Files
TheRock/compiler/amd-llvm/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-private-gas.ll
T
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
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

236 lines
12 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -passes='require<libcall-lowering-info>,atomic-expand' %s | FileCheck -check-prefixes=GFX1200 %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 -passes='require<libcall-lowering-info>,atomic-expand' %s | FileCheck -check-prefixes=GFX1250 %s
define void @system_atomic_store_unordered_float(ptr addrspace(5) %addr, float %val) {
; GFX1200-LABEL: define void @system_atomic_store_unordered_float(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], float [[VAL:%.*]]) #[[ATTR0:[0-9]+]] {
; GFX1200-NEXT: store float [[VAL]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret void
;
; GFX1250-LABEL: define void @system_atomic_store_unordered_float(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], float [[VAL:%.*]]) #[[ATTR0:[0-9]+]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: store atomic float [[VAL]], ptr [[SCRATCH_ASCAST]] unordered, align 4
; GFX1250-NEXT: ret void
;
store atomic float %val, ptr addrspace(5) %addr unordered, align 4
ret void
}
define void @system_atomic_store_unordered_i32(ptr addrspace(5) %addr, i32 %val) {
; GFX1200-LABEL: define void @system_atomic_store_unordered_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[VAL:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: store i32 [[VAL]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret void
;
; GFX1250-LABEL: define void @system_atomic_store_unordered_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[VAL:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: store atomic i32 [[VAL]], ptr [[SCRATCH_ASCAST]] unordered, align 4
; GFX1250-NEXT: ret void
;
store atomic i32 %val, ptr addrspace(5) %addr unordered, align 4
ret void
}
define void @system_atomic_store_release_i32(ptr addrspace(5) %addr, i32 %val) {
; GFX1200-LABEL: define void @system_atomic_store_release_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[VAL:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: store i32 [[VAL]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret void
;
; GFX1250-LABEL: define void @system_atomic_store_release_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[VAL:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: store atomic i32 [[VAL]], ptr [[SCRATCH_ASCAST]] release, align 4
; GFX1250-NEXT: ret void
;
store atomic i32 %val, ptr addrspace(5) %addr release, align 4
ret void
}
define void @workgroup_atomic_store_release_i32(ptr addrspace(5) %addr, i32 %val) {
; GFX1200-LABEL: define void @workgroup_atomic_store_release_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[VAL:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: store i32 [[VAL]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret void
;
; GFX1250-LABEL: define void @workgroup_atomic_store_release_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[VAL:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: store atomic i32 [[VAL]], ptr [[SCRATCH_ASCAST]] syncscope("workgroup") release, align 4
; GFX1250-NEXT: ret void
;
store atomic i32 %val, ptr addrspace(5) %addr syncscope("workgroup") release, align 4
ret void
}
define float @system_atomic_load_unordered_float(ptr addrspace(5) %addr) {
; GFX1200-LABEL: define float @system_atomic_load_unordered_float(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[VAL:%.*]] = load float, ptr addrspace(5) [[ADDR]], align 4, !invariant.load [[META0:![0-9]+]], !nontemporal [[META1:![0-9]+]]
; GFX1200-NEXT: ret float [[VAL]]
;
; GFX1250-LABEL: define float @system_atomic_load_unordered_float(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = load atomic float, ptr [[SCRATCH_ASCAST]] unordered, align 4, !invariant.load [[META0:![0-9]+]], !nontemporal [[META1:![0-9]+]]
; GFX1250-NEXT: ret float [[VAL]]
;
%val = load atomic float, ptr addrspace(5) %addr unordered, align 4, !invariant.load !1, !nontemporal !0
ret float %val
}
define i32 @system_atomic_load_unordered_i32(ptr addrspace(5) %addr) {
; GFX1200-LABEL: define i32 @system_atomic_load_unordered_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[VAL:%.*]] = load i32, ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret i32 [[VAL]]
;
; GFX1250-LABEL: define i32 @system_atomic_load_unordered_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = load atomic i32, ptr [[SCRATCH_ASCAST]] unordered, align 4
; GFX1250-NEXT: ret i32 [[VAL]]
;
%val = load atomic i32, ptr addrspace(5) %addr unordered, align 4
ret i32 %val
}
define i32 @system_atomic_load_acquire_i32(ptr addrspace(5) %addr) {
; GFX1200-LABEL: define i32 @system_atomic_load_acquire_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[VAL:%.*]] = load i32, ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret i32 [[VAL]]
;
; GFX1250-LABEL: define i32 @system_atomic_load_acquire_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = load atomic i32, ptr [[SCRATCH_ASCAST]] acquire, align 4
; GFX1250-NEXT: ret i32 [[VAL]]
;
%val = load atomic i32, ptr addrspace(5) %addr acquire, align 4
ret i32 %val
}
define i32 @workgroup_atomic_load_acquire_i32(ptr addrspace(5) %addr) {
; GFX1200-LABEL: define i32 @workgroup_atomic_load_acquire_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[VAL:%.*]] = load i32, ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret i32 [[VAL]]
;
; GFX1250-LABEL: define i32 @workgroup_atomic_load_acquire_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = load atomic i32, ptr [[SCRATCH_ASCAST]] syncscope("workgroup") acquire, align 4
; GFX1250-NEXT: ret i32 [[VAL]]
;
%val = load atomic i32, ptr addrspace(5) %addr syncscope("workgroup") acquire, align 4
ret i32 %val
}
define i32 @system_atomic_cmpxchg_acq_rel_acquire_i32(ptr addrspace(5) %addr, i32 %old, i32 %in) {
; GFX1200-LABEL: define i32 @system_atomic_cmpxchg_acq_rel_acquire_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[OLD:%.*]], i32 [[IN:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], [[OLD]]
; GFX1200-NEXT: [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[IN]], i32 [[TMP1]]
; GFX1200-NEXT: store i32 [[TMP3]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: [[TMP4:%.*]] = insertvalue { i32, i1 } poison, i32 [[TMP1]], 0
; GFX1200-NEXT: [[TMP5:%.*]] = insertvalue { i32, i1 } [[TMP4]], i1 [[TMP2]], 1
; GFX1200-NEXT: [[RES:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; GFX1200-NEXT: ret i32 [[RES]]
;
; GFX1250-LABEL: define i32 @system_atomic_cmpxchg_acq_rel_acquire_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[OLD:%.*]], i32 [[IN:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = cmpxchg volatile ptr [[SCRATCH_ASCAST]], i32 [[OLD]], i32 [[IN]] acq_rel acquire, align 4, !nontemporal [[META1]]
; GFX1250-NEXT: [[RES:%.*]] = extractvalue { i32, i1 } [[VAL]], 0
; GFX1250-NEXT: ret i32 [[RES]]
;
%val = cmpxchg volatile ptr addrspace(5) %addr, i32 %old, i32 %in acq_rel acquire, !nontemporal !0
%res = extractvalue { i32, i1 } %val, 0
ret i32 %res
}
define i32 @system_atomicrmw_xchg_acq_rel_i32(ptr addrspace(5) %addr, i32 %in) {
; GFX1200-LABEL: define i32 @system_atomicrmw_xchg_acq_rel_i32(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[IN:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: store i32 [[IN]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret i32 [[TMP1]]
;
; GFX1250-LABEL: define i32 @system_atomicrmw_xchg_acq_rel_i32(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], i32 [[IN:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = atomicrmw volatile xchg ptr [[SCRATCH_ASCAST]], i32 [[IN]] acq_rel, align 4
; GFX1250-NEXT: ret i32 [[VAL]]
;
%val = atomicrmw volatile xchg ptr addrspace(5) %addr, i32 %in acq_rel
ret i32 %val
}
define i16 @system_atomicrmw_xchg_acq_rel_i16(ptr addrspace(5) %addr, i16 %in) {
; GFX1200-LABEL: define i16 @system_atomicrmw_xchg_acq_rel_i16(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], i16 [[IN:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[TMP1:%.*]] = load i16, ptr addrspace(5) [[ADDR]], align 2
; GFX1200-NEXT: store i16 [[IN]], ptr addrspace(5) [[ADDR]], align 2
; GFX1200-NEXT: ret i16 [[TMP1]]
;
; GFX1250-LABEL: define i16 @system_atomicrmw_xchg_acq_rel_i16(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], i16 [[IN:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = atomicrmw volatile xchg ptr [[SCRATCH_ASCAST]], i16 [[IN]] acq_rel, align 2
; GFX1250-NEXT: ret i16 [[VAL]]
;
%val = atomicrmw volatile xchg ptr addrspace(5) %addr, i16 %in acq_rel
ret i16 %val
}
define half @system_atomicrmw_fmax_acq_rel_half(ptr addrspace(5) %addr, half %in) {
; GFX1200-LABEL: define half @system_atomicrmw_fmax_acq_rel_half(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], half [[IN:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[TMP1:%.*]] = load half, ptr addrspace(5) [[ADDR]], align 2
; GFX1200-NEXT: [[TMP2:%.*]] = call half @llvm.maxnum.f16(half [[TMP1]], half [[IN]])
; GFX1200-NEXT: store half [[TMP2]], ptr addrspace(5) [[ADDR]], align 2
; GFX1200-NEXT: ret half [[TMP1]]
;
; GFX1250-LABEL: define half @system_atomicrmw_fmax_acq_rel_half(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], half [[IN:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = atomicrmw volatile fmax ptr [[SCRATCH_ASCAST]], half [[IN]] acq_rel, align 2
; GFX1250-NEXT: ret half [[VAL]]
;
%val = atomicrmw volatile fmax ptr addrspace(5) %addr, half %in acq_rel
ret half %val
}
define float @system_atomicrmw_fminimum_acq_rel_float(ptr addrspace(5) %addr, float %in) {
; GFX1200-LABEL: define float @system_atomicrmw_fminimum_acq_rel_float(
; GFX1200-SAME: ptr addrspace(5) [[ADDR:%.*]], float [[IN:%.*]]) #[[ATTR0]] {
; GFX1200-NEXT: [[TMP1:%.*]] = load float, ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: [[TMP2:%.*]] = call float @llvm.minimum.f32(float [[TMP1]], float [[IN]])
; GFX1200-NEXT: store float [[TMP2]], ptr addrspace(5) [[ADDR]], align 4
; GFX1200-NEXT: ret float [[TMP1]]
;
; GFX1250-LABEL: define float @system_atomicrmw_fminimum_acq_rel_float(
; GFX1250-SAME: ptr addrspace(5) [[ADDR:%.*]], float [[IN:%.*]]) #[[ATTR0]] {
; GFX1250-NEXT: [[SCRATCH_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[ADDR]] to ptr
; GFX1250-NEXT: [[VAL:%.*]] = atomicrmw volatile fminimum ptr [[SCRATCH_ASCAST]], float [[IN]] acq_rel, align 4, !nontemporal [[META1]]
; GFX1250-NEXT: ret float [[VAL]]
;
%val = atomicrmw volatile fminimum ptr addrspace(5) %addr, float %in acq_rel, !nontemporal !0
ret float %val
}
!0 = !{}
!1 = !{i32 1}
;.
; GFX1200: [[META0]] = !{i32 1}
; GFX1200: [[META1]] = !{}
;.
; GFX1250: [[META0]] = !{i32 1}
; GFX1250: [[META1]] = !{}
;.