[PATCH] D152613: [AMDGPU] Remove integer division in VOPD checks

Stanislav Mekhanoshin via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Sat Jun 10 01:26:17 PDT 2023


rampitec created this revision.
rampitec added reviewers: foad, Joe_Nash.
Herald added subscribers: StephenFan, kerbowa, hiraditya, tpr, dstuttard, yaxunl, jvesely, kzhuravl, arsenm.
Herald added a project: All.
rampitec requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

There is no way any compiler can simplify this division, while
the check is done rather often.


https://reviews.llvm.org/D152613

Files:
  llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
  llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h


Index: llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
===================================================================
--- llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -560,8 +560,9 @@
   MAX_OPR_NUM = DST_NUM + MAX_SRC_NUM
 };
 
-// Number of VGPR banks per VOPD component operand.
-constexpr unsigned BANKS_NUM[] = {2, 4, 4, 2};
+// LSB mask for VGPR banks per VOPD component operand.
+// 4 banks result in a mask 3, setting 2 lower bits.
+constexpr unsigned VOPD_VGPR_BANK_MASKS[] = {1, 3, 3, 1};
 
 enum ComponentIndex : unsigned { X = 0, Y = 1 };
 constexpr unsigned COMPONENTS[] = {ComponentIndex::X, ComponentIndex::Y};
Index: llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -577,9 +577,10 @@
 
   unsigned CompOprIdx;
   for (CompOprIdx = 0; CompOprIdx < Component::MAX_OPR_NUM; ++CompOprIdx) {
-    unsigned BanksNum = BANKS_NUM[CompOprIdx];
+    unsigned BanksMasks = VOPD_VGPR_BANK_MASKS[CompOprIdx];
     if (OpXRegs[CompOprIdx] && OpYRegs[CompOprIdx] &&
-        (OpXRegs[CompOprIdx] % BanksNum == OpYRegs[CompOprIdx] % BanksNum))
+        ((OpXRegs[CompOprIdx] & BanksMasks) ==
+         (OpYRegs[CompOprIdx] & BanksMasks)))
       return CompOprIdx;
   }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D152613.530183.patch
Type: text/x-patch
Size: 1403 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230610/010bc2af/attachment.bin>


More information about the llvm-commits mailing list