derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
422 lines
12 KiB
YAML
422 lines
12 KiB
YAML
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
|
|
# RUN: llc -mtriple aarch64-unknown-unknown -run-pass=instruction-select -global-isel-abort=1 -verify-machineinstrs %s -o - | FileCheck %s
|
|
#
|
|
# Check folding a G_SHL into a G_BRCOND which has been matched as a TB(N)Z.
|
|
...
|
|
---
|
|
name: fold_shl
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: fold_shl
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64all = COPY $x0
|
|
; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr32all = COPY %copy.sub_32
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr32 = COPY [[COPY]]
|
|
; CHECK-NEXT: TBNZW [[COPY1]], 2, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
|
|
; tbnz (shl x, 1), 3 == tbnz x, 2
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 1
|
|
%fold_me:gpr(i64) = G_SHL %copy, %fold_cst
|
|
|
|
%and:gpr(i64) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
...
|
|
---
|
|
name: dont_fold_shl_1
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: dont_fold_shl_1
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64 = COPY $x0
|
|
; CHECK-NEXT: %fold_me:gpr64 = UBFMXri %copy, 59, 58
|
|
; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr32all = COPY %fold_me.sub_32
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr32 = COPY [[COPY]]
|
|
; CHECK-NEXT: TBNZW [[COPY1]], 3, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
|
|
; 5 > 3, so we cannot do the transformation as above.
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 5
|
|
%fold_me:gpr(i64) = G_SHL %copy, %fold_cst
|
|
|
|
%and:gpr(i64) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
...
|
|
---
|
|
name: dont_fold_shl_2
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: dont_fold_shl_2
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64 = COPY $x0
|
|
; CHECK-NEXT: %fold_cst:gpr64 = MOVi64imm -5
|
|
; CHECK-NEXT: %fold_me:gpr64 = LSLVXr %copy, %fold_cst
|
|
; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr32all = COPY %fold_me.sub_32
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr32 = COPY [[COPY]]
|
|
; CHECK-NEXT: TBNZW [[COPY1]], 3, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
|
|
; Same case as above, except we wrap around.
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 -5
|
|
%fold_me:gpr(i64) = G_SHL %copy, %fold_cst
|
|
|
|
%and:gpr(i64) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: dont_fold_shl_3
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: dont_fold_shl_3
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64 = COPY $x0
|
|
; CHECK-NEXT: %shl:gpr64 = UBFMXri %copy, 62, 61
|
|
; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr32all = COPY %shl.sub_32
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr32 = COPY [[COPY]]
|
|
; CHECK-NEXT: TBNZW [[COPY1]], 3, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: %second_use:gpr64sp = ORRXri %shl, 8000
|
|
; CHECK-NEXT: $x0 = COPY %second_use
|
|
; CHECK-NEXT: RET_ReallyLR implicit $x0
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 2
|
|
|
|
; Don't walk past the G_SHL when it's used more than once.
|
|
%shl:gpr(i64) = G_SHL %copy, %fold_cst
|
|
%and:gpr(i64) = G_AND %shl, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
|
|
bb.1:
|
|
%second_use:gpr(i64) = G_OR %shl, %bit
|
|
$x0 = COPY %second_use
|
|
RET_ReallyLR implicit $x0
|
|
|
|
...
|
|
---
|
|
name: fold_ashr_in_range
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: fold_ashr_in_range
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64all = COPY $x0
|
|
; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr32all = COPY %copy.sub_32
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr32 = COPY [[COPY]]
|
|
; CHECK-NEXT: TBNZW [[COPY1]], 4, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
|
|
; tb(n)z (ashr x, c), b == tbz(x, b + c) when b+c <= the size of the type.
|
|
; In this case, we should get 1 + 3 = 4 as the bit number.
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 1
|
|
%fold_me:gpr(i64) = G_ASHR %copy, %fold_cst
|
|
|
|
%and:gpr(i64) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: fold_ashr_msb_1
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: fold_ashr_msb_1
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr32 = COPY $w0
|
|
; CHECK-NEXT: TBNZW %copy, 31, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i32) = COPY $w0
|
|
%bit:gpr(i32) = G_CONSTANT i32 8
|
|
%zero:gpr(i32) = G_CONSTANT i32 0
|
|
|
|
; We should get a TBNZW with a 31 as the bit.
|
|
%fold_cst:gpr(i32) = G_CONSTANT i32 1234
|
|
%fold_me:gpr(i32) = G_ASHR %copy, %fold_cst
|
|
|
|
%and:gpr(i32) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i32), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: fold_ashr_msb_2
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: fold_ashr_msb_2
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64 = COPY $x0
|
|
; CHECK-NEXT: TBNZX %copy, 63, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
|
|
; We should get a TBNZX with a 63 as the bit.
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 1234
|
|
%fold_me:gpr(i64) = G_ASHR %copy, %fold_cst
|
|
|
|
%and:gpr(i64) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: fold_lshr
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: fold_lshr
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr32 = COPY $w0
|
|
; CHECK-NEXT: TBNZW %copy, 4, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i32) = COPY $w0
|
|
%bit:gpr(i32) = G_CONSTANT i32 8
|
|
%zero:gpr(i32) = G_CONSTANT i32 0
|
|
|
|
; We should get 4 as the test bit.
|
|
%fold_cst:gpr(i32) = G_CONSTANT i32 1
|
|
%fold_me:gpr(i32) = G_LSHR %copy, %fold_cst
|
|
|
|
%and:gpr(i32) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i32), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: fold_lshr_2
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: fold_lshr_2
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr64 = COPY $x0
|
|
; CHECK-NEXT: TBNZX %copy, 32, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i64) = COPY $x0
|
|
%bit:gpr(i64) = G_CONSTANT i64 8
|
|
%zero:gpr(i64) = G_CONSTANT i64 0
|
|
|
|
; We're testing a i64.
|
|
; 3 + 29 = 32, which is less than 63, so we can fold.
|
|
%fold_cst:gpr(i64) = G_CONSTANT i64 29
|
|
%fold_me:gpr(i64) = G_LSHR %copy, %fold_cst
|
|
|
|
%and:gpr(i64) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i64), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: dont_fold_lshr
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: dont_fold_lshr
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr32 = COPY $w0
|
|
; CHECK-NEXT: %fold_cst:gpr32 = MOVi32imm 29
|
|
; CHECK-NEXT: %fold_me:gpr32 = LSRVWr %copy, %fold_cst
|
|
; CHECK-NEXT: TBNZW %fold_me, 3, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i32) = COPY $w0
|
|
%bit:gpr(i32) = G_CONSTANT i32 8
|
|
%zero:gpr(i32) = G_CONSTANT i32 0
|
|
|
|
; We're testing a i32.
|
|
; 3 + 29 = 32, which is greater than 31, so we don't fold.
|
|
%fold_cst:gpr(i32) = G_CONSTANT i32 29
|
|
%fold_me:gpr(i32) = G_LSHR %copy, %fold_cst
|
|
|
|
%and:gpr(i32) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i32), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|
|
|
|
...
|
|
---
|
|
name: lshr_negative
|
|
alignment: 4
|
|
legalized: true
|
|
regBankSelected: true
|
|
body: |
|
|
; CHECK-LABEL: name: lshr_negative
|
|
; CHECK: bb.0:
|
|
; CHECK-NEXT: successors: %bb.0(0x40000000), %bb.1(0x40000000)
|
|
; CHECK-NEXT: liveins: $x0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: %copy:gpr32 = COPY $w0
|
|
; CHECK-NEXT: TBNZW %copy, 2, %bb.1
|
|
; CHECK-NEXT: B %bb.0
|
|
; CHECK-NEXT: {{ $}}
|
|
; CHECK-NEXT: bb.1:
|
|
; CHECK-NEXT: RET_ReallyLR
|
|
bb.0:
|
|
successors: %bb.0, %bb.1
|
|
liveins: $x0
|
|
%copy:gpr(i32) = COPY $w0
|
|
%bit:gpr(i32) = G_CONSTANT i32 8
|
|
%zero:gpr(i32) = G_CONSTANT i32 0
|
|
|
|
; Constant becomes very large and wraps around. Since it's larger than the
|
|
; bit width, that means the LSHR is poison, so we can still fold.
|
|
%fold_cst:gpr(i32) = G_CONSTANT i32 -1
|
|
%fold_me:gpr(i32) = G_LSHR %copy, %fold_cst
|
|
|
|
%and:gpr(i32) = G_AND %fold_me, %bit
|
|
%cmp:gpr(i32) = G_ICMP intpred(ne), %and(i32), %zero
|
|
G_BRCOND %cmp, %bb.1
|
|
G_BR %bb.0
|
|
bb.1:
|
|
RET_ReallyLR
|