Files
TheRock/compiler/amd-llvm/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.load.monitor.gfx1250.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

247 lines
12 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GFX1250,GFX1250-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GFX1250,GFX1250-GISEL %s
declare i32 @llvm.amdgcn.global.load.monitor.b32.i32(ptr addrspace(1), i32, metadata)
declare <2 x i32> @llvm.amdgcn.global.load.monitor.b64.v2i32(ptr addrspace(1), i32, metadata)
declare <4 x i32> @llvm.amdgcn.global.load.monitor.b128.v4i32(ptr addrspace(1), i32, metadata)
declare i32 @llvm.amdgcn.flat.load.monitor.b32.i32(ptr, i32, metadata)
declare <2 x i32> @llvm.amdgcn.flat.load.monitor.b64.v2i32(ptr, i32, metadata)
declare <4 x i32> @llvm.amdgcn.flat.load.monitor.b128.v4i32(ptr, i32, metadata)
define amdgpu_ps void @global_load_monitor_b32_vaddr_relaxed_sys(ptr addrspace(1) %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: global_load_monitor_b32_vaddr_relaxed_sys:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: global_load_monitor_b32 v0, v[0:1], off offset:32 scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b32 v[2:3], v0, off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr i64, ptr addrspace(1) %addr, i32 4
%val = call i32 @llvm.amdgcn.global.load.monitor.b32.i32(ptr addrspace(1) %gep, i32 0, metadata !0)
store i32 %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b32_saddr_relaxed_sys(ptr addrspace(1) inreg %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: global_load_monitor_b32_saddr_relaxed_sys:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: v_mov_b32_e32 v2, 0
; GFX1250-NEXT: global_load_monitor_b32 v2, v2, s[0:1] offset:32 scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b32 v[0:1], v2, off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr i64, ptr addrspace(1) %addr, i32 4
%val = call i32 @llvm.amdgcn.global.load.monitor.b32.i32(ptr addrspace(1) %gep, i32 0, metadata !0)
store i32 %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b64_vaddr_acquire_agent(ptr addrspace(1) %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: global_load_monitor_b64_vaddr_acquire_agent:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: global_load_monitor_b64 v[0:1], v[0:1], off offset:32 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b64 v[2:3], v[0:1], off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr i64, ptr addrspace(1) %addr, i32 4
%val = call <2 x i32> @llvm.amdgcn.global.load.monitor.b64.v2i32(ptr addrspace(1) %gep, i32 2, metadata !1)
store <2 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b64_saddr_acquire_agent(ptr addrspace(1) inreg %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: global_load_monitor_b64_saddr_acquire_agent:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: v_mov_b32_e32 v2, 0
; GFX1250-NEXT: global_load_monitor_b64 v[2:3], v2, s[0:1] offset:32 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b64 v[0:1], v[2:3], off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr i64, ptr addrspace(1) %addr, i32 4
%val = call <2 x i32> @llvm.amdgcn.global.load.monitor.b64.v2i32(ptr addrspace(1) %gep, i32 2, metadata !1)
store <2 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b128_vaddr_seq_cst_workgroup(ptr addrspace(1) %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: global_load_monitor_b128_vaddr_seq_cst_workgroup:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: global_load_monitor_b128 v[4:7], v[0:1], off offset:32
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b128 v[2:3], v[4:7], off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr i64, ptr addrspace(1) %addr, i32 4
%val = call <4 x i32> @llvm.amdgcn.global.load.monitor.b128.v4i32(ptr addrspace(1) %gep, i32 5, metadata !2)
store <4 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b128_saddr_seq_cst_workgroup(ptr addrspace(1) inreg %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: global_load_monitor_b128_saddr_seq_cst_workgroup:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: v_mov_b32_e32 v2, 0
; GFX1250-NEXT: global_load_monitor_b128 v[2:5], v2, s[0:1] offset:32
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b128 v[0:1], v[2:5], off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr i64, ptr addrspace(1) %addr, i32 4
%val = call <4 x i32> @llvm.amdgcn.global.load.monitor.b128.v4i32(ptr addrspace(1) %gep, i32 5, metadata !2)
store <4 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @flat_load_monitor_b32_seq_cst_sys(ptr %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: flat_load_monitor_b32_seq_cst_sys:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: flat_load_monitor_b32 v0, v[0:1] offset:32 scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b32 v[2:3], v0, off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr inbounds i64, ptr addrspace(0) %addr, i32 4
%val = call i32 @llvm.amdgcn.flat.load.monitor.b32.i32(ptr addrspace(0) %gep, i32 5, metadata !0)
store i32 %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @flat_load_monitor_b64_seq_cst_agent(ptr %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: flat_load_monitor_b64_seq_cst_agent:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: flat_load_monitor_b64 v[0:1], v[0:1] offset:32 scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b64 v[2:3], v[0:1], off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr inbounds i64, ptr addrspace(0) %addr, i32 4
%val = call <2 x i32> @llvm.amdgcn.flat.load.monitor.b64.v2i32(ptr addrspace(0) %gep, i32 5, metadata !1)
store <2 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @flat_load_monitor_b128_acquire_sys(ptr %addr, ptr addrspace(1) %use) {
; GFX1250-LABEL: flat_load_monitor_b128_acquire_sys:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: flat_load_monitor_b128 v[4:7], v[0:1] offset:32 scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b128 v[2:3], v[4:7], off
; GFX1250-NEXT: s_endpgm
entry:
%gep = getelementptr inbounds i64, ptr addrspace(0) %addr, i32 4
%val = call <4 x i32> @llvm.amdgcn.flat.load.monitor.b128.v4i32(ptr addrspace(0) %gep, i32 2, metadata !0)
store <4 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b32_saddr_scale_offset_acquire_agent(ptr addrspace(1) inreg %addr, ptr addrspace(1) %use, i32 %idx) {
; GFX1250-LABEL: global_load_monitor_b32_saddr_scale_offset_acquire_agent:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: global_load_monitor_b32 v2, v2, s[0:1] scale_offset scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_DEV
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b32 v[0:1], v2, off
; GFX1250-NEXT: s_endpgm
entry:
%idxprom = sext i32 %idx to i64
%gep = getelementptr i32, ptr addrspace(1) %addr, i64 %idxprom
%val = call i32 @llvm.amdgcn.global.load.monitor.b32.i32(ptr addrspace(1) %gep, i32 2, metadata !1)
store i32 %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b64_saddr_scale_offset_acquire_workgroup(ptr addrspace(1) inreg %addr, ptr addrspace(1) %use, i32 %idx) {
; GFX1250-LABEL: global_load_monitor_b64_saddr_scale_offset_acquire_workgroup:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: global_load_monitor_b64 v[2:3], v2, s[0:1] scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b64 v[0:1], v[2:3], off
; GFX1250-NEXT: s_endpgm
entry:
%idxprom = sext i32 %idx to i64
%gep = getelementptr i64, ptr addrspace(1) %addr, i64 %idxprom
%val = call <2 x i32> @llvm.amdgcn.global.load.monitor.b64.v2i32(ptr addrspace(1) %gep, i32 2, metadata !2)
store <2 x i32> %val, ptr addrspace(1) %use
ret void
}
define amdgpu_ps void @global_load_monitor_b64_saddr_no_scale_offset_seq_cst_sys(ptr addrspace(1) inreg %addr, ptr addrspace(1) %use, i32 %idx) {
; GFX1250-LABEL: global_load_monitor_b64_saddr_no_scale_offset_seq_cst_sys:
; GFX1250: ; %bb.0: ; %entry
; GFX1250-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-NEXT: v_nop
; GFX1250-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-NEXT: v_ashrrev_i32_e32 v3, 31, v2
; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT: v_lshl_add_u64 v[2:3], v[2:3], 2, s[0:1]
; GFX1250-NEXT: global_load_monitor_b64 v[2:3], v[2:3], off scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_inv scope:SCOPE_SYS
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: global_store_b64 v[0:1], v[2:3], off
; GFX1250-NEXT: s_endpgm
entry:
%idxprom = sext i32 %idx to i64
%gep = getelementptr i32, ptr addrspace(1) %addr, i64 %idxprom
%val = call <2 x i32> @llvm.amdgcn.global.load.monitor.b64.v2i32(ptr addrspace(1) %gep, i32 5, metadata !0)
store <2 x i32> %val, ptr addrspace(1) %use
ret void
}
!0 = !{ !"" }
!1 = !{ !"agent" }
!2 = !{ !"workgroup" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX1250-GISEL: {{.*}}
; GFX1250-SDAG: {{.*}}