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
232 lines
9.0 KiB
LLVM
232 lines
9.0 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SDAG %s
|
|
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GISEL %s
|
|
|
|
; Load argument depends on waitcnt which should be skipped.
|
|
define amdgpu_kernel void @call_memory_arg_load(ptr addrspace(3) %ptr, i32) #0 {
|
|
; GCN-LABEL: call_memory_arg_load:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_load_dword s6, s[6:7], 0x0
|
|
; GCN-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; GCN-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; GCN-NEXT: s_add_u32 s0, s0, s11
|
|
; GCN-NEXT: s_addc_u32 s1, s1, 0
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_mov_b32_e32 v0, s6
|
|
; GCN-NEXT: ds_read_b32 v0, v0
|
|
; GCN-NEXT: s_getpc_b64 s[8:9]
|
|
; GCN-NEXT: s_add_u32 s8, s8, func@rel32@lo+4
|
|
; GCN-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12
|
|
; GCN-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; GCN-NEXT: s_mov_b32 s32, 0
|
|
; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; GCN-NEXT: s_endpgm
|
|
%vgpr = load volatile i32, ptr addrspace(3) %ptr
|
|
call void @func(i32 %vgpr)
|
|
ret void
|
|
}
|
|
|
|
; Memory waitcnt with no register dependence on the call
|
|
define amdgpu_kernel void @call_memory_no_dep(ptr addrspace(1) %ptr, i32) #0 {
|
|
; SDAG-LABEL: call_memory_no_dep:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_load_dwordx2 s[6:7], s[6:7], 0x0
|
|
; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; SDAG-NEXT: s_add_u32 s0, s0, s11
|
|
; SDAG-NEXT: s_addc_u32 s1, s1, 0
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, 0
|
|
; SDAG-NEXT: s_getpc_b64 s[8:9]
|
|
; SDAG-NEXT: s_add_u32 s8, s8, func@rel32@lo+4
|
|
; SDAG-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12
|
|
; SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SDAG-NEXT: global_store_dword v0, v0, s[6:7]
|
|
; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, 0
|
|
; SDAG-NEXT: s_mov_b32 s32, 0
|
|
; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; SDAG-NEXT: s_endpgm
|
|
;
|
|
; GISEL-LABEL: call_memory_no_dep:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_load_dwordx2 s[6:7], s[6:7], 0x0
|
|
; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; GISEL-NEXT: s_add_u32 s0, s0, s11
|
|
; GISEL-NEXT: s_addc_u32 s1, s1, 0
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GISEL-NEXT: s_getpc_b64 s[8:9]
|
|
; GISEL-NEXT: s_add_u32 s8, s8, func@rel32@lo+4
|
|
; GISEL-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12
|
|
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GISEL-NEXT: global_store_dword v0, v0, s[6:7]
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; GISEL-NEXT: s_mov_b32 s32, 0
|
|
; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; GISEL-NEXT: s_endpgm
|
|
store i32 0, ptr addrspace(1) %ptr
|
|
call void @func(i32 0)
|
|
ret void
|
|
}
|
|
|
|
; Should not wait after the call before memory
|
|
define amdgpu_kernel void @call_no_wait_after_call(ptr addrspace(1) %ptr, i32) #0 {
|
|
; SDAG-LABEL: call_no_wait_after_call:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; SDAG-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0
|
|
; SDAG-NEXT: s_add_u32 s0, s0, s11
|
|
; SDAG-NEXT: s_addc_u32 s1, s1, 0
|
|
; SDAG-NEXT: s_getpc_b64 s[8:9]
|
|
; SDAG-NEXT: s_add_u32 s8, s8, func@rel32@lo+4
|
|
; SDAG-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12
|
|
; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, 0
|
|
; SDAG-NEXT: s_mov_b32 s32, 0
|
|
; SDAG-NEXT: v_mov_b32_e32 v40, 0
|
|
; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; SDAG-NEXT: global_store_dword v40, v40, s[34:35]
|
|
; SDAG-NEXT: s_endpgm
|
|
;
|
|
; GISEL-LABEL: call_no_wait_after_call:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; GISEL-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0
|
|
; GISEL-NEXT: s_add_u32 s0, s0, s11
|
|
; GISEL-NEXT: s_addc_u32 s1, s1, 0
|
|
; GISEL-NEXT: s_getpc_b64 s[8:9]
|
|
; GISEL-NEXT: s_add_u32 s8, s8, func@rel32@lo+4
|
|
; GISEL-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; GISEL-NEXT: s_mov_b32 s32, 0
|
|
; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GISEL-NEXT: global_store_dword v0, v0, s[34:35]
|
|
; GISEL-NEXT: s_endpgm
|
|
call void @func(i32 0)
|
|
store i32 0, ptr addrspace(1) %ptr
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @call_no_wait_after_call_return_val(ptr addrspace(1) %ptr, i32) #0 {
|
|
; SDAG-LABEL: call_no_wait_after_call_return_val:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; SDAG-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0
|
|
; SDAG-NEXT: s_add_u32 s0, s0, s11
|
|
; SDAG-NEXT: s_addc_u32 s1, s1, 0
|
|
; SDAG-NEXT: s_getpc_b64 s[8:9]
|
|
; SDAG-NEXT: s_add_u32 s8, s8, func.return@rel32@lo+4
|
|
; SDAG-NEXT: s_addc_u32 s9, s9, func.return@rel32@hi+12
|
|
; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, 0
|
|
; SDAG-NEXT: s_mov_b32 s32, 0
|
|
; SDAG-NEXT: v_mov_b32_e32 v40, 0
|
|
; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; SDAG-NEXT: global_store_dword v40, v0, s[34:35]
|
|
; SDAG-NEXT: s_endpgm
|
|
;
|
|
; GISEL-LABEL: call_no_wait_after_call_return_val:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; GISEL-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0
|
|
; GISEL-NEXT: s_add_u32 s0, s0, s11
|
|
; GISEL-NEXT: s_addc_u32 s1, s1, 0
|
|
; GISEL-NEXT: s_getpc_b64 s[8:9]
|
|
; GISEL-NEXT: s_add_u32 s8, s8, func.return@rel32@lo+4
|
|
; GISEL-NEXT: s_addc_u32 s9, s9, func.return@rel32@hi+12
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; GISEL-NEXT: s_mov_b32 s32, 0
|
|
; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; GISEL-NEXT: v_mov_b32_e32 v1, 0
|
|
; GISEL-NEXT: global_store_dword v1, v0, s[34:35]
|
|
; GISEL-NEXT: s_endpgm
|
|
%rv = call i32 @func.return(i32 0)
|
|
store i32 %rv, ptr addrspace(1) %ptr
|
|
ret void
|
|
}
|
|
|
|
; Need to wait for the address dependency
|
|
define amdgpu_kernel void @call_got_load(ptr addrspace(1) %ptr, i32) #0 {
|
|
; SDAG-LABEL: call_got_load:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; SDAG-NEXT: s_add_u32 s0, s0, s11
|
|
; SDAG-NEXT: s_addc_u32 s1, s1, 0
|
|
; SDAG-NEXT: s_getpc_b64 s[6:7]
|
|
; SDAG-NEXT: s_add_u32 s6, s6, got.func@gotpcrel32@lo+4
|
|
; SDAG-NEXT: s_addc_u32 s7, s7, got.func@gotpcrel32@hi+12
|
|
; SDAG-NEXT: s_load_dwordx2 s[8:9], s[6:7], 0x0
|
|
; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; SDAG-NEXT: v_mov_b32_e32 v0, 0
|
|
; SDAG-NEXT: s_mov_b32 s32, 0
|
|
; SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; SDAG-NEXT: s_endpgm
|
|
;
|
|
; GISEL-LABEL: call_got_load:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11
|
|
; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0
|
|
; GISEL-NEXT: s_add_u32 s0, s0, s11
|
|
; GISEL-NEXT: s_addc_u32 s1, s1, 0
|
|
; GISEL-NEXT: s_getpc_b64 s[6:7]
|
|
; GISEL-NEXT: s_add_u32 s6, s6, got.func@gotpcrel32@lo+4
|
|
; GISEL-NEXT: s_addc_u32 s7, s7, got.func@gotpcrel32@hi+12
|
|
; GISEL-NEXT: s_load_dwordx2 s[8:9], s[6:7], 0x0
|
|
; GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5]
|
|
; GISEL-NEXT: s_mov_b32 s32, 0
|
|
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9]
|
|
; GISEL-NEXT: s_endpgm
|
|
call void @got.func(i32 0)
|
|
ret void
|
|
}
|
|
|
|
; Need to wait for the address dependency
|
|
define void @tailcall_got_load(ptr addrspace(1) %ptr, i32) #0 {
|
|
; GCN-LABEL: tailcall_got_load:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: s_getpc_b64 s[4:5]
|
|
; GCN-NEXT: s_add_u32 s4, s4, got.func@gotpcrel32@lo+4
|
|
; GCN-NEXT: s_addc_u32 s5, s5, got.func@gotpcrel32@hi+12
|
|
; GCN-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
|
|
; GCN-NEXT: v_mov_b32_e32 v0, 0
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: s_setpc_b64 s[4:5]
|
|
tail call void @got.func(i32 0)
|
|
ret void
|
|
}
|
|
|
|
; No need to wait for the load.
|
|
define void @tail_call_memory_arg_load(ptr addrspace(3) %ptr, i32) #0 {
|
|
; GCN-LABEL: tail_call_memory_arg_load:
|
|
; GCN: ; %bb.0:
|
|
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GCN-NEXT: ds_read_b32 v0, v0
|
|
; GCN-NEXT: s_getpc_b64 s[4:5]
|
|
; GCN-NEXT: s_add_u32 s4, s4, func@rel32@lo+4
|
|
; GCN-NEXT: s_addc_u32 s5, s5, func@rel32@hi+12
|
|
; GCN-NEXT: s_setpc_b64 s[4:5]
|
|
%vgpr = load volatile i32, ptr addrspace(3) %ptr
|
|
tail call void @func(i32 %vgpr)
|
|
ret void
|
|
}
|
|
|
|
declare hidden void @func(i32) #0
|
|
declare hidden i32 @func.return(i32) #0
|
|
declare void @got.func(i32) #0
|
|
|
|
attributes #0 = { nounwind "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-cluster-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-cluster-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-cluster-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" }
|
|
|