Skip to content

Commit 7a6945f

Browse files
sushgokhsushgokh
authored
[AArch64][SLP] Add NFC test cases for floating point reductions (#106507)
A successive patch would be added to fix some of the tests. Pull request: #106507
1 parent 04d3a6b commit 7a6945f

File tree

2 files changed

+880
-0
lines changed

2 files changed

+880
-0
lines changed

‎llvm/test/Analysis/CostModel/AArch64/reduce-fadd.ll

Lines changed: 42 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -7,8 +7,11 @@ target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
77

88
define void @strict_fp_reductions() {
99
; CHECK-LABEL: 'strict_fp_reductions'
10+
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
1011
; CHECK-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
1112
; CHECK-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
13+
; CHECK-NEXT: Cost Model: Found an estimated cost of 76 for instruction: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
14+
; CHECK-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
1215
; CHECK-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
1316
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
1417
; CHECK-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
@@ -18,8 +21,11 @@ define void @strict_fp_reductions() {
1821
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
1922
;
2023
; FP16-LABEL: 'strict_fp_reductions'
24+
; FP16-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
2125
; FP16-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
2226
; FP16-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
27+
; FP16-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
28+
; FP16-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
2329
; FP16-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
2430
; FP16-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
2531
; FP16-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
@@ -29,8 +35,11 @@ define void @strict_fp_reductions() {
2935
; FP16-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
3036
;
3137
; BF16-LABEL: 'strict_fp_reductions'
38+
; BF16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
3239
; BF16-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
3340
; BF16-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
41+
; BF16-NEXT: Cost Model: Found an estimated cost of 76 for instruction: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
42+
; BF16-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
3443
; BF16-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
3544
; BF16-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
3645
; BF16-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
@@ -39,8 +48,11 @@ define void @strict_fp_reductions() {
3948
; BF16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
4049
; BF16-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
4150
;
51+
%fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)
4252
%fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)
4353
%fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)
54+
%fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)
55+
%fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)
4456
%fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)
4557
%fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)
4658
%fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)
@@ -54,12 +66,18 @@ define void @strict_fp_reductions() {
5466

5567
define void @fast_fp_reductions() {
5668
; CHECK-LABEL: 'fast_fp_reductions'
69+
; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
70+
; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
5771
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
5872
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
5973
; CHECK-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
6074
; CHECK-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
75+
; CHECK-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
76+
; CHECK-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
6177
; CHECK-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)
6278
; CHECK-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)
79+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
80+
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
6381
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
6482
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
6583
; CHECK-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
@@ -77,12 +95,18 @@ define void @fast_fp_reductions() {
7795
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
7896
;
7997
; FP16-LABEL: 'fast_fp_reductions'
98+
; FP16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
99+
; FP16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
80100
; FP16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
81101
; FP16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
82102
; FP16-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
83103
; FP16-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
104+
; FP16-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
105+
; FP16-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
84106
; FP16-NEXT: Cost Model: Found an estimated cost of 35 for instruction: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)
85107
; FP16-NEXT: Cost Model: Found an estimated cost of 35 for instruction: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)
108+
; FP16-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
109+
; FP16-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
86110
; FP16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
87111
; FP16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
88112
; FP16-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
@@ -100,12 +124,18 @@ define void @fast_fp_reductions() {
100124
; FP16-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
101125
;
102126
; BF16-LABEL: 'fast_fp_reductions'
127+
; BF16-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
128+
; BF16-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
103129
; BF16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
104130
; BF16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
105131
; BF16-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
106132
; BF16-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
133+
; BF16-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
134+
; BF16-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
107135
; BF16-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)
108136
; BF16-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)
137+
; BF16-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
138+
; BF16-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
109139
; BF16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
110140
; BF16-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
111141
; BF16-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
@@ -122,15 +152,24 @@ define void @fast_fp_reductions() {
122152
; BF16-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
123153
; BF16-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
124154
;
155+
%fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)
156+
%fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)
157+
125158
%fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)
126159
%fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)
127160

128161
%fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)
129162
%fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)
130163

164+
%fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)
165+
%fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)
166+
131167
%fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0.0, <11 x half> undef)
132168
%fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0.0, <13 x half> undef)
133169

170+
%fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)
171+
%fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)
172+
134173
%fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)
135174
%fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)
136175

@@ -158,11 +197,14 @@ define void @fast_fp_reductions() {
158197
declare bfloat @llvm.vector.reduce.fadd.v4f8(bfloat, <4 x bfloat>)
159198
declare fp128 @llvm.vector.reduce.fadd.v4f128(fp128, <4 x fp128>)
160199

200+
declare half @llvm.vector.reduce.fadd.v2f16(half, <2 x half>)
161201
declare half @llvm.vector.reduce.fadd.v4f16(half, <4 x half>)
162202
declare half @llvm.vector.reduce.fadd.v8f16(half, <8 x half>)
203+
declare half @llvm.vector.reduce.fadd.v16f16(half, <16 x half>)
163204
declare half @llvm.vector.reduce.fadd.v11f16(half, <11 x half>)
164205
declare half @llvm.vector.reduce.fadd.v13f16(half, <13 x half>)
165206

207+
declare float @llvm.vector.reduce.fadd.v2f32(float, <2 x float>)
166208
declare float @llvm.vector.reduce.fadd.v4f32(float, <4 x float>)
167209
declare float @llvm.vector.reduce.fadd.v8f32(float, <8 x float>)
168210
declare float @llvm.vector.reduce.fadd.v13f32(float, <13 x float>)

0 commit comments

Comments
 (0)