Skip to content

Commit 032c1c0

Browse files
fhahnfhahn
committed
!fixup address comments, thanks
1 parent 85d4624 commit 032c1c0

File tree

3 files changed

+48
-40
lines changed

3 files changed

+48
-40
lines changed

‎llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -3110,8 +3110,9 @@ bool VPReplicateRecipe::shouldPack() const {
31103110
});
31113111
}
31123112

3113-
/// Returns true if \p Ptr is a pointer computation for which the legacy cost
3114-
/// model computes a SCEV expression when computing the address cost.
3113+
/// Returns a SCEV expression for \p Ptr if it is a pointer computation for
3114+
/// which the legacy cost model computes a SCEV expression when computing the
3115+
/// address cost. Otherwise return nullptr.
31153116
static const SCEV *getAddressAccessSCEV(const VPValue *Ptr, ScalarEvolution &SE,
31163117
const Loop *L) {
31173118
auto *PtrR = Ptr->getDefiningRecipe();
@@ -3296,7 +3297,7 @@ InstructionCost VPReplicateRecipe::computeCost(ElementCount VF,
32963297
Type *PtrTy = isSingleScalar() ? ScalarPtrTy : toVectorTy(ScalarPtrTy, VF);
32973298

32983299
const SCEV *PtrSCEV = getAddressAccessSCEV(PtrOp, Ctx.SE, Ctx.L);
3299-
if (PtrSCEV && isa<SCEVCouldNotCompute>(PtrSCEV))
3300+
if (isa_and_nonnull<SCEVCouldNotCompute>(PtrSCEV))
33003301
break;
33013302
InstructionCost ScalarCost =
33023303
ScalarMemOpCost + Ctx.TTI.getAddressComputationCost(

‎llvm/lib/Transforms/Vectorize/VPlanUtils.cpp‎

Lines changed: 1 addition & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -98,8 +98,7 @@ const SCEV *vputils::getSCEVExprForVPValue(const VPValue *V,
9898
const SCEV *Start = getSCEVExprForVPValue(R->getOperand(0), SE, L);
9999
const SCEV *IV = getSCEVExprForVPValue(R->getOperand(1), SE, L);
100100
const SCEV *Scale = getSCEVExprForVPValue(R->getOperand(2), SE, L);
101-
if (isa<SCEVCouldNotCompute>(Start) || isa<SCEVCouldNotCompute>(IV) ||
102-
isa<SCEVCouldNotCompute>(Scale))
101+
if (any_of(ArrayRef({Start, IV, Scale}), IsaPred<SCEVCouldNotCompute>))
103102
return SE.getCouldNotCompute();
104103

105104
return SE.getAddExpr(SE.getTruncateOrSignExtend(Start, IV->getType()),
@@ -167,7 +166,6 @@ const SCEV *vputils::getSCEVExprForVPValue(const VPValue *V,
167166
// non-negative, we can use nuw.
168167
return SE.getAddExpr(Base, Offset);
169168
})
170-
171169
.Default([&SE](const VPRecipeBase *) { return SE.getCouldNotCompute(); });
172170
}
173171

‎llvm/test/Transforms/LoopVectorize/AArch64/replicating-load-store-costs.ll‎

Lines changed: 43 additions & 34 deletions
Original file line numberDiff line numberDiff line change
@@ -463,21 +463,21 @@ define void @test_prefer_vector_addressing(ptr %start, ptr %ms, ptr noalias %src
463463
; CHECK-NEXT: [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP11]]
464464
; CHECK-NEXT: [[NEXT_GEP4:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP12]]
465465
; CHECK-NEXT: [[NEXT_GEP5:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP13]]
466-
; CHECK-NEXT: [[TMP14:%.*]] = load i64, ptr [[NEXT_GEP]], align 1, !tbaa [[LONG_LONG_TBAA12:![0-9]+]]
467-
; CHECK-NEXT: [[TMP15:%.*]] = load i64, ptr [[NEXT_GEP3]], align 1, !tbaa [[LONG_LONG_TBAA12]]
468-
; CHECK-NEXT: [[TMP16:%.*]] = load i64, ptr [[NEXT_GEP4]], align 1, !tbaa [[LONG_LONG_TBAA12]]
469-
; CHECK-NEXT: [[TMP17:%.*]] = load i64, ptr [[NEXT_GEP5]], align 1, !tbaa [[LONG_LONG_TBAA12]]
466+
; CHECK-NEXT: [[TMP14:%.*]] = load i64, ptr [[NEXT_GEP]], align 1, !tbaa [[LONG_LONG_TBAA14:![0-9]+]]
467+
; CHECK-NEXT: [[TMP15:%.*]] = load i64, ptr [[NEXT_GEP3]], align 1, !tbaa [[LONG_LONG_TBAA14]]
468+
; CHECK-NEXT: [[TMP16:%.*]] = load i64, ptr [[NEXT_GEP4]], align 1, !tbaa [[LONG_LONG_TBAA14]]
469+
; CHECK-NEXT: [[TMP17:%.*]] = load i64, ptr [[NEXT_GEP5]], align 1, !tbaa [[LONG_LONG_TBAA14]]
470470
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP14]]
471471
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP15]]
472472
; CHECK-NEXT: [[TMP20:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP16]]
473473
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP17]]
474-
; CHECK-NEXT: store i32 0, ptr [[TMP18]], align 4, !tbaa [[INT_TBAA17:![0-9]+]]
475-
; CHECK-NEXT: store i32 0, ptr [[TMP19]], align 4, !tbaa [[INT_TBAA17]]
476-
; CHECK-NEXT: store i32 0, ptr [[TMP20]], align 4, !tbaa [[INT_TBAA17]]
477-
; CHECK-NEXT: store i32 0, ptr [[TMP21]], align 4, !tbaa [[INT_TBAA17]]
474+
; CHECK-NEXT: store i32 0, ptr [[TMP18]], align 4, !tbaa [[INT_TBAA19:![0-9]+]]
475+
; CHECK-NEXT: store i32 0, ptr [[TMP19]], align 4, !tbaa [[INT_TBAA19]]
476+
; CHECK-NEXT: store i32 0, ptr [[TMP20]], align 4, !tbaa [[INT_TBAA19]]
477+
; CHECK-NEXT: store i32 0, ptr [[TMP21]], align 4, !tbaa [[INT_TBAA19]]
478478
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
479479
; CHECK-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
480-
; CHECK-NEXT: br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
480+
; CHECK-NEXT: br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
481481
; CHECK: [[MIDDLE_BLOCK]]:
482482
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP6]], [[N_VEC]]
483483
; CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
@@ -543,36 +543,45 @@ exit:
543543
define double @test_scalarization_cost_for_load_of_address(ptr %src.0, ptr %src.1, ptr %src.2) {
544544
; CHECK-LABEL: define double @test_scalarization_cost_for_load_of_address(
545545
; CHECK-SAME: ptr [[SRC_0:%.*]], ptr [[SRC_1:%.*]], ptr [[SRC_2:%.*]]) {
546-
; CHECK-NEXT: [[ENTRY:.*]]:
547-
; CHECK-NEXT: br label %[[LOOP:.*]]
548-
; CHECK: [[LOOP]]:
549-
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
550-
; CHECK-NEXT: [[RED:%.*]] = phi double [ 3.000000e+00, %[[ENTRY]] ], [ [[RED_NEXT:%.*]], %[[LOOP]] ]
546+
; CHECK-NEXT: [[ENTRY:.*:]]
547+
; CHECK-NEXT: br label %[[VECTOR_PH:.*]]
548+
; CHECK: [[VECTOR_PH]]:
549+
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
550+
; CHECK: [[VECTOR_BODY]]:
551+
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
552+
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi double [ 3.000000e+00, %[[VECTOR_PH]] ], [ [[TMP21:%.*]], %[[VECTOR_BODY]] ]
553+
; CHECK-NEXT: [[IV:%.*]] = add i64 [[INDEX]], 0
554+
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
551555
; CHECK-NEXT: [[GEP_0:%.*]] = getelementptr [[T:%.*]], ptr [[SRC_0]], i64 [[IV]]
552-
; CHECK-NEXT: [[L_0:%.*]] = load double, ptr [[GEP_0]], align 8
553-
; CHECK-NEXT: [[GEP_8:%.*]] = getelementptr i8, ptr [[GEP_0]], i64 8
554-
; CHECK-NEXT: [[L_1:%.*]] = load double, ptr [[GEP_8]], align 8
555-
; CHECK-NEXT: [[GEP_16:%.*]] = getelementptr i8, ptr [[GEP_0]], i64 16
556-
; CHECK-NEXT: [[L_2:%.*]] = load double, ptr [[GEP_16]], align 8
557-
; CHECK-NEXT: [[MUL_0:%.*]] = fmul double [[L_0]], 3.000000e+00
558-
; CHECK-NEXT: [[MUL_1:%.*]] = fmul double [[L_1]], 3.000000e+00
559-
; CHECK-NEXT: [[MUL_2:%.*]] = fmul double [[L_2]], 3.000000e+00
560-
; CHECK-NEXT: [[ADD_0:%.*]] = fadd double [[MUL_0]], [[MUL_1]]
561-
; CHECK-NEXT: [[ADD_1:%.*]] = fadd double [[ADD_0]], [[MUL_2]]
556+
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <6 x double>, ptr [[GEP_0]], align 8
557+
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <6 x double> [[WIDE_VEC]], <6 x double> poison, <2 x i32> <i32 0, i32 3>
558+
; CHECK-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <6 x double> [[WIDE_VEC]], <6 x double> poison, <2 x i32> <i32 1, i32 4>
559+
; CHECK-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <6 x double> [[WIDE_VEC]], <6 x double> poison, <2 x i32> <i32 2, i32 5>
560+
; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x double> [[STRIDED_VEC]], splat (double 3.000000e+00)
561+
; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> [[STRIDED_VEC1]], splat (double 3.000000e+00)
562+
; CHECK-NEXT: [[TMP5:%.*]] = fmul <2 x double> [[STRIDED_VEC2]], splat (double 3.000000e+00)
563+
; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x double> [[TMP3]], [[TMP4]]
564+
; CHECK-NEXT: [[TMP7:%.*]] = fadd <2 x double> [[TMP6]], [[TMP5]]
562565
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr double, ptr [[SRC_1]], i64 [[IV]]
563-
; CHECK-NEXT: [[L:%.*]] = load double, ptr [[GEP_SRC]], align 8
564-
; CHECK-NEXT: [[MUL256_US:%.*]] = fmul double [[ADD_1]], [[L]]
565-
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr [[T_2:%.*]], ptr [[SRC_2]], i64 [[IV]]
566+
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[GEP_SRC]], align 8
567+
; CHECK-NEXT: [[TMP9:%.*]] = fmul <2 x double> [[TMP7]], [[WIDE_LOAD]]
568+
; CHECK-NEXT: [[GEP_SRC_2:%.*]] = getelementptr [[T_2:%.*]], ptr [[SRC_2]], i64 [[IV]]
569+
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr [[T_2]], ptr [[SRC_2]], i64 [[TMP1]]
570+
; CHECK-NEXT: [[GEP_72:%.*]] = getelementptr i8, ptr [[GEP_SRC_2]], i64 72
566571
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP11]], i64 72
572+
; CHECK-NEXT: [[L_P_2:%.*]] = load ptr, ptr [[GEP_72]], align 8
567573
; CHECK-NEXT: [[TMP15:%.*]] = load ptr, ptr [[TMP13]], align 8
574+
; CHECK-NEXT: [[LV:%.*]] = load double, ptr [[L_P_2]], align 8
568575
; CHECK-NEXT: [[TMP17:%.*]] = load double, ptr [[TMP15]], align 8
569-
; CHECK-NEXT: [[RED_NEXT]] = tail call double @llvm.fmuladd.f64(double [[MUL256_US]], double [[TMP17]], double [[RED]])
570-
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
571-
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV]], 1
572-
; CHECK-NEXT: br i1 [[EC]], label %[[SCALAR_PH:.*]], label %[[LOOP]]
573-
; CHECK: [[SCALAR_PH]]:
574-
; CHECK-NEXT: [[RED_NEXT_LCSSA:%.*]] = phi double [ [[RED_NEXT]], %[[LOOP]] ]
575-
; CHECK-NEXT: ret double [[RED_NEXT_LCSSA]]
576+
; CHECK-NEXT: [[TMP18:%.*]] = insertelement <2 x double> poison, double [[LV]], i32 0
577+
; CHECK-NEXT: [[TMP19:%.*]] = insertelement <2 x double> [[TMP18]], double [[TMP17]], i32 1
578+
; CHECK-NEXT: [[TMP20:%.*]] = fmul <2 x double> [[TMP9]], [[TMP19]]
579+
; CHECK-NEXT: [[TMP21]] = call double @llvm.vector.reduce.fadd.v2f64(double [[VEC_PHI]], <2 x double> [[TMP20]])
580+
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
581+
; CHECK-NEXT: br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
582+
; CHECK: [[MIDDLE_BLOCK]]:
583+
; CHECK-NEXT: br [[EXIT:label %.*]]
584+
; CHECK: [[SCALAR_PH:.*:]]
576585
;
577586
entry:
578587
br label %loop

0 commit comments

Comments
 (0)