com/llvm/llvm-project/commit/bab513bedcf5c7582a24c97976042af4dba19c01
  Author: Lukas Sommer <lukas.sommer at amd.com>
  Date:   2026-09-10 (Thu, 10 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIRegisterInfo.h
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.256bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.320bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.576bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.640bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.704bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.768bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.832bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.896bit.ll
    M llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.960bit.ll
    M llvm/test/CodeGen/AMDGPU/av_spill_cross_bb_usage.mir
    M llvm/test/CodeGen/AMDGPU/function-resource-usage.ll
    M llvm/test/CodeGen/AMDGPU/indirect-call.ll
    M llvm/test/CodeGen/AMDGPU/ran-out-of-sgprs-allocation-failure.mir
    M llvm/test/CodeGen/AMDGPU/split-liverange-overlapping-copies.mir
    M llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask-phi-extend.ll
    M llvm/test/CodeGen/AMDGPU/unspill-vgpr-after-rewrite-vgpr-mfma.ll
    M llvm/test/CodeGen/AMDGPU/unstructured-cfg-def-use-issue.ll
    M llvm/test/CodeGen/AMDGPU/whole-wave-register-copy.ll

  Log Message:
  -----------
  [AMDGPU] Use new CSR cost calculation (#219220)

So far, AMDGPU backend relied on the legacy calculation for the cost of
the first use of a callee-save register. This legacy path scales the
entry block frequency by a fixed value, making a comparison with
alternatives, e.g., rematerialization, difficult due to different
scales. One example for that is the work in
https://github.com/llvm/llvm-project/pull/206756, where comparison of
rematerialization cost and first use of a CSR is hard to compare due to
the different scales.

Change to instead use the new calculation path with different target
hooks for better comparability.

As the new scale is different, the cost was changed. The new cost value
accounts for various factors impacting CSR cost.

This change (with different cost value) was originally part of
https://github.com/llvm/llvm-project/pull/202007. Another part of that
PR was already extracted into
https://github.com/llvm/llvm-project/pull/214857.

---------

Signed-off-by: Lukas Sommer <lukas.sommer at amd.com>



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