llvm-for-llvmta/test/CodeGen/AArch64/GlobalISel/legalize-fma.mir

234 lines
12 KiB
Plaintext
Raw Permalink Normal View History

2022-04-25 10:02:23 +02:00
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc %s -verify-machineinstrs -mtriple=aarch64-unknown-unknown -run-pass=legalizer -simplify-mir -aarch64-neon-syntax=apple -mattr=-fullfp16 -o - | FileCheck %s --check-prefix=NO-FP16
# RUN: llc %s -verify-machineinstrs -mtriple=aarch64-unknown-unknown -run-pass=legalizer -simplify-mir -aarch64-neon-syntax=apple -mattr=+fullfp16 -o - | FileCheck %s --check-prefix=FP16
...
---
name: test_v4f16.fma
alignment: 4
tracksRegLiveness: true
machineFunctionInfo: {}
body: |
bb.0:
liveins: $d0, $d1, $d2
; NO-FP16-LABEL: name: test_v4f16.fma
; NO-FP16: liveins: $d0, $d1, $d2
; NO-FP16: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0
; NO-FP16: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $d1
; NO-FP16: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $d2
; NO-FP16: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
; NO-FP16: [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
; NO-FP16: [[UV8:%[0-9]+]]:_(s16), [[UV9:%[0-9]+]]:_(s16), [[UV10:%[0-9]+]]:_(s16), [[UV11:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY2]](<4 x s16>)
; NO-FP16: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
; NO-FP16: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16)
; NO-FP16: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV8]](s16)
; NO-FP16: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FPEXT]], [[FPEXT1]], [[FPEXT2]]
; NO-FP16: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA]](s32)
; NO-FP16: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
; NO-FP16: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16)
; NO-FP16: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV9]](s16)
; NO-FP16: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FPEXT3]], [[FPEXT4]], [[FPEXT5]]
; NO-FP16: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA1]](s32)
; NO-FP16: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
; NO-FP16: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16)
; NO-FP16: [[FPEXT8:%[0-9]+]]:_(s32) = G_FPEXT [[UV10]](s16)
; NO-FP16: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FPEXT6]], [[FPEXT7]], [[FPEXT8]]
; NO-FP16: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA2]](s32)
; NO-FP16: [[FPEXT9:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
; NO-FP16: [[FPEXT10:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16)
; NO-FP16: [[FPEXT11:%[0-9]+]]:_(s32) = G_FPEXT [[UV11]](s16)
; NO-FP16: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FPEXT9]], [[FPEXT10]], [[FPEXT11]]
; NO-FP16: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA3]](s32)
; NO-FP16: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16)
; NO-FP16: $d0 = COPY [[BUILD_VECTOR]](<4 x s16>)
; NO-FP16: RET_ReallyLR implicit $d0
; FP16-LABEL: name: test_v4f16.fma
; FP16: liveins: $d0, $d1, $d2
; FP16: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0
; FP16: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $d1
; FP16: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $d2
; FP16: [[FMA:%[0-9]+]]:_(<4 x s16>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; FP16: $d0 = COPY [[FMA]](<4 x s16>)
; FP16: RET_ReallyLR implicit $d0
%0:_(<4 x s16>) = COPY $d0
%1:_(<4 x s16>) = COPY $d1
%2:_(<4 x s16>) = COPY $d2
%3:_(<4 x s16>) = G_FMA %0, %1, %2
$d0 = COPY %3(<4 x s16>)
RET_ReallyLR implicit $d0
...
---
name: test_v8f16.fma
alignment: 4
tracksRegLiveness: true
machineFunctionInfo: {}
body: |
bb.0:
liveins: $q0, $q1, $q2
; NO-FP16-LABEL: name: test_v8f16.fma
; NO-FP16: liveins: $q0, $q1, $q2
; NO-FP16: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0
; NO-FP16: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $q1
; NO-FP16: [[COPY2:%[0-9]+]]:_(<8 x s16>) = COPY $q2
; NO-FP16: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<8 x s16>)
; NO-FP16: [[UV8:%[0-9]+]]:_(s16), [[UV9:%[0-9]+]]:_(s16), [[UV10:%[0-9]+]]:_(s16), [[UV11:%[0-9]+]]:_(s16), [[UV12:%[0-9]+]]:_(s16), [[UV13:%[0-9]+]]:_(s16), [[UV14:%[0-9]+]]:_(s16), [[UV15:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<8 x s16>)
; NO-FP16: [[UV16:%[0-9]+]]:_(s16), [[UV17:%[0-9]+]]:_(s16), [[UV18:%[0-9]+]]:_(s16), [[UV19:%[0-9]+]]:_(s16), [[UV20:%[0-9]+]]:_(s16), [[UV21:%[0-9]+]]:_(s16), [[UV22:%[0-9]+]]:_(s16), [[UV23:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY2]](<8 x s16>)
; NO-FP16: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
; NO-FP16: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV8]](s16)
; NO-FP16: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV16]](s16)
; NO-FP16: [[FMA:%[0-9]+]]:_(s32) = G_FMA [[FPEXT]], [[FPEXT1]], [[FPEXT2]]
; NO-FP16: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA]](s32)
; NO-FP16: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
; NO-FP16: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV9]](s16)
; NO-FP16: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV17]](s16)
; NO-FP16: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FPEXT3]], [[FPEXT4]], [[FPEXT5]]
; NO-FP16: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA1]](s32)
; NO-FP16: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
; NO-FP16: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV10]](s16)
; NO-FP16: [[FPEXT8:%[0-9]+]]:_(s32) = G_FPEXT [[UV18]](s16)
; NO-FP16: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FPEXT6]], [[FPEXT7]], [[FPEXT8]]
; NO-FP16: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA2]](s32)
; NO-FP16: [[FPEXT9:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
; NO-FP16: [[FPEXT10:%[0-9]+]]:_(s32) = G_FPEXT [[UV11]](s16)
; NO-FP16: [[FPEXT11:%[0-9]+]]:_(s32) = G_FPEXT [[UV19]](s16)
; NO-FP16: [[FMA3:%[0-9]+]]:_(s32) = G_FMA [[FPEXT9]], [[FPEXT10]], [[FPEXT11]]
; NO-FP16: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA3]](s32)
; NO-FP16: [[FPEXT12:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16)
; NO-FP16: [[FPEXT13:%[0-9]+]]:_(s32) = G_FPEXT [[UV12]](s16)
; NO-FP16: [[FPEXT14:%[0-9]+]]:_(s32) = G_FPEXT [[UV20]](s16)
; NO-FP16: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FPEXT12]], [[FPEXT13]], [[FPEXT14]]
; NO-FP16: [[FPTRUNC4:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA4]](s32)
; NO-FP16: [[FPEXT15:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16)
; NO-FP16: [[FPEXT16:%[0-9]+]]:_(s32) = G_FPEXT [[UV13]](s16)
; NO-FP16: [[FPEXT17:%[0-9]+]]:_(s32) = G_FPEXT [[UV21]](s16)
; NO-FP16: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FPEXT15]], [[FPEXT16]], [[FPEXT17]]
; NO-FP16: [[FPTRUNC5:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA5]](s32)
; NO-FP16: [[FPEXT18:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16)
; NO-FP16: [[FPEXT19:%[0-9]+]]:_(s32) = G_FPEXT [[UV14]](s16)
; NO-FP16: [[FPEXT20:%[0-9]+]]:_(s32) = G_FPEXT [[UV22]](s16)
; NO-FP16: [[FMA6:%[0-9]+]]:_(s32) = G_FMA [[FPEXT18]], [[FPEXT19]], [[FPEXT20]]
; NO-FP16: [[FPTRUNC6:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA6]](s32)
; NO-FP16: [[FPEXT21:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16)
; NO-FP16: [[FPEXT22:%[0-9]+]]:_(s32) = G_FPEXT [[UV15]](s16)
; NO-FP16: [[FPEXT23:%[0-9]+]]:_(s32) = G_FPEXT [[UV23]](s16)
; NO-FP16: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FPEXT21]], [[FPEXT22]], [[FPEXT23]]
; NO-FP16: [[FPTRUNC7:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMA7]](s32)
; NO-FP16: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16), [[FPTRUNC4]](s16), [[FPTRUNC5]](s16), [[FPTRUNC6]](s16), [[FPTRUNC7]](s16)
; NO-FP16: $q0 = COPY [[BUILD_VECTOR]](<8 x s16>)
; NO-FP16: RET_ReallyLR implicit $q0
; FP16-LABEL: name: test_v8f16.fma
; FP16: liveins: $q0, $q1, $q2
; FP16: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0
; FP16: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $q1
; FP16: [[COPY2:%[0-9]+]]:_(<8 x s16>) = COPY $q2
; FP16: [[FMA:%[0-9]+]]:_(<8 x s16>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; FP16: $q0 = COPY [[FMA]](<8 x s16>)
; FP16: RET_ReallyLR implicit $q0
%0:_(<8 x s16>) = COPY $q0
%1:_(<8 x s16>) = COPY $q1
%2:_(<8 x s16>) = COPY $q2
%3:_(<8 x s16>) = G_FMA %0, %1, %2
$q0 = COPY %3(<8 x s16>)
RET_ReallyLR implicit $q0
...
---
name: test_v2f32.fma
alignment: 4
tracksRegLiveness: true
machineFunctionInfo: {}
body: |
bb.0:
liveins: $d0, $d1, $d2
; NO-FP16-LABEL: name: test_v2f32.fma
; NO-FP16: liveins: $d0, $d1, $d2
; NO-FP16: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0
; NO-FP16: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
; NO-FP16: [[COPY2:%[0-9]+]]:_(<2 x s32>) = COPY $d2
; NO-FP16: [[FMA:%[0-9]+]]:_(<2 x s32>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; NO-FP16: $d0 = COPY [[FMA]](<2 x s32>)
; NO-FP16: RET_ReallyLR implicit $d0
; FP16-LABEL: name: test_v2f32.fma
; FP16: liveins: $d0, $d1, $d2
; FP16: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0
; FP16: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1
; FP16: [[COPY2:%[0-9]+]]:_(<2 x s32>) = COPY $d2
; FP16: [[FMA:%[0-9]+]]:_(<2 x s32>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; FP16: $d0 = COPY [[FMA]](<2 x s32>)
; FP16: RET_ReallyLR implicit $d0
%0:_(<2 x s32>) = COPY $d0
%1:_(<2 x s32>) = COPY $d1
%2:_(<2 x s32>) = COPY $d2
%3:_(<2 x s32>) = G_FMA %0, %1, %2
$d0 = COPY %3(<2 x s32>)
RET_ReallyLR implicit $d0
...
---
name: test_v4f32.fma
alignment: 4
tracksRegLiveness: true
machineFunctionInfo: {}
body: |
bb.0:
liveins: $q0, $q1, $q2
; NO-FP16-LABEL: name: test_v4f32.fma
; NO-FP16: liveins: $q0, $q1, $q2
; NO-FP16: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; NO-FP16: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; NO-FP16: [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $q2
; NO-FP16: [[FMA:%[0-9]+]]:_(<4 x s32>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; NO-FP16: $q0 = COPY [[FMA]](<4 x s32>)
; NO-FP16: RET_ReallyLR implicit $q0
; FP16-LABEL: name: test_v4f32.fma
; FP16: liveins: $q0, $q1, $q2
; FP16: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; FP16: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; FP16: [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $q2
; FP16: [[FMA:%[0-9]+]]:_(<4 x s32>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; FP16: $q0 = COPY [[FMA]](<4 x s32>)
; FP16: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%2:_(<4 x s32>) = COPY $q2
%3:_(<4 x s32>) = G_FMA %0, %1, %2
$q0 = COPY %3(<4 x s32>)
RET_ReallyLR implicit $q0
...
---
name: test_v2f64.fma
alignment: 4
tracksRegLiveness: true
machineFunctionInfo: {}
body: |
bb.0:
liveins: $q0, $q1, $q2
; NO-FP16-LABEL: name: test_v2f64.fma
; NO-FP16: liveins: $q0, $q1, $q2
; NO-FP16: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0
; NO-FP16: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1
; NO-FP16: [[COPY2:%[0-9]+]]:_(<2 x s64>) = COPY $q2
; NO-FP16: [[FMA:%[0-9]+]]:_(<2 x s64>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; NO-FP16: $q0 = COPY [[FMA]](<2 x s64>)
; NO-FP16: RET_ReallyLR implicit $q0
; FP16-LABEL: name: test_v2f64.fma
; FP16: liveins: $q0, $q1, $q2
; FP16: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0
; FP16: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1
; FP16: [[COPY2:%[0-9]+]]:_(<2 x s64>) = COPY $q2
; FP16: [[FMA:%[0-9]+]]:_(<2 x s64>) = G_FMA [[COPY]], [[COPY1]], [[COPY2]]
; FP16: $q0 = COPY [[FMA]](<2 x s64>)
; FP16: RET_ReallyLR implicit $q0
%0:_(<2 x s64>) = COPY $q0
%1:_(<2 x s64>) = COPY $q1
%2:_(<2 x s64>) = COPY $q2
%3:_(<2 x s64>) = G_FMA %0, %1, %2
$q0 = COPY %3(<2 x s64>)
RET_ReallyLR implicit $q0