[llvm] [LV] Return Invalid from getLegacyCost when instruction cost forced. (PR #154543)
Kerry McLaughlin via llvm-commits
llvm-commits at lists.llvm.org
Tue Aug 26 01:53:01 PDT 2025
https://github.com/kmclaughlin-arm updated https://github.com/llvm/llvm-project/pull/154543
>From f1e98a60121587b5458fb3526856de85a653b9e5 Mon Sep 17 00:00:00 2001
From: Kerry McLaughlin <kerry.mclaughlin at arm.com>
Date: Wed, 20 Aug 2025 12:55:11 +0000
Subject: [PATCH 1/4] [LV] Return Invalid from getLegacyCost when instruction
cost forced.
LoopVectorizationCostModel::expectedCost will only override the cost
returned by getInstructionCost when valid. This patch ensures we do
the same in VPCostContext::getLegacyCost, avoiding the "VPlan cost
model and legacy cost model disagreed" assert in the included test.
---
.../Transforms/Vectorize/LoopVectorize.cpp | 5 ++--
.../AArch64/force-inst-cost-invalid.ll | 24 +++++++++++++++++++
2 files changed, 27 insertions(+), 2 deletions(-)
create mode 100644 llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 70f884016d08c..2d86b2b27f5f6 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -6760,9 +6760,10 @@ void LoopVectorizationPlanner::plan(ElementCount UserVF, unsigned UserIC) {
InstructionCost VPCostContext::getLegacyCost(Instruction *UI,
ElementCount VF) const {
- if (ForceTargetInstructionCost.getNumOccurrences())
+ InstructionCost Cost = CM.getInstructionCost(UI, VF);
+ if (Cost.isValid() && ForceTargetInstructionCost.getNumOccurrences())
return InstructionCost(ForceTargetInstructionCost.getNumOccurrences());
- return CM.getInstructionCost(UI, VF);
+ return Cost;
}
bool VPCostContext::isLegacyUniformAfterVectorization(Instruction *I,
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll b/llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll
new file mode 100644
index 0000000000000..52a72014392d7
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll
@@ -0,0 +1,24 @@
+; REQUIRES: asserts
+; RUN: opt < %s -passes=loop-vectorize -force-target-instruction-cost=1 -debug-only=loop-vectorize -S -disable-output 2>&1 | FileCheck %s
+target triple = "aarch64-linux-gnu"
+
+define i32 @invalid_legacy_cost(i64 %N) #0 {
+; CHECK: LV: Checking a loop in 'invalid_legacy_cost
+; CHECK: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %0 = alloca i8, i64 0, align 16
+entry:
+ br label %for.body
+
+for.body:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+ %0 = alloca i8, i64 0, align 16
+ %arrayidx = getelementptr ptr, ptr null, i64 %iv
+ store ptr %0, ptr %arrayidx, align 8
+ %iv.next = add i64 %iv, 1
+ %exitcond.not = icmp eq i64 %iv, %N
+ br i1 %exitcond.not, label %for.end, label %for.body
+
+for.end:
+ ret i32 0
+}
+
+attributes #0 = { "target-features"="+neon,+sve" vscale_range(1,16) }
>From e9ccb54b4c5d20331a62c1b393a9d36ff45cabef Mon Sep 17 00:00:00 2001
From: Kerry McLaughlin <kerry.mclaughlin at arm.com>
Date: Wed, 20 Aug 2025 14:18:49 +0000
Subject: [PATCH 2/4] - Replace getNumOccurrences() with
ForceTargetInstructionCost
---
llvm/lib/Transforms/Vectorize/LoopVectorize.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 2d86b2b27f5f6..1bfe57edd97f1 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -6762,7 +6762,7 @@ InstructionCost VPCostContext::getLegacyCost(Instruction *UI,
ElementCount VF) const {
InstructionCost Cost = CM.getInstructionCost(UI, VF);
if (Cost.isValid() && ForceTargetInstructionCost.getNumOccurrences())
- return InstructionCost(ForceTargetInstructionCost.getNumOccurrences());
+ return InstructionCost(ForceTargetInstructionCost);
return Cost;
}
>From 4b4d4a5724f1462ea3e34e72b4b55398975d7404 Mon Sep 17 00:00:00 2001
From: Kerry McLaughlin <kerry.mclaughlin at arm.com>
Date: Wed, 20 Aug 2025 16:12:25 +0000
Subject: [PATCH 3/4] - Moved new test into force-target-instruction-cost.ll -
Set -force-target-instruction-cost=100
---
.../AArch64/force-inst-cost-invalid.ll | 24 -------
.../AArch64/force-target-instruction-cost.ll | 69 +++++++++++++++++--
2 files changed, 63 insertions(+), 30 deletions(-)
delete mode 100644 llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll b/llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll
deleted file mode 100644
index 52a72014392d7..0000000000000
--- a/llvm/test/Transforms/LoopVectorize/AArch64/force-inst-cost-invalid.ll
+++ /dev/null
@@ -1,24 +0,0 @@
-; REQUIRES: asserts
-; RUN: opt < %s -passes=loop-vectorize -force-target-instruction-cost=1 -debug-only=loop-vectorize -S -disable-output 2>&1 | FileCheck %s
-target triple = "aarch64-linux-gnu"
-
-define i32 @invalid_legacy_cost(i64 %N) #0 {
-; CHECK: LV: Checking a loop in 'invalid_legacy_cost
-; CHECK: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %0 = alloca i8, i64 0, align 16
-entry:
- br label %for.body
-
-for.body:
- %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
- %0 = alloca i8, i64 0, align 16
- %arrayidx = getelementptr ptr, ptr null, i64 %iv
- store ptr %0, ptr %arrayidx, align 8
- %iv.next = add i64 %iv, 1
- %exitcond.not = icmp eq i64 %iv, %N
- br i1 %exitcond.not, label %for.end, label %for.body
-
-for.end:
- ret i32 0
-}
-
-attributes #0 = { "target-features"="+neon,+sve" vscale_range(1,16) }
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
index ad184bec2ac7b..34c0b58bec230 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -p loop-vectorize -force-target-instruction-cost=1 -S %s | FileCheck %s
+; RUN: opt -p loop-vectorize -force-target-instruction-cost=100 -S %s | FileCheck %s
target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128-Fn32"
target triple = "arm64-apple-macosx14.0.0"
@@ -65,19 +65,17 @@ define void @test_iv_cost(ptr %ptr.start, i8 %a, i64 %b) {
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[START]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]]
; CHECK: [[VECTOR_MAIN_LOOP_ITER_CHECK]]:
-; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[START]], 32
+; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[START]], 16
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
-; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[START]], 32
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[START]], 16
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[START]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[PTR_START]], i64 [[INDEX]]
-; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 16
; CHECK-NEXT: store <16 x i8> zeroinitializer, ptr [[NEXT_GEP1]], align 1
-; CHECK-NEXT: store <16 x i8> zeroinitializer, ptr [[TMP3]], align 1
-; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
@@ -366,6 +364,63 @@ exit:
ret void
}
+define void @invalid_legacy_cost(i64 %N, ptr %x) #0 {
+; CHECK-LABEL: define void @invalid_legacy_cost(
+; CHECK-SAME: i64 [[N:%.*]], ptr [[X:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: [[ENTRY:.*]]:
+; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N]], 1
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 2
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK: [[VECTOR_PH]]:
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 2
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
+; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
+; CHECK: [[VECTOR_BODY]]:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP5:%.*]] = alloca i8, i64 0, align 16
+; CHECK-NEXT: [[TMP6:%.*]] = alloca i8, i64 0, align 16
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x ptr> poison, ptr [[TMP5]], i32 0
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x ptr> [[TMP7]], ptr [[TMP6]], i32 1
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr ptr, ptr [[X]], i64 [[INDEX]]
+; CHECK-NEXT: store <2 x ptr> [[TMP8]], ptr [[TMP9]], align 8
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT: br label %[[FOR_BODY:.*]]
+; CHECK: [[FOR_BODY]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP12:%.*]] = alloca i8, i64 0, align 16
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr ptr, ptr [[X]], i64 [[IV]]
+; CHECK-NEXT: store ptr [[TMP12]], ptr [[ARRAYIDX]], align 8
+; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV]], [[N]]
+; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK: [[FOR_END]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %for.body
+
+for.body:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+ %0 = alloca i8, i64 0, align 16
+ %arrayidx = getelementptr ptr, ptr %x, i64 %iv
+ store ptr %0, ptr %arrayidx, align 8
+ %iv.next = add i64 %iv, 1
+ %exitcond.not = icmp eq i64 %iv, %N
+ br i1 %exitcond.not, label %for.end, label %for.body
+
+for.end:
+ ret void
+}
+
+attributes #0 = { "target-features"="+neon,+sve" vscale_range(1,16) }
+
declare void @llvm.assume(i1 noundef)
declare i64 @llvm.umin.i64(i64, i64)
;.
@@ -392,4 +447,6 @@ declare i64 @llvm.umin.i64(i64, i64)
; CHECK: [[META20]] = !{[[META13]]}
; CHECK: [[LOOP21]] = distinct !{[[LOOP21]], [[META1]], [[META2]]}
; CHECK: [[LOOP22]] = distinct !{[[LOOP22]], [[META1]]}
+; CHECK: [[LOOP23]] = distinct !{[[LOOP23]], [[META1]], [[META2]]}
+; CHECK: [[LOOP24]] = distinct !{[[LOOP24]], [[META2]], [[META1]]}
;.
>From 2f14b65ee4ff64e7e6d768b655a8b9a9bddb9346 Mon Sep 17 00:00:00 2001
From: Kerry McLaughlin <kerry.mclaughlin at arm.com>
Date: Tue, 26 Aug 2025 08:47:29 +0000
Subject: [PATCH 4/4] - Use -force-target-instruction-cost=10
---
.../LoopVectorize/AArch64/force-target-instruction-cost.ll | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
index 34c0b58bec230..708967e3d13af 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
@@ -1,5 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -p loop-vectorize -force-target-instruction-cost=100 -S %s | FileCheck %s
+; RUN: opt -p loop-vectorize -force-target-instruction-cost=10 -S %s | FileCheck %s
target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128-Fn32"
target triple = "arm64-apple-macosx14.0.0"
More information about the llvm-commits
mailing list