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
419 lines
17 KiB
LLVM
419 lines
17 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefixes=GFX6-SDAG %s
|
|
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefixes=GFX6-GISEL %s
|
|
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GFX8-SDAG %s
|
|
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GFX8-GISEL %s
|
|
|
|
declare float @llvm.fabs.f32(float) #0
|
|
declare float @llvm.copysign.f32(float, float) #0
|
|
declare double @llvm.fabs.f64(double) #0
|
|
declare i32 @llvm.amdgcn.frexp.exp.i32.f32(float) #0
|
|
declare i32 @llvm.amdgcn.frexp.exp.i32.f64(double) #0
|
|
|
|
define amdgpu_kernel void @s_test_frexp_exp_f32(ptr addrspace(1) %out, float %src) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_frexp_exp_f32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dword s6, s[4:5], 0xb
|
|
; GFX6-SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s6
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_frexp_exp_f32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dword s3, s[4:5], 0xb
|
|
; GFX6-GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f32_e32 v0, s3
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_frexp_exp_f32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dword s6, s[4:5], 0x2c
|
|
; GFX8-SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s6
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_frexp_exp_f32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dword s3, s[4:5], 0x2c
|
|
; GFX8-GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f32_e32 v0, s3
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f32(float %src)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @s_test_fabs_frexp_exp_f32(ptr addrspace(1) %out, float %src) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_fabs_frexp_exp_f32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dword s6, s[4:5], 0xb
|
|
; GFX6-SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s6
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_fabs_frexp_exp_f32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dword s3, s[4:5], 0xb
|
|
; GFX6-GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f32_e64 v0, |s3|
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_fabs_frexp_exp_f32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dword s6, s[4:5], 0x2c
|
|
; GFX8-SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s6
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_fabs_frexp_exp_f32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dword s3, s[4:5], 0x2c
|
|
; GFX8-GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f32_e64 v0, |s3|
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%fabs.src = call float @llvm.fabs.f32(float %src)
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f32(float %fabs.src)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @s_test_fneg_fabs_frexp_exp_f32(ptr addrspace(1) %out, float %src) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_fneg_fabs_frexp_exp_f32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dword s6, s[4:5], 0xb
|
|
; GFX6-SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s6
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_fneg_fabs_frexp_exp_f32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dword s3, s[4:5], 0xb
|
|
; GFX6-GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f32_e64 v0, -|s3|
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_fneg_fabs_frexp_exp_f32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dword s6, s[4:5], 0x2c
|
|
; GFX8-SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s6
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_fneg_fabs_frexp_exp_f32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dword s3, s[4:5], 0x2c
|
|
; GFX8-GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f32_e64 v0, -|s3|
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%fabs.src = call float @llvm.fabs.f32(float %src)
|
|
%fneg.fabs.src = fneg float %fabs.src
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f32(float %fneg.fabs.src)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @s_test_copysign_frexp_exp_f32(ptr addrspace(1) %out, float %src, float %sign) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_copysign_frexp_exp_f32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: s_mov_b64 s[4:5], s[2:3]
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s4
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_copysign_frexp_exp_f32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: s_and_b32 s4, s2, 0x7fffffff
|
|
; GFX6-GISEL-NEXT: s_and_b32 s3, s3, 0x80000000
|
|
; GFX6-GISEL-NEXT: s_or_b32 s4, s4, s3
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f32_e32 v0, s4
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_copysign_frexp_exp_f32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: s_mov_b64 s[4:5], s[2:3]
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, s4
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_copysign_frexp_exp_f32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: s_and_b32 s4, s2, 0x7fffffff
|
|
; GFX8-GISEL-NEXT: s_and_b32 s3, s3, 0x80000000
|
|
; GFX8-GISEL-NEXT: s_or_b32 s4, s4, s3
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f32_e32 v0, s4
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%copysign = call float @llvm.copysign.f32(float %src, float %sign)
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f32(float %copysign)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @s_test_frexp_exp_f64(ptr addrspace(1) %out, double %src) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_frexp_exp_f64:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s6, -1
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s4, s0
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s5, s1
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[4:7], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_frexp_exp_f64:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_frexp_exp_f64:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_frexp_exp_f64:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f64(double %src)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @s_test_fabs_frexp_exp_f64(ptr addrspace(1) %out, double %src) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_fabs_frexp_exp_f64:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s6, -1
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s4, s0
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s5, s1
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[4:7], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_fabs_frexp_exp_f64:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f64_e64 v0, |s[2:3]|
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_fabs_frexp_exp_f64:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_fabs_frexp_exp_f64:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f64_e64 v0, |s[2:3]|
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%fabs.src = call double @llvm.fabs.f64(double %src)
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f64(double %fabs.src)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @s_test_fneg_fabs_frexp_exp_f64(ptr addrspace(1) %out, double %src) #1 {
|
|
; GFX6-SDAG-LABEL: s_test_fneg_fabs_frexp_exp_f64:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s6, -1
|
|
; GFX6-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s4, s0
|
|
; GFX6-SDAG-NEXT: s_mov_b32 s5, s1
|
|
; GFX6-SDAG-NEXT: buffer_store_dword v0, off, s[4:7], 0
|
|
; GFX6-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX6-GISEL-LABEL: s_test_fneg_fabs_frexp_exp_f64:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
|
|
; GFX6-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f64_e64 v0, -|s[2:3]|
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX6-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX6-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX6-GISEL-NEXT: s_endpgm
|
|
;
|
|
; GFX8-SDAG-LABEL: s_test_fneg_fabs_frexp_exp_f64:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, s[2:3]
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-SDAG-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-SDAG-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-SDAG-NEXT: s_endpgm
|
|
;
|
|
; GFX8-GISEL-LABEL: s_test_fneg_fabs_frexp_exp_f64:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GFX8-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f64_e64 v0, -|s[2:3]|
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s2, -1
|
|
; GFX8-GISEL-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX8-GISEL-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX8-GISEL-NEXT: s_endpgm
|
|
%fabs.src = call double @llvm.fabs.f64(double %src)
|
|
%fneg.fabs.src = fneg double %fabs.src
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f64(double %fneg.fabs.src)
|
|
store i32 %frexp.exp, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps i32 @v_test_frexp_exp_f32(float %src) {
|
|
; GFX6-SDAG-LABEL: v_test_frexp_exp_f32:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, v0
|
|
; GFX6-SDAG-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX6-SDAG-NEXT: ; return to shader part epilog
|
|
;
|
|
; GFX6-GISEL-LABEL: v_test_frexp_exp_f32:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f32_e32 v0, v0
|
|
; GFX6-GISEL-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX6-GISEL-NEXT: ; return to shader part epilog
|
|
;
|
|
; GFX8-SDAG-LABEL: v_test_frexp_exp_f32:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f32_e32 v0, v0
|
|
; GFX8-SDAG-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX8-SDAG-NEXT: ; return to shader part epilog
|
|
;
|
|
; GFX8-GISEL-LABEL: v_test_frexp_exp_f32:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f32_e32 v0, v0
|
|
; GFX8-GISEL-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX8-GISEL-NEXT: ; return to shader part epilog
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f32(float %src)
|
|
ret i32 %frexp.exp
|
|
}
|
|
|
|
define amdgpu_ps i32 @v_test_frexp_exp_f64(double %src) {
|
|
; GFX6-SDAG-LABEL: v_test_frexp_exp_f64:
|
|
; GFX6-SDAG: ; %bb.0:
|
|
; GFX6-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, v[0:1]
|
|
; GFX6-SDAG-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX6-SDAG-NEXT: ; return to shader part epilog
|
|
;
|
|
; GFX6-GISEL-LABEL: v_test_frexp_exp_f64:
|
|
; GFX6-GISEL: ; %bb.0:
|
|
; GFX6-GISEL-NEXT: v_frexp_exp_i32_f64_e32 v0, v[0:1]
|
|
; GFX6-GISEL-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX6-GISEL-NEXT: ; return to shader part epilog
|
|
;
|
|
; GFX8-SDAG-LABEL: v_test_frexp_exp_f64:
|
|
; GFX8-SDAG: ; %bb.0:
|
|
; GFX8-SDAG-NEXT: v_frexp_exp_i32_f64_e32 v0, v[0:1]
|
|
; GFX8-SDAG-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX8-SDAG-NEXT: ; return to shader part epilog
|
|
;
|
|
; GFX8-GISEL-LABEL: v_test_frexp_exp_f64:
|
|
; GFX8-GISEL: ; %bb.0:
|
|
; GFX8-GISEL-NEXT: v_frexp_exp_i32_f64_e32 v0, v[0:1]
|
|
; GFX8-GISEL-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX8-GISEL-NEXT: ; return to shader part epilog
|
|
%frexp.exp = call i32 @llvm.amdgcn.frexp.exp.i32.f64(double %src)
|
|
ret i32 %frexp.exp
|
|
}
|
|
|
|
attributes #0 = { nounwind readnone }
|
|
attributes #1 = { nounwind }
|