derek/TheRock
1
0 Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
Files
main
TheRock/compiler/amd-llvm/llvm/test/CodeGen/AMDGPU/scale-offset-scratch.ll
T
derek.holloway 7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
Details
CodeQL / Analyze (python) (push) Canceled after 0s
Details
pre-commit / pre-commit (push) Canceled after 0s
Details
Multi-Arch CI / setup (push) Canceled after 0s
Details
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Details
Setup (Multi-Arch) / setup (push) Canceled after 0s
Details
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Details
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
Details
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Details
Multi-Arch CI / CI Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Details
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
Details
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

552 lines
24 KiB
LLVM
Raw Permalink Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-FAKE16 %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1250 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-FAKE16 %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-REAL16 %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1250 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-REAL16 %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1310 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX13,GFX13-FAKE16 %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1310 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX13,GFX13-FAKE16 %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1310 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX13,GFX13-REAL16 %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1310 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX13,GFX13-REAL16 %s
define amdgpu_ps float @scratch_load_b32_alloca_idxprom(i32 %idx) {
; GFX1250-LABEL: scratch_load_b32_alloca_idxprom:
; 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: scratch_load_b32 v0, v0, off scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b32_alloca_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%p = alloca [64 x i32], align 4, addrspace(5)
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds float, ptr addrspace(5) %p, i64 %idxprom
%ret = load float, ptr addrspace(5) %arrayidx, align 4
ret float %ret
}
define amdgpu_ps float @scratch_load_b32_idxprom(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b32_idxprom:
; 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: scratch_load_b32 v0, v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b32_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = sext i32 %idx to i64
%arrayidx = getelementptr inbounds float, ptr addrspace(5) %p, i64 %idxprom
%ret = load float, ptr addrspace(5) %arrayidx, align 4
ret float %ret
}
define amdgpu_ps float @scratch_load_b32_idx32(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b32_idx32:
; 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: scratch_load_b32 v0, v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b32_idx32:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%arrayidx = getelementptr inbounds float, ptr addrspace(5) %p, i32 %idx
%ret = load float, ptr addrspace(5) %arrayidx, align 4
ret float %ret
}
define amdgpu_ps float @scratch_load_b32_idxprom_wrong_stride(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b32_idxprom_wrong_stride:
; 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_lshlrev_b32_e32 v0, 3, v0
; GFX1250-NEXT: scratch_load_b32 v0, v0, s0
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b32_idxprom_wrong_stride:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: v_lshlrev_b32_e32 v0, 3, v0
; GFX13-NEXT: scratch_load_b32 v0, v0, s0
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds <2 x float>, ptr addrspace(5) %p, i64 %idxprom
%ret = load float, ptr addrspace(5) %arrayidx, align 4
ret float %ret
}
define amdgpu_ps float @scratch_load_b16_idxprom_ioffset(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b16_idxprom_ioffset:
; 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: scratch_load_u16 v0, v0, s0 offset:32 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b16_idxprom_ioffset:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_u16 v0, v0, s0 offset:32 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = sext i32 %idx to i64
%idxadd = add i64 %idxprom, 16
%arrayidx = getelementptr inbounds i16, ptr addrspace(5) %p, i64 %idxadd
%ld = load i16, ptr addrspace(5) %arrayidx, align 2
%ret.i32 = zext i16 %ld to i32
%ret = bitcast i32 %ret.i32 to float
ret float %ret
}
define amdgpu_ps <2 x float> @scratch_load_b64_idxprom(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b64_idxprom:
; 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: scratch_load_b64 v[0:1], v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b64_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b64 v[0:1], v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds <2 x float>, ptr addrspace(5) %p, i64 %idxprom
%ret = load <2 x float>, ptr addrspace(5) %arrayidx, align 4
ret <2 x float> %ret
}
define amdgpu_ps <3 x float> @scratch_load_b96_idxprom(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b96_idxprom:
; 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: scratch_load_b96 v[0:2], v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b96_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b96 v[0:2], v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds [3 x float], ptr addrspace(5) %p, i64 %idxprom
%ret = load <3 x float>, ptr addrspace(5) %arrayidx, align 4
ret <3 x float> %ret
}
define amdgpu_ps <3 x float> @scratch_load_b96_idxpromi_ioffset(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b96_idxpromi_ioffset:
; 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: scratch_load_b96 v[0:2], v0, s0 offset:192 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b96_idxpromi_ioffset:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b96 v[0:2], v0, s0 offset:192 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = zext i32 %idx to i64
%idxadd = add i64 %idxprom, 16
%arrayidx = getelementptr inbounds [3 x float], ptr addrspace(5) %p, i64 %idxadd
%ret = load <3 x float>, ptr addrspace(5) %arrayidx, align 4
ret <3 x float> %ret
}
define amdgpu_ps <4 x float> @scratch_load_b128_idxprom(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_load_b128_idxprom:
; 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: scratch_load_b128 v[0:3], v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b128_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b128 v[0:3], v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds <4 x float>, ptr addrspace(5) %p, i64 %idxprom
%ret = load <4 x float>, ptr addrspace(5) %arrayidx, align 4
ret <4 x float> %ret
}
define amdgpu_ps float @scratch_load_b32_idxprom_range(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b32_idxprom_range:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_b32 v0, v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b32_idxprom_range:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_b32 v0, v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds float, ptr addrspace(5) %p, i64 %idxprom
%ret = load float, ptr addrspace(5) %arrayidx, align 4
ret float %ret
}
define amdgpu_ps float @scratch_load_b32_idxprom_range_ioffset(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b32_idxprom_range_ioffset:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_b32 v0, v0, s0 offset:64 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b32_idxprom_range_ioffset:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_b32 v0, v0, s0 offset:64 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%idxadd = add i64 %idxprom, 16
%arrayidx = getelementptr inbounds float, ptr addrspace(5) %p, i64 %idxadd
%ret = load float, ptr addrspace(5) %arrayidx, align 4
ret float %ret
}
define amdgpu_ps float @scratch_load_b8_idxprom_range_ioffset(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b8_idxprom_range_ioffset:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_u8 v0, v0, s0 offset:16
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b8_idxprom_range_ioffset:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_u8 v0, v0, s0 offset:16
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%idxadd = add i64 %idxprom, 16
%arrayidx = getelementptr inbounds i8, ptr addrspace(5) %p, i64 %idxadd
%ld = load i8, ptr addrspace(5) %arrayidx
%ret.i32 = zext i8 %ld to i32
%ret = bitcast i32 %ret.i32 to float
ret float %ret
}
define amdgpu_ps float @scratch_load_b16_idxprom_range(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b16_idxprom_range:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_u16 v0, v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b16_idxprom_range:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_u16 v0, v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds i16, ptr addrspace(5) %p, i64 %idxprom
%ld = load i16, ptr addrspace(5) %arrayidx, align 2
%ret.i32 = zext i16 %ld to i32
%ret = bitcast i32 %ret.i32 to float
ret float %ret
}
define amdgpu_ps float @scratch_load_b16_idxprom_range_ioffset(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b16_idxprom_range_ioffset:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_u16 v0, v0, s0 offset:32 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b16_idxprom_range_ioffset:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_u16 v0, v0, s0 offset:32 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%idxadd = add i64 %idxprom, 16
%arrayidx = getelementptr inbounds i16, ptr addrspace(5) %p, i64 %idxadd
%ld = load i16, ptr addrspace(5) %arrayidx, align 2
%ret.i32 = zext i16 %ld to i32
%ret = bitcast i32 %ret.i32 to float
ret float %ret
}
define amdgpu_ps <2 x float> @scratch_load_b64_idxprom_range(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b64_idxprom_range:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_b64 v[0:1], v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b64_idxprom_range:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_b64 v[0:1], v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds <2 x float>, ptr addrspace(5) %p, i64 %idxprom
%ret = load <2 x float>, ptr addrspace(5) %arrayidx, align 4
ret <2 x float> %ret
}
; Multiplication is unsigned here, so we cannot match it.
define amdgpu_ps <3 x float> @scratch_load_b96_idxprom_range(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b96_idxprom_range:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_b96 v[0:2], v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b96_idxprom_range:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_b96 v[0:2], v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = sext i32 %idx to i64
%arrayidx = getelementptr inbounds [3 x float], ptr addrspace(5) %p, i64 %idxprom
%ret = load <3 x float>, ptr addrspace(5) %arrayidx, align 4
ret <3 x float> %ret
}
define amdgpu_ps <3 x float> @scratch_load_b96_idxprom_range_ioffset(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b96_idxprom_range_ioffset:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_b96 v[0:2], v0, s0 offset:192 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b96_idxprom_range_ioffset:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_b96 v[0:2], v0, s0 offset:192 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = sext i32 %idx to i64
%idxadd = add i64 %idxprom, 16
%arrayidx = getelementptr inbounds [3 x float], ptr addrspace(5) %p, i64 %idxadd
%ret = load <3 x float>, ptr addrspace(5) %arrayidx, align 4
ret <3 x float> %ret
}
define amdgpu_ps <4 x float> @scratch_load_b128_idxprom_range(ptr addrspace(5) align 4 inreg %p, ptr addrspace(5) align 4 %pp) {
; GFX1250-LABEL: scratch_load_b128_idxprom_range:
; 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: scratch_load_b32 v0, v0, off
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: scratch_load_b128 v[0:3], v0, s0 scale_offset
; GFX1250-NEXT: s_wait_loadcnt 0x0
; GFX1250-NEXT: ; return to shader part epilog
;
; GFX13-LABEL: scratch_load_b128_idxprom_range:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: scratch_load_b32 v0, v0, off
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: scratch_load_b128 v[0:3], v0, s0 scale_offset
; GFX13-NEXT: s_wait_loadcnt 0x0
; GFX13-NEXT: ; return to shader part epilog
entry:
%idx = load i32, ptr addrspace(5) %pp, align 4, !range !0
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds <4 x float>, ptr addrspace(5) %p, i64 %idxprom
%ret = load <4 x float>, ptr addrspace(5) %arrayidx, align 4
ret <4 x float> %ret
}
define amdgpu_ps void @scratch_store_b32_idxprom(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_store_b32_idxprom:
; 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 v1, 1.0
; GFX1250-NEXT: scratch_store_b32 v0, v1, s0 scale_offset
; GFX1250-NEXT: s_endpgm
;
; GFX13-LABEL: scratch_store_b32_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: v_mov_b32_e32 v1, 1.0
; GFX13-NEXT: scratch_store_b32 v0, v1, s0 scale_offset
; GFX13-NEXT: s_endpgm
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds float, ptr addrspace(5) %p, i64 %idxprom
store float 1.0, ptr addrspace(5) %arrayidx, align 4
ret void
}
define amdgpu_ps void @scratch_store_b16_idxprom(ptr addrspace(5) align 2 inreg %p, i32 %idx) {
; GFX1250-FAKE16-LABEL: scratch_store_b16_idxprom:
; GFX1250-FAKE16: ; %bb.0: ; %entry
; GFX1250-FAKE16-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-FAKE16-NEXT: v_nop
; GFX1250-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-FAKE16-NEXT: v_mov_b32_e32 v1, 1
; GFX1250-FAKE16-NEXT: scratch_store_b16 v0, v1, s0 scale_offset
; GFX1250-FAKE16-NEXT: s_endpgm
;
; GFX1250-REAL16-LABEL: scratch_store_b16_idxprom:
; GFX1250-REAL16: ; %bb.0: ; %entry
; GFX1250-REAL16-NEXT: global_prefetch_b8 v0, s[0:1] scope:SCOPE_SE
; GFX1250-REAL16-NEXT: v_nop
; GFX1250-REAL16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
; GFX1250-REAL16-NEXT: v_mov_b16_e32 v1.l, 1
; GFX1250-REAL16-NEXT: scratch_store_b16 v0, v1, s0 scale_offset
; GFX1250-REAL16-NEXT: s_endpgm
;
; GFX13-FAKE16-LABEL: scratch_store_b16_idxprom:
; GFX13-FAKE16: ; %bb.0: ; %entry
; GFX13-FAKE16-NEXT: v_mov_b32_e32 v1, 1
; GFX13-FAKE16-NEXT: scratch_store_b16 v0, v1, s0 scale_offset
; GFX13-FAKE16-NEXT: s_endpgm
;
; GFX13-REAL16-LABEL: scratch_store_b16_idxprom:
; GFX13-REAL16: ; %bb.0: ; %entry
; GFX13-REAL16-NEXT: v_mov_b16_e32 v1.l, 1
; GFX13-REAL16-NEXT: scratch_store_b16 v0, v1, s0 scale_offset
; GFX13-REAL16-NEXT: s_endpgm
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds i16, ptr addrspace(5) %p, i64 %idxprom
store i16 1, ptr addrspace(5) %arrayidx, align 2
ret void
}
define amdgpu_ps void @scratch_store_b64_idxprom(ptr addrspace(5) align 4 inreg %p, i32 %idx) {
; GFX1250-LABEL: scratch_store_b64_idxprom:
; 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_b64_e32 v[2:3], 1.0
; GFX1250-NEXT: scratch_store_b64 v0, v[2:3], s0 scale_offset
; GFX1250-NEXT: s_endpgm
;
; GFX13-LABEL: scratch_store_b64_idxprom:
; GFX13: ; %bb.0: ; %entry
; GFX13-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0x3ff00000
; GFX13-NEXT: scratch_store_b64 v0, v[1:2], s0 scale_offset
; GFX13-NEXT: s_endpgm
entry:
%idxprom = zext i32 %idx to i64
%arrayidx = getelementptr inbounds double, ptr addrspace(5) %p, i64 %idxprom
store double 1.0, ptr addrspace(5) %arrayidx, align 4
ret void
}
!0 = !{i32 0, i32 1024}
Reference in new issue View Git Blame Copy Permalink