[MemCpyOpt][profcheck] Set unknown branch weights for certain selects

[mtrofin]

Issue #147390

[mtrofin]

• [MemCpyOpt][profcheck] Set unknown branch weights for certain selects #167597 👈 (View in Graphite)
• [MergeICmp][profcheck] Propagate profile info #167594
• [NFC] Generalize the arithmetic type for getDisjunctionWeights #167593
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Mircea Trofin (mtrofin)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/167597.diff

2 Files Affected:

• (modified) llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp (+3)
• (modified) llvm/test/Transforms/MemCpyOpt/memset-memcpy-dbgloc.ll (+5-3)

diff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 08be5df9872b7..5d11286e2cd7f 100644
--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -47,6 +47,7 @@
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/PassManager.h"
+#include "llvm/IR/ProfDataUtils.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/User.h"
 #include "llvm/IR/Value.h"
@@ -1366,6 +1367,8 @@ bool MemCpyOptPass::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
   Value *SizeDiff = Builder.CreateSub(DestSize, SrcSize);
   Value *MemsetLen = Builder.CreateSelect(
       Ule, ConstantInt::getNullValue(DestSize->getType()), SizeDiff);
+  if (auto *SI = dyn_cast<SelectInst>(MemsetLen))
+    setExplicitlyUnknownBranchWeightsIfProfiled(*SI, DEBUG_TYPE);
   Instruction *NewMemSet =
       Builder.CreateMemSet(Builder.CreatePtrAdd(Dest, SrcSize),
                            MemSet->getOperand(1), MemsetLen, Alignment);
diff --git a/llvm/test/Transforms/MemCpyOpt/memset-memcpy-dbgloc.ll b/llvm/test/Transforms/MemCpyOpt/memset-memcpy-dbgloc.ll
index 3f577f09ada1e..edd64e5efe1aa 100644
--- a/llvm/test/Transforms/MemCpyOpt/memset-memcpy-dbgloc.ll
+++ b/llvm/test/Transforms/MemCpyOpt/memset-memcpy-dbgloc.ll
@@ -7,14 +7,14 @@ target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
 declare void @llvm.memset.p0.i64(ptr nocapture, i8, i64, i1)
 declare void @llvm.memcpy.p0.p0.i64(ptr nocapture, ptr nocapture readonly, i64, i1)
 
-define void @test_constant(i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) !dbg !5 {
+define void @test_constant(i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) !dbg !5 !prof !14 {
 ; CHECK-LABEL: define void @test_constant(
-; CHECK-SAME: i64 [[SRC_SIZE:%.*]], ptr [[DST:%.*]], i64 [[DST_SIZE:%.*]], i8 [[C:%.*]]) !dbg [[DBG5:![0-9]+]] {
+; CHECK-SAME: i64 [[SRC_SIZE:%.*]], ptr [[DST:%.*]], i64 [[DST_SIZE:%.*]], i8 [[C:%.*]]) !dbg [[DBG5:![0-9]+]] !prof {{.*}} {
 ; CHECK-NEXT:    [[NON_ZERO:%.*]] = icmp ne i64 [[SRC_SIZE]], 0
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[NON_ZERO]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule i64 [[DST_SIZE]], [[SRC_SIZE]], !dbg [[DBG11:![0-9]+]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub i64 [[DST_SIZE]], [[SRC_SIZE]], !dbg [[DBG11]]
-; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP1]], i64 0, i64 [[TMP2]], !dbg [[DBG11]]
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP1]], i64 0, i64 [[TMP2]], !dbg [[DBG11]], !prof [[SELPROF:![0-9]+]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[DST]], i64 [[SRC_SIZE]], !dbg [[DBG11]]
 ; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 1 [[TMP4]], i8 [[C]], i64 [[TMP3]], i1 false), !dbg [[DBG11]]
 ; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr [[DST]], ptr @C, i64 [[SRC_SIZE]], i1 false), !dbg [[DBG12:![0-9]+]]
@@ -29,6 +29,7 @@ define void @test_constant(i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) !dbg !
 
 ; Validate that the memset is mapped to DILocation for the original memset.
 ; CHECK: [[DBG11]] = !DILocation(line: 1,
+; CHECK: [[SELPROF]] = !{!"unknown", !"memcpyopt"}
 ; CHECK: [[DBG12]] = !DILocation(line: 2,
 ; CHECK: [[DBG13]] = !DILocation(line: 3,
 
@@ -50,3 +51,4 @@ define void @test_constant(i64 %src_size, ptr %dst, i64 %dst_size, i8 %c) !dbg !
 !11 = !DILocation(line: 1, column: 1, scope: !5)
 !12 = !DILocation(line: 2, column: 1, scope: !5)
 !13 = !DILocation(line: 3, column: 1, scope: !5)
+!14 = !{!"function_entry_count", i32 10}
\ No newline at end of file

[teresajohnson]

lgtm with one question below

[teresajohnson]

Should this have a TODO or FIXME? I.e. is there a hope that we would be able to put something other than unknown here?

[mtrofin]

Chatted offline with @xur-llvm, actually we do record a histogram of frequency of various kinds of sizes, so some kind of analysis would be possible. Let's see how many of these selects make it in CodeGenPrepare (we have the DEBUG_TYPE to help with that). My suspicion is that we (1) don't do a good VP propagation for memops in the first place, and (2) some of these memops are synthetic (from intrinsics). Currently the profcheck work is about branches (br, switch, select), but we should get back to this.

[mtrofin]

Merge activity

• Nov 14, 6:35 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Nov 14, 6:36 PM UTC: @mtrofin merged this pull request with Graphite.