THU-DSP-LAB
/
llvm-project
mirror of https://github.com/THU-DSP-LAB/llvm-project.git
2
0
Fork 1
Code Issues Packages Projects Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/mad-mix.ll

1909 lines
82 KiB
LLVM
Raw Permalink Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX900,SDAG-GFX900 %s
; RUN: llc -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX906,SDAG-GFX906 %s
; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=VI,SDAG-VI %s
; RUN: llc -march=amdgcn -mcpu=hawaii -verify-machineinstrs < %s | FileCheck -check-prefixes=CI,SDAG-CI %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX900,GISEL-GFX900 %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX906,GISEL-GFX906 %s
; RUN: llc -global-isel -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=VI,GISEL-VI %s
; RUN: llc -global-isel -march=amdgcn -mcpu=hawaii -verify-machineinstrs < %s | FileCheck -check-prefixes=CI,GISEL-CI %s
define float @v_mad_mix_f32_f16lo_f16lo_f16lo(half %src0, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v3, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; VI-NEXT: v_mac_f32_e32 v0, v3, v1
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_f16hi_f16hi_f16hi_int(i32 %src0, i32 %src1, i32 %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_int:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_int:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_int:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_mac_f32_e32 v0, v3, v1
; VI-NEXT: s_setpc_b64 s[30:31]
;
; CI-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_int:
; CI: ; %bb.0:
; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; CI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; CI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; CI-NEXT: v_cvt_f32_f16_e32 v3, v0
; CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; CI-NEXT: v_mac_f32_e32 v0, v3, v1
; CI-NEXT: s_setpc_b64 s[30:31]
%src0.hi = lshr i32 %src0, 16
%src1.hi = lshr i32 %src1, 16
%src2.hi = lshr i32 %src2, 16
%src0.i16 = trunc i32 %src0.hi to i16
%src1.i16 = trunc i32 %src1.hi to i16
%src2.i16 = trunc i32 %src2.hi to i16
%src0.fp16 = bitcast i16 %src0.i16 to half
%src1.fp16 = bitcast i16 %src1.i16 to half
%src2.fp16 = bitcast i16 %src2.i16 to half
%src0.ext = fpext half %src0.fp16 to float
%src1.ext = fpext half %src1.fp16 to float
%src2.ext = fpext half %src2.fp16 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_f16hi_f16hi_f16hi_elt(<2 x half> %src0, <2 x half> %src1, <2 x half> %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_elt:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_elt:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_elt:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_mac_f32_e32 v0, v3, v1
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_elt:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v1, v3, v5
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16hi_f16hi_f16hi_elt:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v3
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v5
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.hi = extractelement <2 x half> %src0, i32 1
%src1.hi = extractelement <2 x half> %src1, i32 1
%src2.hi = extractelement <2 x half> %src2, i32 1
%src0.ext = fpext half %src0.hi to float
%src1.ext = fpext half %src1.hi to float
%src2.ext = fpext half %src2.hi to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define <2 x float> @v_mad_mix_v2f32(<2 x half> %src0, <2 x half> %src1, <2 x half> %src2) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_v2f32:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: v_mad_mix_f32 v3, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; SDAG-GFX900-NEXT: v_mov_b32_e32 v1, v3
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_v2f32:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: v_fma_mix_f32 v3, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; SDAG-GFX906-NEXT: v_mov_b32_e32 v1, v3
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_v2f32:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v4, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v6, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; SDAG-VI-NEXT: v_mac_f32_e32 v1, v3, v5
; SDAG-VI-NEXT: v_mac_f32_e32 v0, v4, v6
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_v2f32:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v6, v3
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v3, v5
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v5, v6
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v6, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v0, v4
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v4, v6
; SDAG-CI-NEXT: v_mac_f32_e32 v3, v1, v5
; SDAG-CI-NEXT: v_mov_b32_e32 v1, v3
; SDAG-CI-NEXT: v_mac_f32_e32 v0, v4, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_v2f32:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: v_mad_mix_f32 v3, v0, v1, v2 op_sel_hi:[1,1,1]
; GISEL-GFX900-NEXT: v_mad_mix_f32 v1, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; GISEL-GFX900-NEXT: v_mov_b32_e32 v0, v3
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_v2f32:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: v_fma_mix_f32 v3, v0, v1, v2 op_sel_hi:[1,1,1]
; GISEL-GFX906-NEXT: v_fma_mix_f32 v1, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1]
; GISEL-GFX906-NEXT: v_mov_b32_e32 v0, v3
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_v2f32:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v3, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v5, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v3, v5
; GISEL-VI-NEXT: v_mac_f32_e32 v1, v4, v6
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_v2f32:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v6, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v7, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v4
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v5
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v6, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v1, v7, v3
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext <2 x half> %src0 to <2 x float>
%src1.ext = fpext <2 x half> %src1 to <2 x float>
%src2.ext = fpext <2 x half> %src2 to <2 x float>
%result = tail call <2 x float> @llvm.fmuladd.v2f32(<2 x float> %src0.ext, <2 x float> %src1.ext, <2 x float> %src2.ext)
ret <2 x float> %result
}
define <2 x float> @v_mad_mix_v2f32_shuffle(<2 x half> %src0, <2 x half> %src1, <2 x half> %src2) #0 {
; GFX900-LABEL: v_mad_mix_v2f32_shuffle:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v3, v0, v1, v2 op_sel:[1,0,1] op_sel_hi:[1,1,1]
; GFX900-NEXT: v_mad_mix_f32 v1, v0, v1, v2 op_sel:[0,1,1] op_sel_hi:[1,1,1]
; GFX900-NEXT: v_mov_b32_e32 v0, v3
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_v2f32_shuffle:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v3, v0, v1, v2 op_sel:[1,0,1] op_sel_hi:[1,1,1]
; GFX906-NEXT: v_fma_mix_f32 v1, v0, v1, v2 op_sel:[0,1,1] op_sel_hi:[1,1,1]
; GFX906-NEXT: v_mov_b32_e32 v0, v3
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_v2f32_shuffle:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_e32 v4, v0
; VI-NEXT: v_cvt_f32_f16_e32 v0, v1
; VI-NEXT: v_cvt_f32_f16_sdwa v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_mad_f32 v0, v3, v0, v2
; VI-NEXT: v_mac_f32_e32 v2, v4, v1
; VI-NEXT: v_mov_b32_e32 v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_v2f32_shuffle:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v4, v5
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v5, v1
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v1, v4
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v4, v5
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v5, v0
; SDAG-CI-NEXT: v_mad_f32 v0, v4, v2, v1
; SDAG-CI-NEXT: v_mac_f32_e32 v1, v5, v3
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_v2f32_shuffle:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GISEL-CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GISEL-CI-NEXT: v_or_b32_e32 v0, v1, v0
; GISEL-CI-NEXT: v_lshlrev_b32_e32 v1, 16, v5
; GISEL-CI-NEXT: v_and_b32_e32 v4, 0xffff, v4
; GISEL-CI-NEXT: v_or_b32_e32 v1, v1, v4
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v4, 16, v0
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v4, v4
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v5, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v3
; GISEL-CI-NEXT: v_mad_f32 v0, v4, v0, v1
; GISEL-CI-NEXT: v_mac_f32_e32 v1, v5, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.shuf = shufflevector <2 x half> %src0, <2 x half> undef, <2 x i32> <i32 1, i32 0>
%src1.shuf = shufflevector <2 x half> %src1, <2 x half> undef, <2 x i32> <i32 0, i32 1>
%src2.shuf = shufflevector <2 x half> %src2, <2 x half> undef, <2 x i32> <i32 1, i32 1>
%src0.ext = fpext <2 x half> %src0.shuf to <2 x float>
%src1.ext = fpext <2 x half> %src1.shuf to <2 x float>
%src2.ext = fpext <2 x half> %src2.shuf to <2 x float>
%result = tail call <2 x float> @llvm.fmuladd.v2f32(<2 x float> %src0.ext, <2 x float> %src1.ext, <2 x float> %src2.ext)
ret <2 x float> %result
}
define float @v_mad_mix_f32_negf16lo_f16lo_f16lo(half %src0, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_negf16lo_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, -v0, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_negf16lo_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, -v0, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_negf16lo_f16lo_f16lo:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-VI-NEXT: v_mad_f32 v0, -v0, v1, v2
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_negf16lo_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, -v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_negf16lo_f16lo_f16lo:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e64 v3, -v0
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_negf16lo_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e64 v3, -v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%src0.ext.neg = fneg float %src0.ext
%result = tail call float @llvm.fmuladd.f32(float %src0.ext.neg, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_absf16lo_f16lo_f16lo(half %src0, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_absf16lo_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, |v0|, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_absf16lo_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, |v0|, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_absf16lo_f16lo_f16lo:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; VI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_absf16lo_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_absf16lo_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%src0.ext.abs = call float @llvm.fabs.f32(float %src0.ext)
%result = tail call float @llvm.fmuladd.f32(float %src0.ext.abs, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_negabsf16lo_f16lo_f16lo(half %src0, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_negabsf16lo_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, -|v0|, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_negabsf16lo_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, -|v0|, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_negabsf16lo_f16lo_f16lo:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; VI-NEXT: v_mad_f32 v0, -|v0|, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_negabsf16lo_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, -|v0|, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_negabsf16lo_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_mad_f32 v0, -|v0|, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%src0.ext.abs = call float @llvm.fabs.f32(float %src0.ext)
%src0.ext.neg.abs = fneg float %src0.ext.abs
%result = tail call float @llvm.fmuladd.f32(float %src0.ext.neg.abs, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f32(half %src0, half %src1, float %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_negf32(half %src0, half %src1, float %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_negf32:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, -v2 op_sel_hi:[1,1,0]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_negf32:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, -v2 op_sel_hi:[1,1,0]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_negf32:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, -v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_negf32:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, -v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_negf32:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, -v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.neg = fneg float %src2
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.neg)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_absf32(half %src0, half %src1, float %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_absf32:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, |v2| op_sel_hi:[1,1,0]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_absf32:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, |v2| op_sel_hi:[1,1,0]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_absf32:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, |v2|
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_absf32:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, |v2|
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_absf32:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, |v2|
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.abs = call float @llvm.fabs.f32(float %src2)
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.abs)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_negabsf32(half %src0, half %src1, float %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_negabsf32:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, -|v2| op_sel_hi:[1,1,0]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_negabsf32:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, -|v2| op_sel_hi:[1,1,0]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_negabsf32:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, -|v2|
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_negabsf32:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, -|v2|
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_negabsf32:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, -|v2|
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.abs = call float @llvm.fabs.f32(float %src2)
%src2.neg.abs = fneg float %src2.abs
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.neg.abs)
ret float %result
}
; TODO: Fold inline immediates. Need to be careful because it is an
; f16 inline immediate that may be converted to f32, not an actual f32
; inline immediate.
define float @v_mad_mix_f32_f16lo_f16lo_f32imm1(half %src0, half %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 1.0
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 1.0
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, 1.0
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, 1.0
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: v_mov_b32_e32 v2, 1.0
; GISEL-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: v_mov_b32_e32 v2, 1.0
; GISEL-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imm1:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, 1.0
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float 1.0)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f32imminv2pi(half %src0, half %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 0.15915494
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 0.15915494
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, 0.15915494
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_madak_f32 v0, v0, v1, 0x3e22f983
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: v_mov_b32_e32 v2, 0.15915494
; GISEL-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: v_mov_b32_e32 v2, 0.15915494
; GISEL-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32imminv2pi:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mov_b32_e32 v0, 0x3e22f983
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v2, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float 0x3FC45F3060000000)
ret float %result
}
; Attempt to break inline immediate folding. If the operand is
; interpreted as f32, the inline immediate is really the f16 inline
; imm value converted to f32.
; fpext f16 1/2pi = 0x3e230000
; f32 1/2pi = 0x3e22f983
define float @v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi(half %src0, half %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 0x3e230000
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 0x3e230000
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_madak_f32 v0, v0, v1, 0x3e230000
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_madak_f32 v0, v0, v1, 0x3e230000
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: v_mov_b32_e32 v2, 0x3e230000
; GISEL-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: v_mov_b32_e32 v2, 0x3e230000
; GISEL-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_mov_b32_e32 v0, 0x3e230000
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v2, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imminv2pi:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mov_b32_e32 v0, 0x3e230000
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v2, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2 = fpext half 0xH3118 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_cvtf16imm63(half %src0, half %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 0x367c0000
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 0x367c0000
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_madak_f32 v0, v0, v1, 0x367c0000
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_madak_f32 v0, v0, v1, 0x367c0000
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: v_mov_b32_e32 v2, 0x367c0000
; GISEL-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: v_mov_b32_e32 v2, 0x367c0000
; GISEL-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_mov_b32_e32 v0, 0x367c0000
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v2, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_cvtf16imm63:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mov_b32_e32 v0, 0x367c0000
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v2, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2 = fpext half 0xH003F to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2)
ret float %result
}
define <2 x float> @v_mad_mix_v2f32_f32imm1(<2 x half> %src0, <2 x half> %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_v2f32_f32imm1:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 1.0
; SDAG-GFX900-NEXT: v_mad_mix_f32 v2, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: v_mov_b32_e32 v1, v2
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_v2f32_f32imm1:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 1.0
; SDAG-GFX906-NEXT: v_fma_mix_f32 v2, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: v_mov_b32_e32 v1, v2
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_v2f32_f32imm1:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v3, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_mad_f32 v0, v0, v3, 1.0
; SDAG-VI-NEXT: v_mad_f32 v1, v2, v1, 1.0
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_v2f32_f32imm1:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v2, 1.0
; SDAG-CI-NEXT: v_mad_f32 v1, v1, v3, 1.0
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_v2f32_f32imm1:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: s_mov_b32 s4, 1.0
; GISEL-GFX900-NEXT: v_mad_mix_f32 v2, v0, v1, s4 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: v_mad_mix_f32 v1, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: v_mov_b32_e32 v0, v2
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_v2f32_f32imm1:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: s_mov_b32 s4, 1.0
; GISEL-GFX906-NEXT: v_fma_mix_f32 v2, v0, v1, s4 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: v_fma_mix_f32 v1, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: v_mov_b32_e32 v0, v2
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_v2f32_f32imm1:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_mad_f32 v0, v2, v0, 1.0
; GISEL-VI-NEXT: v_mad_f32 v1, v3, v1, 1.0
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_v2f32_f32imm1:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v2, 1.0
; GISEL-CI-NEXT: v_mad_f32 v1, v1, v3, 1.0
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext <2 x half> %src0 to <2 x float>
%src1.ext = fpext <2 x half> %src1 to <2 x float>
%result = tail call <2 x float> @llvm.fmuladd.v2f32(<2 x float> %src0.ext, <2 x float> %src1.ext, <2 x float> <float 1.0, float 1.0>)
ret <2 x float> %result
}
define <2 x float> @v_mad_mix_v2f32_cvtf16imminv2pi(<2 x half> %src0, <2 x half> %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 0x3e230000
; SDAG-GFX900-NEXT: v_mad_mix_f32 v2, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: v_mov_b32_e32 v1, v2
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 0x3e230000
; SDAG-GFX906-NEXT: v_fma_mix_f32 v2, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: v_mov_b32_e32 v1, v2
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v3, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_mov_b32_e32 v1, 0x3e230000
; SDAG-VI-NEXT: v_madak_f32 v0, v0, v3, 0x3e230000
; SDAG-VI-NEXT: v_mac_f32_e32 v1, v2, v4
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v4, v1
; SDAG-CI-NEXT: v_mov_b32_e32 v1, 0x3e230000
; SDAG-CI-NEXT: v_madak_f32 v0, v0, v2, 0x3e230000
; SDAG-CI-NEXT: v_mac_f32_e32 v1, v4, v3
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: s_mov_b32 s4, 0x3e230000
; GISEL-GFX900-NEXT: v_mad_mix_f32 v2, v0, v1, s4 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: v_mad_mix_f32 v1, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: v_mov_b32_e32 v0, v2
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: s_mov_b32 s4, 0x3e230000
; GISEL-GFX906-NEXT: v_fma_mix_f32 v2, v0, v1, s4 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: v_fma_mix_f32 v1, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: v_mov_b32_e32 v0, v2
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: s_mov_b32 s4, 0x3e230000
; GISEL-VI-NEXT: v_mad_f32 v0, v2, v0, s4
; GISEL-VI-NEXT: v_mad_f32 v1, v3, v1, s4
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_v2f32_cvtf16imminv2pi:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; GISEL-CI-NEXT: s_mov_b32 s4, 0x3e230000
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v2, s4
; GISEL-CI-NEXT: v_mad_f32 v1, v1, v3, s4
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext <2 x half> %src0 to <2 x float>
%src1.ext = fpext <2 x half> %src1 to <2 x float>
%src2 = fpext <2 x half> <half 0xH3118, half 0xH3118> to <2 x float>
%result = tail call <2 x float> @llvm.fmuladd.v2f32(<2 x float> %src0.ext, <2 x float> %src1.ext, <2 x float> %src2)
ret <2 x float> %result
}
define <2 x float> @v_mad_mix_v2f32_f32imminv2pi(<2 x half> %src0, <2 x half> %src1) #0 {
; SDAG-GFX900-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; SDAG-GFX900: ; %bb.0:
; SDAG-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX900-NEXT: s_mov_b32 s4, 0.15915494
; SDAG-GFX900-NEXT: v_mad_mix_f32 v2, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX900-NEXT: v_mov_b32_e32 v1, v2
; SDAG-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-GFX906-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; SDAG-GFX906: ; %bb.0:
; SDAG-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-GFX906-NEXT: s_mov_b32 s4, 0.15915494
; SDAG-GFX906-NEXT: v_fma_mix_f32 v2, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, s4 op_sel_hi:[1,1,0]
; SDAG-GFX906-NEXT: v_mov_b32_e32 v1, v2
; SDAG-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v3, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_mad_f32 v0, v0, v3, 0.15915494
; SDAG-VI-NEXT: v_mad_f32 v1, v2, v1, 0.15915494
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v4, v1
; SDAG-CI-NEXT: v_mov_b32_e32 v1, 0x3e22f983
; SDAG-CI-NEXT: v_madak_f32 v0, v0, v2, 0x3e22f983
; SDAG-CI-NEXT: v_mac_f32_e32 v1, v4, v3
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX900-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; GISEL-GFX900: ; %bb.0:
; GISEL-GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX900-NEXT: s_mov_b32 s4, 0.15915494
; GISEL-GFX900-NEXT: v_mad_mix_f32 v2, v0, v1, s4 op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: v_mad_mix_f32 v1, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; GISEL-GFX900-NEXT: v_mov_b32_e32 v0, v2
; GISEL-GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX906-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; GISEL-GFX906: ; %bb.0:
; GISEL-GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-GFX906-NEXT: s_mov_b32 s4, 0.15915494
; GISEL-GFX906-NEXT: v_fma_mix_f32 v2, v0, v1, s4 op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: v_fma_mix_f32 v1, v0, v1, s4 op_sel:[1,1,0] op_sel_hi:[1,1,0]
; GISEL-GFX906-NEXT: v_mov_b32_e32 v0, v2
; GISEL-GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v2, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_mad_f32 v0, v2, v0, 0.15915494
; GISEL-VI-NEXT: v_mad_f32 v1, v3, v1, 0.15915494
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_v2f32_f32imminv2pi:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v3
; GISEL-CI-NEXT: s_mov_b32 s4, 0x3e22f983
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v2, s4
; GISEL-CI-NEXT: v_mad_f32 v1, v1, v3, s4
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext <2 x half> %src0 to <2 x float>
%src1.ext = fpext <2 x half> %src1 to <2 x float>
%src2 = fpext <2 x half> <half 0xH3118, half 0xH3118> to <2 x float>
%result = tail call <2 x float> @llvm.fmuladd.v2f32(<2 x float> %src0.ext, <2 x float> %src1.ext, <2 x float> <float 0x3FC45F3060000000, float 0x3FC45F3060000000>)
ret <2 x float> %result
}
define float @v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt(<2 x half> %src0, <2 x half> %src1, <2 x half> %src2) #0 {
; GFX900-LABEL: v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1] clamp
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel:[1,1,1] op_sel_hi:[1,1,1] clamp
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_sdwa v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_mad_f32 v0, v0, v1, v2 clamp
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v1, v3, v5 clamp
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_clamp_f32_f16hi_f16hi_f16hi_elt:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v3
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v5
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, v2 clamp
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.hi = extractelement <2 x half> %src0, i32 1
%src1.hi = extractelement <2 x half> %src1, i32 1
%src2.hi = extractelement <2 x half> %src2, i32 1
%src0.ext = fpext half %src0.hi to float
%src1.ext = fpext half %src1.hi to float
%src2.ext = fpext half %src2.hi to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
%max = call float @llvm.maxnum.f32(float %result, float 0.0)
%clamp = call float @llvm.minnum.f32(float %max, float 1.0)
ret float %clamp
}
define float @no_mix_simple(float %src0, float %src1, float %src2) #0 {
; GFX900-LABEL: no_mix_simple:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_f32 v0, v0, v1, v2
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: no_mix_simple:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_f32 v0, v0, v1, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: no_mix_simple:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_mad_f32 v0, v0, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; CI-LABEL: no_mix_simple:
; CI: ; %bb.0:
; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CI-NEXT: v_mad_f32 v0, v0, v1, v2
; CI-NEXT: s_setpc_b64 s[30:31]
%result = call float @llvm.fmuladd.f32(float %src0, float %src1, float %src2)
ret float %result
}
define float @no_mix_simple_fabs(float %src0, float %src1, float %src2) #0 {
; GFX900-LABEL: no_mix_simple_fabs:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_f32 v0, |v0|, v1, v2
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: no_mix_simple_fabs:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_f32 v0, |v0|, v1, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: no_mix_simple_fabs:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; CI-LABEL: no_mix_simple_fabs:
; CI: ; %bb.0:
; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; CI-NEXT: s_setpc_b64 s[30:31]
%src0.fabs = call float @llvm.fabs.f32(float %src0)
%result = call float @llvm.fmuladd.f32(float %src0.fabs, float %src1, float %src2)
ret float %result
}
; FIXME(DAG): Should abe able to select in this case.
; All sources are converted from f16, so it doesn't matter
; v_mad_mix_f32 flushes.
define float @v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals(half %src0, half %src1, half %src2) #1 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX900-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX900-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX900-NEXT: v_fma_f32 v0, v0, v1, v2
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; VI-NEXT: v_mul_f32_e32 v0, v0, v1
; VI-NEXT: v_add_f32_e32 v0, v0, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_fma_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_fma_f32 v0, v0, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f32_denormals(half %src0, half %src1, float %src2) #1 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX900-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX900-NEXT: v_fma_f32 v0, v0, v1, v2
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mul_f32_e32 v0, v0, v1
; VI-NEXT: v_add_f32_e32 v0, v0, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_fma_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_fma_f32 v0, v0, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2)
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals_fmulfadd(half %src0, half %src1, half %src2) #1 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals_fmulfadd:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX900-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX900-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX900-NEXT: v_mul_f32_e32 v0, v0, v1
; GFX900-NEXT: v_add_f32_e32 v0, v0, v2
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals_fmulfadd:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX906-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX906-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX906-NEXT: v_mul_f32_e32 v0, v0, v1
; GFX906-NEXT: v_add_f32_e32 v0, v0, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals_fmulfadd:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; VI-NEXT: v_mul_f32_e32 v0, v0, v1
; VI-NEXT: v_add_f32_e32 v0, v0, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals_fmulfadd:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mul_f32_e32 v0, v0, v1
; SDAG-CI-NEXT: v_add_f32_e32 v0, v0, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_denormals_fmulfadd:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_mul_f32_e32 v0, v0, v1
; GISEL-CI-NEXT: v_add_f32_e32 v0, v0, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%mul = fmul float %src0.ext, %src1.ext
%result = fadd float %mul, %src2.ext
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f32_denormals_fmulfadd(half %src0, half %src1, float %src2) #1 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals_fmulfadd:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX900-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX900-NEXT: v_mul_f32_e32 v0, v0, v1
; GFX900-NEXT: v_add_f32_e32 v0, v0, v2
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals_fmulfadd:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX906-NEXT: v_cvt_f32_f16_e32 v1, v1
; GFX906-NEXT: v_mul_f32_e32 v0, v0, v1
; GFX906-NEXT: v_add_f32_e32 v0, v0, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals_fmulfadd:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mul_f32_e32 v0, v0, v1
; VI-NEXT: v_add_f32_e32 v0, v0, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals_fmulfadd:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mul_f32_e32 v0, v0, v1
; SDAG-CI-NEXT: v_add_f32_e32 v0, v0, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_denormals_fmulfadd:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mul_f32_e32 v0, v0, v1
; GISEL-CI-NEXT: v_add_f32_e32 v0, v0, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%mul = fmul float %src0.ext, %src1.ext
%result = fadd float %mul, %src2
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f16lo_f32_flush_fmulfadd(half %src0, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_flush_fmulfadd:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_flush_fmulfadd:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_flush_fmulfadd:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v3, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; VI-NEXT: v_mac_f32_e32 v0, v3, v1
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_flush_fmulfadd:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f16lo_f32_flush_fmulfadd:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%mul = fmul contract float %src0.ext, %src1.ext
%result = fadd contract float %mul, %src2.ext
ret float %result
}
define float @v_mad_mix_f32_f16lo_f16lo_f32_flush_fmulfadd(half %src0, half %src1, float %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_flush_fmulfadd:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_flush_fmulfadd:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, v0, v1, v2 op_sel_hi:[1,1,0]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_flush_fmulfadd:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_mad_f32 v0, v0, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_flush_fmulfadd:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_f16lo_f16lo_f32_flush_fmulfadd:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%mul = fmul contract float %src0.ext, %src1.ext
%result = fadd contract float %mul, %src2
ret float %result
}
define float @v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo(i32 %src0.arg, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, -v0, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, -v0, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-VI-NEXT: v_mad_f32 v0, -v0, v1, v2
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_cvt_f32_f16_e32 v0, v0
; SDAG-CI-NEXT: v_mad_f32 v0, -v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_cvt_f32_f16_e64 v3, -v0
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_negprecvtf16lo_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_cvt_f32_f16_e64 v3, -v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.arg.bc = bitcast i32 %src0.arg to <2 x half>
%src0 = extractelement <2 x half> %src0.arg.bc, i32 0
%src0.neg = fneg half %src0
%src0.ext = fpext half %src0.neg to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
; %src0.ext.neg = fneg float %src0.ext
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
; Make sure we don't fold pre-cvt fneg if we already have a fabs
define float @v_mad_mix_f32_precvtnegf16hi_abs_f16lo_f16lo(i32 %src0.arg, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_precvtnegf16hi_abs_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: s_mov_b32 s4, 0x8000
; GFX900-NEXT: v_xor_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX900-NEXT: v_mad_mix_f32 v0, |v0|, v1, v2 op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_precvtnegf16hi_abs_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: s_mov_b32 s4, 0x8000
; GFX906-NEXT: v_xor_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX906-NEXT: v_fma_mix_f32 v0, |v0|, v1, v2 op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_precvtnegf16hi_abs_f16lo_f16lo:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_sdwa v0, -v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; VI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_precvtnegf16hi_abs_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e64 v0, |v0|
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_precvtnegf16hi_abs_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e64 v0, -v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v2, v2
; GISEL-CI-NEXT: v_mad_f32 v0, |v0|, v1, v2
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.arg.bc = bitcast i32 %src0.arg to <2 x half>
%src0 = extractelement <2 x half> %src0.arg.bc, i32 1
%src0.neg = fneg half %src0
%src0.ext = fpext half %src0.neg to float
%src0.ext.abs = call float @llvm.fabs.f32(float %src0.ext)
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext.abs, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_precvtabsf16hi_f16lo_f16lo(i32 %src0.arg, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_precvtabsf16hi_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, |v0|, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_precvtabsf16hi_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, |v0|, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: v_mad_mix_f32_precvtabsf16hi_f16lo_f16lo:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cvt_f32_f16_sdwa v3, |v0| dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; VI-NEXT: v_mac_f32_e32 v0, v3, v1
; VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_precvtabsf16hi_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e64 v0, |v0|
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_precvtabsf16hi_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e64 v3, |v0|
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.arg.bc = bitcast i32 %src0.arg to <2 x half>
%src0 = extractelement <2 x half> %src0.arg.bc, i32 1
%src0.abs = call half @llvm.fabs.f16(half %src0)
%src0.ext = fpext half %src0.abs to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo(i32 %src0.arg, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, -v0, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, -v0, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-VI-NEXT: v_mad_f32 v0, -v0, v1, v2
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e64 v0, -v0
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_preextractfneg_f16hi_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.arg.bc = bitcast i32 %src0.arg to <2 x half>
%fneg = fneg <2 x half> %src0.arg.bc
%src0 = extractelement <2 x half> %fneg, i32 1
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo(i32 %src0.arg, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, |v0|, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, |v0|, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v3, |v0| dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; SDAG-VI-NEXT: v_mac_f32_e32 v0, v3, v1
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e64 v0, |v0|
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_preextractfabs_f16hi_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.arg.bc = bitcast i32 %src0.arg to <2 x half>
%fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %src0.arg.bc)
%src0 = extractelement <2 x half> %fabs, i32 1
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
define float @v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo(i32 %src0.arg, half %src1, half %src2) #0 {
; GFX900-LABEL: v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo:
; GFX900: ; %bb.0:
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX900-NEXT: v_mad_mix_f32 v0, -|v0|, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX900-NEXT: s_setpc_b64 s[30:31]
;
; GFX906-LABEL: v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_fma_mix_f32 v0, -|v0|, v1, v2 op_sel:[1,0,0] op_sel_hi:[1,1,1]
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-VI-LABEL: v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo:
; SDAG-VI: ; %bb.0:
; SDAG-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-VI-NEXT: v_cvt_f32_f16_sdwa v0, |v0| dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SDAG-VI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SDAG-VI-NEXT: v_mad_f32 v0, -v0, v1, v2
; SDAG-VI-NEXT: s_setpc_b64 s[30:31]
;
; SDAG-CI-LABEL: v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo:
; SDAG-CI: ; %bb.0:
; SDAG-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SDAG-CI-NEXT: v_cvt_f32_f16_e64 v0, -|v0|
; SDAG-CI-NEXT: v_mad_f32 v0, v0, v1, v2
; SDAG-CI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-VI-LABEL: v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo:
; GISEL-VI: ; %bb.0:
; GISEL-VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-VI-NEXT: v_or_b32_e32 v0, 0x80008000, v0
; GISEL-VI-NEXT: v_cvt_f32_f16_sdwa v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-VI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-VI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-VI-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-CI-LABEL: v_mad_mix_f32_preextractfabsfneg_f16hi_f16lo_f16lo:
; GISEL-CI: ; %bb.0:
; GISEL-CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-CI-NEXT: v_or_b32_e32 v0, 0x80008000, v0
; GISEL-CI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v3, v0
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v1, v1
; GISEL-CI-NEXT: v_cvt_f32_f16_e32 v0, v2
; GISEL-CI-NEXT: v_mac_f32_e32 v0, v3, v1
; GISEL-CI-NEXT: s_setpc_b64 s[30:31]
%src0.arg.bc = bitcast i32 %src0.arg to <2 x half>
%fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %src0.arg.bc)
%fneg.fabs = fneg <2 x half> %fabs
%src0 = extractelement <2 x half> %fneg.fabs, i32 1
%src0.ext = fpext half %src0 to float
%src1.ext = fpext half %src1 to float
%src2.ext = fpext half %src2 to float
%result = tail call float @llvm.fmuladd.f32(float %src0.ext, float %src1.ext, float %src2.ext)
ret float %result
}
declare half @llvm.fabs.f16(half) #2
declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #2
declare float @llvm.fabs.f32(float) #2
declare float @llvm.minnum.f32(float, float) #2
declare float @llvm.maxnum.f32(float, float) #2
declare float @llvm.fmuladd.f32(float, float, float) #2
declare <2 x float> @llvm.fmuladd.v2f32(<2 x float>, <2 x float>, <2 x float>) #2
attributes #0 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind "denormal-fp-math-f32"="ieee,ieee" }
attributes #2 = { nounwind readnone speculatable }
Reference in New Issue View Git Blame Copy Permalink