[llvm] [AArch64] Runtime-unroll small load/store loops for Apple Silicon CPUs. (PR #118317)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Dec 2 08:10:13 PST 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-aarch64
Author: Florian Hahn (fhahn)
<details>
<summary>Changes</summary>
Add initial heuristics to selectively enable runtime unrolling for loops
where doing so is expected to be highly beneficial on Apple Silicon
CPUs.
To start with, we try to runtime-unroll small, single block loops, if they
have load/store dependencies, to expose more parallel memory access streams [1]
and to improve instruction delivery [2].
We also explicitly avoid runtime-unrolling for loop structures that may
limit the expected gains from runtime unrolling. Such loops include
loops with complex control flow (aren't innermost loops, have multiple
exits, have a large number of blocks), trip count expansion is
expensive and are expected to execute a small number of iterations.
Note that the heuristics here may be overly conservative and we err on the
side of avoiding runtime unrolling rather than unroll excessively. They are
all subject to further refinement.
[1] 4.6.10 in Apple Silicon CPU Optimization Guide
[2] 4.4.4 in Apple Silicon CPU Optimization Guide
(Depends on https://github.com/llvm/llvm-project/pull/118316 for TTI
changes, which are included in this PR for now)
---
Full diff: https://github.com/llvm/llvm-project/pull/118317.diff
9 Files Affected:
- (modified) llvm/include/llvm/Analysis/TargetTransformInfo.h (+3)
- (modified) llvm/include/llvm/Transforms/Utils/UnrollLoop.h (+2-1)
- (modified) llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp (+90)
- (modified) llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp (+3)
- (modified) llvm/lib/Transforms/Utils/LoopUnroll.cpp (+5-4)
- (modified) llvm/lib/Transforms/Utils/LoopUnrollAndJam.cpp (+3-1)
- (modified) llvm/lib/Transforms/Utils/LoopUnrollRuntime.cpp (+4-3)
- (modified) llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling.ll (+78-4)
- (modified) llvm/unittests/Transforms/Utils/UnrollLoopTest.cpp (+1-1)
``````````diff
diff --git a/llvm/include/llvm/Analysis/TargetTransformInfo.h b/llvm/include/llvm/Analysis/TargetTransformInfo.h
index 89231e23e388a7..aa530b54c5c6d3 100644
--- a/llvm/include/llvm/Analysis/TargetTransformInfo.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfo.h
@@ -615,6 +615,9 @@ class TargetTransformInfo {
unsigned MaxIterationsCountToAnalyze;
/// Don't disable runtime unroll for the loops which were vectorized.
bool UnrollVectorizedLoop = false;
+ /// Don't allow runtime unrolling if expanding the trip count takes more
+ /// than SCEVExpansionBudget.
+ unsigned SCEVExpansionBudget;
};
/// Get target-customized preferences for the generic loop unrolling
diff --git a/llvm/include/llvm/Transforms/Utils/UnrollLoop.h b/llvm/include/llvm/Transforms/Utils/UnrollLoop.h
index 797c082333a76c..8cf17ced458c82 100644
--- a/llvm/include/llvm/Transforms/Utils/UnrollLoop.h
+++ b/llvm/include/llvm/Transforms/Utils/UnrollLoop.h
@@ -75,6 +75,7 @@ struct UnrollLoopOptions {
bool UnrollRemainder;
bool ForgetAllSCEV;
const Instruction *Heart = nullptr;
+ unsigned SCEVExpansionBudget;
};
LoopUnrollResult UnrollLoop(Loop *L, UnrollLoopOptions ULO, LoopInfo *LI,
@@ -90,7 +91,7 @@ bool UnrollRuntimeLoopRemainder(
bool UseEpilogRemainder, bool UnrollRemainder, bool ForgetAllSCEV,
LoopInfo *LI, ScalarEvolution *SE, DominatorTree *DT, AssumptionCache *AC,
const TargetTransformInfo *TTI, bool PreserveLCSSA,
- Loop **ResultLoop = nullptr);
+ unsigned SCEVExpansionBudget, Loop **ResultLoop = nullptr);
LoopUnrollResult UnrollAndJamLoop(Loop *L, unsigned Count, unsigned TripCount,
unsigned TripMultiple, bool UnrollRemainder,
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 5b333d33cffd52..035854b8a40293 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -3989,6 +3989,90 @@ getFalkorUnrollingPreferences(Loop *L, ScalarEvolution &SE,
}
}
+/// For Apple CPUs, we want to runtime-unroll loops to make better use if the
+/// OOO engine's wide instruction window and various predictors.
+static void
+getAppleRuntimeUnrollPreferences(Loop *L, ScalarEvolution &SE,
+ TargetTransformInfo::UnrollingPreferences &UP,
+ AArch64TTIImpl &TTI) {
+ // Limit loops with structure that is highly likely to benefit from runtime
+ // unrolling; that is we exclude outer loops, loops with multiple exits and
+ // many blocks (i.e. likely with complex control flow). Note that the
+ // heuristics here may be overly conservative and we err on the side of
+ // avoiding runtime unrolling rather than unroll excessively. They are all
+ // subject to further refinement.
+ if (!L->isInnermost() || !L->getExitBlock() || L->getNumBlocks() > 8)
+ return;
+
+ const SCEV *BTC = SE.getBackedgeTakenCount(L);
+ if (isa<SCEVConstant>(BTC) || isa<SCEVCouldNotCompute>(BTC) ||
+ (SE.getSmallConstantMaxTripCount(L) > 0 &&
+ SE.getSmallConstantMaxTripCount(L) <= 32))
+ return;
+ if (findStringMetadataForLoop(L, "llvm.loop.isvectorized"))
+ return;
+
+ int64_t Size = 0;
+ for (auto *BB : L->getBlocks()) {
+ for (auto &I : *BB) {
+ if (!isa<IntrinsicInst>(&I) && isa<CallBase>(&I))
+ return;
+ SmallVector<const Value *, 4> Operands(I.operand_values());
+ Size +=
+ *TTI.getInstructionCost(&I, Operands, TTI::TCK_CodeSize).getValue();
+ }
+ }
+
+ // Limit to loops with trip counts that are cheap to expand.
+ UP.SCEVExpansionBudget = 1;
+
+ // Try to unroll small, single block loops, if they have load/store
+ // dependencies, to expose more parallel memory access streams.
+ if (L->getHeader() != L->getLoopLatch() || Size > 8)
+ return;
+
+ SmallPtrSet<const SCEV *, 8> LoadPtrs;
+ SmallPtrSet<const SCEV *, 8> StorePtrs;
+ SmallPtrSet<Value *, 8> LoadedValues;
+ SmallVector<StoreInst *> Stores;
+ for (auto *BB : L->blocks()) {
+ for (auto &I : *BB) {
+ Value *Ptr = getLoadStorePointerOperand(&I);
+ if (!Ptr)
+ continue;
+ const SCEV *PtrSCEV = SE.getSCEV(Ptr);
+ if (SE.isLoopInvariant(PtrSCEV, L))
+ continue;
+ if (isa<LoadInst>(&I)) {
+ LoadPtrs.insert(PtrSCEV);
+ LoadedValues.insert(&I);
+ } else {
+ Stores.push_back(cast<StoreInst>(&I));
+ StorePtrs.insert(PtrSCEV);
+ }
+ }
+ }
+
+ // Try to find an unroll count that maximizes the use of the instruction
+ // window.
+ unsigned UC = std::max(16ll / Size, 2ll);
+ unsigned BestUC = 0;
+ while (UC <= 8 && UC * Size <= 48) {
+ if ((UC * Size % 16) == 0 || (BestUC * Size % 16) < (UC * Size % 16) % 16) {
+ BestUC = UC;
+ }
+ UC++;
+ }
+
+ if (BestUC == 0 || none_of(Stores, [&LoadedValues](StoreInst *SI) {
+ return LoadedValues.contains(SI->getOperand(0));
+ }))
+ return;
+
+ UP.Runtime = true;
+ UP.DefaultUnrollRuntimeCount = BestUC;
+}
+
void AArch64TTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
TTI::UnrollingPreferences &UP,
OptimizationRemarkEmitter *ORE) {
@@ -4010,6 +4094,12 @@ void AArch64TTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
EnableFalkorHWPFUnrollFix)
getFalkorUnrollingPreferences(L, SE, UP);
+ if (ST->getProcFamily() == AArch64Subtarget::AppleA14 ||
+ ST->getProcFamily() == AArch64Subtarget::AppleA15 ||
+ ST->getProcFamily() == AArch64Subtarget::AppleA16 ||
+ ST->getProcFamily() == AArch64Subtarget::AppleM4)
+ getAppleRuntimeUnrollPreferences(L, SE, UP, *this);
+
// Scan the loop: don't unroll loops with calls as this could prevent
// inlining. Don't unroll vector loops either, as they don't benefit much from
// unrolling.
diff --git a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
index 953748f07af028..260cc72c3188d2 100644
--- a/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopUnrollPass.cpp
@@ -57,6 +57,7 @@
#include "llvm/Transforms/Utils/LoopPeel.h"
#include "llvm/Transforms/Utils/LoopSimplify.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
+#include "llvm/Transforms/Utils/ScalarEvolutionExpander.h"
#include "llvm/Transforms/Utils/SizeOpts.h"
#include "llvm/Transforms/Utils/UnrollLoop.h"
#include <algorithm>
@@ -218,6 +219,7 @@ TargetTransformInfo::UnrollingPreferences llvm::gatherUnrollingPreferences(
UP.UnrollAndJam = false;
UP.UnrollAndJamInnerLoopThreshold = 60;
UP.MaxIterationsCountToAnalyze = UnrollMaxIterationsCountToAnalyze;
+ UP.SCEVExpansionBudget = SCEVCheapExpansionBudget;
// Override with any target specific settings
TTI.getUnrollingPreferences(L, SE, UP, &ORE);
@@ -1349,6 +1351,7 @@ tryToUnrollLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, ScalarEvolution &SE,
ULO.Runtime = UP.Runtime;
ULO.ForgetAllSCEV = ForgetAllSCEV;
ULO.Heart = getLoopConvergenceHeart(L);
+ ULO.SCEVExpansionBudget = UP.SCEVExpansionBudget;
LoopUnrollResult UnrollResult = UnrollLoop(
L, ULO, LI, &SE, &DT, &AC, &TTI, &ORE, PreserveLCSSA, &RemainderLoop, AA);
if (UnrollResult == LoopUnrollResult::Unmodified)
diff --git a/llvm/lib/Transforms/Utils/LoopUnroll.cpp b/llvm/lib/Transforms/Utils/LoopUnroll.cpp
index 85d4415970a00a..b11d92836a998f 100644
--- a/llvm/lib/Transforms/Utils/LoopUnroll.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUnroll.cpp
@@ -60,6 +60,7 @@
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Transforms/Utils/LoopSimplify.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
+#include "llvm/Transforms/Utils/ScalarEvolutionExpander.h"
#include "llvm/Transforms/Utils/SimplifyIndVar.h"
#include "llvm/Transforms/Utils/UnrollLoop.h"
#include "llvm/Transforms/Utils/ValueMapper.h"
@@ -589,10 +590,10 @@ llvm::UnrollLoop(Loop *L, UnrollLoopOptions ULO, LoopInfo *LI,
: isEpilogProfitable(L);
if (ULO.Runtime &&
- !UnrollRuntimeLoopRemainder(L, ULO.Count, ULO.AllowExpensiveTripCount,
- EpilogProfitability, ULO.UnrollRemainder,
- ULO.ForgetAllSCEV, LI, SE, DT, AC, TTI,
- PreserveLCSSA, RemainderLoop)) {
+ !UnrollRuntimeLoopRemainder(
+ L, ULO.Count, ULO.AllowExpensiveTripCount, EpilogProfitability,
+ ULO.UnrollRemainder, ULO.ForgetAllSCEV, LI, SE, DT, AC, TTI,
+ PreserveLCSSA, ULO.SCEVExpansionBudget, RemainderLoop)) {
if (ULO.Force)
ULO.Runtime = false;
else {
diff --git a/llvm/lib/Transforms/Utils/LoopUnrollAndJam.cpp b/llvm/lib/Transforms/Utils/LoopUnrollAndJam.cpp
index 2c2400d9dd7a82..49209e33f2d1dd 100644
--- a/llvm/lib/Transforms/Utils/LoopUnrollAndJam.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUnrollAndJam.cpp
@@ -48,6 +48,7 @@
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Cloning.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
+#include "llvm/Transforms/Utils/ScalarEvolutionExpander.h"
#include "llvm/Transforms/Utils/UnrollLoop.h"
#include "llvm/Transforms/Utils/ValueMapper.h"
#include <assert.h>
@@ -241,7 +242,8 @@ llvm::UnrollAndJamLoop(Loop *L, unsigned Count, unsigned TripCount,
if (!UnrollRuntimeLoopRemainder(L, Count, /*AllowExpensiveTripCount*/ false,
/*UseEpilogRemainder*/ true,
UnrollRemainder, /*ForgetAllSCEV*/ false,
- LI, SE, DT, AC, TTI, true, EpilogueLoop)) {
+ LI, SE, DT, AC, TTI, true,
+ SCEVCheapExpansionBudget, EpilogueLoop)) {
LLVM_DEBUG(dbgs() << "Won't unroll-and-jam; remainder loop could not be "
"generated when assuming runtime trip count\n");
return LoopUnrollResult::Unmodified;
diff --git a/llvm/lib/Transforms/Utils/LoopUnrollRuntime.cpp b/llvm/lib/Transforms/Utils/LoopUnrollRuntime.cpp
index 0179d350b9e0e7..b0bc55cd64c377 100644
--- a/llvm/lib/Transforms/Utils/LoopUnrollRuntime.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUnrollRuntime.cpp
@@ -582,7 +582,8 @@ bool llvm::UnrollRuntimeLoopRemainder(
Loop *L, unsigned Count, bool AllowExpensiveTripCount,
bool UseEpilogRemainder, bool UnrollRemainder, bool ForgetAllSCEV,
LoopInfo *LI, ScalarEvolution *SE, DominatorTree *DT, AssumptionCache *AC,
- const TargetTransformInfo *TTI, bool PreserveLCSSA, Loop **ResultLoop) {
+ const TargetTransformInfo *TTI, bool PreserveLCSSA,
+ unsigned SCEVExpansionBudget, Loop **ResultLoop) {
LLVM_DEBUG(dbgs() << "Trying runtime unrolling on Loop: \n");
LLVM_DEBUG(L->dump());
LLVM_DEBUG(UseEpilogRemainder ? dbgs() << "Using epilog remainder.\n"
@@ -672,8 +673,8 @@ bool llvm::UnrollRuntimeLoopRemainder(
const DataLayout &DL = Header->getDataLayout();
SCEVExpander Expander(*SE, DL, "loop-unroll");
if (!AllowExpensiveTripCount &&
- Expander.isHighCostExpansion(TripCountSC, L, SCEVCheapExpansionBudget,
- TTI, PreHeaderBR)) {
+ Expander.isHighCostExpansion(TripCountSC, L, SCEVExpansionBudget, TTI,
+ PreHeaderBR)) {
LLVM_DEBUG(dbgs() << "High cost for expanding trip count scev!\n");
return false;
}
diff --git a/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling.ll b/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling.ll
index deacec795fb03a..d27d5e74e28f2e 100644
--- a/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling.ll
+++ b/llvm/test/Transforms/LoopUnroll/AArch64/apple-unrolling.ll
@@ -12,17 +12,91 @@ define void @small_load_store_loop(ptr %src, ptr %dst, i64 %N, i64 %scale) {
; APPLE-LABEL: define void @small_load_store_loop(
; APPLE-SAME: ptr [[SRC:%.*]], ptr [[DST:%.*]], i64 [[N:%.*]], i64 [[SCALE:%.*]]) #[[ATTR0:[0-9]+]] {
; APPLE-NEXT: [[ENTRY:.*]]:
+; APPLE-NEXT: [[TMP0:%.*]] = add i64 [[N]], -1
+; APPLE-NEXT: [[XTRAITER:%.*]] = and i64 [[N]], 7
+; APPLE-NEXT: [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 7
+; APPLE-NEXT: br i1 [[TMP1]], label %[[EXIT_UNR_LCSSA:.*]], label %[[ENTRY_NEW:.*]]
+; APPLE: [[ENTRY_NEW]]:
+; APPLE-NEXT: [[UNROLL_ITER:%.*]] = sub i64 [[N]], [[XTRAITER]]
; APPLE-NEXT: br label %[[LOOP:.*]]
; APPLE: [[LOOP]]:
-; APPLE-NEXT: [[IV_EPIL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT_EPIL:%.*]], %[[LOOP]] ]
+; APPLE-NEXT: [[IV_EPIL:%.*]] = phi i64 [ 0, %[[ENTRY_NEW]] ], [ [[IV_NEXT_7:%.*]], %[[LOOP]] ]
+; APPLE-NEXT: [[NITER:%.*]] = phi i64 [ 0, %[[ENTRY_NEW]] ], [ [[NITER_NEXT_7:%.*]], %[[LOOP]] ]
; APPLE-NEXT: [[SCALED_IV_EPIL:%.*]] = mul nuw nsw i64 [[IV_EPIL]], [[SCALE]]
; APPLE-NEXT: [[GEP_SRC_EPIL:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_EPIL]]
; APPLE-NEXT: [[L_EPIL:%.*]] = load float, ptr [[GEP_SRC_EPIL]], align 4
; APPLE-NEXT: [[GEP_DST_EPIL:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_EPIL]]
; APPLE-NEXT: store float [[L_EPIL]], ptr [[GEP_DST_EPIL]], align 4
-; APPLE-NEXT: [[IV_NEXT_EPIL]] = add nuw nsw i64 [[IV_EPIL]], 1
-; APPLE-NEXT: [[EC_EPIL:%.*]] = icmp eq i64 [[IV_NEXT_EPIL]], [[N]]
-; APPLE-NEXT: br i1 [[EC_EPIL]], label %[[EXIT:.*]], label %[[LOOP]]
+; APPLE-NEXT: [[IV_NEXT_EPIL:%.*]] = add nuw nsw i64 [[IV_EPIL]], 1
+; APPLE-NEXT: [[SCALED_IV_1:%.*]] = mul nuw nsw i64 [[IV_NEXT_EPIL]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_1:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_1]]
+; APPLE-NEXT: [[L_1:%.*]] = load float, ptr [[GEP_SRC_1]], align 4
+; APPLE-NEXT: [[GEP_DST_1:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_EPIL]]
+; APPLE-NEXT: store float [[L_1]], ptr [[GEP_DST_1]], align 4
+; APPLE-NEXT: [[IV_NEXT_1:%.*]] = add nuw nsw i64 [[IV_EPIL]], 2
+; APPLE-NEXT: [[SCALED_IV_2:%.*]] = mul nuw nsw i64 [[IV_NEXT_1]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_2:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_2]]
+; APPLE-NEXT: [[L_2:%.*]] = load float, ptr [[GEP_SRC_2]], align 4
+; APPLE-NEXT: [[GEP_DST_2:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_1]]
+; APPLE-NEXT: store float [[L_2]], ptr [[GEP_DST_2]], align 4
+; APPLE-NEXT: [[IV_NEXT_2:%.*]] = add nuw nsw i64 [[IV_EPIL]], 3
+; APPLE-NEXT: [[SCALED_IV_3:%.*]] = mul nuw nsw i64 [[IV_NEXT_2]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_3:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_3]]
+; APPLE-NEXT: [[L_3:%.*]] = load float, ptr [[GEP_SRC_3]], align 4
+; APPLE-NEXT: [[GEP_DST_3:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_2]]
+; APPLE-NEXT: store float [[L_3]], ptr [[GEP_DST_3]], align 4
+; APPLE-NEXT: [[IV_NEXT_3:%.*]] = add nuw nsw i64 [[IV_EPIL]], 4
+; APPLE-NEXT: [[SCALED_IV_4:%.*]] = mul nuw nsw i64 [[IV_NEXT_3]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_4:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_4]]
+; APPLE-NEXT: [[L_4:%.*]] = load float, ptr [[GEP_SRC_4]], align 4
+; APPLE-NEXT: [[GEP_DST_4:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_3]]
+; APPLE-NEXT: store float [[L_4]], ptr [[GEP_DST_4]], align 4
+; APPLE-NEXT: [[IV_NEXT_4:%.*]] = add nuw nsw i64 [[IV_EPIL]], 5
+; APPLE-NEXT: [[SCALED_IV_5:%.*]] = mul nuw nsw i64 [[IV_NEXT_4]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_5]]
+; APPLE-NEXT: [[L_5:%.*]] = load float, ptr [[GEP_SRC_5]], align 4
+; APPLE-NEXT: [[GEP_DST_5:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_4]]
+; APPLE-NEXT: store float [[L_5]], ptr [[GEP_DST_5]], align 4
+; APPLE-NEXT: [[IV_NEXT_5:%.*]] = add nuw nsw i64 [[IV_EPIL]], 6
+; APPLE-NEXT: [[SCALED_IV_6:%.*]] = mul nuw nsw i64 [[IV_NEXT_5]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_6:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_6]]
+; APPLE-NEXT: [[L_6:%.*]] = load float, ptr [[GEP_SRC_6]], align 4
+; APPLE-NEXT: [[GEP_DST_6:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_5]]
+; APPLE-NEXT: store float [[L_6]], ptr [[GEP_DST_6]], align 4
+; APPLE-NEXT: [[IV_NEXT_6:%.*]] = add nuw nsw i64 [[IV_EPIL]], 7
+; APPLE-NEXT: [[SCALED_IV_7:%.*]] = mul nuw nsw i64 [[IV_NEXT_6]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_7:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_7]]
+; APPLE-NEXT: [[L_7:%.*]] = load float, ptr [[GEP_SRC_7]], align 4
+; APPLE-NEXT: [[GEP_DST_7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_NEXT_6]]
+; APPLE-NEXT: store float [[L_7]], ptr [[GEP_DST_7]], align 4
+; APPLE-NEXT: [[IV_NEXT_7]] = add nuw nsw i64 [[IV_EPIL]], 8
+; APPLE-NEXT: [[NITER_NEXT_7]] = add i64 [[NITER]], 8
+; APPLE-NEXT: [[NITER_NCMP_7:%.*]] = icmp eq i64 [[NITER_NEXT_7]], [[UNROLL_ITER]]
+; APPLE-NEXT: br i1 [[NITER_NCMP_7]], label %[[EXIT_UNR_LCSSA_LOOPEXIT:.*]], label %[[LOOP]]
+; APPLE: [[EXIT_UNR_LCSSA_LOOPEXIT]]:
+; APPLE-NEXT: [[IV_UNR_PH:%.*]] = phi i64 [ [[IV_NEXT_7]], %[[LOOP]] ]
+; APPLE-NEXT: br label %[[EXIT_UNR_LCSSA]]
+; APPLE: [[EXIT_UNR_LCSSA]]:
+; APPLE-NEXT: [[IV_UNR:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_UNR_PH]], %[[EXIT_UNR_LCSSA_LOOPEXIT]] ]
+; APPLE-NEXT: [[LCMP_MOD:%.*]] = icmp ne i64 [[XTRAITER]], 0
+; APPLE-NEXT: br i1 [[LCMP_MOD]], label %[[LOOP_EPIL_PREHEADER:.*]], label %[[EXIT:.*]]
+; APPLE: [[LOOP_EPIL_PREHEADER]]:
+; APPLE-NEXT: br label %[[LOOP_EPIL:.*]]
+; APPLE: [[LOOP_EPIL]]:
+; APPLE-NEXT: [[IV_EPIL1:%.*]] = phi i64 [ [[IV_UNR]], %[[LOOP_EPIL_PREHEADER]] ], [ [[IV_NEXT_EPIL1:%.*]], %[[LOOP_EPIL]] ]
+; APPLE-NEXT: [[EPIL_ITER:%.*]] = phi i64 [ 0, %[[LOOP_EPIL_PREHEADER]] ], [ [[EPIL_ITER_NEXT:%.*]], %[[LOOP_EPIL]] ]
+; APPLE-NEXT: [[SCALED_IV_EPIL1:%.*]] = mul nuw nsw i64 [[IV_EPIL1]], [[SCALE]]
+; APPLE-NEXT: [[GEP_SRC_EPIL1:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 [[SCALED_IV_EPIL1]]
+; APPLE-NEXT: [[L_EPIL1:%.*]] = load float, ptr [[GEP_SRC_EPIL1]], align 4
+; APPLE-NEXT: [[GEP_DST_EPIL1:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 [[IV_EPIL1]]
+; APPLE-NEXT: store float [[L_EPIL1]], ptr [[GEP_DST_EPIL1]], align 4
+; APPLE-NEXT: [[IV_NEXT_EPIL1]] = add nuw nsw i64 [[IV_EPIL1]], 1
+; APPLE-NEXT: [[EC_EPIL:%.*]] = icmp eq i64 [[IV_NEXT_EPIL1]], [[N]]
+; APPLE-NEXT: [[EPIL_ITER_NEXT]] = add i64 [[EPIL_ITER]], 1
+; APPLE-NEXT: [[EPIL_ITER_CMP:%.*]] = icmp ne i64 [[EPIL_ITER_NEXT]], [[XTRAITER]]
+; APPLE-NEXT: br i1 [[EPIL_ITER_CMP]], label %[[LOOP_EPIL]], label %[[EXIT_EPILOG_LCSSA:.*]], !llvm.loop [[LOOP0:![0-9]+]]
+; APPLE: [[EXIT_EPILOG_LCSSA]]:
+; APPLE-NEXT: br label %[[EXIT]]
; APPLE: [[EXIT]]:
; APPLE-NEXT: ret void
;
diff --git a/llvm/unittests/Transforms/Utils/UnrollLoopTest.cpp b/llvm/unittests/Transforms/Utils/UnrollLoopTest.cpp
index 241d55bd624576..b49e37d9eee984 100644
--- a/llvm/unittests/Transforms/Utils/UnrollLoopTest.cpp
+++ b/llvm/unittests/Transforms/Utils/UnrollLoopTest.cpp
@@ -73,6 +73,6 @@ while.end: ; preds = %while.cond
bool ret =
UnrollRuntimeLoopRemainder(L, 4, true, false, false, false, &LI, &SE, &DT,
- &AC, /*TTI=*/nullptr, PreserveLCSSA);
+ &AC, /*TTI=*/nullptr, PreserveLCSSA, 4);
EXPECT_FALSE(ret);
}
``````````
</details>
https://github.com/llvm/llvm-project/pull/118317
More information about the llvm-commits
mailing list