[profcheck] Disable verification of selects on vector conditions. #167973
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Member
| @llvm/pr-subscribers-pgo Author: Mircea Trofin (mtrofin) ChangesWe don't currently support profile metadata on selects where the condition is a vector. Issue #147390 Patch is 66.16 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/167973.diff 3 Files Affected:
diff --git a/llvm/lib/Transforms/Utils/ProfileVerify.cpp b/llvm/lib/Transforms/Utils/ProfileVerify.cpp index c578b4b839258..149c0879edcdd 100644 --- a/llvm/lib/Transforms/Utils/ProfileVerify.cpp +++ b/llvm/lib/Transforms/Utils/ProfileVerify.cpp @@ -102,9 +102,11 @@ bool ProfileInjector::inject() { for (auto &BB : F) { if (AnnotateSelect) { for (auto &I : BB) { - if (isa<SelectInst>(I) && !I.getMetadata(LLVMContext::MD_prof)) - setBranchWeights(I, {SelectTrueWeight, SelectFalseWeight}, - /*IsExpected=*/false); + if (auto *SI = dyn_cast<SelectInst>(&I)) + if (!SI->getCondition()->getType()->isVectorTy() && + !I.getMetadata(LLVMContext::MD_prof)) + setBranchWeights(I, {SelectTrueWeight, SelectFalseWeight}, + /*IsExpected=*/false); } } auto *Term = getTerminatorBenefitingFromMDProf(BB); @@ -185,9 +187,11 @@ PreservedAnalyses ProfileVerifierPass::run(Function &F, for (const auto &BB : F) { if (AnnotateSelect) { for (const auto &I : BB) - if (isa<SelectInst>(I) && !I.getMetadata(LLVMContext::MD_prof)) - F.getContext().emitError( - "Profile verification failed: select annotation missing"); + if (auto *SI = dyn_cast<SelectInst>(&I)) + if (!SI->getCondition()->getType()->isVectorTy() && + !I.getMetadata(LLVMContext::MD_prof)) + F.getContext().emitError( + "Profile verification failed: select annotation missing"); } if (const auto *Term = ProfileInjector::getTerminatorBenefitingFromMDProf(BB)) diff --git a/llvm/test/Transforms/PGOProfile/profcheck-select.ll b/llvm/test/Transforms/PGOProfile/profcheck-select.ll index 74bcb3f52428b..e6b3ddd42fcb0 100644 --- a/llvm/test/Transforms/PGOProfile/profcheck-select.ll +++ b/llvm/test/Transforms/PGOProfile/profcheck-select.ll @@ -15,7 +15,11 @@ ; RUN: not opt -passes=prof-verify %t/verify-missing.ll 2>&1 | FileCheck %t/verify-missing.ll ; verify we can disable it. It's sufficient to see opt not failing. -; RUN: opt -passes=prof-verify -profcheck-annotate-select=0 %t/verify-missing.ll +; RUN: opt -passes=prof-verify -profcheck-annotate-select=0 --disable-output %t/verify-missing.ll + +; verify vector selects without profiles are OK. It's sufficient opt doesn't fail. +; RUN: opt -passes=prof-verify --disable-output %t/verify-vec.ll + ;--- inject.ll declare void @foo(i32 %a); @@ -24,8 +28,16 @@ define void @bar(i1 %c) { call void @foo(i32 %v) ret void } + +define <2 x i32> @vec(<2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2) { + %r = select <2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2 + ret <2 x i32> %r +} + ; CHECK-LABEL: @bar ; CHECK: %v = select i1 %c, i32 1, i32 2, !prof !1 +; CHECK-LABEL: @vec +; CHECK-NOT: select {{.*}} !prof ; CHECK: !0 = !{!"function_entry_count", i64 1000} ; CHECK: !1 = !{!"branch_weights", i32 2, i32 3} @@ -64,4 +76,10 @@ define void @bar(i1 %c) !prof !0 { ret void } !0 = !{!"function_entry_count", i64 1000} -; CHECK: Profile verification failed: select annotation missing \ No newline at end of file +; CHECK: Profile verification failed: select annotation missing + +;--- verify-vec.ll +define <2 x i32> @vec(<2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2) !prof !{!"function_entry_count", i32 10} { + %r = select <2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2 + ret <2 x i32> %r +} diff --git a/llvm/utils/profcheck-xfail.txt b/llvm/utils/profcheck-xfail.txt index 10a7b62229b8d..04fc387425159 100644 --- a/llvm/utils/profcheck-xfail.txt +++ b/llvm/utils/profcheck-xfail.txt @@ -1,679 +1,649 @@ -Bitcode/fcmp-fast.ll -Bitcode/flags.ll -CodeGen/AArch64/cgdata-merge-local.ll -CodeGen/AArch64/llvm-masked-gather-legal-for-sve.ll -CodeGen/AArch64/llvm-masked-scatter-legal-for-sve.ll -CodeGen/AArch64/selectopt-cast.ll -CodeGen/AArch64/selectopt.ll -CodeGen/ARM/loopvectorize_pr33804.ll -CodeGen/ARM/sjljeh-swifterror.ll -CodeGen/Hexagon/autohvx/interleave.ll -CodeGen/Hexagon/loop-idiom/hexagon-memmove1.ll -CodeGen/Hexagon/loop-idiom/hexagon-memmove2.ll -CodeGen/Hexagon/loop-idiom/memmove-rt-check.ll -CodeGen/Hexagon/masked_gather.ll -CodeGen/NVPTX/lower-ctor-dtor.ll -CodeGen/RISCV/zmmul.ll -CodeGen/WebAssembly/memory-interleave.ll -CodeGen/X86/AMX/amx-low-intrinsics.ll -CodeGen/X86/masked_gather_scatter.ll -CodeGen/X86/nocfivalue.ll -DebugInfo/AArch64/ir-outliner.ll -DebugInfo/assignment-tracking/X86/hotcoldsplit.ll -DebugInfo/Generic/block-asan.ll -DebugInfo/X86/asan_debug_info.ll -LTO/X86/diagnostic-handler-remarks-with-hotness.ll -Other/optimization-remarks-auto.ll -Other/X86/debugcounter-partiallyinlinelibcalls.ll -Transforms/AtomicExpand/ARM/atomic-expansion-v7.ll -Transforms/AtomicExpand/SPARC/partword.ll -Transforms/Attributor/align.ll -Transforms/Attributor/ArgumentPromotion/2008-02-01-ReturnAttrs.ll -Transforms/Attributor/ArgumentPromotion/2008-07-02-array-indexing.ll -Transforms/Attributor/ArgumentPromotion/2008-09-07-CGUpdate.ll -Transforms/Attributor/ArgumentPromotion/2008-09-08-CGUpdateSelfEdge.ll -Transforms/Attributor/ArgumentPromotion/aggregate-promote.ll -Transforms/Attributor/ArgumentPromotion/alignment.ll -Transforms/Attributor/ArgumentPromotion/alloca-as.ll -Transforms/Attributor/ArgumentPromotion/array.ll -Transforms/Attributor/ArgumentPromotion/attrs.ll -Transforms/Attributor/ArgumentPromotion/basictest.ll -Transforms/Attributor/ArgumentPromotion/byval-2.ll -Transforms/Attributor/ArgumentPromotion/byval.ll -Transforms/Attributor/ArgumentPromotion/chained.ll -Transforms/Attributor/ArgumentPromotion/control-flow2.ll -Transforms/Attributor/ArgumentPromotion/crash.ll -Transforms/Attributor/ArgumentPromotion/dbg.ll -Transforms/Attributor/ArgumentPromotion/fp80.ll -Transforms/Attributor/ArgumentPromotion/live_called_from_dead_2.ll -Transforms/Attributor/ArgumentPromotion/live_called_from_dead.ll -Transforms/Attributor/ArgumentPromotion/nonzero-address-spaces.ll -Transforms/Attributor/ArgumentPromotion/pr27568.ll -Transforms/Attributor/ArgumentPromotion/profile.ll -Transforms/Attributor/ArgumentPromotion/reserve-tbaa.ll -Transforms/Attributor/ArgumentPromotion/tail.ll -Transforms/Attributor/ArgumentPromotion/X86/attributes.ll -Transforms/Attributor/ArgumentPromotion/X86/min-legal-vector-width.ll -Transforms/Attributor/callgraph.ll -Transforms/Attributor/call-simplify-pointer-info.ll -Transforms/Attributor/internal-noalias.ll -Transforms/Attributor/IPConstantProp/2009-09-24-byval-ptr.ll -Transforms/Attributor/IPConstantProp/recursion.ll -Transforms/Attributor/IPConstantProp/remove-call-inst.ll -Transforms/Attributor/liveness.ll -Transforms/Attributor/memory_locations.ll -Transforms/Attributor/misc_crash.ll -Transforms/Attributor/noalias.ll -Transforms/Attributor/nodelete.ll -Transforms/Attributor/nofpclass.ll -Transforms/Attributor/nounwind.ll -Transforms/Attributor/potential.ll -Transforms/Attributor/range.ll -Transforms/Attributor/reduced/clear_cached_analysis_for_deleted_functions.ll -Transforms/Attributor/reduced/register_benchmark_test.ll -Transforms/Attributor/undefined_behavior.ll -Transforms/Attributor/value-simplify-gpu.ll -Transforms/Attributor/value-simplify.ll -Transforms/Attributor/value-simplify-local-remote.ll -Transforms/Attributor/value-simplify-pointer-info.ll -Transforms/Attributor/value-simplify-reachability.ll -Transforms/BlockExtractor/extract-blocks.ll -Transforms/BlockExtractor/extract-blocks-with-groups.ll -Transforms/CodeExtractor/2004-03-13-LoopExtractorCrash.ll -Transforms/CodeExtractor/2004-03-14-DominanceProblem.ll -Transforms/CodeExtractor/2004-03-14-NoSwitchSupport.ll -Transforms/CodeExtractor/2004-03-17-MissedLiveIns.ll -Transforms/CodeExtractor/2004-03-17-UpdatePHIsOutsideRegion.ll -Transforms/CodeExtractor/extract-assume.ll -Transforms/CodeExtractor/input-value-debug.ll -Transforms/CodeExtractor/LoopExtractor_alloca.ll -Transforms/CodeExtractor/LoopExtractor_crash.ll -Transforms/CodeExtractor/LoopExtractor_infinite.ll -Transforms/CodeExtractor/LoopExtractor.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div-constant-numerator.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div-not-exact.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div-special-cases.ll -Transforms/CodeGenPrepare/X86/vec-shift-inseltpoison.ll -Transforms/CodeGenPrepare/X86/vec-shift.ll -Transforms/Coroutines/coro-await-suspend-lower-invoke.ll -Transforms/Coroutines/coro-await-suspend-lower.ll -Transforms/Coroutines/coro-byval-param.ll -Transforms/Coroutines/coro-catchswitch-cleanuppad.ll -Transforms/Coroutines/coro-debug-coro-frame.ll -Transforms/Coroutines/coro-debug-dbg.values.ll -Transforms/Coroutines/coro-debug-dbg.values-not_used_in_frame.ll -Transforms/Coroutines/coro-debug-frame-variable.ll -Transforms/Coroutines/coro-debug-O2.ll -Transforms/Coroutines/coro-debug-spill-dbg.declare.ll -Transforms/Coroutines/coro-materialize.ll -Transforms/Coroutines/coro-noop.ll -Transforms/Coroutines/coro-noop-pacbti.ll -Transforms/Coroutines/coro-only-destroy-when-complete.ll -Transforms/Coroutines/coro-spill-suspend.ll -Transforms/Coroutines/coro-split-00.ll -Transforms/Coroutines/coro-split-alloc.ll -Transforms/Coroutines/coro-split-dbg-labels.ll -Transforms/Coroutines/coro-split-final-suspend.ll -Transforms/Coroutines/coro-split-hidden.ll -Transforms/Coroutines/coro-split-musttail10.ll -Transforms/Coroutines/coro-split-musttail12.ll -Transforms/Coroutines/coro-split-musttail13.ll -Transforms/Coroutines/coro-split-musttail1.ll -Transforms/Coroutines/coro-split-musttail2.ll -Transforms/Coroutines/coro-split-musttail3.ll -Transforms/Coroutines/coro-split-musttail4.ll -Transforms/Coroutines/coro-split-musttail5.ll -Transforms/Coroutines/coro-split-musttail6.ll -Transforms/Coroutines/coro-split-musttail7.ll -Transforms/Coroutines/coro-split-musttail8.ll -Transforms/Coroutines/coro-split-musttail9.ll -Transforms/Coroutines/coro-split-musttail-chain-pgo-counter-promo.ll -Transforms/Coroutines/coro-split-musttail.ll -Transforms/Coroutines/coro-split-musttail-ppc64le.ll -Transforms/Coroutines/coro-split-noinline.ll -Transforms/Coroutines/no-suspend.ll -Transforms/CorrelatedValuePropagation/cond-at-use.ll -Transforms/CorrelatedValuePropagation/urem-expansion.ll -Transforms/CorrelatedValuePropagation/urem.ll -Transforms/CrossDSOCFI/basic.ll -Transforms/CrossDSOCFI/cfi_functions.ll -Transforms/CrossDSOCFI/thumb.ll -Transforms/ExpandLargeDivRem/X86/sdiv129.ll -Transforms/ExpandLargeDivRem/X86/srem129.ll -Transforms/ExpandLargeDivRem/X86/udiv129.ll -Transforms/ExpandLargeDivRem/X86/urem129.ll -Transforms/ExpandLargeDivRem/X86/vector.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-fptosi129.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-fptoui129.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-si129tofp.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-ui129tofp.ll -Transforms/ExpandMemCmp/AArch64/memcmp.ll -Transforms/ExpandMemCmp/X86/memcmp.ll -Transforms/ExpandMemCmp/X86/memcmp-x32.ll -Transforms/ExpandVariadics/expand-va-intrinsic-split-linkage.ll -Transforms/ExpandVariadics/expand-va-intrinsic-split-simple.ll -Transforms/ExpandVariadics/intrinsics.ll -Transforms/FixIrreducible/basic.ll -Transforms/FixIrreducible/bug45623.ll -Transforms/FixIrreducible/callbr.ll -Transforms/FixIrreducible/nested.ll -Transforms/FixIrreducible/switch.ll -Transforms/GCOVProfiling/atomic-counter.ll -Transforms/GCOVProfiling/exit-block.ll -Transforms/GCOVProfiling/function-numbering.ll -Transforms/GCOVProfiling/global-ctor.ll -Transforms/GCOVProfiling/kcfi.ll -Transforms/GCOVProfiling/kcfi-normalize.ll -Transforms/GCOVProfiling/linezero.ll -Transforms/GCOVProfiling/linkagename.ll -Transforms/GCOVProfiling/module-flags.ll -Transforms/GCOVProfiling/noprofile.ll -Transforms/GCOVProfiling/nosanitize.ll -Transforms/GCOVProfiling/reset.ll -Transforms/GCOVProfiling/split-indirectbr-critical-edges.ll -Transforms/GCOVProfiling/three-element-mdnode.ll -Transforms/GCOVProfiling/version.ll -Transforms/GlobalOpt/disable-globals-aa.ll -Transforms/GlobalOpt/global-demotion.ll -Transforms/GlobalOpt/globalvar-code-model.ll -Transforms/GlobalOpt/large-int-crash.ll -Transforms/GlobalOpt/shrink-global-to-bool-check-debug.ll -Transforms/GlobalOpt/shrink-global-to-bool-opaque-ptrs.ll -Transforms/GVN/debugloc-load-select.ll -Transforms/GVN/load-through-select-dbg.ll -Transforms/GVN/masked-load-store.ll -Transforms/GVN/masked-load-store-no-mem-dep.ll -Transforms/GVN/opaque-ptr.ll -Transforms/GVN/pr69301.ll -Transforms/GVN/pre-invalid-prof-metadata.ll -Transforms/GVN/PRE/pre-load-through-select.ll -Transforms/GVN/PRE/pre-loop-load-through-select.ll -Transforms/HotColdSplit/addr-taken.ll -Transforms/HotColdSplit/apply-noreturn-bonus.ll -Transforms/HotColdSplit/assumption-cache-invalidation.ll -Transforms/HotColdSplit/delete-use-without-def-dbg-val.ll -Transforms/HotColdSplit/duplicate-phi-preds-crash.ll -Transforms/HotColdSplit/eh-pads.ll -Transforms/HotColdSplit/forward-dfs-reaches-marked-block.ll -Transforms/HotColdSplit/invalid-dbg-assign.ll -Transforms/HotColdSplit/lifetime-markers-on-inputs-1.ll -Transforms/HotColdSplit/lifetime-markers-on-inputs-2.ll -Transforms/HotColdSplit/lifetime-markers-on-inputs-3.ll -Transforms/HotColdSplit/minsize.ll -Transforms/HotColdSplit/multiple-exits.ll -Transforms/HotColdSplit/musttail.ll -Transforms/HotColdSplit/noreturn.ll -Transforms/HotColdSplit/outline-cold-asm.ll -Transforms/HotColdSplit/outline-disjoint-diamonds.ll -Transforms/HotColdSplit/outline-if-then-else.ll -Transforms/HotColdSplit/outline-inner-region.ll -Transforms/HotColdSplit/outline-multiple-entry-region.ll -Transforms/HotColdSplit/outline-outer-region.ll -Transforms/HotColdSplit/outline-while-loop.ll -Transforms/HotColdSplit/phi-with-distinct-outlined-values.ll -Transforms/HotColdSplit/region-overlap.ll -Transforms/HotColdSplit/resume.ll -Transforms/HotColdSplit/retain-section.ll -Transforms/HotColdSplit/section-splitting-custom.ll -Transforms/HotColdSplit/section-splitting-default.ll -Transforms/HotColdSplit/split-cold-2.ll -Transforms/HotColdSplit/split-out-dbg-label.ll -Transforms/HotColdSplit/split-out-dbg-val-of-arg.ll -Transforms/HotColdSplit/split-phis-in-exit-blocks.ll -Transforms/HotColdSplit/split-static-profile.ll -Transforms/HotColdSplit/stale-assume-in-original-func.ll -Transforms/HotColdSplit/succ-block-with-self-edge.ll -Transforms/HotColdSplit/swifterror.ll -Transforms/HotColdSplit/transfer-debug-info.ll -Transforms/HotColdSplit/unwind.ll -Transforms/HotColdSplit/update-split-loop-metadata.ll -Transforms/IndirectBrExpand/basic.ll -Transforms/IndVarSimplify/debugloc-rem-subst.ll -Transforms/IndVarSimplify/eliminate-rem.ll -Transforms/IndVarSimplify/invalidate-modified-lcssa-phi.ll -Transforms/IndVarSimplify/pr45835.ll -Transforms/IndVarSimplify/preserving-debugloc-rem-div.ll -Transforms/InstCombine/2004-09-20-BadLoadCombine.ll -Transforms/InstCombine/2005-04-07-UDivSelectCrash.ll -Transforms/InstCombine/AArch64/sve-intrinsic-sel.ll -Transforms/InstCombine/AArch64/sve-intrinsic-simplify-binop.ll -Transforms/InstCombine/AArch64/sve-intrinsic-simplify-shift.ll -Transforms/InstCombine/add-mask.ll -Transforms/InstCombine/add-shl-mul-umax.ll -Transforms/InstCombine/and2.ll -Transforms/InstCombine/and-fcmp.ll -Transforms/InstCombine/and.ll -Transforms/InstCombine/and-or-icmps.ll -Transforms/InstCombine/apint-div1.ll -Transforms/InstCombine/apint-div2.ll -Transforms/InstCombine/ashr-demand.ll -Transforms/InstCombine/atomic.ll -Transforms/InstCombine/binop-cast.ll -Transforms/InstCombine/binop-select-cast-of-select-cond.ll -Transforms/InstCombine/binop-select.ll -Transforms/InstCombine/bit-checks.ll -Transforms/InstCombine/bitreverse.ll -Transforms/InstCombine/branch.ll -Transforms/InstCombine/builtin-dynamic-object-size.ll -Transforms/InstCombine/canonicalize-clamp-like-pattern-between-negative-and-positive-thresholds.ll -Transforms/InstCombine/canonicalize-clamp-like-pattern-between-zero-and-positive-threshold.ll -Transforms/InstCombine/cast-mul-select.ll -Transforms/InstCombine/clamp-to-minmax.ll -Transforms/InstCombine/conditional-negation.ll -Transforms/InstCombine/cttz.ll -Transforms/InstCombine/debuginfo-invert.ll -Transforms/InstCombine/demorgan.ll -Transforms/InstCombine/div.ll -Transforms/InstCombine/div-shift.ll -Transforms/InstCombine/fabs.ll -Transforms/InstCombine/fcmp-select.ll -Transforms/InstCombine/ffs-1.ll -Transforms/InstCombine/ffs-i16.ll -Transforms/InstCombine/fmul-bool.ll -Transforms/InstCombine/fmul.ll -Transforms/InstCombine/fneg.ll -Transforms/InstCombine/fold-ctpop-of-not.ll -Transforms/InstCombine/fold-ext-eq-c-with-op.ll -Transforms/InstCombine/free-inversion.ll -Transforms/InstCombine/icmp-and-lowbit-mask.ll -Transforms/InstCombine/icmp.ll -Transforms/InstCombine/icmp-mul-and.ll -Transforms/InstCombine/icmp-of-and-x.ll -Transforms/InstCombine/icmp-of-or-x.ll -Transforms/InstCombine/icmp-select-implies-common-op.ll -Transforms/InstCombine/icmp-select.ll -Transforms/InstCombine/icmp-with-selects.ll -Transforms/InstCombine/intrinsic-select.ll -Transforms/InstCombine/ldexp-ext.ll -Transforms/InstCombine/ldexp.ll -Transforms/InstCombine/load-bitcast-select.ll -Transforms/InstCombine/load.ll -Transforms/InstCombine/load-select.ll -Transforms/InstCombine/loadstore-metadata.ll -Transforms/InstCombine/logical-select-inseltpoison.ll -Transforms/InstCombine/logical-select.ll -Transforms/InstCombine/lshr.ll -Transforms/InstCombine/masked_intrinsics-inseltpoison.ll -Transforms/InstCombine/masked_intrinsics.ll -Transforms/InstCombine/memchr-11.ll -Transforms/InstCombine/memchr-2.ll -Transforms/InstCombine/memchr-3.ll -Transforms/InstCombine/memchr-6.ll -Transforms/InstCombine/memchr-7.ll -Transforms/InstCombine/memchr-9.ll -Transforms/InstCombine/memchr.ll -Transforms/InstCombine/mem-gep-zidx.ll -Transforms/InstCombine/memrchr-3.ll -Transforms/InstCombine/memrchr-4.ll -Transforms/InstCombine/minmax-fold.ll -Transforms/InstCombine/minmax-fp.ll -Transforms/InstCombine/minmax-intrinsics.ll -Transforms/InstCombine/mul-inseltpoison.ll -Transforms/InstCombine/mul.ll -Transforms/InstCombine/mul-masked-bits.ll -Transforms/InstCombine/mul-pow2.ll -Transforms/InstCombine/multiple-uses-load-bitcast-select.ll -Transforms/InstCombine/narrow.ll -Transforms/InstCombine/negated-bitmask.ll -Transforms/InstCombine/nested-select.ll -Transforms/InstCombine/not.ll -Transforms/InstCombine/or-bitmask.ll -Transforms/InstCombine/or-fcmp.ll -Transforms/InstCombine/or.ll -Transforms/InstCombine/pow-1.ll -Transforms/InstCombine/pow-3.ll -Transforms/InstCombine/pow-sqrt.ll -Transforms/InstCombine/pull-conditional-binop-through-shift.ll -Transforms/InstCombine/rem.ll -Transforms/InstCombine/sdiv-canonicalize.ll -Transforms/InstCombine/sdiv-guard.ll -Transforms/InstCombine/select-and-or.ll -Transforms/InstCombine/select-cmp-br.ll -Transforms/InstCombine/select-cmp.ll -Transforms/InstCombine/select-factorize.ll -Transforms/InstCombine/select_frexp.ll -Transforms/InstCombine/select.ll -Transforms/InstCombine/select-min-max.ll -Transforms/InstCombine/select-of-symmetric-selects.ll -Transforms/InstCombine/select-select.ll -Transforms/InstCombine/shift.ll -Transforms/InstCombine/shuffle-select-narrow-inseltpoison.ll -Transforms/InstCombine/shuffle-select-narrow.ll -Transforms/InstCombine/simplify-demanded-fpclass.ll -Transforms/InstCombine/sink-not-into-another-hand-of-logical-and.ll -Transforms/InstCombine/sink-not-into-another-hand-of-logical-or.ll -Transforms/InstCombine/sink-not-into-logical-and.ll -Transforms/InstCombine/sink-not-into-logical-or.ll -Transforms/InstCombine/strchr-1.ll -Transforms/InstCombine/strchr-3.ll -Transforms/InstCombine/strlen-1.ll -Transforms/InstCombine/strrchr-3.ll -Transforms/InstCombine/sub-ashr-and-to-icmp-select.ll -Transforms/InstCombine/sub-ashr-or-to-icmp-select.ll -Transforms/InstCombine/sub-xor-cmp.ll -Transforms/InstCombine/truncating-saturate.ll -Transforms/InstCombine/unordered-fcmp-select.ll -Transforms/InstCombine/urem-via-cmp-select.ll -Transforms/InstCombine/vec_sext.ll -Transforms/InstCombine/vector-urem.ll -Transforms/InstCombine/wcslen-1.ll -Transforms/InstCombine/wcslen-3.ll -Transforms/InstCombine/X86/blend_x86.ll -Transforms/InstCombine/X86/x86-avx512-inseltpoison.ll -Transforms/InstCombine/X86/x86-avx512.ll -Transforms/InstCombine/xor-and-or.ll -Transforms/InstComb... [truncated] |
Member
| @llvm/pr-subscribers-llvm-transforms Author: Mircea Trofin (mtrofin) ChangesWe don't currently support profile metadata on selects where the condition is a vector. Issue #147390 Patch is 66.16 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/167973.diff 3 Files Affected:
diff --git a/llvm/lib/Transforms/Utils/ProfileVerify.cpp b/llvm/lib/Transforms/Utils/ProfileVerify.cpp index c578b4b839258..149c0879edcdd 100644 --- a/llvm/lib/Transforms/Utils/ProfileVerify.cpp +++ b/llvm/lib/Transforms/Utils/ProfileVerify.cpp @@ -102,9 +102,11 @@ bool ProfileInjector::inject() { for (auto &BB : F) { if (AnnotateSelect) { for (auto &I : BB) { - if (isa<SelectInst>(I) && !I.getMetadata(LLVMContext::MD_prof)) - setBranchWeights(I, {SelectTrueWeight, SelectFalseWeight}, - /*IsExpected=*/false); + if (auto *SI = dyn_cast<SelectInst>(&I)) + if (!SI->getCondition()->getType()->isVectorTy() && + !I.getMetadata(LLVMContext::MD_prof)) + setBranchWeights(I, {SelectTrueWeight, SelectFalseWeight}, + /*IsExpected=*/false); } } auto *Term = getTerminatorBenefitingFromMDProf(BB); @@ -185,9 +187,11 @@ PreservedAnalyses ProfileVerifierPass::run(Function &F, for (const auto &BB : F) { if (AnnotateSelect) { for (const auto &I : BB) - if (isa<SelectInst>(I) && !I.getMetadata(LLVMContext::MD_prof)) - F.getContext().emitError( - "Profile verification failed: select annotation missing"); + if (auto *SI = dyn_cast<SelectInst>(&I)) + if (!SI->getCondition()->getType()->isVectorTy() && + !I.getMetadata(LLVMContext::MD_prof)) + F.getContext().emitError( + "Profile verification failed: select annotation missing"); } if (const auto *Term = ProfileInjector::getTerminatorBenefitingFromMDProf(BB)) diff --git a/llvm/test/Transforms/PGOProfile/profcheck-select.ll b/llvm/test/Transforms/PGOProfile/profcheck-select.ll index 74bcb3f52428b..e6b3ddd42fcb0 100644 --- a/llvm/test/Transforms/PGOProfile/profcheck-select.ll +++ b/llvm/test/Transforms/PGOProfile/profcheck-select.ll @@ -15,7 +15,11 @@ ; RUN: not opt -passes=prof-verify %t/verify-missing.ll 2>&1 | FileCheck %t/verify-missing.ll ; verify we can disable it. It's sufficient to see opt not failing. -; RUN: opt -passes=prof-verify -profcheck-annotate-select=0 %t/verify-missing.ll +; RUN: opt -passes=prof-verify -profcheck-annotate-select=0 --disable-output %t/verify-missing.ll + +; verify vector selects without profiles are OK. It's sufficient opt doesn't fail. +; RUN: opt -passes=prof-verify --disable-output %t/verify-vec.ll + ;--- inject.ll declare void @foo(i32 %a); @@ -24,8 +28,16 @@ define void @bar(i1 %c) { call void @foo(i32 %v) ret void } + +define <2 x i32> @vec(<2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2) { + %r = select <2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2 + ret <2 x i32> %r +} + ; CHECK-LABEL: @bar ; CHECK: %v = select i1 %c, i32 1, i32 2, !prof !1 +; CHECK-LABEL: @vec +; CHECK-NOT: select {{.*}} !prof ; CHECK: !0 = !{!"function_entry_count", i64 1000} ; CHECK: !1 = !{!"branch_weights", i32 2, i32 3} @@ -64,4 +76,10 @@ define void @bar(i1 %c) !prof !0 { ret void } !0 = !{!"function_entry_count", i64 1000} -; CHECK: Profile verification failed: select annotation missing \ No newline at end of file +; CHECK: Profile verification failed: select annotation missing + +;--- verify-vec.ll +define <2 x i32> @vec(<2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2) !prof !{!"function_entry_count", i32 10} { + %r = select <2 x i1> %c, <2 x i32> %v1, <2 x i32> %v2 + ret <2 x i32> %r +} diff --git a/llvm/utils/profcheck-xfail.txt b/llvm/utils/profcheck-xfail.txt index 10a7b62229b8d..04fc387425159 100644 --- a/llvm/utils/profcheck-xfail.txt +++ b/llvm/utils/profcheck-xfail.txt @@ -1,679 +1,649 @@ -Bitcode/fcmp-fast.ll -Bitcode/flags.ll -CodeGen/AArch64/cgdata-merge-local.ll -CodeGen/AArch64/llvm-masked-gather-legal-for-sve.ll -CodeGen/AArch64/llvm-masked-scatter-legal-for-sve.ll -CodeGen/AArch64/selectopt-cast.ll -CodeGen/AArch64/selectopt.ll -CodeGen/ARM/loopvectorize_pr33804.ll -CodeGen/ARM/sjljeh-swifterror.ll -CodeGen/Hexagon/autohvx/interleave.ll -CodeGen/Hexagon/loop-idiom/hexagon-memmove1.ll -CodeGen/Hexagon/loop-idiom/hexagon-memmove2.ll -CodeGen/Hexagon/loop-idiom/memmove-rt-check.ll -CodeGen/Hexagon/masked_gather.ll -CodeGen/NVPTX/lower-ctor-dtor.ll -CodeGen/RISCV/zmmul.ll -CodeGen/WebAssembly/memory-interleave.ll -CodeGen/X86/AMX/amx-low-intrinsics.ll -CodeGen/X86/masked_gather_scatter.ll -CodeGen/X86/nocfivalue.ll -DebugInfo/AArch64/ir-outliner.ll -DebugInfo/assignment-tracking/X86/hotcoldsplit.ll -DebugInfo/Generic/block-asan.ll -DebugInfo/X86/asan_debug_info.ll -LTO/X86/diagnostic-handler-remarks-with-hotness.ll -Other/optimization-remarks-auto.ll -Other/X86/debugcounter-partiallyinlinelibcalls.ll -Transforms/AtomicExpand/ARM/atomic-expansion-v7.ll -Transforms/AtomicExpand/SPARC/partword.ll -Transforms/Attributor/align.ll -Transforms/Attributor/ArgumentPromotion/2008-02-01-ReturnAttrs.ll -Transforms/Attributor/ArgumentPromotion/2008-07-02-array-indexing.ll -Transforms/Attributor/ArgumentPromotion/2008-09-07-CGUpdate.ll -Transforms/Attributor/ArgumentPromotion/2008-09-08-CGUpdateSelfEdge.ll -Transforms/Attributor/ArgumentPromotion/aggregate-promote.ll -Transforms/Attributor/ArgumentPromotion/alignment.ll -Transforms/Attributor/ArgumentPromotion/alloca-as.ll -Transforms/Attributor/ArgumentPromotion/array.ll -Transforms/Attributor/ArgumentPromotion/attrs.ll -Transforms/Attributor/ArgumentPromotion/basictest.ll -Transforms/Attributor/ArgumentPromotion/byval-2.ll -Transforms/Attributor/ArgumentPromotion/byval.ll -Transforms/Attributor/ArgumentPromotion/chained.ll -Transforms/Attributor/ArgumentPromotion/control-flow2.ll -Transforms/Attributor/ArgumentPromotion/crash.ll -Transforms/Attributor/ArgumentPromotion/dbg.ll -Transforms/Attributor/ArgumentPromotion/fp80.ll -Transforms/Attributor/ArgumentPromotion/live_called_from_dead_2.ll -Transforms/Attributor/ArgumentPromotion/live_called_from_dead.ll -Transforms/Attributor/ArgumentPromotion/nonzero-address-spaces.ll -Transforms/Attributor/ArgumentPromotion/pr27568.ll -Transforms/Attributor/ArgumentPromotion/profile.ll -Transforms/Attributor/ArgumentPromotion/reserve-tbaa.ll -Transforms/Attributor/ArgumentPromotion/tail.ll -Transforms/Attributor/ArgumentPromotion/X86/attributes.ll -Transforms/Attributor/ArgumentPromotion/X86/min-legal-vector-width.ll -Transforms/Attributor/callgraph.ll -Transforms/Attributor/call-simplify-pointer-info.ll -Transforms/Attributor/internal-noalias.ll -Transforms/Attributor/IPConstantProp/2009-09-24-byval-ptr.ll -Transforms/Attributor/IPConstantProp/recursion.ll -Transforms/Attributor/IPConstantProp/remove-call-inst.ll -Transforms/Attributor/liveness.ll -Transforms/Attributor/memory_locations.ll -Transforms/Attributor/misc_crash.ll -Transforms/Attributor/noalias.ll -Transforms/Attributor/nodelete.ll -Transforms/Attributor/nofpclass.ll -Transforms/Attributor/nounwind.ll -Transforms/Attributor/potential.ll -Transforms/Attributor/range.ll -Transforms/Attributor/reduced/clear_cached_analysis_for_deleted_functions.ll -Transforms/Attributor/reduced/register_benchmark_test.ll -Transforms/Attributor/undefined_behavior.ll -Transforms/Attributor/value-simplify-gpu.ll -Transforms/Attributor/value-simplify.ll -Transforms/Attributor/value-simplify-local-remote.ll -Transforms/Attributor/value-simplify-pointer-info.ll -Transforms/Attributor/value-simplify-reachability.ll -Transforms/BlockExtractor/extract-blocks.ll -Transforms/BlockExtractor/extract-blocks-with-groups.ll -Transforms/CodeExtractor/2004-03-13-LoopExtractorCrash.ll -Transforms/CodeExtractor/2004-03-14-DominanceProblem.ll -Transforms/CodeExtractor/2004-03-14-NoSwitchSupport.ll -Transforms/CodeExtractor/2004-03-17-MissedLiveIns.ll -Transforms/CodeExtractor/2004-03-17-UpdatePHIsOutsideRegion.ll -Transforms/CodeExtractor/extract-assume.ll -Transforms/CodeExtractor/input-value-debug.ll -Transforms/CodeExtractor/LoopExtractor_alloca.ll -Transforms/CodeExtractor/LoopExtractor_crash.ll -Transforms/CodeExtractor/LoopExtractor_infinite.ll -Transforms/CodeExtractor/LoopExtractor.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div-constant-numerator.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div-not-exact.ll -Transforms/CodeGenPrepare/NVPTX/bypass-slow-div-special-cases.ll -Transforms/CodeGenPrepare/X86/vec-shift-inseltpoison.ll -Transforms/CodeGenPrepare/X86/vec-shift.ll -Transforms/Coroutines/coro-await-suspend-lower-invoke.ll -Transforms/Coroutines/coro-await-suspend-lower.ll -Transforms/Coroutines/coro-byval-param.ll -Transforms/Coroutines/coro-catchswitch-cleanuppad.ll -Transforms/Coroutines/coro-debug-coro-frame.ll -Transforms/Coroutines/coro-debug-dbg.values.ll -Transforms/Coroutines/coro-debug-dbg.values-not_used_in_frame.ll -Transforms/Coroutines/coro-debug-frame-variable.ll -Transforms/Coroutines/coro-debug-O2.ll -Transforms/Coroutines/coro-debug-spill-dbg.declare.ll -Transforms/Coroutines/coro-materialize.ll -Transforms/Coroutines/coro-noop.ll -Transforms/Coroutines/coro-noop-pacbti.ll -Transforms/Coroutines/coro-only-destroy-when-complete.ll -Transforms/Coroutines/coro-spill-suspend.ll -Transforms/Coroutines/coro-split-00.ll -Transforms/Coroutines/coro-split-alloc.ll -Transforms/Coroutines/coro-split-dbg-labels.ll -Transforms/Coroutines/coro-split-final-suspend.ll -Transforms/Coroutines/coro-split-hidden.ll -Transforms/Coroutines/coro-split-musttail10.ll -Transforms/Coroutines/coro-split-musttail12.ll -Transforms/Coroutines/coro-split-musttail13.ll -Transforms/Coroutines/coro-split-musttail1.ll -Transforms/Coroutines/coro-split-musttail2.ll -Transforms/Coroutines/coro-split-musttail3.ll -Transforms/Coroutines/coro-split-musttail4.ll -Transforms/Coroutines/coro-split-musttail5.ll -Transforms/Coroutines/coro-split-musttail6.ll -Transforms/Coroutines/coro-split-musttail7.ll -Transforms/Coroutines/coro-split-musttail8.ll -Transforms/Coroutines/coro-split-musttail9.ll -Transforms/Coroutines/coro-split-musttail-chain-pgo-counter-promo.ll -Transforms/Coroutines/coro-split-musttail.ll -Transforms/Coroutines/coro-split-musttail-ppc64le.ll -Transforms/Coroutines/coro-split-noinline.ll -Transforms/Coroutines/no-suspend.ll -Transforms/CorrelatedValuePropagation/cond-at-use.ll -Transforms/CorrelatedValuePropagation/urem-expansion.ll -Transforms/CorrelatedValuePropagation/urem.ll -Transforms/CrossDSOCFI/basic.ll -Transforms/CrossDSOCFI/cfi_functions.ll -Transforms/CrossDSOCFI/thumb.ll -Transforms/ExpandLargeDivRem/X86/sdiv129.ll -Transforms/ExpandLargeDivRem/X86/srem129.ll -Transforms/ExpandLargeDivRem/X86/udiv129.ll -Transforms/ExpandLargeDivRem/X86/urem129.ll -Transforms/ExpandLargeDivRem/X86/vector.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-fptosi129.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-fptoui129.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-si129tofp.ll -Transforms/ExpandLargeFpConvert/X86/expand-large-fp-convert-ui129tofp.ll -Transforms/ExpandMemCmp/AArch64/memcmp.ll -Transforms/ExpandMemCmp/X86/memcmp.ll -Transforms/ExpandMemCmp/X86/memcmp-x32.ll -Transforms/ExpandVariadics/expand-va-intrinsic-split-linkage.ll -Transforms/ExpandVariadics/expand-va-intrinsic-split-simple.ll -Transforms/ExpandVariadics/intrinsics.ll -Transforms/FixIrreducible/basic.ll -Transforms/FixIrreducible/bug45623.ll -Transforms/FixIrreducible/callbr.ll -Transforms/FixIrreducible/nested.ll -Transforms/FixIrreducible/switch.ll -Transforms/GCOVProfiling/atomic-counter.ll -Transforms/GCOVProfiling/exit-block.ll -Transforms/GCOVProfiling/function-numbering.ll -Transforms/GCOVProfiling/global-ctor.ll -Transforms/GCOVProfiling/kcfi.ll -Transforms/GCOVProfiling/kcfi-normalize.ll -Transforms/GCOVProfiling/linezero.ll -Transforms/GCOVProfiling/linkagename.ll -Transforms/GCOVProfiling/module-flags.ll -Transforms/GCOVProfiling/noprofile.ll -Transforms/GCOVProfiling/nosanitize.ll -Transforms/GCOVProfiling/reset.ll -Transforms/GCOVProfiling/split-indirectbr-critical-edges.ll -Transforms/GCOVProfiling/three-element-mdnode.ll -Transforms/GCOVProfiling/version.ll -Transforms/GlobalOpt/disable-globals-aa.ll -Transforms/GlobalOpt/global-demotion.ll -Transforms/GlobalOpt/globalvar-code-model.ll -Transforms/GlobalOpt/large-int-crash.ll -Transforms/GlobalOpt/shrink-global-to-bool-check-debug.ll -Transforms/GlobalOpt/shrink-global-to-bool-opaque-ptrs.ll -Transforms/GVN/debugloc-load-select.ll -Transforms/GVN/load-through-select-dbg.ll -Transforms/GVN/masked-load-store.ll -Transforms/GVN/masked-load-store-no-mem-dep.ll -Transforms/GVN/opaque-ptr.ll -Transforms/GVN/pr69301.ll -Transforms/GVN/pre-invalid-prof-metadata.ll -Transforms/GVN/PRE/pre-load-through-select.ll -Transforms/GVN/PRE/pre-loop-load-through-select.ll -Transforms/HotColdSplit/addr-taken.ll -Transforms/HotColdSplit/apply-noreturn-bonus.ll -Transforms/HotColdSplit/assumption-cache-invalidation.ll -Transforms/HotColdSplit/delete-use-without-def-dbg-val.ll -Transforms/HotColdSplit/duplicate-phi-preds-crash.ll -Transforms/HotColdSplit/eh-pads.ll -Transforms/HotColdSplit/forward-dfs-reaches-marked-block.ll -Transforms/HotColdSplit/invalid-dbg-assign.ll -Transforms/HotColdSplit/lifetime-markers-on-inputs-1.ll -Transforms/HotColdSplit/lifetime-markers-on-inputs-2.ll -Transforms/HotColdSplit/lifetime-markers-on-inputs-3.ll -Transforms/HotColdSplit/minsize.ll -Transforms/HotColdSplit/multiple-exits.ll -Transforms/HotColdSplit/musttail.ll -Transforms/HotColdSplit/noreturn.ll -Transforms/HotColdSplit/outline-cold-asm.ll -Transforms/HotColdSplit/outline-disjoint-diamonds.ll -Transforms/HotColdSplit/outline-if-then-else.ll -Transforms/HotColdSplit/outline-inner-region.ll -Transforms/HotColdSplit/outline-multiple-entry-region.ll -Transforms/HotColdSplit/outline-outer-region.ll -Transforms/HotColdSplit/outline-while-loop.ll -Transforms/HotColdSplit/phi-with-distinct-outlined-values.ll -Transforms/HotColdSplit/region-overlap.ll -Transforms/HotColdSplit/resume.ll -Transforms/HotColdSplit/retain-section.ll -Transforms/HotColdSplit/section-splitting-custom.ll -Transforms/HotColdSplit/section-splitting-default.ll -Transforms/HotColdSplit/split-cold-2.ll -Transforms/HotColdSplit/split-out-dbg-label.ll -Transforms/HotColdSplit/split-out-dbg-val-of-arg.ll -Transforms/HotColdSplit/split-phis-in-exit-blocks.ll -Transforms/HotColdSplit/split-static-profile.ll -Transforms/HotColdSplit/stale-assume-in-original-func.ll -Transforms/HotColdSplit/succ-block-with-self-edge.ll -Transforms/HotColdSplit/swifterror.ll -Transforms/HotColdSplit/transfer-debug-info.ll -Transforms/HotColdSplit/unwind.ll -Transforms/HotColdSplit/update-split-loop-metadata.ll -Transforms/IndirectBrExpand/basic.ll -Transforms/IndVarSimplify/debugloc-rem-subst.ll -Transforms/IndVarSimplify/eliminate-rem.ll -Transforms/IndVarSimplify/invalidate-modified-lcssa-phi.ll -Transforms/IndVarSimplify/pr45835.ll -Transforms/IndVarSimplify/preserving-debugloc-rem-div.ll -Transforms/InstCombine/2004-09-20-BadLoadCombine.ll -Transforms/InstCombine/2005-04-07-UDivSelectCrash.ll -Transforms/InstCombine/AArch64/sve-intrinsic-sel.ll -Transforms/InstCombine/AArch64/sve-intrinsic-simplify-binop.ll -Transforms/InstCombine/AArch64/sve-intrinsic-simplify-shift.ll -Transforms/InstCombine/add-mask.ll -Transforms/InstCombine/add-shl-mul-umax.ll -Transforms/InstCombine/and2.ll -Transforms/InstCombine/and-fcmp.ll -Transforms/InstCombine/and.ll -Transforms/InstCombine/and-or-icmps.ll -Transforms/InstCombine/apint-div1.ll -Transforms/InstCombine/apint-div2.ll -Transforms/InstCombine/ashr-demand.ll -Transforms/InstCombine/atomic.ll -Transforms/InstCombine/binop-cast.ll -Transforms/InstCombine/binop-select-cast-of-select-cond.ll -Transforms/InstCombine/binop-select.ll -Transforms/InstCombine/bit-checks.ll -Transforms/InstCombine/bitreverse.ll -Transforms/InstCombine/branch.ll -Transforms/InstCombine/builtin-dynamic-object-size.ll -Transforms/InstCombine/canonicalize-clamp-like-pattern-between-negative-and-positive-thresholds.ll -Transforms/InstCombine/canonicalize-clamp-like-pattern-between-zero-and-positive-threshold.ll -Transforms/InstCombine/cast-mul-select.ll -Transforms/InstCombine/clamp-to-minmax.ll -Transforms/InstCombine/conditional-negation.ll -Transforms/InstCombine/cttz.ll -Transforms/InstCombine/debuginfo-invert.ll -Transforms/InstCombine/demorgan.ll -Transforms/InstCombine/div.ll -Transforms/InstCombine/div-shift.ll -Transforms/InstCombine/fabs.ll -Transforms/InstCombine/fcmp-select.ll -Transforms/InstCombine/ffs-1.ll -Transforms/InstCombine/ffs-i16.ll -Transforms/InstCombine/fmul-bool.ll -Transforms/InstCombine/fmul.ll -Transforms/InstCombine/fneg.ll -Transforms/InstCombine/fold-ctpop-of-not.ll -Transforms/InstCombine/fold-ext-eq-c-with-op.ll -Transforms/InstCombine/free-inversion.ll -Transforms/InstCombine/icmp-and-lowbit-mask.ll -Transforms/InstCombine/icmp.ll -Transforms/InstCombine/icmp-mul-and.ll -Transforms/InstCombine/icmp-of-and-x.ll -Transforms/InstCombine/icmp-of-or-x.ll -Transforms/InstCombine/icmp-select-implies-common-op.ll -Transforms/InstCombine/icmp-select.ll -Transforms/InstCombine/icmp-with-selects.ll -Transforms/InstCombine/intrinsic-select.ll -Transforms/InstCombine/ldexp-ext.ll -Transforms/InstCombine/ldexp.ll -Transforms/InstCombine/load-bitcast-select.ll -Transforms/InstCombine/load.ll -Transforms/InstCombine/load-select.ll -Transforms/InstCombine/loadstore-metadata.ll -Transforms/InstCombine/logical-select-inseltpoison.ll -Transforms/InstCombine/logical-select.ll -Transforms/InstCombine/lshr.ll -Transforms/InstCombine/masked_intrinsics-inseltpoison.ll -Transforms/InstCombine/masked_intrinsics.ll -Transforms/InstCombine/memchr-11.ll -Transforms/InstCombine/memchr-2.ll -Transforms/InstCombine/memchr-3.ll -Transforms/InstCombine/memchr-6.ll -Transforms/InstCombine/memchr-7.ll -Transforms/InstCombine/memchr-9.ll -Transforms/InstCombine/memchr.ll -Transforms/InstCombine/mem-gep-zidx.ll -Transforms/InstCombine/memrchr-3.ll -Transforms/InstCombine/memrchr-4.ll -Transforms/InstCombine/minmax-fold.ll -Transforms/InstCombine/minmax-fp.ll -Transforms/InstCombine/minmax-intrinsics.ll -Transforms/InstCombine/mul-inseltpoison.ll -Transforms/InstCombine/mul.ll -Transforms/InstCombine/mul-masked-bits.ll -Transforms/InstCombine/mul-pow2.ll -Transforms/InstCombine/multiple-uses-load-bitcast-select.ll -Transforms/InstCombine/narrow.ll -Transforms/InstCombine/negated-bitmask.ll -Transforms/InstCombine/nested-select.ll -Transforms/InstCombine/not.ll -Transforms/InstCombine/or-bitmask.ll -Transforms/InstCombine/or-fcmp.ll -Transforms/InstCombine/or.ll -Transforms/InstCombine/pow-1.ll -Transforms/InstCombine/pow-3.ll -Transforms/InstCombine/pow-sqrt.ll -Transforms/InstCombine/pull-conditional-binop-through-shift.ll -Transforms/InstCombine/rem.ll -Transforms/InstCombine/sdiv-canonicalize.ll -Transforms/InstCombine/sdiv-guard.ll -Transforms/InstCombine/select-and-or.ll -Transforms/InstCombine/select-cmp-br.ll -Transforms/InstCombine/select-cmp.ll -Transforms/InstCombine/select-factorize.ll -Transforms/InstCombine/select_frexp.ll -Transforms/InstCombine/select.ll -Transforms/InstCombine/select-min-max.ll -Transforms/InstCombine/select-of-symmetric-selects.ll -Transforms/InstCombine/select-select.ll -Transforms/InstCombine/shift.ll -Transforms/InstCombine/shuffle-select-narrow-inseltpoison.ll -Transforms/InstCombine/shuffle-select-narrow.ll -Transforms/InstCombine/simplify-demanded-fpclass.ll -Transforms/InstCombine/sink-not-into-another-hand-of-logical-and.ll -Transforms/InstCombine/sink-not-into-another-hand-of-logical-or.ll -Transforms/InstCombine/sink-not-into-logical-and.ll -Transforms/InstCombine/sink-not-into-logical-or.ll -Transforms/InstCombine/strchr-1.ll -Transforms/InstCombine/strchr-3.ll -Transforms/InstCombine/strlen-1.ll -Transforms/InstCombine/strrchr-3.ll -Transforms/InstCombine/sub-ashr-and-to-icmp-select.ll -Transforms/InstCombine/sub-ashr-or-to-icmp-select.ll -Transforms/InstCombine/sub-xor-cmp.ll -Transforms/InstCombine/truncating-saturate.ll -Transforms/InstCombine/unordered-fcmp-select.ll -Transforms/InstCombine/urem-via-cmp-select.ll -Transforms/InstCombine/vec_sext.ll -Transforms/InstCombine/vector-urem.ll -Transforms/InstCombine/wcslen-1.ll -Transforms/InstCombine/wcslen-3.ll -Transforms/InstCombine/X86/blend_x86.ll -Transforms/InstCombine/X86/x86-avx512-inseltpoison.ll -Transforms/InstCombine/X86/x86-avx512.ll -Transforms/InstCombine/xor-and-or.ll -Transforms/InstComb... [truncated] |
mtrofin force-pushed the users/mtrofin/11-13-_profcheck_disable_verification_of_selects_on_vector_conditions branch from 167be33 to e781a5a Compare 
boomanaiden154 approved these changes Nov 14, 2025
Contributor
boomanaiden154 left a comment
boomanaiden154 left a comment There was a problem hiding this comment.
LGTM, some minor nits.
| setBranchWeights(I, {SelectTrueWeight, SelectFalseWeight}, | ||
| /*IsExpected=*/false); | ||
| if (auto *SI = dyn_cast<SelectInst>(&I)) | ||
| if (!SI->getCondition()->getType()->isVectorTy() && |
Contributor
There was a problem hiding this comment.
Probably more readable if we have a if (!SI->...->isVectorTy() { continue; }.
Member Author
There was a problem hiding this comment.
would need more instructions, what am I missing.
| "Profile verification failed: select annotation missing"); | ||
| if (auto *SI = dyn_cast<SelectInst>(&I)) | ||
| if (!SI->getCondition()->getType()->isVectorTy() && | ||
| !I.getMetadata(LLVMContext::MD_prof)) |
| | ||
| ; verify we can disable it. It's sufficient to see opt not failing. | ||
| ; RUN: opt -passes=prof-verify -profcheck-annotate-select=0 %t/verify-missing.ll | ||
| ; RUN: opt -passes=prof-verify -profcheck-annotate-select=0 --disable-output %t/verify-missing.ll |
Contributor
There was a problem hiding this comment.
Why are we disabling output? It should just go nowhere.
Member Author
There was a problem hiding this comment.
when you run the lit test at the command line, it gets printed out.
Member Author
mtrofin deleted the users/mtrofin/11-13-_profcheck_disable_verification_of_selects_on_vector_conditions branch 
Collaborator
| LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/157/builds/42200 Here is the relevant piece of the build log for the reference |
boomanaiden154 added a commit to boomanaiden154/llvm-project that referenced this pull request Nov 14, 2025
I think this is quite a bit more readable than the nested conditionals. From review feedback that was not addressed precommitn in llvm#167973.
boomanaiden154 added a commit that referenced this pull request Nov 14, 2025
I think this is quite a bit more readable than the nested conditionals. From review feedback that was not addressed precommitn in #167973.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Add this suggestion to a batch that can be applied as a single commit. This suggestion is invalid because no changes were made to the code. Suggestions cannot be applied while the pull request is closed. Suggestions cannot be applied while viewing a subset of changes. Only one suggestion per line can be applied in a batch. Add this suggestion to a batch that can be applied as a single commit. Applying suggestions on deleted lines is not supported. You must change the existing code in this line in order to create a valid suggestion. Outdated suggestions cannot be applied. This suggestion has been applied or marked resolved. Suggestions cannot be applied from pending reviews. Suggestions cannot be applied on multi-line comments. Suggestions cannot be applied while the pull request is queued to merge. Suggestion cannot be applied right now. Please check back later.
We don't currently support profile metadata on selects where the condition is a vector.
Issue #147390