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
169 lines
7.2 KiB
LLVM
169 lines
7.2 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --filter-out "s_setreg_imm32_b32" --version 6
|
|
; RUN: llc < %s -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 | FileCheck %s -check-prefixes=GCN,GFX12,GFX12-SDAG
|
|
; RUN: llc < %s -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 | FileCheck %s -check-prefixes=GCN,GFX12,GFX12-GISEL
|
|
; RUN: llc < %s -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 | FileCheck %s -check-prefixes=GCN,GFX1250,GFX1250-SDAG
|
|
; RUN: llc < %s -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1250 | FileCheck %s -check-prefixes=GCN,GFX1250,GFX1250-GISEL
|
|
|
|
declare float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data)
|
|
declare float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
|
|
|
|
define amdgpu_cs void @flat_atomic_fmin_num_f32_noret(ptr %ptr, float %data) {
|
|
; GCN-LABEL: flat_atomic_fmin_num_f32_noret:
|
|
; GCN: ; %bb.0:
|
|
; GCN: flat_atomic_min_num_f32 v[0:1], v2
|
|
; GCN: s_endpgm
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_cs void @flat_atomic_fmax_num_f32_noret(ptr %ptr, float %data) {
|
|
; GCN-LABEL: flat_atomic_fmax_num_f32_noret:
|
|
; GCN: ; %bb.0:
|
|
; GCN: flat_atomic_max_num_f32 v[0:1], v2
|
|
; GCN: s_endpgm
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_cs float @flat_atomic_fmin_num_f32_rtn(float %data, ptr %ptr) {
|
|
; GFX12-LABEL: flat_atomic_fmin_num_f32_rtn:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12: flat_atomic_min_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
|
|
; GFX12: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12: ; return to shader part epilog
|
|
;
|
|
; GFX1250-SDAG-LABEL: flat_atomic_fmin_num_f32_rtn:
|
|
; GFX1250-SDAG: ; %bb.0:
|
|
; GFX1250-SDAG: v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
|
|
; GFX1250-SDAG: flat_atomic_min_num_f32 v0, v[2:3], v0 th:TH_ATOMIC_RETURN
|
|
; GFX1250-SDAG: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-SDAG: ; return to shader part epilog
|
|
;
|
|
; GFX1250-GISEL-LABEL: flat_atomic_fmin_num_f32_rtn:
|
|
; GFX1250-GISEL: ; %bb.0:
|
|
; GFX1250-GISEL: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v5, v2
|
|
; GFX1250-GISEL: flat_atomic_min_num_f32 v0, v[4:5], v0 th:TH_ATOMIC_RETURN
|
|
; GFX1250-GISEL: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-GISEL: ; return to shader part epilog
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret float %ret
|
|
}
|
|
|
|
define amdgpu_cs float @flat_atomic_fmax_num_f32_rtn(float %data, ptr %ptr) {
|
|
; GFX12-LABEL: flat_atomic_fmax_num_f32_rtn:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12: flat_atomic_max_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
|
|
; GFX12: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12: ; return to shader part epilog
|
|
;
|
|
; GFX1250-SDAG-LABEL: flat_atomic_fmax_num_f32_rtn:
|
|
; GFX1250-SDAG: ; %bb.0:
|
|
; GFX1250-SDAG: v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
|
|
; GFX1250-SDAG: flat_atomic_max_num_f32 v0, v[2:3], v0 th:TH_ATOMIC_RETURN
|
|
; GFX1250-SDAG: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-SDAG: ; return to shader part epilog
|
|
;
|
|
; GFX1250-GISEL-LABEL: flat_atomic_fmax_num_f32_rtn:
|
|
; GFX1250-GISEL: ; %bb.0:
|
|
; GFX1250-GISEL: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v5, v2
|
|
; GFX1250-GISEL: flat_atomic_max_num_f32 v0, v[4:5], v0 th:TH_ATOMIC_RETURN
|
|
; GFX1250-GISEL: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-GISEL: ; return to shader part epilog
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret float %ret
|
|
}
|
|
|
|
define amdgpu_ps void @flat_atomic_fmin_num_f32_noret_saddr(ptr inreg %ptr, float %data) {
|
|
; GFX12-SDAG-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
|
|
; GFX12-SDAG: ; %bb.0:
|
|
; GFX12-SDAG: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
|
|
; GFX12-SDAG: flat_atomic_min_num_f32 v[1:2], v0
|
|
; GFX12-SDAG: s_endpgm
|
|
;
|
|
; GFX12-GISEL-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
|
|
; GFX12-GISEL: ; %bb.0:
|
|
; GFX12-GISEL: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX12-GISEL: flat_atomic_min_num_f32 v[1:2], v0
|
|
; GFX12-GISEL: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: flat_atomic_fmin_num_f32_noret_saddr:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250: v_mov_b32_e32 v1, 0
|
|
; GFX1250: flat_atomic_min_num_f32 v1, v0, s[0:1]
|
|
; GFX1250: s_endpgm
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @flat_atomic_fmax_num_f32_noret_saddr(ptr inreg %ptr, float %data) {
|
|
; GFX12-SDAG-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
|
|
; GFX12-SDAG: ; %bb.0:
|
|
; GFX12-SDAG: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
|
|
; GFX12-SDAG: flat_atomic_max_num_f32 v[1:2], v0
|
|
; GFX12-SDAG: s_endpgm
|
|
;
|
|
; GFX12-GISEL-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
|
|
; GFX12-GISEL: ; %bb.0:
|
|
; GFX12-GISEL: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX12-GISEL: flat_atomic_max_num_f32 v[1:2], v0
|
|
; GFX12-GISEL: s_endpgm
|
|
;
|
|
; GFX1250-LABEL: flat_atomic_fmax_num_f32_noret_saddr:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250: v_mov_b32_e32 v1, 0
|
|
; GFX1250: flat_atomic_max_num_f32 v1, v0, s[0:1]
|
|
; GFX1250: s_endpgm
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps float @flat_atomic_fmin_num_f32_rtn_saddr(ptr inreg %ptr, float %data) {
|
|
; GFX12-SDAG-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
|
|
; GFX12-SDAG: ; %bb.0:
|
|
; GFX12-SDAG: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
|
|
; GFX12-SDAG: flat_atomic_min_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
|
|
; GFX12-SDAG: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-SDAG: ; return to shader part epilog
|
|
;
|
|
; GFX12-GISEL-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
|
|
; GFX12-GISEL: ; %bb.0:
|
|
; GFX12-GISEL: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX12-GISEL: flat_atomic_min_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
|
|
; GFX12-GISEL: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-GISEL: ; return to shader part epilog
|
|
;
|
|
; GFX1250-LABEL: flat_atomic_fmin_num_f32_rtn_saddr:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250: v_mov_b32_e32 v1, 0
|
|
; GFX1250: flat_atomic_min_num_f32 v0, v1, v0, s[0:1] th:TH_ATOMIC_RETURN
|
|
; GFX1250: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250: ; return to shader part epilog
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmin.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret float %ret
|
|
}
|
|
|
|
define amdgpu_ps float @flat_atomic_fmax_num_f32_rtn_saddr(ptr inreg %ptr, float %data) {
|
|
; GFX12-SDAG-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
|
|
; GFX12-SDAG: ; %bb.0:
|
|
; GFX12-SDAG: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
|
|
; GFX12-SDAG: flat_atomic_max_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
|
|
; GFX12-SDAG: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-SDAG: ; return to shader part epilog
|
|
;
|
|
; GFX12-GISEL-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
|
|
; GFX12-GISEL: ; %bb.0:
|
|
; GFX12-GISEL: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX12-GISEL: flat_atomic_max_num_f32 v0, v[1:2], v0 th:TH_ATOMIC_RETURN
|
|
; GFX12-GISEL: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-GISEL: ; return to shader part epilog
|
|
;
|
|
; GFX1250-LABEL: flat_atomic_fmax_num_f32_rtn_saddr:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250: v_mov_b32_e32 v1, 0
|
|
; GFX1250: flat_atomic_max_num_f32 v0, v1, v0, s[0:1] th:TH_ATOMIC_RETURN
|
|
; GFX1250: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250: ; return to shader part epilog
|
|
%ret = call float @llvm.amdgcn.flat.atomic.fmax.num.f32.p1.f32(ptr %ptr, float %data)
|
|
ret float %ret
|
|
}
|