[llvm] [LLVM][CostModel][SVE] Return InvalidCost for bfloat scalable vector ordered arithmetic reductions. (PR #202569)
via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 9 03:22:44 PDT 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-aarch64
Author: Paul Walker (paulwalker-arm)
<details>
<summary>Changes</summary>
Also extends existing FP vector.reduce cost-model tests to cover all intrinsics and vector types from "vscale x 1" to vectors that are 4x type legal.
---
Patch is 163.23 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/202569.diff
3 Files Affected:
- (modified) llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp (+1-1)
- (modified) llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll (-76)
- (added) llvm/test/Analysis/CostModel/AArch64/sve-vector-reduce-fp.ll (+1137)
``````````diff
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 4a55109bd2b23..ad84fe8cdaec7 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -5789,7 +5789,7 @@ AArch64TTIImpl::getArithmeticReductionCost(unsigned Opcode, VectorType *ValTy,
return BaseCost + FixedVTy->getNumElements();
}
- if (Opcode != Instruction::FAdd)
+ if (Opcode != Instruction::FAdd || ValTy->getElementType()->isBFloatTy())
return InstructionCost::getInvalid();
auto *VTy = cast<ScalableVectorType>(ValTy);
diff --git a/llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll b/llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
index a7bf82a62ea04..17a90510f8a98 100644
--- a/llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/sve-intrinsics.ll
@@ -177,15 +177,6 @@ define void @reductions(<vscale x 4 x i32> %v0, <vscale x 4 x i64> %v1, <vscale
; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of Invalid for: %smax_nxv1i64 = call i64 @llvm.vector.reduce.smax.nxv1i64(<vscale x 1 x i64> poison)
; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 2 for: %smax_nxv4i32 = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> %v0)
; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 3 for: %smax_nxv4i64 = call i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64> %v1)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f32 = call fast float @llvm.vector.reduce.fadd.nxv1f32(float 0.000000e+00, <vscale x 1 x float> poison)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f32 = call fast float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v2)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv4f64 = call fast double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v3)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of Invalid for: %fmin_nxv1f32 = call fast float @llvm.vector.reduce.fmin.nxv1f32(<vscale x 1 x float> poison)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 2 for: %fmin_nxv4f32 = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> %v2)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 3 for: %fmin_nxv4f64 = call fast double @llvm.vector.reduce.fmin.nxv4f64(<vscale x 4 x double> %v3)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of Invalid for: %fmax_nxv1f32 = call fast float @llvm.vector.reduce.fmax.nxv1f32(<vscale x 1 x float> poison)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv4f32 = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> %v2)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of 3 for: %fmax_nxv4f64 = call fast double @llvm.vector.reduce.fmax.nxv4f64(<vscale x 4 x double> %v3)
; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; CHECK-VSCALE-2-LABEL: 'reductions'
@@ -216,15 +207,6 @@ define void @reductions(<vscale x 4 x i32> %v0, <vscale x 4 x i64> %v1, <vscale
; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of Invalid for: %smax_nxv1i64 = call i64 @llvm.vector.reduce.smax.nxv1i64(<vscale x 1 x i64> poison)
; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 2 for: %smax_nxv4i32 = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> %v0)
; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 3 for: %smax_nxv4i64 = call i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64> %v1)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f32 = call fast float @llvm.vector.reduce.fadd.nxv1f32(float 0.000000e+00, <vscale x 1 x float> poison)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f32 = call fast float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v2)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv4f64 = call fast double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v3)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of Invalid for: %fmin_nxv1f32 = call fast float @llvm.vector.reduce.fmin.nxv1f32(<vscale x 1 x float> poison)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 2 for: %fmin_nxv4f32 = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> %v2)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 3 for: %fmin_nxv4f64 = call fast double @llvm.vector.reduce.fmin.nxv4f64(<vscale x 4 x double> %v3)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of Invalid for: %fmax_nxv1f32 = call fast float @llvm.vector.reduce.fmax.nxv1f32(<vscale x 1 x float> poison)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv4f32 = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> %v2)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of 3 for: %fmax_nxv4f64 = call fast double @llvm.vector.reduce.fmax.nxv4f64(<vscale x 4 x double> %v3)
; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; TYPE_BASED_ONLY-LABEL: 'reductions'
@@ -255,15 +237,6 @@ define void @reductions(<vscale x 4 x i32> %v0, <vscale x 4 x i64> %v1, <vscale
; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of Invalid for: %smax_nxv1i64 = call i64 @llvm.vector.reduce.smax.nxv1i64(<vscale x 1 x i64> poison)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 2 for: %smax_nxv4i32 = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> %v0)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 3 for: %smax_nxv4i64 = call i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64> %v1)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f32 = call fast float @llvm.vector.reduce.fadd.nxv1f32(float 0.000000e+00, <vscale x 1 x float> poison)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f32 = call fast float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v2)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv4f64 = call fast double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v3)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of Invalid for: %fmin_nxv1f32 = call fast float @llvm.vector.reduce.fmin.nxv1f32(<vscale x 1 x float> poison)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 2 for: %fmin_nxv4f32 = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> %v2)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 3 for: %fmin_nxv4f64 = call fast double @llvm.vector.reduce.fmin.nxv4f64(<vscale x 4 x double> %v3)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of Invalid for: %fmax_nxv1f32 = call fast float @llvm.vector.reduce.fmax.nxv1f32(<vscale x 1 x float> poison)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv4f32 = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> %v2)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of 3 for: %fmax_nxv4f64 = call fast double @llvm.vector.reduce.fmax.nxv4f64(<vscale x 4 x double> %v3)
; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
%add_nxv1i32 = call i32 @llvm.vector.reduce.add.nxv1i32(<vscale x 1 x i32> poison)
@@ -294,46 +267,6 @@ define void @reductions(<vscale x 4 x i32> %v0, <vscale x 4 x i64> %v1, <vscale
%smax_nxv4i32 = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> %v0)
%smax_nxv4i64 = call i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64> %v1)
- %fadd_nxv1f32 = call fast float @llvm.vector.reduce.fadd.nxv1f32(float 0.0, <vscale x 1 x float> poison)
- %fadd_nxv4f32 = call fast float @llvm.vector.reduce.fadd.nxv4f32(float 0.0, <vscale x 4 x float> %v2)
- %fadd_nxv4f64 = call fast double @llvm.vector.reduce.fadd.nxv4f64(double 0.0, <vscale x 4 x double> %v3)
- %fmin_nxv1f32 = call fast float @llvm.vector.reduce.fmin.nxv1f32(<vscale x 1 x float> poison)
- %fmin_nxv4f32 = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> %v2)
- %fmin_nxv4f64 = call fast double @llvm.vector.reduce.fmin.nxv4f64(<vscale x 4 x double> %v3)
- %fmax_nxv1f32 = call fast float @llvm.vector.reduce.fmax.nxv1f32(<vscale x 1 x float> poison)
- %fmax_nxv4f32 = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> %v2)
- %fmax_nxv4f64 = call fast double @llvm.vector.reduce.fmax.nxv4f64(<vscale x 4 x double> %v3)
-
- ret void
-}
-
-define void @strict_fp_reductions(<vscale x 4 x float> %v0, <vscale x 4 x double> %v1) {
-; CHECK-VSCALE-1-LABEL: 'strict_fp_reductions'
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_nxv4f32 = call float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v0)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_nxv4f64 = call double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v1)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of Invalid for: %fmul_nxv4f32 = call float @llvm.vector.reduce.fmul.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v0)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of Invalid for: %fmul_nxv4f64 = call double @llvm.vector.reduce.fmul.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v1)
-; CHECK-VSCALE-1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
-;
-; CHECK-VSCALE-2-LABEL: 'strict_fp_reductions'
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:8 Lat:24 SizeLat:8 for: %fadd_nxv4f32 = call float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v0)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:8 Lat:24 SizeLat:8 for: %fadd_nxv4f64 = call double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v1)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of Invalid for: %fmul_nxv4f32 = call float @llvm.vector.reduce.fmul.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v0)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of Invalid for: %fmul_nxv4f64 = call double @llvm.vector.reduce.fmul.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v1)
-; CHECK-VSCALE-2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
-;
-; TYPE_BASED_ONLY-LABEL: 'strict_fp_reductions'
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_nxv4f32 = call float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v0)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_nxv4f64 = call double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v1)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of Invalid for: %fmul_nxv4f32 = call float @llvm.vector.reduce.fmul.nxv4f32(float 0.000000e+00, <vscale x 4 x float> %v0)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of Invalid for: %fmul_nxv4f64 = call double @llvm.vector.reduce.fmul.nxv4f64(double 0.000000e+00, <vscale x 4 x double> %v1)
-; TYPE_BASED_ONLY-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
-;
- %fadd_nxv4f32 = call float @llvm.vector.reduce.fadd.nxv4f32(float 0.0, <vscale x 4 x float> %v0)
- %fadd_nxv4f64 = call double @llvm.vector.reduce.fadd.nxv4f64(double 0.0, <vscale x 4 x double> %v1)
- %fmul_nxv4f32 = call float @llvm.vector.reduce.fmul.nxv4f32(float 0.0, <vscale x 4 x float> %v0)
- %fmul_nxv4f64 = call double @llvm.vector.reduce.fmul.nxv4f64(double 0.0, <vscale x 4 x double> %v1)
-
ret void
}
@@ -355,15 +288,6 @@ declare i32 @llvm.vector.reduce.umax.nxv4i32(<vscale x 4 x i32>)
declare i64 @llvm.vector.reduce.umax.nxv4i64(<vscale x 4 x i64>)
declare i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32>)
declare i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64>)
-declare float @llvm.vector.reduce.fadd.nxv4f32(float, <vscale x 4 x float>)
-declare double @llvm.vector.reduce.fadd.nxv4f64(double, <vscale x 4 x double>)
-declare float @llvm.vector.reduce.fmul.nxv4f32(float, <vscale x 4 x float>)
-declare double @llvm.vector.reduce.fmul.nxv4f64(double, <vscale x 4 x double>)
-declare float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float>)
-declare double @llvm.vector.reduce.fmin.nxv4f64(<vscale x 4 x double>)
-declare float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float>)
-declare double @llvm.vector.reduce.fmax.nxv4f64(<vscale x 4 x double>)
-
define void @count_zeroes(<vscale x 4 x i32> %A) {
; CHECK-VSCALE-1-LABEL: 'count_zeroes'
diff --git a/llvm/test/Analysis/CostModel/AArch64/sve-vector-reduce-fp.ll b/llvm/test/Analysis/CostModel/AArch64/sve-vector-reduce-fp.ll
new file mode 100644
index 0000000000000..adb5895804716
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/AArch64/sve-vector-reduce-fp.ll
@@ -0,0 +1,1137 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output -S -mattr=+sve | FileCheck %s --check-prefixes=CHECK,NOBF16
+; RUN: opt < %s -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output -S -mattr=+sve,+bf16 | FileCheck %s --check-prefixes=CHECK,BF16
+; RUN: opt < %s -passes="print<cost-model>" -cost-kind=all 2>&1 -intrinsic-cost-strategy=type-based-intrinsic-cost -disable-output -S -mattr=+sve | FileCheck %s --check-prefix=TYPE_BASED
+
+target triple = "aarch64-unknown-linux-gnu"
+
+define void @reductions() {
+; CHECK-LABEL: 'reductions'
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f16 = call fast half @llvm.vector.reduce.fadd.nxv1f16(half poison, <vscale x 1 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv2f16 = call fast half @llvm.vector.reduce.fadd.nxv2f16(half poison, <vscale x 2 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f16 = call fast half @llvm.vector.reduce.fadd.nxv4f16(half poison, <vscale x 4 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv8f16 = call fast half @llvm.vector.reduce.fadd.nxv8f16(half poison, <vscale x 8 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv1616 = call fast half @llvm.vector.reduce.fadd.nxv16f16(half poison, <vscale x 16 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:5 Lat:11 SizeLat:5 for: %fadd_nxv3216 = call fast half @llvm.vector.reduce.fadd.nxv32f16(half poison, <vscale x 32 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f32 = call fast float @llvm.vector.reduce.fadd.nxv1f32(float poison, <vscale x 1 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv2f32 = call fast float @llvm.vector.reduce.fadd.nxv2f32(float poison, <vscale x 2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f32 = call fast float @llvm.vector.reduce.fadd.nxv4f32(float poison, <vscale x 4 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv8f32 = call fast float @llvm.vector.reduce.fadd.nxv8f32(float poison, <vscale x 8 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:5 Lat:11 SizeLat:5 for: %fadd_nxv16f32 = call fast float @llvm.vector.reduce.fadd.nxv16f32(float poison, <vscale x 16 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f64 = call fast double @llvm.vector.reduce.fadd.nxv1f64(double poison, <vscale x 1 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv2f64 = call fast double @llvm.vector.reduce.fadd.nxv2f64(double poison, <vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv4f64 = call fast double @llvm.vector.reduce.fadd.nxv4f64(double poison, <vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:5 Lat:11 SizeLat:5 for: %fadd_nxv8f64 = call fast double @llvm.vector.reduce.fadd.nxv8f64(double poison, <vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f16_identity = call fast half @llvm.vector.reduce.fadd.nxv1f16(half 0.000000e+00, <vscale x 1 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv2f16_identity = call fast half @llvm.vector.reduce.fadd.nxv2f16(half 0.000000e+00, <vscale x 2 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f16_identity = call fast half @llvm.vector.reduce.fadd.nxv4f16(half 0.000000e+00, <vscale x 4 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv8f16_identity = call fast half @llvm.vector.reduce.fadd.nxv8f16(half 0.000000e+00, <vscale x 8 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv1616_identity = call fast half @llvm.vector.reduce.fadd.nxv16f16(half 0.000000e+00, <vscale x 16 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:5 Lat:11 SizeLat:5 for: %fadd_nxv3216_identity = call fast half @llvm.vector.reduce.fadd.nxv32f16(half 0.000000e+00, <vscale x 32 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f32_identity = call fast float @llvm.vector.reduce.fadd.nxv1f32(float 0.000000e+00, <vscale x 1 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv2f32_identity = call fast float @llvm.vector.reduce.fadd.nxv2f32(float 0.000000e+00, <vscale x 2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv4f32_identity = call fast float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv8f32_identity = call fast float @llvm.vector.reduce.fadd.nxv8f32(float 0.000000e+00, <vscale x 8 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:5 Lat:11 SizeLat:5 for: %fadd_nxv16f32_identity = call fast float @llvm.vector.reduce.fadd.nxv16f32(float 0.000000e+00, <vscale x 16 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fadd_nxv1f64_identity = call fast double @llvm.vector.reduce.fadd.nxv1f64(double 0.000000e+00, <vscale x 1 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_nxv2f64_identity = call fast double @llvm.vector.reduce.fadd.nxv2f64(double 0.000000e+00, <vscale x 2 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:5 SizeLat:3 for: %fadd_nxv4f64_identity = call fast double @llvm.vector.reduce.fadd.nxv4f64(double 0.000000e+00, <vscale x 4 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of RThru:5 CodeSize:5 Lat:11 SizeLat:5 for: %fadd_nxv8f64_identity = call fast double @llvm.vector.reduce.fadd.nxv8f64(double 0.000000e+00, <vscale x 8 x double> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fmax_nxv1f16 = call fast half @llvm.vector.reduce.fmax.nxv1f16(<vscale x 1 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv2f16 = call fast half @llvm.vector.reduce.fmax.nxv2f16(<vscale x 2 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv4f16 = call fast half @llvm.vector.reduce.fmax.nxv4f16(<vscale x 4 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv8f16 = call fast half @llvm.vector.reduce.fmax.nxv8f16(<vscale x 8 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 3 for: %fmax_nxv16f16 = call fast half @llvm.vector.reduce.fmax.nxv16f16(<vscale x 16 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of 5 for: %fmax_nxv32f16 = call fast half @llvm.vector.reduce.fmax.nxv32f16(<vscale x 32 x half> poison)
+; CHECK-NEXT: Cost Model: Found costs of Invalid for: %fmax_nxv1f32 = call fast float @llvm.vector.reduce.fmax.nxv1f32(<vscale x 1 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv2f32 = call fast float @llvm.vector.reduce.fmax.nxv2f32(<vscale x 2 x float> poison)
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %fmax_nxv4f32 = call fast float @llvm.vector.reduce.fmax...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/202569
More information about the llvm-commits
mailing list