Skip to content

Commit cd6844c

Browse files
lei137lei137
authored
[PowerPC][NFC] autogen mma tc checks via update_cc_test_checks (#108584)
Update mma tests in prep for changes needed in a followup patch for #107229. Checks for ``clang/test/CodeGen/PowerPC/builtins-ppc-pair-mma-types.c`` seem to have been manually upated to rename temp variables even though it says checks was auto generated. Regenerate via script. Add noopt checks for ``clang/test/CodeGen/PowerPC/builtins-ppc-build-pair-mma.c``.
1 parent 82987bd commit cd6844c

File tree

2 files changed

+170
-102
lines changed

2 files changed

+170
-102
lines changed

‎clang/test/CodeGen/PowerPC/builtins-ppc-build-pair-mma.c

Lines changed: 68 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3,6 +3,8 @@
33
// RUN: -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK-LE
44
// RUN: %clang_cc1 -O3 -triple powerpc64-unknown-unknown -target-cpu pwr10 \
55
// RUN: -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK-BE
6+
// RUN: %clang_cc1 -O0 -triple powerpc64le-unknown-unknown -target-cpu pwr10 \
7+
// RUN: -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK-LE-NOOPT
68

79
// CHECK-LE-LABEL: @test1(
810
// CHECK-LE-NEXT: entry:
@@ -16,6 +18,42 @@
1618
// CHECK-BE-NEXT: store <512 x i1> [[TMP0]], ptr [[RESP:%.*]], align 64, !tbaa [[TBAA2:![0-9]+]]
1719
// CHECK-BE-NEXT: ret void
1820
//
21+
// CHECK-LE-NOOPT-LABEL: @test1(
22+
// CHECK-LE-NOOPT-NEXT: entry:
23+
// CHECK-LE-NOOPT-NEXT: [[VQP_ADDR:%.*]] = alloca ptr, align 8
24+
// CHECK-LE-NOOPT-NEXT: [[VPP_ADDR:%.*]] = alloca ptr, align 8
25+
// CHECK-LE-NOOPT-NEXT: [[VC1_ADDR:%.*]] = alloca <16 x i8>, align 16
26+
// CHECK-LE-NOOPT-NEXT: [[VC2_ADDR:%.*]] = alloca <16 x i8>, align 16
27+
// CHECK-LE-NOOPT-NEXT: [[VC3_ADDR:%.*]] = alloca <16 x i8>, align 16
28+
// CHECK-LE-NOOPT-NEXT: [[VC4_ADDR:%.*]] = alloca <16 x i8>, align 16
29+
// CHECK-LE-NOOPT-NEXT: [[RESP_ADDR:%.*]] = alloca ptr, align 8
30+
// CHECK-LE-NOOPT-NEXT: [[VQ:%.*]] = alloca <512 x i1>, align 64
31+
// CHECK-LE-NOOPT-NEXT: [[VP:%.*]] = alloca <256 x i1>, align 32
32+
// CHECK-LE-NOOPT-NEXT: [[RES:%.*]] = alloca <512 x i1>, align 64
33+
// CHECK-LE-NOOPT-NEXT: store ptr [[VQP:%.*]], ptr [[VQP_ADDR]], align 8
34+
// CHECK-LE-NOOPT-NEXT: store ptr [[VPP:%.*]], ptr [[VPP_ADDR]], align 8
35+
// CHECK-LE-NOOPT-NEXT: store <16 x i8> [[VC1:%.*]], ptr [[VC1_ADDR]], align 16
36+
// CHECK-LE-NOOPT-NEXT: store <16 x i8> [[VC2:%.*]], ptr [[VC2_ADDR]], align 16
37+
// CHECK-LE-NOOPT-NEXT: store <16 x i8> [[VC3:%.*]], ptr [[VC3_ADDR]], align 16
38+
// CHECK-LE-NOOPT-NEXT: store <16 x i8> [[VC4:%.*]], ptr [[VC4_ADDR]], align 16
39+
// CHECK-LE-NOOPT-NEXT: store ptr [[RESP:%.*]], ptr [[RESP_ADDR]], align 8
40+
// CHECK-LE-NOOPT-NEXT: [[TMP0:%.*]] = load ptr, ptr [[VQP_ADDR]], align 8
41+
// CHECK-LE-NOOPT-NEXT: [[TMP1:%.*]] = load <512 x i1>, ptr [[TMP0]], align 64
42+
// CHECK-LE-NOOPT-NEXT: store <512 x i1> [[TMP1]], ptr [[VQ]], align 64
43+
// CHECK-LE-NOOPT-NEXT: [[TMP2:%.*]] = load ptr, ptr [[VPP_ADDR]], align 8
44+
// CHECK-LE-NOOPT-NEXT: [[TMP3:%.*]] = load <256 x i1>, ptr [[TMP2]], align 32
45+
// CHECK-LE-NOOPT-NEXT: store <256 x i1> [[TMP3]], ptr [[VP]], align 32
46+
// CHECK-LE-NOOPT-NEXT: [[TMP4:%.*]] = load <16 x i8>, ptr [[VC1_ADDR]], align 16
47+
// CHECK-LE-NOOPT-NEXT: [[TMP5:%.*]] = load <16 x i8>, ptr [[VC2_ADDR]], align 16
48+
// CHECK-LE-NOOPT-NEXT: [[TMP6:%.*]] = load <16 x i8>, ptr [[VC3_ADDR]], align 16
49+
// CHECK-LE-NOOPT-NEXT: [[TMP7:%.*]] = load <16 x i8>, ptr [[VC4_ADDR]], align 16
50+
// CHECK-LE-NOOPT-NEXT: [[TMP8:%.*]] = call <512 x i1> @llvm.ppc.mma.assemble.acc(<16 x i8> [[TMP7]], <16 x i8> [[TMP6]], <16 x i8> [[TMP5]], <16 x i8> [[TMP4]])
51+
// CHECK-LE-NOOPT-NEXT: store <512 x i1> [[TMP8]], ptr [[RES]], align 64
52+
// CHECK-LE-NOOPT-NEXT: [[TMP9:%.*]] = load <512 x i1>, ptr [[RES]], align 64
53+
// CHECK-LE-NOOPT-NEXT: [[TMP10:%.*]] = load ptr, ptr [[RESP_ADDR]], align 8
54+
// CHECK-LE-NOOPT-NEXT: store <512 x i1> [[TMP9]], ptr [[TMP10]], align 64
55+
// CHECK-LE-NOOPT-NEXT: ret void
56+
//
1957
void test1(unsigned char *vqp, unsigned char *vpp, vector unsigned char vc1, vector unsigned char vc2,
2058
vector unsigned char vc3, vector unsigned char vc4, unsigned char *resp) {
2159
__vector_quad vq = *((__vector_quad *)vqp);
@@ -37,6 +75,36 @@ void test1(unsigned char *vqp, unsigned char *vpp, vector unsigned char vc1, vec
3775
// CHECK-BE-NEXT: store <256 x i1> [[TMP0]], ptr [[RESP:%.*]], align 32, !tbaa [[TBAA6:![0-9]+]]
3876
// CHECK-BE-NEXT: ret void
3977
//
78+
// CHECK-LE-NOOPT-LABEL: @test2(
79+
// CHECK-LE-NOOPT-NEXT: entry:
80+
// CHECK-LE-NOOPT-NEXT: [[VQP_ADDR:%.*]] = alloca ptr, align 8
81+
// CHECK-LE-NOOPT-NEXT: [[VPP_ADDR:%.*]] = alloca ptr, align 8
82+
// CHECK-LE-NOOPT-NEXT: [[VC1_ADDR:%.*]] = alloca <16 x i8>, align 16
83+
// CHECK-LE-NOOPT-NEXT: [[VC2_ADDR:%.*]] = alloca <16 x i8>, align 16
84+
// CHECK-LE-NOOPT-NEXT: [[RESP_ADDR:%.*]] = alloca ptr, align 8
85+
// CHECK-LE-NOOPT-NEXT: [[VQ:%.*]] = alloca <512 x i1>, align 64
86+
// CHECK-LE-NOOPT-NEXT: [[VP:%.*]] = alloca <256 x i1>, align 32
87+
// CHECK-LE-NOOPT-NEXT: [[RES:%.*]] = alloca <256 x i1>, align 32
88+
// CHECK-LE-NOOPT-NEXT: store ptr [[VQP:%.*]], ptr [[VQP_ADDR]], align 8
89+
// CHECK-LE-NOOPT-NEXT: store ptr [[VPP:%.*]], ptr [[VPP_ADDR]], align 8
90+
// CHECK-LE-NOOPT-NEXT: store <16 x i8> [[VC1:%.*]], ptr [[VC1_ADDR]], align 16
91+
// CHECK-LE-NOOPT-NEXT: store <16 x i8> [[VC2:%.*]], ptr [[VC2_ADDR]], align 16
92+
// CHECK-LE-NOOPT-NEXT: store ptr [[RESP:%.*]], ptr [[RESP_ADDR]], align 8
93+
// CHECK-LE-NOOPT-NEXT: [[TMP0:%.*]] = load ptr, ptr [[VQP_ADDR]], align 8
94+
// CHECK-LE-NOOPT-NEXT: [[TMP1:%.*]] = load <512 x i1>, ptr [[TMP0]], align 64
95+
// CHECK-LE-NOOPT-NEXT: store <512 x i1> [[TMP1]], ptr [[VQ]], align 64
96+
// CHECK-LE-NOOPT-NEXT: [[TMP2:%.*]] = load ptr, ptr [[VPP_ADDR]], align 8
97+
// CHECK-LE-NOOPT-NEXT: [[TMP3:%.*]] = load <256 x i1>, ptr [[TMP2]], align 32
98+
// CHECK-LE-NOOPT-NEXT: store <256 x i1> [[TMP3]], ptr [[VP]], align 32
99+
// CHECK-LE-NOOPT-NEXT: [[TMP4:%.*]] = load <16 x i8>, ptr [[VC1_ADDR]], align 16
100+
// CHECK-LE-NOOPT-NEXT: [[TMP5:%.*]] = load <16 x i8>, ptr [[VC2_ADDR]], align 16
101+
// CHECK-LE-NOOPT-NEXT: [[TMP6:%.*]] = call <256 x i1> @llvm.ppc.vsx.assemble.pair(<16 x i8> [[TMP5]], <16 x i8> [[TMP4]])
102+
// CHECK-LE-NOOPT-NEXT: store <256 x i1> [[TMP6]], ptr [[RES]], align 64
103+
// CHECK-LE-NOOPT-NEXT: [[TMP7:%.*]] = load <256 x i1>, ptr [[RES]], align 32
104+
// CHECK-LE-NOOPT-NEXT: [[TMP8:%.*]] = load ptr, ptr [[RESP_ADDR]], align 8
105+
// CHECK-LE-NOOPT-NEXT: store <256 x i1> [[TMP7]], ptr [[TMP8]], align 32
106+
// CHECK-LE-NOOPT-NEXT: ret void
107+
//
40108
void test2(unsigned char *vqp, unsigned char *vpp, vector unsigned char vc1,
41109
vector unsigned char vc2, unsigned char *resp) {
42110
__vector_quad vq = *((__vector_quad *)vqp);

0 commit comments

Comments
 (0)