OLD | NEW |
1 ; This file checks that Subzero generates code in accordance with the | 1 ; This file checks that Subzero generates code in accordance with the |
2 ; calling convention for vectors. | 2 ; calling convention for vectors. |
3 | 3 |
4 ; RUN: %llvm2ice -O2 --verbose none %s | FileCheck %s | 4 ; TODO(jvoung): remove -ffunction-sections. Sadly... diassembling some |
5 ; RUN: %llvm2ice -Om1 --verbose none %s | FileCheck --check-prefix=OPTM1 %s | 5 ; "ret" within a section are broken in llvm-objdump for LLVM 3.4, and 3.5. |
6 ; RUN: %llvm2ice -O2 --verbose none %s \ | 6 ; It was only fixed in 3.6. |
7 ; RUN: | llvm-mc -triple=i686-none-nacl -x86-asm-syntax=intel -filetype=obj | 7 |
8 ; RUN: %llvm2ice -Om1 --verbose none %s \ | 8 ; RUN: %llvm2ice -O2 --verbose none -ffunction-sections %s \ |
9 ; RUN: | llvm-mc -triple=i686-none-nacl -x86-asm-syntax=intel -filetype=obj | 9 ; RUN: | llvm-mc -triple=i686-none-nacl -x86-asm-syntax=intel -filetype=obj \ |
| 10 ; RUN: | llvm-objdump -d --symbolize -x86-asm-syntax=intel - | FileCheck %s |
| 11 ; RUN: %llvm2ice -Om1 --verbose none -ffunction-sections %s \ |
| 12 ; RUN: | llvm-mc -triple=i686-none-nacl -x86-asm-syntax=intel -filetype=obj \ |
| 13 ; RUN: | llvm-objdump -d --symbolize -x86-asm-syntax=intel - \ |
| 14 ; RUN: | FileCheck --check-prefix=OPTM1 %s |
10 ; RUN: %llvm2ice --verbose none %s | FileCheck --check-prefix=ERRORS %s | 15 ; RUN: %llvm2ice --verbose none %s | FileCheck --check-prefix=ERRORS %s |
11 ; RUN: %llvm2iceinsts %s | %szdiff %s | FileCheck --check-prefix=DUMP %s | 16 ; RUN: %llvm2iceinsts %s | %szdiff %s | FileCheck --check-prefix=DUMP %s |
12 ; RUN: %llvm2iceinsts --pnacl %s | %szdiff %s \ | 17 ; RUN: %llvm2iceinsts --pnacl %s | %szdiff %s \ |
13 ; RUN: | FileCheck --check-prefix=DUMP %s | 18 ; RUN: | FileCheck --check-prefix=DUMP %s |
14 | 19 |
15 ; The first five functions test that vectors are moved from their | 20 ; The first five functions test that vectors are moved from their |
16 ; correct argument location to xmm0. | 21 ; correct argument location to xmm0. |
17 | 22 |
18 define <4 x float> @test_returning_arg0(<4 x float> %arg0, <4 x float> %arg1, <4
x float> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5) { | 23 define <4 x float> @test_returning_arg0(<4 x float> %arg0, <4 x float> %arg1, <4
x float> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5) { |
19 entry: | 24 entry: |
(...skipping 44 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
64 ; OPTM1-LABEL: test_returning_arg3: | 69 ; OPTM1-LABEL: test_returning_arg3: |
65 ; OPTM1: movups xmmword ptr [[LOC:.*]], xmm3 | 70 ; OPTM1: movups xmmword ptr [[LOC:.*]], xmm3 |
66 ; OPTM1: movups xmm0, xmmword ptr [[LOC]] | 71 ; OPTM1: movups xmm0, xmmword ptr [[LOC]] |
67 ; OPTM1: ret | 72 ; OPTM1: ret |
68 } | 73 } |
69 | 74 |
70 define <4 x float> @test_returning_arg4(<4 x float> %arg0, <4 x float> %arg1, <4
x float> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5) { | 75 define <4 x float> @test_returning_arg4(<4 x float> %arg0, <4 x float> %arg1, <4
x float> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5) { |
71 entry: | 76 entry: |
72 ret <4 x float> %arg4 | 77 ret <4 x float> %arg4 |
73 ; CHECK-LABEL: test_returning_arg4: | 78 ; CHECK-LABEL: test_returning_arg4: |
74 ; CHECK: movups xmm0, xmmword ptr [esp+4] | 79 ; CHECK: movups xmm0, xmmword ptr [esp + 4] |
75 ; CHECK: ret | 80 ; CHECK: ret |
76 | 81 |
77 ; OPTM1-LABEL: test_returning_arg4: | 82 ; OPTM1-LABEL: test_returning_arg4: |
78 ; OPTM1: movups xmm0, xmmword ptr {{.*}} | 83 ; OPTM1: movups xmm0, xmmword ptr {{.*}} |
79 ; OPTM1: ret | 84 ; OPTM1: ret |
80 } | 85 } |
81 | 86 |
82 ; The next five functions check that xmm arguments are handled | 87 ; The next five functions check that xmm arguments are handled |
83 ; correctly when interspersed with stack arguments in the argument | 88 ; correctly when interspersed with stack arguments in the argument |
84 ; list. | 89 ; list. |
(...skipping 47 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
132 ; OPTM1-LABEL: test_returning_interspersed_arg3: | 137 ; OPTM1-LABEL: test_returning_interspersed_arg3: |
133 ; OPTM1: movups xmmword ptr [[LOC:.*]], xmm3 | 138 ; OPTM1: movups xmmword ptr [[LOC:.*]], xmm3 |
134 ; OPTM1: movups xmm0, xmmword ptr [[LOC]] | 139 ; OPTM1: movups xmm0, xmmword ptr [[LOC]] |
135 ; OPTM1: ret | 140 ; OPTM1: ret |
136 } | 141 } |
137 | 142 |
138 define <4 x float> @test_returning_interspersed_arg4(i32 %i32arg0, double %doubl
earg0, <4 x float> %arg0, <4 x float> %arg1, i32 %i32arg1, <4 x float> %arg2, do
uble %doublearg1, <4 x float> %arg3, i32 %i32arg2, double %doublearg2, float %fl
oatarg0, <4 x float> %arg4, <4 x float> %arg5, float %floatarg1) { | 143 define <4 x float> @test_returning_interspersed_arg4(i32 %i32arg0, double %doubl
earg0, <4 x float> %arg0, <4 x float> %arg1, i32 %i32arg1, <4 x float> %arg2, do
uble %doublearg1, <4 x float> %arg3, i32 %i32arg2, double %doublearg2, float %fl
oatarg0, <4 x float> %arg4, <4 x float> %arg5, float %floatarg1) { |
139 entry: | 144 entry: |
140 ret <4 x float> %arg4 | 145 ret <4 x float> %arg4 |
141 ; CHECK-LABEL: test_returning_interspersed_arg4: | 146 ; CHECK-LABEL: test_returning_interspersed_arg4: |
142 ; CHECK: movups xmm0, xmmword ptr [esp+52] | 147 ; CHECK: movups xmm0, xmmword ptr [esp + 52] |
143 ; CHECK: ret | 148 ; CHECK: ret |
144 | 149 |
145 ; OPTM1-LABEL: test_returning_interspersed_arg4: | 150 ; OPTM1-LABEL: test_returning_interspersed_arg4: |
146 ; OPTM1: movups xmm0, xmmword ptr {{.*}} | 151 ; OPTM1: movups xmm0, xmmword ptr {{.*}} |
147 ; OPTM1: ret | 152 ; OPTM1: ret |
148 } | 153 } |
149 | 154 |
150 ; Test that vectors are passed correctly as arguments to a function. | 155 ; Test that vectors are passed correctly as arguments to a function. |
151 | 156 |
152 declare void @VectorArgs(<4 x float>, <4 x float>, <4 x float>, <4 x float>, <4
x float>, <4 x float>) | 157 declare void @VectorArgs(<4 x float>, <4 x float>, <4 x float>, <4 x float>, <4
x float>, <4 x float>) |
153 | 158 |
154 declare void @killXmmRegisters() | 159 declare void @killXmmRegisters() |
155 | 160 |
156 define void @test_passing_vectors(<4 x float> %arg0, <4 x float> %arg1, <4 x flo
at> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5, <4 x float>
%arg6, <4 x float> %arg7, <4 x float> %arg8, <4 x float> %arg9) { | 161 define void @test_passing_vectors(<4 x float> %arg0, <4 x float> %arg1, <4 x flo
at> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5, <4 x float>
%arg6, <4 x float> %arg7, <4 x float> %arg8, <4 x float> %arg9) { |
157 entry: | 162 entry: |
158 ; Kills XMM registers so that no in-arg lowering code interferes | 163 ; Kills XMM registers so that no in-arg lowering code interferes |
159 ; with the test. | 164 ; with the test. |
160 call void @killXmmRegisters() | 165 call void @killXmmRegisters() |
161 call void @VectorArgs(<4 x float> %arg9, <4 x float> %arg8, <4 x float> %arg7,
<4 x float> %arg6, <4 x float> %arg5, <4 x float> %arg4) | 166 call void @VectorArgs(<4 x float> %arg9, <4 x float> %arg8, <4 x float> %arg7,
<4 x float> %arg6, <4 x float> %arg5, <4 x float> %arg4) |
162 ret void | 167 ret void |
163 ; CHECK-LABEL: test_passing_vectors: | 168 ; CHECK-LABEL: test_passing_vectors: |
164 ; CHECK: sub esp, 32 | 169 ; CHECK: sub esp, 32 |
165 ; CHECK: movups [[ARG5:.*]], xmmword ptr [esp+64] | 170 ; CHECK: movups [[ARG5:.*]], xmmword ptr [esp + 64] |
166 ; CHECK: movups xmmword ptr [esp], [[ARG5]] | 171 ; CHECK: movups xmmword ptr [esp], [[ARG5]] |
167 ; CHECK: movups [[ARG6:.*]], xmmword ptr [esp+48] | 172 ; CHECK: movups [[ARG6:.*]], xmmword ptr [esp + 48] |
168 ; CHECK: movups xmmword ptr [esp+16], [[ARG6]] | 173 ; CHECK: movups xmmword ptr [esp + 16], [[ARG6]] |
169 ; CHECK: movups xmm0, xmmword ptr [esp+128] | 174 ; CHECK: movups xmm0, xmmword ptr [esp + 128] |
170 ; CHECK: movups xmm1, xmmword ptr [esp+112] | 175 ; CHECK: movups xmm1, xmmword ptr [esp + 112] |
171 ; CHECK: movups xmm2, xmmword ptr [esp+96] | 176 ; CHECK: movups xmm2, xmmword ptr [esp + 96] |
172 ; CHECK: movups xmm3, xmmword ptr [esp+80] | 177 ; CHECK: movups xmm3, xmmword ptr [esp + 80] |
173 ; CHECK: call VectorArgs | 178 ; CHECK: call -4 |
174 ; CHECK-NEXT: add esp, 32 | 179 ; CHECK-NEXT: add esp, 32 |
175 ; CHECK: ret | 180 ; CHECK: ret |
176 | 181 |
177 ; OPTM1-LABEL: test_passing_vectors: | 182 ; OPTM1-LABEL: test_passing_vectors: |
178 ; OPTM1: sub esp, 32 | 183 ; OPTM1: sub esp, 32 |
179 ; OPTM1: movups [[ARG5:.*]], xmmword ptr {{.*}} | 184 ; OPTM1: movups [[ARG5:.*]], xmmword ptr {{.*}} |
180 ; OPTM1: movups xmmword ptr [esp], [[ARG5]] | 185 ; OPTM1: movups xmmword ptr [esp], [[ARG5]] |
181 ; OPTM1: movups [[ARG6:.*]], xmmword ptr {{.*}} | 186 ; OPTM1: movups [[ARG6:.*]], xmmword ptr {{.*}} |
182 ; OPTM1: movups xmmword ptr [esp+16], [[ARG6]] | 187 ; OPTM1: movups xmmword ptr [esp + 16], [[ARG6]] |
183 ; OPTM1: movups xmm0, xmmword ptr {{.*}} | 188 ; OPTM1: movups xmm0, xmmword ptr {{.*}} |
184 ; OPTM1: movups xmm1, xmmword ptr {{.*}} | 189 ; OPTM1: movups xmm1, xmmword ptr {{.*}} |
185 ; OPTM1: movups xmm2, xmmword ptr {{.*}} | 190 ; OPTM1: movups xmm2, xmmword ptr {{.*}} |
186 ; OPTM1: movups xmm3, xmmword ptr {{.*}} | 191 ; OPTM1: movups xmm3, xmmword ptr {{.*}} |
187 ; OPTM1: call VectorArgs | 192 ; OPTM1: call -4 |
188 ; OPTM1-NEXT: add esp, 32 | 193 ; OPTM1-NEXT: add esp, 32 |
189 ; OPTM1: ret | 194 ; OPTM1: ret |
190 } | 195 } |
191 | 196 |
192 declare void @InterspersedVectorArgs(<4 x float>, i64, <4 x float>, i64, <4 x fl
oat>, float, <4 x float>, double, <4 x float>, i32, <4 x float>) | 197 declare void @InterspersedVectorArgs(<4 x float>, i64, <4 x float>, i64, <4 x fl
oat>, float, <4 x float>, double, <4 x float>, i32, <4 x float>) |
193 | 198 |
194 define void @test_passing_vectors_interspersed(<4 x float> %arg0, <4 x float> %a
rg1, <4 x float> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5,
<4 x float> %arg6, <4 x float> %arg7, <4 x float> %arg8, <4 x float> %arg9) { | 199 define void @test_passing_vectors_interspersed(<4 x float> %arg0, <4 x float> %a
rg1, <4 x float> %arg2, <4 x float> %arg3, <4 x float> %arg4, <4 x float> %arg5,
<4 x float> %arg6, <4 x float> %arg7, <4 x float> %arg8, <4 x float> %arg9) { |
195 entry: | 200 entry: |
196 ; Kills XMM registers so that no in-arg lowering code interferes | 201 ; Kills XMM registers so that no in-arg lowering code interferes |
197 ; with the test. | 202 ; with the test. |
198 call void @killXmmRegisters() | 203 call void @killXmmRegisters() |
199 call void @InterspersedVectorArgs(<4 x float> %arg9, i64 0, <4 x float> %arg8,
i64 1, <4 x float> %arg7, float 2.000000e+00, <4 x float> %arg6, double 3.00000
0e+00, <4 x float> %arg5, i32 4, <4 x float> %arg4) | 204 call void @InterspersedVectorArgs(<4 x float> %arg9, i64 0, <4 x float> %arg8,
i64 1, <4 x float> %arg7, float 2.000000e+00, <4 x float> %arg6, double 3.00000
0e+00, <4 x float> %arg5, i32 4, <4 x float> %arg4) |
200 ret void | 205 ret void |
201 ; CHECK-LABEL: test_passing_vectors_interspersed: | 206 ; CHECK-LABEL: test_passing_vectors_interspersed: |
202 ; CHECK: sub esp, 80 | 207 ; CHECK: sub esp, 80 |
203 ; CHECK: movups [[ARG9:.*]], xmmword ptr [esp+112] | 208 ; CHECK: movups [[ARG9:.*]], xmmword ptr [esp + 112] |
204 ; CHECK: movups xmmword ptr [esp+32], [[ARG9]] | 209 ; CHECK: movups xmmword ptr [esp + 32], [[ARG9]] |
205 ; CHECK: movups [[ARG11:.*]], xmmword ptr [esp+96] | 210 ; CHECK: movups [[ARG11:.*]], xmmword ptr [esp + 96] |
206 ; CHECK: movups xmmword ptr [esp+64], [[ARG11]] | 211 ; CHECK: movups xmmword ptr [esp + 64], [[ARG11]] |
207 ; CHECK: movups xmm0, xmmword ptr [esp+176] | 212 ; CHECK: movups xmm0, xmmword ptr [esp + 176] |
208 ; CHECK: movups xmm1, xmmword ptr [esp+160] | 213 ; CHECK: movups xmm1, xmmword ptr [esp + 160] |
209 ; CHECK: movups xmm2, xmmword ptr [esp+144] | 214 ; CHECK: movups xmm2, xmmword ptr [esp + 144] |
210 ; CHECK: movups xmm3, xmmword ptr [esp+128] | 215 ; CHECK: movups xmm3, xmmword ptr [esp + 128] |
211 ; CHECK: call InterspersedVectorArgs | 216 ; CHECK: call -4 |
212 ; CHECK-NEXT: add esp, 80 | 217 ; CHECK-NEXT: add esp, 80 |
213 ; CHECK: ret | 218 ; CHECK: ret |
214 | 219 |
215 ; OPTM1-LABEL: test_passing_vectors_interspersed: | 220 ; OPTM1-LABEL: test_passing_vectors_interspersed: |
216 ; OPTM1: sub esp, 80 | 221 ; OPTM1: sub esp, 80 |
217 ; OPTM1: movups [[ARG9:.*]], xmmword ptr {{.*}} | 222 ; OPTM1: movups [[ARG9:.*]], xmmword ptr {{.*}} |
218 ; OPTM1: movups xmmword ptr [esp+32], [[ARG9]] | 223 ; OPTM1: movups xmmword ptr [esp + 32], [[ARG9]] |
219 ; OPTM1: movups [[ARG11:.*]], xmmword ptr {{.*}} | 224 ; OPTM1: movups [[ARG11:.*]], xmmword ptr {{.*}} |
220 ; OPTM1: movups xmmword ptr [esp+64], [[ARG11]] | 225 ; OPTM1: movups xmmword ptr [esp + 64], [[ARG11]] |
221 ; OPTM1: movups xmm0, xmmword ptr {{.*}} | 226 ; OPTM1: movups xmm0, xmmword ptr {{.*}} |
222 ; OPTM1: movups xmm1, xmmword ptr {{.*}} | 227 ; OPTM1: movups xmm1, xmmword ptr {{.*}} |
223 ; OPTM1: movups xmm2, xmmword ptr {{.*}} | 228 ; OPTM1: movups xmm2, xmmword ptr {{.*}} |
224 ; OPTM1: movups xmm3, xmmword ptr {{.*}} | 229 ; OPTM1: movups xmm3, xmmword ptr {{.*}} |
225 ; OPTM1: call InterspersedVectorArgs | 230 ; OPTM1: call -4 |
226 ; OPTM1-NEXT: add esp, 80 | 231 ; OPTM1-NEXT: add esp, 80 |
227 ; OPTM1: ret | 232 ; OPTM1: ret |
228 } | 233 } |
229 | 234 |
230 ; Test that a vector returned from a function is recognized to be in | 235 ; Test that a vector returned from a function is recognized to be in |
231 ; xmm0. | 236 ; xmm0. |
232 | 237 |
233 declare <4 x float> @VectorReturn(<4 x float> %arg0) | 238 declare <4 x float> @VectorReturn(<4 x float> %arg0) |
234 | 239 |
235 define void @test_receiving_vectors(<4 x float> %arg0) { | 240 define void @test_receiving_vectors(<4 x float> %arg0) { |
236 entry: | 241 entry: |
237 %result = call <4 x float> @VectorReturn(<4 x float> %arg0) | 242 %result = call <4 x float> @VectorReturn(<4 x float> %arg0) |
238 %result2 = call <4 x float> @VectorReturn(<4 x float> %result) | 243 %result2 = call <4 x float> @VectorReturn(<4 x float> %result) |
239 ret void | 244 ret void |
240 ; CHECK-LABEL: test_receiving_vectors: | 245 ; CHECK-LABEL: test_receiving_vectors: |
241 ; CHECK: call VectorReturn | 246 ; CHECK: call -4 |
242 ; CHECK-NOT: movups xmm0 | 247 ; CHECK-NOT: movups xmm0 |
243 ; CHECK: call VectorReturn | 248 ; CHECK: call -4 |
244 ; CHECK: ret | 249 ; CHECK: ret |
245 | 250 |
246 ; OPTM1-LABEL: test_receiving_vectors: | 251 ; OPTM1-LABEL: test_receiving_vectors: |
247 ; OPTM1: call VectorReturn | 252 ; OPTM1: call -4 |
248 ; OPTM1: movups {{.*}}, xmm0 | 253 ; OPTM1: movups {{.*}}, xmm0 |
249 ; OPTM1: movups xmm0, {{.*}} | 254 ; OPTM1: movups xmm0, {{.*}} |
250 ; OPTM1: call VectorReturn | 255 ; OPTM1: call -4 |
251 ; OPTM1: ret | 256 ; OPTM1: ret |
252 } | 257 } |
253 | 258 |
254 ; ERRORS-NOT: ICE translation error | 259 ; ERRORS-NOT: ICE translation error |
255 ; DUMP-NOT: SZ | 260 ; DUMP-NOT: SZ |
OLD | NEW |