Files
TheRock/compiler/amd-llvm/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cooperative.atomic-workgroup.ll
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

498 lines
22 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GFX1250,GFX1250-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GFX1250,GFX1250-GISEL %s
define i32 @test_flat_amdgcn_cooperative_atomic_load_32x4B_relaxed(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_32x4B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b32 v0, v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call i32 @llvm.amdgcn.cooperative.atomic.load.32x4B.p0(ptr %addr, i32 0, metadata !0)
ret i32 %0
}
define <2 x i32> @test_flat_amdgcn_cooperative_atomic_load_16x8B_relaxed(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_16x8B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b64 v[0:1], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <2 x i32> @llvm.amdgcn.cooperative.atomic.load.16x8B.p0(ptr %addr, i32 0, metadata !0)
ret <2 x i32> %0
}
define <4 x i32> @test_flat_amdgcn_cooperative_atomic_load_8x16B_relaxed(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_8x16B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b128 v[0:3], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <4 x i32> @llvm.amdgcn.cooperative.atomic.load.8x16B.p0(ptr %addr, i32 0, metadata !0)
ret <4 x i32> %0
}
define void @test_flat_amdgcn_cooperative_atomic_store_32x4B_relaxed(ptr noundef %addr, i32 noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_32x4B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.32x4B.p0(ptr %addr, i32 %val, i32 0, metadata !0)
ret void
}
define void @test_flat_amdgcn_cooperative_atomic_store_16x8B_relaxed(ptr noundef %addr, <2 x i32> noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_16x8B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_store_b64 v[0:1], v[2:3] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.16x8B.p0(ptr %addr, <2 x i32> %val, i32 0, metadata !0)
ret void
}
define void @test_flat_amdgcn_cooperative_atomic_store_8x16B_relaxed(ptr noundef %addr, <4 x i32> noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_8x16B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_store_b128 v[0:1], v[2:5] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.8x16B.p0(ptr %addr, <4 x i32> %val, i32 0, metadata !0)
ret void
}
define i32 @test_flat_amdgcn_cooperative_atomic_load_32x4B_acquire(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_32x4B_acquire:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b32 v0, v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call i32 @llvm.amdgcn.cooperative.atomic.load.32x4B.p0(ptr %addr, i32 2, metadata !0)
ret i32 %0
}
define <2 x i32> @test_flat_amdgcn_cooperative_atomic_load_16x8B_acquire(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_16x8B_acquire:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b64 v[0:1], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <2 x i32> @llvm.amdgcn.cooperative.atomic.load.16x8B.p0(ptr %addr, i32 2, metadata !0)
ret <2 x i32> %0
}
define <4 x i32> @test_flat_amdgcn_cooperative_atomic_load_8x16B_acquire(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_8x16B_acquire:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b128 v[0:3], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <4 x i32> @llvm.amdgcn.cooperative.atomic.load.8x16B.p0(ptr %addr, i32 2, metadata !0)
ret <4 x i32> %0
}
define void @test_flat_amdgcn_cooperative_atomic_store_32x4B_release(ptr noundef %addr, i32 noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_32x4B_release:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.32x4B.p0(ptr %addr, i32 %val, i32 3, metadata !0)
ret void
}
define void @test_flat_amdgcn_cooperative_atomic_store_16x8B_release(ptr noundef %addr, <2 x i32> noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_16x8B_release:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b64 v[0:1], v[2:3] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.16x8B.p0(ptr %addr, <2 x i32> %val, i32 3, metadata !0)
ret void
}
define void @test_flat_amdgcn_cooperative_atomic_store_8x16B_release(ptr noundef %addr, <4 x i32> noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_8x16B_release:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b128 v[0:1], v[2:5] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.8x16B.p0(ptr %addr, <4 x i32> %val, i32 3, metadata !0)
ret void
}
define i32 @test_flat_amdgcn_cooperative_atomic_load_32x4B_seq_cst(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_32x4B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_load_b32 v0, v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call i32 @llvm.amdgcn.cooperative.atomic.load.32x4B.p0(ptr %addr, i32 5, metadata !0)
ret i32 %0
}
define <2 x i32> @test_flat_amdgcn_cooperative_atomic_load_16x8B_seq_cst(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_16x8B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_load_b64 v[0:1], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <2 x i32> @llvm.amdgcn.cooperative.atomic.load.16x8B.p0(ptr %addr, i32 5, metadata !0)
ret <2 x i32> %0
}
define <4 x i32> @test_flat_amdgcn_cooperative_atomic_load_8x16B_seq_cst(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_load_8x16B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_load_b128 v[0:3], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <4 x i32> @llvm.amdgcn.cooperative.atomic.load.8x16B.p0(ptr %addr, i32 5, metadata !0)
ret <4 x i32> %0
}
define void @test_flat_amdgcn_cooperative_atomic_store_32x4B_seq_cst(ptr noundef %addr, i32 noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_32x4B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.32x4B.p0(ptr %addr, i32 %val, i32 5, metadata !0)
ret void
}
define void @test_flat_amdgcn_cooperative_atomic_store_16x8B_seq_cst(ptr noundef %addr, <2 x i32> noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_16x8B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b64 v[0:1], v[2:3] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.16x8B.p0(ptr %addr, <2 x i32> %val, i32 5, metadata !0)
ret void
}
define void @test_flat_amdgcn_cooperative_atomic_store_8x16B_seq_cst(ptr noundef %addr, <4 x i32> noundef %val) {
; GFX1250-LABEL: test_flat_amdgcn_cooperative_atomic_store_8x16B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b128 v[0:1], v[2:5] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.8x16B.p0(ptr %addr, <4 x i32> %val, i32 5, metadata !0)
ret void
}
define i32 @test_one_as_flat_amdgcn_cooperative_atomic_load_32x4B_relaxed(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_32x4B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b32 v0, v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call i32 @llvm.amdgcn.cooperative.atomic.load.32x4B.p0(ptr %addr, i32 0, metadata !1)
ret i32 %0
}
define <2 x i32> @test_one_as_flat_amdgcn_cooperative_atomic_load_16x8B_relaxed(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_16x8B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b64 v[0:1], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <2 x i32> @llvm.amdgcn.cooperative.atomic.load.16x8B.p0(ptr %addr, i32 0, metadata !1)
ret <2 x i32> %0
}
define <4 x i32> @test_one_as_flat_amdgcn_cooperative_atomic_load_8x16B_relaxed(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_8x16B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b128 v[0:3], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <4 x i32> @llvm.amdgcn.cooperative.atomic.load.8x16B.p0(ptr %addr, i32 0, metadata !1)
ret <4 x i32> %0
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_32x4B_relaxed(ptr noundef %addr, i32 noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_32x4B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.32x4B.p0(ptr %addr, i32 %val, i32 0, metadata !1)
ret void
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_16x8B_relaxed(ptr noundef %addr, <2 x i32> noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_16x8B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_store_b64 v[0:1], v[2:3] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.16x8B.p0(ptr %addr, <2 x i32> %val, i32 0, metadata !1)
ret void
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_8x16B_relaxed(ptr noundef %addr, <4 x i32> noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_8x16B_relaxed:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_store_b128 v[0:1], v[2:5] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.8x16B.p0(ptr %addr, <4 x i32> %val, i32 0, metadata !1)
ret void
}
define i32 @test_one_as_flat_amdgcn_cooperative_atomic_load_32x4B_acquire(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_32x4B_acquire:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b32 v0, v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call i32 @llvm.amdgcn.cooperative.atomic.load.32x4B.p0(ptr %addr, i32 2, metadata !1)
ret i32 %0
}
define <2 x i32> @test_one_as_flat_amdgcn_cooperative_atomic_load_16x8B_acquire(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_16x8B_acquire:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b64 v[0:1], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <2 x i32> @llvm.amdgcn.cooperative.atomic.load.16x8B.p0(ptr %addr, i32 2, metadata !1)
ret <2 x i32> %0
}
define <4 x i32> @test_one_as_flat_amdgcn_cooperative_atomic_load_8x16B_acquire(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_8x16B_acquire:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: flat_load_b128 v[0:3], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <4 x i32> @llvm.amdgcn.cooperative.atomic.load.8x16B.p0(ptr %addr, i32 2, metadata !1)
ret <4 x i32> %0
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_32x4B_release(ptr noundef %addr, i32 noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_32x4B_release:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.32x4B.p0(ptr %addr, i32 %val, i32 3, metadata !1)
ret void
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_16x8B_release(ptr noundef %addr, <2 x i32> noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_16x8B_release:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b64 v[0:1], v[2:3] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.16x8B.p0(ptr %addr, <2 x i32> %val, i32 3, metadata !1)
ret void
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_8x16B_release(ptr noundef %addr, <4 x i32> noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_8x16B_release:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b128 v[0:1], v[2:5] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.8x16B.p0(ptr %addr, <4 x i32> %val, i32 3, metadata !1)
ret void
}
define i32 @test_one_as_flat_amdgcn_cooperative_atomic_load_32x4B_seq_cst(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_32x4B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_load_b32 v0, v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call i32 @llvm.amdgcn.cooperative.atomic.load.32x4B.p0(ptr %addr, i32 5, metadata !1)
ret i32 %0
}
define <2 x i32> @test_one_as_flat_amdgcn_cooperative_atomic_load_16x8B_seq_cst(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_16x8B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_load_b64 v[0:1], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <2 x i32> @llvm.amdgcn.cooperative.atomic.load.16x8B.p0(ptr %addr, i32 5, metadata !1)
ret <2 x i32> %0
}
define <4 x i32> @test_one_as_flat_amdgcn_cooperative_atomic_load_8x16B_seq_cst(ptr noundef readonly %addr) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_load_8x16B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_load_b128 v[0:3], v[0:1] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
%0 = tail call <4 x i32> @llvm.amdgcn.cooperative.atomic.load.8x16B.p0(ptr %addr, i32 5, metadata !1)
ret <4 x i32> %0
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_32x4B_seq_cst(ptr noundef %addr, i32 noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_32x4B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.32x4B.p0(ptr %addr, i32 %val, i32 5, metadata !1)
ret void
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_16x8B_seq_cst(ptr noundef %addr, <2 x i32> noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_16x8B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b64 v[0:1], v[2:3] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.16x8B.p0(ptr %addr, <2 x i32> %val, i32 5, metadata !1)
ret void
}
define void @test_one_as_flat_amdgcn_cooperative_atomic_store_8x16B_seq_cst(ptr noundef %addr, <4 x i32> noundef %val) {
; GFX1250-LABEL: test_one_as_flat_amdgcn_cooperative_atomic_store_8x16B_seq_cst:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: s_wait_kmcnt 0x0
; GFX1250-NEXT: s_wait_storecnt 0x0
; GFX1250-NEXT: flat_store_b128 v[0:1], v[2:5] scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_dscnt 0x0
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
entry:
tail call void @llvm.amdgcn.cooperative.atomic.store.8x16B.p0(ptr %addr, <4 x i32> %val, i32 5, metadata !1)
ret void
}
!0 = !{ !"workgroup" }
!1 = !{ !"workgroup-one-as" }
;; GFX1250: {{.*}}
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX1250-GISEL: {{.*}}
; GFX1250-SDAG: {{.*}}