[LV] Pre-commit test for #151664 #151671
Merged
[LV] Pre-commit test for #151664 #151671
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
Hoisted vector instructions are costed incorrectly.
llvmbot
added
llvm:analysis
Member
|
@llvm/pr-subscribers-llvm-transforms Author: Ramkumar Ramachandra (artagnon) ChangesHoisted vector instructions are costed incorrectly. Full diff: https://github.com/llvm/llvm-project/pull/151671.diff 3 Files Affected:
diff --git a/llvm/test/Analysis/CostModel/AArch64/min-max.ll b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
index b824f5309adc1..3c052b1eeea82 100644
--- a/llvm/test/Analysis/CostModel/AArch64/min-max.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
@@ -377,6 +377,66 @@ define void @maximum() {
ret void
}
+define void @minimumnum() {
+; CHECK-LABEL: 'minimumnum'
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %1 = call <2 x float> @llvm.minimumnum.v2f32(<2 x float> poison, <2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %2 = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> poison, <4 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 10 for: %3 = call <8 x float> @llvm.minimumnum.v8f32(<8 x float> poison, <8 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 20 for: %4 = call <16 x float> @llvm.minimumnum.v16f32(<16 x float> poison, <16 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %5 = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> poison, <2 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 10 for: %6 = call <4 x double> @llvm.minimumnum.v4f64(<4 x double> poison, <4 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 20 for: %7 = call <8 x double> @llvm.minimumnum.v8f64(<8 x double> poison, <8 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 40 for: %8 = call <16 x double> @llvm.minimumnum.v16f64(<16 x double> poison, <16 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <2 x float> @llvm.minimumnum.v2f32(<2 x float> poison, <2 x float> poison)
+ call <4 x float> @llvm.minimumnum.v4f32(<4 x float> poison, <4 x float> poison)
+ call <8 x float> @llvm.minimumnum.v8f32(<8 x float> poison, <8 x float> poison)
+ call <16 x float> @llvm.minimumnum.v16f32(<16 x float> poison, <16 x float> poison)
+ call <2 x double> @llvm.minimumnum.v2f64(<2 x double> poison, <2 x double> poison)
+ call <4 x double> @llvm.minimumnum.v4f64(<4 x double> poison, <4 x double> poison)
+ call <8 x double> @llvm.minimumnum.v8f64(<8 x double> poison, <8 x double> poison)
+ call <16 x double> @llvm.minimumnum.v16f64(<16 x double> poison, <16 x double> poison)
+ ret void
+}
+
+define void @minimumnum_fp16() {
+; CHECK-NOF16-LABEL: 'minimumnum_fp16'
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:4 CodeSize:3 Lat:4 SizeLat:4 for: %1 = call <2 x half> @llvm.minimumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %2 = call <4 x half> @llvm.minimumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:15 Lat:22 SizeLat:22 for: %3 = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:44 CodeSize:30 Lat:44 SizeLat:44 for: %4 = call <16 x half> @llvm.minimumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+; CHECK-F16-LABEL: 'minimumnum_fp16'
+; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %1 = call <2 x half> @llvm.minimumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %2 = call <4 x half> @llvm.minimumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %3 = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of 10 for: %4 = call <16 x half> @llvm.minimumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <2 x half> @llvm.minimumnum.v2f16(<2 x half> poison, <2 x half> poison)
+ call <4 x half> @llvm.minimumnum.v4f16(<4 x half> poison, <4 x half> poison)
+ call <8 x half> @llvm.minimumnum.v8f16(<8 x half> poison, <8 x half> poison)
+ call <16 x half> @llvm.minimumnum.v16f16(<16 x half> poison, <16 x half> poison)
+ ret void
+}
+
+define void @minimumnum_bf16() {
+; CHECK-LABEL: 'minimumnum_bf16'
+; CHECK-NEXT: Cost Model: Found costs of RThru:4 CodeSize:3 Lat:4 SizeLat:4 for: %1 = call <2 x bfloat> @llvm.minimumnum.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %2 = call <4 x bfloat> @llvm.minimumnum.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:22 CodeSize:15 Lat:22 SizeLat:22 for: %3 = call <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:44 CodeSize:30 Lat:44 SizeLat:44 for: %4 = call <16 x bfloat> @llvm.minimumnum.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <2 x bfloat> @llvm.minimumnum.v2fb16(<2 x bfloat> poison, <2 x bfloat> poison)
+ call <4 x bfloat> @llvm.minimumnum.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+ call <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+ call <16 x bfloat> @llvm.minimumnum.v6bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+ ret void
+}
+
declare i8 @llvm.umin.i8(i8, i8)
declare i16 @llvm.umin.i16(i16, i16)
declare i32 @llvm.umin.i32(i32, i32)
@@ -512,3 +572,20 @@ declare <4 x float> @llvm.maximum.v4f32(<4 x float>, <4 x float>)
declare <8 x float> @llvm.maximum.v8f32(<8 x float>, <8 x float>)
declare <2 x double> @llvm.maximum.v2f64(<2 x double>, <2 x double>)
declare <4 x double> @llvm.maximum.v4f64(<4 x double>, <4 x double>)
+
+declare <2 x half> @llvm.minimumnum.v2f16(<2 x half>, <2 x half>)
+declare <4 x half> @llvm.minimumnum.v4f16(<4 x half>, <4 x half>)
+declare <8 x half> @llvm.minimumnum.v8f16(<8 x half>, <8 x half>)
+declare <16 x half> @llvm.minimumnum.v16f16(<16 x half>, <16 x half>)
+declare <2 x bfloat> @llvm.minimumnum.2bf16(<2 x bfloat>, <2 x bfloat>)
+declare <4 x bfloat> @llvm.minimumnum.4bf16(<4 x bfloat>, <4 x bfloat>)
+declare <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat>, <8 x bfloat>)
+declare <16 x bfloat> @llvm.minimumnum.v16bf16(<16 x bfloat>, <16 x bfloat>)
+declare <2 x float> @llvm.minimumnum.v2f32(<2 x float>, <2 x float>)
+declare <4 x float> @llvm.minimumnum.v4f32(<4 x float>, <4 x float>)
+declare <8 x float> @llvm.minimumnum.v8f32(<8 x float>, <8 x float>)
+declare <16 x float> @llvm.minimumnum.v16f32(<16 x float>, <16 x float>)
+declare <2 x double> @llvm.minimumnum.v2f64(<2 x double>, <2 x double>)
+declare <4 x double> @llvm.minimumnum.v4f64(<4 x double>, <4 x double>)
+declare <8 x double> @llvm.minimumnum.nxv8f64(<8 x double>, <8 x double>)
+declare <16 x double> @llvm.minimumnum.nxv16f64(<16 x double>, <16 x double>)
diff --git a/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll b/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll
index 829ce127493ed..73574dd61d75c 100644
--- a/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll
@@ -255,6 +255,60 @@ define void @maximum() {
ret void
}
+define void @minimumnum() {
+; CHECK-LABEL: 'minimumnum'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %1 = call <vscale x 2 x float> @llvm.minimumnum.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %2 = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %3 = call <vscale x 8 x float> @llvm.minimumnum.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %4 = call <vscale x 16 x float> @llvm.minimumnum.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %5 = call <vscale x 2 x double> @llvm.minimumnum.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %6 = call <vscale x 4 x double> @llvm.minimumnum.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %7 = call <vscale x 8 x double> @llvm.minimumnum.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %8 = call <vscale x 16 x double> @llvm.minimumnum.nxv16f64(<vscale x 16 x double> poison, <vscale x 16 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <vscale x 2 x float> @llvm.minimumnum(<vscale x 2 x float> poison, <vscale x 2 x float> poison)
+ call <vscale x 4 x float> @llvm.minimumnum(<vscale x 4 x float> poison, <vscale x 4 x float> poison)
+ call <vscale x 8 x float> @llvm.minimumnum(<vscale x 8 x float> poison, <vscale x 8 x float> poison)
+ call <vscale x 16 x float> @llvm.minimumnum(<vscale x 16 x float> poison, <vscale x 16 x float> poison)
+ call <vscale x 2 x double> @llvm.minimumnum(<vscale x 2 x double> poison, <vscale x 2 x double> poison)
+ call <vscale x 4 x double> @llvm.minimumnum(<vscale x 4 x double> poison, <vscale x 4 x double> poison)
+ call <vscale x 8 x double> @llvm.minimumnum(<vscale x 8 x double> poison, <vscale x 8 x double> poison)
+ call <vscale x 16 x double> @llvm.minimumnum(<vscale x 16 x double> poison, <vscale x 16 x double> poison)
+ ret void
+}
+
+define void @minimumnum_fp16() {
+; CHECK-LABEL: 'minimumnum_fp16'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %1 = call <vscale x 2 x half> @llvm.minimumnum.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %2 = call <vscale x 4 x half> @llvm.minimumnum.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %3 = call <vscale x 8 x half> @llvm.minimumnum.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %4 = call <vscale x 16 x half> @llvm.minimumnum.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <vscale x 2 x half> @llvm.minimumnum(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+ call <vscale x 4 x half> @llvm.minimumnum(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+ call <vscale x 8 x half> @llvm.minimumnum(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+ call <vscale x 16 x half> @llvm.minimumnum(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+ ret void
+}
+
+define void @minimumnum_bf16() {
+; CHECK-LABEL: 'minimumnum_bf16'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %1 = call <vscale x 2 x bfloat> @llvm.minimumnum.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %2 = call <vscale x 4 x bfloat> @llvm.minimumnum.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %3 = call <vscale x 8 x bfloat> @llvm.minimumnum.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %4 = call <vscale x 16 x bfloat> @llvm.minimumnum.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <vscale x 2 x bfloat> @llvm.minimumnum(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+ call <vscale x 4 x bfloat> @llvm.minimumnum(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+ call <vscale x 8 x bfloat> @llvm.minimumnum(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+ call <vscale x 16 x bfloat> @llvm.minimumnum(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+ ret void
+}
+
+
declare <vscale x 4 x i8> @llvm.umin.nxv4i8(<vscale x 4 x i8>, <vscale x 4 x i8>)
declare <vscale x 8 x i8> @llvm.umin.nxv8i8(<vscale x 8 x i8>, <vscale x 8 x i8>)
declare <vscale x 16 x i8> @llvm.umin.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
@@ -354,3 +408,20 @@ declare <vscale x 4 x float> @llvm.maximum.nxv4f32(<vscale x 4 x float>, <vscale
declare <vscale x 8 x float> @llvm.maximum.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>)
declare <vscale x 2 x double> @llvm.maximum.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>)
declare <vscale x 4 x double> @llvm.maximum.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>)
+
+declare <vscale x 2 x half> @llvm.minimumnum.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>)
+declare <vscale x 4 x half> @llvm.minimumnum.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>)
+declare <vscale x 8 x half> @llvm.minimumnum.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 16 x half> @llvm.minimumnum.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>)
+declare <vscale x 2 x bfloat> @llvm.minimumnum.nxv2bf16(<vscale x 2 x bfloat>, <vscale x 2 x bfloat>)
+declare <vscale x 4 x bfloat> @llvm.minimumnum.nxv4bf16(<vscale x 4 x bfloat>, <vscale x 4 x bfloat>)
+declare <vscale x 8 x bfloat> @llvm.minimumnum.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>)
+declare <vscale x 16 x bfloat> @llvm.minimumnum.nxv16bf16(<vscale x 16 x bfloat>, <vscale x 16 x bfloat>)
+declare <vscale x 2 x float> @llvm.minimumnum.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>)
+declare <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 8 x float> @llvm.minimumnum.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>)
+declare <vscale x 16 x float> @llvm.minimumnum.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>)
+declare <vscale x 2 x double> @llvm.minimumnum.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>)
+declare <vscale x 4 x double> @llvm.minimumnum.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>)
+declare <vscale x 8 x double> @llvm.minimumnum.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>)
+declare <vscale x 16 x double> @llvm.minimumnum.nxv16f64(<vscale x 16 x double>, <vscale x 16 x double>)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll b/llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll
new file mode 100644
index 0000000000000..2823712c9143a
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll
@@ -0,0 +1,62 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph" --version 5
+; RUN: opt -passes=loop-vectorize -mtriple=aarch64 -mattr=+sve -S \
+; RUN: -debug-only=loop-vectorize %s 2>&1 | FileCheck %s
+
+; FIXME: Hoisted vector code should be costed with scalable cost.
+; In this example, `<vscale x 4 x float> @llvm.minimumnum` has an invalid cost,
+; and hence should not be produced by LoopVectorize. Unfortunately, what is
+; actually costed is `<4 x float> @llvm.minimumnum`.
+
+; CHECK: LV: Found an estimated cost of 3 for VF 1 For instruction: %res = tail call float @llvm.minimumnum.f32(float 0.000000e+00, float 0.000000e+00)
+define void @cost_hoisted_vector_code(ptr %p) {
+; CHECK-LABEL: define void @cost_hoisted_vector_code(
+; CHECK-SAME: ptr [[P:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8
+; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK: [[VECTOR_PH]]:
+; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 -1, [[TMP3]]
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 -1, [[N_MOD_VF]]
+; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 1, [[N_VEC]]
+; CHECK-NEXT: [[TMP7:%.*]] = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
+; CHECK: [[VECTOR_BODY]]:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[P]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; CHECK-NEXT: [[TMP11:%.*]] = getelementptr float, ptr [[TMP8]], i64 [[TMP10]]
+; CHECK-NEXT: store <vscale x 4 x float> [[TMP7]], ptr [[TMP8]], align 4
+; CHECK-NEXT: store <vscale x 4 x float> [[TMP7]], ptr [[TMP11]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 -1, [[N_VEC]]
+; CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK: [[SCALAR_PH]]:
+;
+entry:
+ br label %loop
+
+loop: ; preds = %loop, %entry
+ %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]
+ %idx = phi i64 [ 0, %entry ], [ %idx.next, %loop ]
+ %res = tail call float @llvm.minimumnum.f32(float 0.0, float 0.0)
+ %gep.p.red = getelementptr float, ptr %p, i64 %idx
+ store float %res, ptr %gep.p.red, align 4
+ %idx.next = add i64 %idx, 1
+ %iv.next = add i64 %iv, 1
+ %exit.cond = icmp eq i64 %iv.next, 0
+ br i1 %exit.cond, label %exit, label %loop
+
+exit: ; preds = %loop
+ ret void
+}
+
+declare float @llvm.minimumnum.f32(float, float)
|
Member
|
@llvm/pr-subscribers-llvm-analysis Author: Ramkumar Ramachandra (artagnon) ChangesHoisted vector instructions are costed incorrectly. Full diff: https://github.com/llvm/llvm-project/pull/151671.diff 3 Files Affected:
diff --git a/llvm/test/Analysis/CostModel/AArch64/min-max.ll b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
index b824f5309adc1..3c052b1eeea82 100644
--- a/llvm/test/Analysis/CostModel/AArch64/min-max.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
@@ -377,6 +377,66 @@ define void @maximum() {
ret void
}
+define void @minimumnum() {
+; CHECK-LABEL: 'minimumnum'
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %1 = call <2 x float> @llvm.minimumnum.v2f32(<2 x float> poison, <2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %2 = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> poison, <4 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 10 for: %3 = call <8 x float> @llvm.minimumnum.v8f32(<8 x float> poison, <8 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 20 for: %4 = call <16 x float> @llvm.minimumnum.v16f32(<16 x float> poison, <16 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %5 = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> poison, <2 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 10 for: %6 = call <4 x double> @llvm.minimumnum.v4f64(<4 x double> poison, <4 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 20 for: %7 = call <8 x double> @llvm.minimumnum.v8f64(<8 x double> poison, <8 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 40 for: %8 = call <16 x double> @llvm.minimumnum.v16f64(<16 x double> poison, <16 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <2 x float> @llvm.minimumnum.v2f32(<2 x float> poison, <2 x float> poison)
+ call <4 x float> @llvm.minimumnum.v4f32(<4 x float> poison, <4 x float> poison)
+ call <8 x float> @llvm.minimumnum.v8f32(<8 x float> poison, <8 x float> poison)
+ call <16 x float> @llvm.minimumnum.v16f32(<16 x float> poison, <16 x float> poison)
+ call <2 x double> @llvm.minimumnum.v2f64(<2 x double> poison, <2 x double> poison)
+ call <4 x double> @llvm.minimumnum.v4f64(<4 x double> poison, <4 x double> poison)
+ call <8 x double> @llvm.minimumnum.v8f64(<8 x double> poison, <8 x double> poison)
+ call <16 x double> @llvm.minimumnum.v16f64(<16 x double> poison, <16 x double> poison)
+ ret void
+}
+
+define void @minimumnum_fp16() {
+; CHECK-NOF16-LABEL: 'minimumnum_fp16'
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:4 CodeSize:3 Lat:4 SizeLat:4 for: %1 = call <2 x half> @llvm.minimumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %2 = call <4 x half> @llvm.minimumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:15 Lat:22 SizeLat:22 for: %3 = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:44 CodeSize:30 Lat:44 SizeLat:44 for: %4 = call <16 x half> @llvm.minimumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; CHECK-NOF16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+; CHECK-F16-LABEL: 'minimumnum_fp16'
+; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %1 = call <2 x half> @llvm.minimumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %2 = call <4 x half> @llvm.minimumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %3 = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of 10 for: %4 = call <16 x half> @llvm.minimumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; CHECK-F16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <2 x half> @llvm.minimumnum.v2f16(<2 x half> poison, <2 x half> poison)
+ call <4 x half> @llvm.minimumnum.v4f16(<4 x half> poison, <4 x half> poison)
+ call <8 x half> @llvm.minimumnum.v8f16(<8 x half> poison, <8 x half> poison)
+ call <16 x half> @llvm.minimumnum.v16f16(<16 x half> poison, <16 x half> poison)
+ ret void
+}
+
+define void @minimumnum_bf16() {
+; CHECK-LABEL: 'minimumnum_bf16'
+; CHECK-NEXT: Cost Model: Found costs of RThru:4 CodeSize:3 Lat:4 SizeLat:4 for: %1 = call <2 x bfloat> @llvm.minimumnum.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %2 = call <4 x bfloat> @llvm.minimumnum.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:22 CodeSize:15 Lat:22 SizeLat:22 for: %3 = call <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:44 CodeSize:30 Lat:44 SizeLat:44 for: %4 = call <16 x bfloat> @llvm.minimumnum.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <2 x bfloat> @llvm.minimumnum.v2fb16(<2 x bfloat> poison, <2 x bfloat> poison)
+ call <4 x bfloat> @llvm.minimumnum.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+ call <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+ call <16 x bfloat> @llvm.minimumnum.v6bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+ ret void
+}
+
declare i8 @llvm.umin.i8(i8, i8)
declare i16 @llvm.umin.i16(i16, i16)
declare i32 @llvm.umin.i32(i32, i32)
@@ -512,3 +572,20 @@ declare <4 x float> @llvm.maximum.v4f32(<4 x float>, <4 x float>)
declare <8 x float> @llvm.maximum.v8f32(<8 x float>, <8 x float>)
declare <2 x double> @llvm.maximum.v2f64(<2 x double>, <2 x double>)
declare <4 x double> @llvm.maximum.v4f64(<4 x double>, <4 x double>)
+
+declare <2 x half> @llvm.minimumnum.v2f16(<2 x half>, <2 x half>)
+declare <4 x half> @llvm.minimumnum.v4f16(<4 x half>, <4 x half>)
+declare <8 x half> @llvm.minimumnum.v8f16(<8 x half>, <8 x half>)
+declare <16 x half> @llvm.minimumnum.v16f16(<16 x half>, <16 x half>)
+declare <2 x bfloat> @llvm.minimumnum.2bf16(<2 x bfloat>, <2 x bfloat>)
+declare <4 x bfloat> @llvm.minimumnum.4bf16(<4 x bfloat>, <4 x bfloat>)
+declare <8 x bfloat> @llvm.minimumnum.v8bf16(<8 x bfloat>, <8 x bfloat>)
+declare <16 x bfloat> @llvm.minimumnum.v16bf16(<16 x bfloat>, <16 x bfloat>)
+declare <2 x float> @llvm.minimumnum.v2f32(<2 x float>, <2 x float>)
+declare <4 x float> @llvm.minimumnum.v4f32(<4 x float>, <4 x float>)
+declare <8 x float> @llvm.minimumnum.v8f32(<8 x float>, <8 x float>)
+declare <16 x float> @llvm.minimumnum.v16f32(<16 x float>, <16 x float>)
+declare <2 x double> @llvm.minimumnum.v2f64(<2 x double>, <2 x double>)
+declare <4 x double> @llvm.minimumnum.v4f64(<4 x double>, <4 x double>)
+declare <8 x double> @llvm.minimumnum.nxv8f64(<8 x double>, <8 x double>)
+declare <16 x double> @llvm.minimumnum.nxv16f64(<16 x double>, <16 x double>)
diff --git a/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll b/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll
index 829ce127493ed..73574dd61d75c 100644
--- a/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/sve-min-max.ll
@@ -255,6 +255,60 @@ define void @maximum() {
ret void
}
+define void @minimumnum() {
+; CHECK-LABEL: 'minimumnum'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %1 = call <vscale x 2 x float> @llvm.minimumnum.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %2 = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %3 = call <vscale x 8 x float> @llvm.minimumnum.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %4 = call <vscale x 16 x float> @llvm.minimumnum.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %5 = call <vscale x 2 x double> @llvm.minimumnum.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %6 = call <vscale x 4 x double> @llvm.minimumnum.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %7 = call <vscale x 8 x double> @llvm.minimumnum.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %8 = call <vscale x 16 x double> @llvm.minimumnum.nxv16f64(<vscale x 16 x double> poison, <vscale x 16 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <vscale x 2 x float> @llvm.minimumnum(<vscale x 2 x float> poison, <vscale x 2 x float> poison)
+ call <vscale x 4 x float> @llvm.minimumnum(<vscale x 4 x float> poison, <vscale x 4 x float> poison)
+ call <vscale x 8 x float> @llvm.minimumnum(<vscale x 8 x float> poison, <vscale x 8 x float> poison)
+ call <vscale x 16 x float> @llvm.minimumnum(<vscale x 16 x float> poison, <vscale x 16 x float> poison)
+ call <vscale x 2 x double> @llvm.minimumnum(<vscale x 2 x double> poison, <vscale x 2 x double> poison)
+ call <vscale x 4 x double> @llvm.minimumnum(<vscale x 4 x double> poison, <vscale x 4 x double> poison)
+ call <vscale x 8 x double> @llvm.minimumnum(<vscale x 8 x double> poison, <vscale x 8 x double> poison)
+ call <vscale x 16 x double> @llvm.minimumnum(<vscale x 16 x double> poison, <vscale x 16 x double> poison)
+ ret void
+}
+
+define void @minimumnum_fp16() {
+; CHECK-LABEL: 'minimumnum_fp16'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %1 = call <vscale x 2 x half> @llvm.minimumnum.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %2 = call <vscale x 4 x half> @llvm.minimumnum.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %3 = call <vscale x 8 x half> @llvm.minimumnum.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %4 = call <vscale x 16 x half> @llvm.minimumnum.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <vscale x 2 x half> @llvm.minimumnum(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+ call <vscale x 4 x half> @llvm.minimumnum(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+ call <vscale x 8 x half> @llvm.minimumnum(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+ call <vscale x 16 x half> @llvm.minimumnum(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+ ret void
+}
+
+define void @minimumnum_bf16() {
+; CHECK-LABEL: 'minimumnum_bf16'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %1 = call <vscale x 2 x bfloat> @llvm.minimumnum.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %2 = call <vscale x 4 x bfloat> @llvm.minimumnum.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %3 = call <vscale x 8 x bfloat> @llvm.minimumnum.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %4 = call <vscale x 16 x bfloat> @llvm.minimumnum.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+ call <vscale x 2 x bfloat> @llvm.minimumnum(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+ call <vscale x 4 x bfloat> @llvm.minimumnum(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+ call <vscale x 8 x bfloat> @llvm.minimumnum(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+ call <vscale x 16 x bfloat> @llvm.minimumnum(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+ ret void
+}
+
+
declare <vscale x 4 x i8> @llvm.umin.nxv4i8(<vscale x 4 x i8>, <vscale x 4 x i8>)
declare <vscale x 8 x i8> @llvm.umin.nxv8i8(<vscale x 8 x i8>, <vscale x 8 x i8>)
declare <vscale x 16 x i8> @llvm.umin.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
@@ -354,3 +408,20 @@ declare <vscale x 4 x float> @llvm.maximum.nxv4f32(<vscale x 4 x float>, <vscale
declare <vscale x 8 x float> @llvm.maximum.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>)
declare <vscale x 2 x double> @llvm.maximum.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>)
declare <vscale x 4 x double> @llvm.maximum.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>)
+
+declare <vscale x 2 x half> @llvm.minimumnum.nxv2f16(<vscale x 2 x half>, <vscale x 2 x half>)
+declare <vscale x 4 x half> @llvm.minimumnum.nxv4f16(<vscale x 4 x half>, <vscale x 4 x half>)
+declare <vscale x 8 x half> @llvm.minimumnum.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>)
+declare <vscale x 16 x half> @llvm.minimumnum.nxv16f16(<vscale x 16 x half>, <vscale x 16 x half>)
+declare <vscale x 2 x bfloat> @llvm.minimumnum.nxv2bf16(<vscale x 2 x bfloat>, <vscale x 2 x bfloat>)
+declare <vscale x 4 x bfloat> @llvm.minimumnum.nxv4bf16(<vscale x 4 x bfloat>, <vscale x 4 x bfloat>)
+declare <vscale x 8 x bfloat> @llvm.minimumnum.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>)
+declare <vscale x 16 x bfloat> @llvm.minimumnum.nxv16bf16(<vscale x 16 x bfloat>, <vscale x 16 x bfloat>)
+declare <vscale x 2 x float> @llvm.minimumnum.nxv2f32(<vscale x 2 x float>, <vscale x 2 x float>)
+declare <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>)
+declare <vscale x 8 x float> @llvm.minimumnum.nxv8f32(<vscale x 8 x float>, <vscale x 8 x float>)
+declare <vscale x 16 x float> @llvm.minimumnum.nxv16f32(<vscale x 16 x float>, <vscale x 16 x float>)
+declare <vscale x 2 x double> @llvm.minimumnum.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>)
+declare <vscale x 4 x double> @llvm.minimumnum.nxv4f64(<vscale x 4 x double>, <vscale x 4 x double>)
+declare <vscale x 8 x double> @llvm.minimumnum.nxv8f64(<vscale x 8 x double>, <vscale x 8 x double>)
+declare <vscale x 16 x double> @llvm.minimumnum.nxv16f64(<vscale x 16 x double>, <vscale x 16 x double>)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll b/llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll
new file mode 100644
index 0000000000000..2823712c9143a
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll
@@ -0,0 +1,62 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^scalar.ph" --version 5
+; RUN: opt -passes=loop-vectorize -mtriple=aarch64 -mattr=+sve -S \
+; RUN: -debug-only=loop-vectorize %s 2>&1 | FileCheck %s
+
+; FIXME: Hoisted vector code should be costed with scalable cost.
+; In this example, `<vscale x 4 x float> @llvm.minimumnum` has an invalid cost,
+; and hence should not be produced by LoopVectorize. Unfortunately, what is
+; actually costed is `<4 x float> @llvm.minimumnum`.
+
+; CHECK: LV: Found an estimated cost of 3 for VF 1 For instruction: %res = tail call float @llvm.minimumnum.f32(float 0.000000e+00, float 0.000000e+00)
+define void @cost_hoisted_vector_code(ptr %p) {
+; CHECK-LABEL: define void @cost_hoisted_vector_code(
+; CHECK-SAME: ptr [[P:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8
+; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK: [[VECTOR_PH]]:
+; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 -1, [[TMP3]]
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 -1, [[N_MOD_VF]]
+; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 1, [[N_VEC]]
+; CHECK-NEXT: [[TMP7:%.*]] = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
+; CHECK: [[VECTOR_BODY]]:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[P]], i64 [[INDEX]]
+; CHECK-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; CHECK-NEXT: [[TMP11:%.*]] = getelementptr float, ptr [[TMP8]], i64 [[TMP10]]
+; CHECK-NEXT: store <vscale x 4 x float> [[TMP7]], ptr [[TMP8]], align 4
+; CHECK-NEXT: store <vscale x 4 x float> [[TMP7]], ptr [[TMP11]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 -1, [[N_VEC]]
+; CHECK-NEXT: br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK: [[SCALAR_PH]]:
+;
+entry:
+ br label %loop
+
+loop: ; preds = %loop, %entry
+ %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]
+ %idx = phi i64 [ 0, %entry ], [ %idx.next, %loop ]
+ %res = tail call float @llvm.minimumnum.f32(float 0.0, float 0.0)
+ %gep.p.red = getelementptr float, ptr %p, i64 %idx
+ store float %res, ptr %gep.p.red, align 4
+ %idx.next = add i64 %idx, 1
+ %iv.next = add i64 %iv, 1
+ %exit.cond = icmp eq i64 %iv.next, 0
+ br i1 %exit.cond, label %exit, label %loop
+
+exit: ; preds = %loop
+ ret void
+}
+
+declare float @llvm.minimumnum.f32(float, float)
|
davemgreen
reviewed
Aug 1, 2025
Collaborator
davemgreen
left a comment
davemgreen
left a comment
There was a problem hiding this comment.
Thanks for adding these cost-model tests, they look like a good addition.
llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll
Outdated
Show resolved
Hide resolved
fhahn
reviewed
Aug 1, 2025
llvm/test/Transforms/LoopVectorize/AArch64/pr151664-cost-hoisted-vector-scalable.ll
Outdated
Show resolved
Hide resolved
davemgreen
approved these changes
Aug 1, 2025
| call <vscale x 4 x half> @llvm.minimumnum(<vscale x 4 x half> poison, <vscale x 4 x half> poison) | ||
| call <vscale x 8 x half> @llvm.minimumnum(<vscale x 8 x half> poison, <vscale x 8 x half> poison) | ||
| call <vscale x 16 x half> @llvm.minimumnum(<vscale x 16 x half> poison, <vscale x 16 x half> poison) | ||
| call <vscale x 2 x bfloat> @llvm.minimumnum(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison) |
Collaborator
There was a problem hiding this comment.
BF16 is actually a different arch extension, but we can fix that later when we add proper costs for it.
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/11/builds/20898 Here is the relevant piece of the build log for the reference |
Contributor
Author
|
Oops, forgot the |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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.
Hoisted vector instructions are costed incorrectly.