1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
2 ; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=amdgpu-simplifylib %s | FileCheck %s
4 target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-p7:160:256:256:32-p8:128:128-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-G1-ni:7:8:9"
6 declare float @_Z4fminff(float, float)
7 declare <2 x float> @_Z4fminDv2_fS_(<2 x float>, <2 x float>)
8 declare <3 x float> @_Z4fminDv3_fS_(<3 x float>, <3 x float>)
9 declare <4 x float> @_Z4fminDv4_fS_(<4 x float>, <4 x float>)
10 declare <8 x float> @_Z4fminDv8_fS_(<8 x float>, <8 x float>)
11 declare <16 x float> @_Z4fminDv16_fS_(<16 x float>, <16 x float>)
12 declare double @_Z4fmindd(double, double)
13 declare <2 x double> @_Z4fminDv2_dS_(<2 x double>, <2 x double>)
14 declare <3 x double> @_Z4fminDv3_dS_(<3 x double>, <3 x double>)
15 declare <4 x double> @_Z4fminDv4_dS_(<4 x double>, <4 x double>)
16 declare <8 x double> @_Z4fminDv8_dS_(<8 x double>, <8 x double>)
17 declare <16 x double> @_Z4fminDv16_dS_(<16 x double>, <16 x double>)
18 declare half @_Z4fminDhDh(half, half)
19 declare <2 x half> @_Z4fminDv2_DhS_(<2 x half>, <2 x half>)
20 declare <3 x half> @_Z4fminDv3_DhS_(<3 x half>, <3 x half>)
21 declare <4 x half> @_Z4fminDv4_DhS_(<4 x half>, <4 x half>)
22 declare <8 x half> @_Z4fminDv8_DhS_(<8 x half>, <8 x half>)
23 declare <16 x half> @_Z4fminDv16_DhS_(<16 x half>, <16 x half>)
25 define float @test_fmin_f32(float %x, float %y) {
26 ; CHECK-LABEL: define float @test_fmin_f32
27 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) {
28 ; CHECK-NEXT: [[FMIN:%.*]] = tail call float @llvm.minnum.f32(float [[X]], float [[Y]])
29 ; CHECK-NEXT: ret float [[FMIN]]
31 %fmin = tail call float @_Z4fminff(float %x, float %y)
35 define float @test_fmin_f32_nnan(float %x, float %y) {
36 ; CHECK-LABEL: define float @test_fmin_f32_nnan
37 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) {
38 ; CHECK-NEXT: [[FMIN:%.*]] = tail call nnan float @llvm.minnum.f32(float [[X]], float [[Y]])
39 ; CHECK-NEXT: ret float [[FMIN]]
41 %fmin = tail call nnan float @_Z4fminff(float %x, float %y)
45 define <2 x float> @test_fmin_v2f32(<2 x float> %x, <2 x float> %y) {
46 ; CHECK-LABEL: define <2 x float> @test_fmin_v2f32
47 ; CHECK-SAME: (<2 x float> [[X:%.*]], <2 x float> [[Y:%.*]]) {
48 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <2 x float> @llvm.minnum.v2f32(<2 x float> [[X]], <2 x float> [[Y]])
49 ; CHECK-NEXT: ret <2 x float> [[FMIN]]
51 %fmin = tail call <2 x float> @_Z4fminDv2_fS_(<2 x float> %x, <2 x float> %y)
55 define <3 x float> @test_fmin_v3f32(<3 x float> %x, <3 x float> %y) {
56 ; CHECK-LABEL: define <3 x float> @test_fmin_v3f32
57 ; CHECK-SAME: (<3 x float> [[X:%.*]], <3 x float> [[Y:%.*]]) {
58 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <3 x float> @llvm.minnum.v3f32(<3 x float> [[X]], <3 x float> [[Y]])
59 ; CHECK-NEXT: ret <3 x float> [[FMIN]]
61 %fmin = tail call <3 x float> @_Z4fminDv3_fS_(<3 x float> %x, <3 x float> %y)
65 define <4 x float> @test_fmin_v4f32(<4 x float> %x, <4 x float> %y) {
66 ; CHECK-LABEL: define <4 x float> @test_fmin_v4f32
67 ; CHECK-SAME: (<4 x float> [[X:%.*]], <4 x float> [[Y:%.*]]) {
68 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <4 x float> @llvm.minnum.v4f32(<4 x float> [[X]], <4 x float> [[Y]])
69 ; CHECK-NEXT: ret <4 x float> [[FMIN]]
71 %fmin = tail call <4 x float> @_Z4fminDv4_fS_(<4 x float> %x, <4 x float> %y)
75 define <8 x float> @test_fmin_v8f32(<8 x float> %x, <8 x float> %y) {
76 ; CHECK-LABEL: define <8 x float> @test_fmin_v8f32
77 ; CHECK-SAME: (<8 x float> [[X:%.*]], <8 x float> [[Y:%.*]]) {
78 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <8 x float> @llvm.minnum.v8f32(<8 x float> [[X]], <8 x float> [[Y]])
79 ; CHECK-NEXT: ret <8 x float> [[FMIN]]
81 %fmin = tail call <8 x float> @_Z4fminDv8_fS_(<8 x float> %x, <8 x float> %y)
85 define <16 x float> @test_fmin_v16f32(<16 x float> %x, <16 x float> %y) {
86 ; CHECK-LABEL: define <16 x float> @test_fmin_v16f32
87 ; CHECK-SAME: (<16 x float> [[X:%.*]], <16 x float> [[Y:%.*]]) {
88 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <16 x float> @llvm.minnum.v16f32(<16 x float> [[X]], <16 x float> [[Y]])
89 ; CHECK-NEXT: ret <16 x float> [[FMIN]]
91 %fmin = tail call <16 x float> @_Z4fminDv16_fS_(<16 x float> %x, <16 x float> %y)
92 ret <16 x float> %fmin
95 define double @test_fmin_f64(double %x, double %y) {
96 ; CHECK-LABEL: define double @test_fmin_f64
97 ; CHECK-SAME: (double [[X:%.*]], double [[Y:%.*]]) {
98 ; CHECK-NEXT: [[FMIN:%.*]] = tail call double @llvm.minnum.f64(double [[X]], double [[Y]])
99 ; CHECK-NEXT: ret double [[FMIN]]
101 %fmin = tail call double @_Z4fmindd(double %x, double %y)
105 define <2 x double> @test_fmin_v2f64(<2 x double> %x, <2 x double> %y) {
106 ; CHECK-LABEL: define <2 x double> @test_fmin_v2f64
107 ; CHECK-SAME: (<2 x double> [[X:%.*]], <2 x double> [[Y:%.*]]) {
108 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <2 x double> @llvm.minnum.v2f64(<2 x double> [[X]], <2 x double> [[Y]])
109 ; CHECK-NEXT: ret <2 x double> [[FMIN]]
111 %fmin = tail call <2 x double> @_Z4fminDv2_dS_(<2 x double> %x, <2 x double> %y)
112 ret <2 x double> %fmin
115 define <3 x double> @test_fmin_v3f64(<3 x double> %x, <3 x double> %y) {
116 ; CHECK-LABEL: define <3 x double> @test_fmin_v3f64
117 ; CHECK-SAME: (<3 x double> [[X:%.*]], <3 x double> [[Y:%.*]]) {
118 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <3 x double> @llvm.minnum.v3f64(<3 x double> [[X]], <3 x double> [[Y]])
119 ; CHECK-NEXT: ret <3 x double> [[FMIN]]
121 %fmin = tail call <3 x double> @_Z4fminDv3_dS_(<3 x double> %x, <3 x double> %y)
122 ret <3 x double> %fmin
125 define <4 x double> @test_fmin_v4f64(<4 x double> %x, <4 x double> %y) {
126 ; CHECK-LABEL: define <4 x double> @test_fmin_v4f64
127 ; CHECK-SAME: (<4 x double> [[X:%.*]], <4 x double> [[Y:%.*]]) {
128 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <4 x double> @llvm.minnum.v4f64(<4 x double> [[X]], <4 x double> [[Y]])
129 ; CHECK-NEXT: ret <4 x double> [[FMIN]]
131 %fmin = tail call <4 x double> @_Z4fminDv4_dS_(<4 x double> %x, <4 x double> %y)
132 ret <4 x double> %fmin
135 define <8 x double> @test_fmin_v8f64(<8 x double> %x, <8 x double> %y) {
136 ; CHECK-LABEL: define <8 x double> @test_fmin_v8f64
137 ; CHECK-SAME: (<8 x double> [[X:%.*]], <8 x double> [[Y:%.*]]) {
138 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <8 x double> @llvm.minnum.v8f64(<8 x double> [[X]], <8 x double> [[Y]])
139 ; CHECK-NEXT: ret <8 x double> [[FMIN]]
141 %fmin = tail call <8 x double> @_Z4fminDv8_dS_(<8 x double> %x, <8 x double> %y)
142 ret <8 x double> %fmin
145 define <16 x double> @test_fmin_v16f64(<16 x double> %x, <16 x double> %y) {
146 ; CHECK-LABEL: define <16 x double> @test_fmin_v16f64
147 ; CHECK-SAME: (<16 x double> [[X:%.*]], <16 x double> [[Y:%.*]]) {
148 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <16 x double> @llvm.minnum.v16f64(<16 x double> [[X]], <16 x double> [[Y]])
149 ; CHECK-NEXT: ret <16 x double> [[FMIN]]
151 %fmin = tail call <16 x double> @_Z4fminDv16_dS_(<16 x double> %x, <16 x double> %y)
152 ret <16 x double> %fmin
155 define half @test_fmin_f16(half %x, half %y) {
156 ; CHECK-LABEL: define half @test_fmin_f16
157 ; CHECK-SAME: (half [[X:%.*]], half [[Y:%.*]]) {
158 ; CHECK-NEXT: [[FMIN:%.*]] = tail call half @llvm.minnum.f16(half [[X]], half [[Y]])
159 ; CHECK-NEXT: ret half [[FMIN]]
161 %fmin = tail call half @_Z4fminDhDh(half %x, half %y)
165 define <2 x half> @test_fmin_v2f16(<2 x half> %x, <2 x half> %y) {
166 ; CHECK-LABEL: define <2 x half> @test_fmin_v2f16
167 ; CHECK-SAME: (<2 x half> [[X:%.*]], <2 x half> [[Y:%.*]]) {
168 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <2 x half> @llvm.minnum.v2f16(<2 x half> [[X]], <2 x half> [[Y]])
169 ; CHECK-NEXT: ret <2 x half> [[FMIN]]
171 %fmin = tail call <2 x half> @_Z4fminDv2_DhS_(<2 x half> %x, <2 x half> %y)
175 define <3 x half> @test_fmin_v3f16(<3 x half> %x, <3 x half> %y) {
176 ; CHECK-LABEL: define <3 x half> @test_fmin_v3f16
177 ; CHECK-SAME: (<3 x half> [[X:%.*]], <3 x half> [[Y:%.*]]) {
178 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <3 x half> @llvm.minnum.v3f16(<3 x half> [[X]], <3 x half> [[Y]])
179 ; CHECK-NEXT: ret <3 x half> [[FMIN]]
181 %fmin = tail call <3 x half> @_Z4fminDv3_DhS_(<3 x half> %x, <3 x half> %y)
185 define <4 x half> @test_fmin_v4f16(<4 x half> %x, <4 x half> %y) {
186 ; CHECK-LABEL: define <4 x half> @test_fmin_v4f16
187 ; CHECK-SAME: (<4 x half> [[X:%.*]], <4 x half> [[Y:%.*]]) {
188 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <4 x half> @llvm.minnum.v4f16(<4 x half> [[X]], <4 x half> [[Y]])
189 ; CHECK-NEXT: ret <4 x half> [[FMIN]]
191 %fmin = tail call <4 x half> @_Z4fminDv4_DhS_(<4 x half> %x, <4 x half> %y)
195 define <8 x half> @test_fmin_v8f16(<8 x half> %x, <8 x half> %y) {
196 ; CHECK-LABEL: define <8 x half> @test_fmin_v8f16
197 ; CHECK-SAME: (<8 x half> [[X:%.*]], <8 x half> [[Y:%.*]]) {
198 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <8 x half> @llvm.minnum.v8f16(<8 x half> [[X]], <8 x half> [[Y]])
199 ; CHECK-NEXT: ret <8 x half> [[FMIN]]
201 %fmin = tail call <8 x half> @_Z4fminDv8_DhS_(<8 x half> %x, <8 x half> %y)
205 define <16 x half> @test_fmin_v16f16(<16 x half> %x, <16 x half> %y) {
206 ; CHECK-LABEL: define <16 x half> @test_fmin_v16f16
207 ; CHECK-SAME: (<16 x half> [[X:%.*]], <16 x half> [[Y:%.*]]) {
208 ; CHECK-NEXT: [[FMIN:%.*]] = tail call <16 x half> @llvm.minnum.v16f16(<16 x half> [[X]], <16 x half> [[Y]])
209 ; CHECK-NEXT: ret <16 x half> [[FMIN]]
211 %fmin = tail call <16 x half> @_Z4fminDv16_DhS_(<16 x half> %x, <16 x half> %y)
212 ret <16 x half> %fmin
215 define float @test_fmin_f32_minsize(float %x, float %y) #0 {
216 ; CHECK-LABEL: define float @test_fmin_f32_minsize
217 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) #[[ATTR0:[0-9]+]] {
218 ; CHECK-NEXT: [[FMIN:%.*]] = tail call float @llvm.minnum.f32(float [[X]], float [[Y]])
219 ; CHECK-NEXT: ret float [[FMIN]]
221 %fmin = tail call float @_Z4fminff(float %x, float %y)
225 define float @test_fmin_f32_nnan_minsize(float %x, float %y) #0 {
226 ; CHECK-LABEL: define float @test_fmin_f32_nnan_minsize
227 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) #[[ATTR0]] {
228 ; CHECK-NEXT: [[FMIN:%.*]] = tail call nnan float @llvm.minnum.f32(float [[X]], float [[Y]])
229 ; CHECK-NEXT: ret float [[FMIN]]
231 %fmin = tail call nnan float @_Z4fminff(float %x, float %y)
235 define float @test_fmin_f32_noinline(float %x, float %y) {
236 ; CHECK-LABEL: define float @test_fmin_f32_noinline
237 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) {
238 ; CHECK-NEXT: [[FMIN:%.*]] = tail call float @_Z4fminff(float [[X]], float [[Y]]) #[[ATTR3:[0-9]+]]
239 ; CHECK-NEXT: ret float [[FMIN]]
241 %fmin = tail call float @_Z4fminff(float %x, float %y) #1
245 define float @test_fmin_f32_nnan_noinline(float %x, float %y) {
246 ; CHECK-LABEL: define float @test_fmin_f32_nnan_noinline
247 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) {
248 ; CHECK-NEXT: [[FMIN:%.*]] = tail call nnan float @_Z4fminff(float [[X]], float [[Y]]) #[[ATTR3]]
249 ; CHECK-NEXT: ret float [[FMIN]]
251 %fmin = tail call nnan float @_Z4fminff(float %x, float %y) #1
255 define float @test_fmin_f32_strictfp(float %x, float %y) #2 {
256 ; CHECK-LABEL: define float @test_fmin_f32_strictfp
257 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) #[[ATTR1:[0-9]+]] {
258 ; CHECK-NEXT: [[FMIN:%.*]] = tail call nnan nsz float @_Z4fminff(float [[X]], float [[Y]]) #[[ATTR1]]
259 ; CHECK-NEXT: ret float [[FMIN]]
261 %fmin = tail call nsz nnan float @_Z4fminff(float %x, float %y) #2
265 define float @test_fmin_f32_fast_nobuiltin(float %x, float %y) {
266 ; CHECK-LABEL: define float @test_fmin_f32_fast_nobuiltin
267 ; CHECK-SAME: (float [[X:%.*]], float [[Y:%.*]]) {
268 ; CHECK-NEXT: [[FMIN:%.*]] = tail call fast float @_Z4fminff(float [[X]], float [[Y]]) #[[ATTR4:[0-9]+]]
269 ; CHECK-NEXT: ret float [[FMIN]]
271 %fmin = tail call fast float @_Z4fminff(float %x, float %y) #3
275 attributes #0 = { minsize }
276 attributes #1 = { noinline }
277 attributes #2 = { strictfp }
278 attributes #3 = { nobuiltin }