llvm-for-llvmta/test/CodeGen/AArch64/GlobalISel/postlegalizer-lowering-uzp.mir

147 lines
4.6 KiB
YAML

# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
#
# Check that we can recognize a shuffle mask for a uzp instruction and produce
# a G_UZP1 or G_UZP2 where appropriate.
#
# RUN: llc -mtriple aarch64 -run-pass=aarch64-postlegalizer-lowering -verify-machineinstrs %s -o - | FileCheck %s
...
---
name: uzp1_v4s32
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $q0, $q1
; CHECK-LABEL: name: uzp1_v4s32
; CHECK: liveins: $q0, $q1
; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK: [[UZP1_:%[0-9]+]]:_(<4 x s32>) = G_UZP1 [[COPY]], [[COPY1]]
; CHECK: $q0 = COPY [[UZP1_]](<4 x s32>)
; CHECK: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(0, 2, 4, 6)
$q0 = COPY %2(<4 x s32>)
RET_ReallyLR implicit $q0
...
---
name: uzp2_v4s32
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $q0, $q1
; CHECK-LABEL: name: uzp2_v4s32
; CHECK: liveins: $q0, $q1
; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK: [[UZP2_:%[0-9]+]]:_(<4 x s32>) = G_UZP2 [[COPY]], [[UZP2_]]
; CHECK: $q0 = COPY [[UZP2_]](<4 x s32>)
; CHECK: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%1:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(1, 3, 5, 7)
$q0 = COPY %1(<4 x s32>)
RET_ReallyLR implicit $q0
...
---
name: no_uzp1
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $q0, $q1
; See isUZPMask: Mask[1] != 2 * i + 0
; CHECK-LABEL: name: no_uzp1
; CHECK: liveins: $q0, $q1
; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK: [[SHUF:%[0-9]+]]:_(<4 x s32>) = G_SHUFFLE_VECTOR [[COPY]](<4 x s32>), [[COPY1]], shufflemask(0, 1, 4, 6)
; CHECK: $q0 = COPY [[SHUF]](<4 x s32>)
; CHECK: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(0, 1, 4, 6)
$q0 = COPY %2(<4 x s32>)
RET_ReallyLR implicit $q0
...
---
name: no_uzp2
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $q0, $q1
; See isUZPMask: Mask[1] != 2 * i + 1
; CHECK-LABEL: name: no_uzp2
; CHECK: liveins: $q0, $q1
; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK: [[SHUF:%[0-9]+]]:_(<4 x s32>) = G_SHUFFLE_VECTOR [[COPY]](<4 x s32>), [[COPY1]], shufflemask(1, 4, 5, 7)
; CHECK: $q0 = COPY [[SHUF]](<4 x s32>)
; CHECK: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(1, 4, 5, 7)
$q0 = COPY %2(<4 x s32>)
RET_ReallyLR implicit $q0
...
---
name: uzp1_undef
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $q0, $q1
; Make sure that we can still produce a uzp1/uzp2 with undef indices.
; CHECK-LABEL: name: uzp1_undef
; CHECK: liveins: $q0, $q1
; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK: [[UZP1_:%[0-9]+]]:_(<4 x s32>) = G_UZP1 [[COPY]], [[COPY1]]
; CHECK: $q0 = COPY [[UZP1_]](<4 x s32>)
; CHECK: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(0, -1, 4, 6)
$q0 = COPY %2(<4 x s32>)
RET_ReallyLR implicit $q0
...
---
name: uzp2_undef
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $q0, $q1
; Make sure that we can still produce a uzp1/uzp2 with undef indices.
; CHECK-LABEL: name: uzp2_undef
; CHECK: liveins: $q0, $q1
; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK: [[UZP2_:%[0-9]+]]:_(<4 x s32>) = G_UZP2 [[COPY]], [[UZP2_]]
; CHECK: $q0 = COPY [[UZP2_]](<4 x s32>)
; CHECK: RET_ReallyLR implicit $q0
%0:_(<4 x s32>) = COPY $q0
%1:_(<4 x s32>) = COPY $q1
%1:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(1, 3, -1, 7)
$q0 = COPY %1(<4 x s32>)
RET_ReallyLR implicit $q0