[llvm] [AMDGPU] Add function attribute to disable TBUFFER combine (PR #156454)
Piotr Sobczak via llvm-commits
llvm-commits at lists.llvm.org
Thu Sep 4 08:24:34 PDT 2025
================
@@ -195,6 +195,9 @@ SIMachineFunctionInfo::SIMachineFunctionInfo(const Function &F,
VGPRForAGPRCopy =
AMDGPU::VGPR_32RegClass.getRegister(ST.getMaxNumVGPRs(F) - 1);
}
+
+ if (F.hasFnAttribute("amdgpu-disable-tbuffer-combine"))
+ setDisableTBufferCombine(true);
----------------
piotrAMD wrote:
I think Matt's suggestion was that the conservative behaviour should be the default, so the tbuffer merging should be disabled by default, and only enabled when an attribute is set.
https://github.com/llvm/llvm-project/pull/156454
More information about the llvm-commits
mailing list