fc0653f31c
Recent upstream trends have moved away from explicitly using `-verify-machineinstrs`, as it's already covered by the expensive checks. This PR removes almost all `-verify-machineinstrs` from tests in `llvm/test/CodeGen/AMDGPU/*.ll`, leaving only those tests where its removal currently causes failures.
177 lines
5.6 KiB
LLVM
177 lines
5.6 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1150 < %s | FileCheck %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1150 -global-isel < %s | FileCheck %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -global-isel < %s | FileCheck %s
|
|
|
|
define amdgpu_vs float @sitofp_i32_to_f32(i32 inreg %val) {
|
|
; CHECK-LABEL: sitofp_i32_to_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_f32_i32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = sitofp i32 %val to float
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs float @uitofp_u32_to_f32(i32 inreg %val) {
|
|
; CHECK-LABEL: uitofp_u32_to_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_f32_u32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = uitofp i32 %val to float
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs i32 @fptosi_f32_to_i32(float inreg %val) {
|
|
; CHECK-LABEL: fptosi_f32_to_i32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_i32_f32 s0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = fptosi float %val to i32
|
|
ret i32 %res
|
|
}
|
|
|
|
define amdgpu_vs i32 @fptoui_f32_to_u32(float inreg %val) {
|
|
; CHECK-LABEL: fptoui_f32_to_u32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_u32_f32 s0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = fptoui float %val to i32
|
|
ret i32 %res
|
|
}
|
|
|
|
define amdgpu_vs float @fpext_f16_to_f32(half inreg %val) {
|
|
; CHECK-LABEL: fpext_f16_to_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_f32_f16 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = fpext half %val to float
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs float @fpext_hif16_to_32(<2 x half> inreg %val) {
|
|
; CHECK-LABEL: fpext_hif16_to_32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_hi_f32_f16 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%hielt = extractelement <2 x half> %val, i32 1
|
|
%res = fpext half %hielt to float
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs half @fptrunc_f32_to_f16(float inreg %val) {
|
|
; CHECK-LABEL: fptrunc_f32_to_f16:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_cvt_f16_f32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = fptrunc float %val to half
|
|
ret half %res
|
|
}
|
|
|
|
define amdgpu_vs float @fceil_f32(float inreg %val) {
|
|
; CHECK-LABEL: fceil_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_ceil_f32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call float @llvm.ceil.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs float @ffloor_f32(float inreg %val) {
|
|
; CHECK-LABEL: ffloor_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_floor_f32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call float @llvm.floor.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs float @ftrunc_f32(float inreg %val) {
|
|
; CHECK-LABEL: ftrunc_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_trunc_f32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call float @llvm.trunc.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs float @frint_f32(float inreg %val) {
|
|
; CHECK-LABEL: frint_f32:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_rndne_f32 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call float @llvm.rint.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
define amdgpu_vs half @fceil_f16(half inreg %val) {
|
|
; CHECK-LABEL: fceil_f16:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_ceil_f16 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call half @llvm.ceil.f16(half %val)
|
|
ret half %res
|
|
}
|
|
|
|
define amdgpu_vs half @ffloor_f16(half inreg %val) {
|
|
; CHECK-LABEL: ffloor_f16:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_floor_f16 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call half @llvm.floor.f16(half %val)
|
|
ret half %res
|
|
}
|
|
|
|
define amdgpu_vs half @ftrunc_f16(half inreg %val) {
|
|
; CHECK-LABEL: ftrunc_f16:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_trunc_f16 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call half @llvm.trunc.f16(half %val)
|
|
ret half %res
|
|
}
|
|
|
|
define amdgpu_vs half @frint_f16(half inreg %val) {
|
|
; CHECK-LABEL: frint_f16:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_rndne_f16 s0, s0
|
|
; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: ; return to shader part epilog
|
|
%res = call half @llvm.rint.f16(half %val)
|
|
ret half %res
|
|
}
|
|
|
|
declare float @llvm.ceil.f32(float)
|
|
declare float @llvm.floor.f32(float)
|
|
declare float @llvm.trunc.f32(float)
|
|
declare float @llvm.rint.f32(float)
|
|
declare half @llvm.ceil.f16(half)
|
|
declare half @llvm.floor.f16(half)
|
|
declare half @llvm.trunc.f16(half)
|
|
declare half @llvm.rint.f16(half)
|