| 1 |
REQUIRES: aarch64-registered-target |
| 2 |
|
| 3 |
## PPR Register Class Initialization Testcase |
| 4 |
## Ideally, we should use PTRUE_{B/H/S/D} instead of FADDV_VPZ_D for an isolated test case; |
| 5 |
## However, exegesis does not yet support PTRUE_{B/H/S/D}. |
| 6 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FADDV_VPZ_D --benchmark-phase=assemble-measured-code 2>&1 |
| 7 |
RUN: llvm-objdump -d %d > %t.s |
| 8 |
RUN: FileCheck %s --check-prefix=PPR_ASM < %t.s |
| 9 |
PPR_ASM: <foo>: |
| 10 |
PPR_ASM: ptrue p{{[0-9]+}}.b |
| 11 |
PPR_ASM-NEXT: mov z{{[0-9]+}}.d, #0x0 |
| 12 |
PPR_ASM-NEXT: faddv d{{[0-9]+}}, p{{[0-9]+}}, z{{[0-9]+}} |
| 13 |
|
| 14 |
## ZPR Register Class Initialization Testcase |
| 15 |
## Ideally, we should use DUP_ZI_{B/H/S/D} instead of FADDV_VPZ_D for an isolated test case; |
| 16 |
## However, exegesis does not yet support DUP_ZI_{B/H/S/D}. |
| 17 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FADDV_VPZ_D --benchmark-phase=assemble-measured-code 2>&1 |
| 18 |
RUN: llvm-objdump -d %d > %t.s |
| 19 |
RUN: FileCheck %s --check-prefix=ZPR_ASM < %t.s |
| 20 |
ZPR_ASM: <foo>: |
| 21 |
ZPR_ASM: ptrue p{{[0-9]+}}.b |
| 22 |
ZPR_ASM-NEXT: mov z{{[0-9]+}}.d, #0x0 |
| 23 |
ZPR_ASM-NEXT: faddv d{{[0-9]+}}, p{{[0-9]+}}, z{{[0-9]+}} |
| 24 |
|
| 25 |
## FPR128 Register Class Initialization Testcase |
| 26 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADDVv16i8v --benchmark-phase=assemble-measured-code 2>&1 |
| 27 |
RUN: llvm-objdump -d %d > %t.s |
| 28 |
RUN: FileCheck %s --check-prefix=FPR128-ASM < %t.s |
| 29 |
FPR128-ASM: <foo>: |
| 30 |
FPR128-ASM: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 31 |
FPR128-ASM-NEXT: addv b{{[0-9]+}}, v{{[0-9]+}}.16b |
| 32 |
|
| 33 |
## FPR64 Register Class Initialization Testcase |
| 34 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADDVv4i16v --benchmark-phase=assemble-measured-code 2>&1 |
| 35 |
RUN: llvm-objdump -d %d > %t.s |
| 36 |
RUN: FileCheck %s --check-prefix=FPR64-ASM < %t.s |
| 37 |
FPR64-ASM: <foo>: |
| 38 |
FPR64-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 39 |
FPR64-ASM-NEXT: addv h{{[0-9]+}}, v{{[0-9]+}}.4h |
| 40 |
|
| 41 |
## FPR32 Register Class Initialization Testcase |
| 42 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FABSSr --benchmark-phase=assemble-measured-code 2>&1 |
| 43 |
RUN: llvm-objdump -d %d > %t.s |
| 44 |
RUN: FileCheck %s --check-prefix=FPR32-ASM < %t.s |
| 45 |
FPR32-ASM: <foo>: |
| 46 |
FPR32-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 47 |
FPR32-ASM-NEXT: fabs s{{[0-9]+}}, s{{[0-9]+}} |
| 48 |
|
| 49 |
|
| 50 |
## FPR16 Register Class Initialization Testcase |
| 51 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FABSHr --benchmark-phase=assemble-measured-code 2>&1 |
| 52 |
RUN: llvm-objdump -d %d > %t.s |
| 53 |
RUN: FileCheck %s --check-prefix=FPR16-ASM < %t.s |
| 54 |
FPR16-ASM: <foo>: |
| 55 |
FPR16-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 56 |
FPR16-ASM-NEXT: fabs h{{[0-9]+}}, h{{[0-9]+}} |
| 57 |
|
| 58 |
## FPR8 Register Class Initialization Testcase |
| 59 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=SQABSv1i8 --benchmark-phase=assemble-measured-code 2>&1 |
| 60 |
RUN: llvm-objdump -d %d > %t.s |
| 61 |
RUN: FileCheck %s --check-prefix=FPR8-ASM < %t.s |
| 62 |
FPR8-ASM: <foo>: |
| 63 |
FPR8-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 64 |
FPR8-ASM-NEXT: sqabs b{{[0-9]+}}, b{{[0-9]+}} |
| 65 |
|
| 66 |
|
| 67 |
## FPCR Register Class Initialization Testcase |
| 68 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=BFCVT --benchmark-phase=assemble-measured-code 2>&1 |
| 69 |
RUN: llvm-objdump -d %d > %t.s |
| 70 |
RUN: FileCheck %s --check-prefix=FPCR-ASM < %t.s |
| 71 |
FPCR-ASM: <foo>: |
| 72 |
FPCR-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 73 |
FPCR-ASM-NEXT: mov x8, #0x0 |
| 74 |
FPCR-ASM-NEXT: msr FPCR, x8 |
| 75 |
FPCR-ASM-NEXT: bfcvt h{{[0-9]+}}, s{{[0-9]+}} |
| 76 |
|
| 77 |
## NZCV Register Class Initialization Testcase |
| 78 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADCSWr --benchmark-phase=assemble-measured-code 2>&1 |
| 79 |
RUN: llvm-objdump -d %d > %t.s |
| 80 |
RUN: FileCheck %s --check-prefix=NZCV-ASM < %t.s |
| 81 |
NZCV-ASM: <foo>: |
| 82 |
NZCV-ASM: mrs x8, NZCV |
| 83 |
NZCV-ASM-NEXT: mov x9, #0xf0000000 |
| 84 |
NZCV-ASM-NEXT: bic x8, x8, x9 |
| 85 |
NZCV-ASM-NEXT: msr NZCV, x8 |
| 86 |
NZCV-ASM-NEXT: adcs w{{[0-9]+}}, w{{[0-9]+}}, w{{[0-9]+}} |
| 87 |
|
| 88 |
## FFR Register Class Initialization Testcase |
| 89 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=GLDFF1B_D --benchmark-phase=assemble-measured-code 2>&1 |
| 90 |
RUN: llvm-objdump -d %d > %t.s |
| 91 |
RUN: FileCheck %s --check-prefix=FFR-ASM < %t.s |
| 92 |
FFR-ASM: <foo>: |
| 93 |
FFR-ASM: setffr |
| 94 |
FFR-ASM-NEXT: ldff1b { z{{[0-9]+}}.d }, p{{[0-9]+}}/z, [x{{[0-9]+}}, z{{[0-9]+}}.d] |
| 95 |
|
| 96 |
# NOTE: Multi-register classes below can wrap around, e.g. { v31.1d, v0.1d } |
| 97 |
# FileCheck cant handle this constraint and '--random-generator-seed=<seed>' |
| 98 |
# can't be used either as the implementation relies on uniform_int_distribution |
| 99 |
# which produces inconsistent results across libc++ implementations. |
| 100 |
|
| 101 |
## WSeqPair Register Class Initialization Testcase |
| 102 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=CASPW --benchmark-phase=assemble-measured-code 2>&1 |
| 103 |
RUN: llvm-objdump -d %d > %t.s |
| 104 |
RUN: FileCheck %s --check-prefix=WSEQPAIR-ASM < %t.s |
| 105 |
WSEQPAIR-ASM: <foo>: |
| 106 |
WSEQPAIR-ASM: mov w{{[0-9]+}}, #0x0 |
| 107 |
WSEQPAIR-ASM-NEXT: mov w{{[0-9]+}}, #0x0 |
| 108 |
WSEQPAIR-ASM: casp w{{[0-9]+}}, w{{[0-9]+}}, w{{[0-9]+}}, w{{[0-9]+}}, [x{{[0-9]+}}] |
| 109 |
|
| 110 |
## XSeqPair Register Class Initialization Testcase |
| 111 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=CASPX --benchmark-phase=assemble-measured-code 2>&1 |
| 112 |
RUN: llvm-objdump -d %d > %t.s |
| 113 |
RUN: FileCheck %s --check-prefix=XSEQPAIR-ASM < %t.s |
| 114 |
XSEQPAIR-ASM: <foo>: |
| 115 |
XSEQPAIR-ASM: mov x{{[0-9]+}}, #0x{{[0-9]+}} |
| 116 |
XSEQPAIR-ASM-NEXT: mov x{{[0-9]+}}, #0x{{[0-9]+}} |
| 117 |
XSEQPAIR-ASM: casp x{{[0-9]+}}, x{{[0-9]+}}, x{{[0-9]+}}, x{{[0-9]+}}, [x{{[0-9]+}}] |
| 118 |
|
| 119 |
## DD Register Class Initialization Testcase |
| 120 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ST1Twov1d_POST --benchmark-phase=assemble-measured-code 2>&1 |
| 121 |
RUN: llvm-objdump -d %d > %t.s |
| 122 |
RUN: FileCheck %s --check-prefix=DD-ASM < %t.s |
| 123 |
DD-ASM: <foo>: |
| 124 |
DD-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 125 |
DD-ASM-NEXT: movi d{{[0-9]+}}, #0000000000000000 |
| 126 |
DD-ASM: st1 { v{{[0-9]+}}.1d, v{{[0-9]+}}.1d }, [x{{[0-9]+}}], x{{[0-9]+}} |
| 127 |
|
| 128 |
## DDD Register Class Initialization Testcase |
| 129 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ST1Threev1d_POST --benchmark-phase=assemble-measured-code 2>&1 |
| 130 |
RUN: llvm-objdump -d %d > %t.s |
| 131 |
RUN: FileCheck %s --check-prefix=DDD-ASM < %t.s |
| 132 |
DDD-ASM: <foo>: |
| 133 |
DDD-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 134 |
DDD-ASM-NEXT: movi d{{[0-9]+}}, #0000000000000000 |
| 135 |
DDD-ASM-NEXT: movi d{{[0-9]+}}, #0000000000000000 |
| 136 |
DDD-ASM: st1 { v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d }, [x{{[0-9]+}}], x{{[0-9]+}} |
| 137 |
|
| 138 |
## DDDD Register Class Initialization Testcase |
| 139 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ST1Fourv1d_POST --benchmark-phase=assemble-measured-code 2>&1 |
| 140 |
RUN: llvm-objdump -d %d > %t.s |
| 141 |
RUN: FileCheck %s --check-prefix=DDDD-ASM < %t.s |
| 142 |
DDDD-ASM: <foo>: |
| 143 |
DDDD-ASM: movi d{{[0-9]+}}, #0000000000000000 |
| 144 |
DDDD-ASM-NEXT: movi d{{[0-9]+}}, #0000000000000000 |
| 145 |
DDDD-ASM-NEXT: movi d{{[0-9]+}}, #0000000000000000 |
| 146 |
DDDD-ASM-NEXT: movi d{{[0-9]+}}, #0000000000000000 |
| 147 |
DDDD-ASM: st1 { v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d, v{{[0-9]+}}.1d }, [x{{[0-9]+}}], x{{[0-9]+}} |
| 148 |
|
| 149 |
## QQ Register Class Initialization Testcase |
| 150 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ST1Twov16b_POST --benchmark-phase=assemble-measured-code 2>&1 |
| 151 |
RUN: llvm-objdump -d %d > %t.s |
| 152 |
RUN: FileCheck %s --check-prefix=QQ-ASM < %t.s |
| 153 |
QQ-ASM: <foo>: |
| 154 |
QQ-ASM: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 155 |
QQ-ASM-NEXT: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 156 |
QQ-ASM: st1 { v{{[0-9]+}}.16b, v{{[0-9]+}}.16b }, [x{{[0-9]+}}], x{{[0-9]+}} |
| 157 |
|
| 158 |
## QQQ Register Class Initialization Testcase |
| 159 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ST1Threev16b_POST --benchmark-phase=assemble-measured-code 2>&1 |
| 160 |
RUN: llvm-objdump -d %d > %t.s |
| 161 |
RUN: FileCheck %s --check-prefix=QQQ-ASM < %t.s |
| 162 |
QQQ-ASM: <foo>: |
| 163 |
QQQ-ASM: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 164 |
QQQ-ASM-NEXT: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 165 |
QQQ-ASM-NEXT: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 166 |
QQQ-ASM: st1 { v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b }, [x{{[0-9]+}}], x{{[0-9]+}} |
| 167 |
|
| 168 |
## QQQQ Register Class Initialization Testcase |
| 169 |
RUN: llvm-exegesis -mtriple=aarch64 -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ST1Fourv16b_POST --benchmark-phase=assemble-measured-code 2>&1 |
| 170 |
RUN: llvm-objdump -d %d > %t.s |
| 171 |
RUN: FileCheck %s --check-prefix=QQQQ-ASM < %t.s |
| 172 |
QQQQ-ASM: <foo>: |
| 173 |
QQQQ-ASM: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 174 |
QQQQ-ASM-NEXT: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 175 |
QQQQ-ASM-NEXT: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 176 |
QQQQ-ASM-NEXT: movi v{{[0-9]+}}.2d, #0000000000000000 |
| 177 |
QQQQ-ASM: st1 { v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b }, [x{{[0-9]+}}], x{{[0-9]+}} |