Members/tobaru/cbc/CbC_llvm: test/CodeGen/AMDGPU/mad-combine.ll annotate

annotate test/CodeGen/AMDGPU/mad-combine.ll @ 128:c347d3398279 default tip

fix

author	mir3636
date	Wed, 06 Dec 2017 14:37:17 +0900
parents	803732b1fca8
children

rev	line source
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	1 ; Make sure we still form mad even when unsafe math or fp-contract is allowed instead of fma.
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	2
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	3 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s \| FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=SI-STD-SAFE -check-prefix=FUNC %s
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	4 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s \| FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=SI-STD-SAFE -check-prefix=FUNC %s
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	5 ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s \| FileCheck -check-prefix=SI -check-prefix=SI-STD -check-prefix=SI-STD-UNSAFE -check-prefix=FUNC %s
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	6
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	7 ; Make sure we don't form mad with denormals
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	8 ; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=+fp32-denormals -fp-contract=fast -verify-machineinstrs < %s \| FileCheck -check-prefix=SI -check-prefix=SI-DENORM -check-prefix=SI-DENORM-FASTFMAF -check-prefix=FUNC %s
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	9 ; RUN: llc -march=amdgcn -mcpu=verde -mattr=+fp32-denormals -fp-contract=fast -verify-machineinstrs < %s \| FileCheck -check-prefix=SI -check-prefix=SI-DENORM -check-prefix=SI-DENORM-SLOWFMAF -check-prefix=FUNC %s
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	10
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	11 declare i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	12 declare float @llvm.fabs.f32(float) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	13 declare float @llvm.fma.f32(float, float, float) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	14 declare float @llvm.fmuladd.f32(float, float, float) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	15
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	16 ; (fadd (fmul x, y), z) -> (fma x, y, z)
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	17 ; FUNC-LABEL: {{^}}combine_to_mad_f32_0:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	18 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	19 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	20 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	21
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	22 ; SI-STD: v_mac_f32_e32 [[C]], [[A]], [[B]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	23
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	24 ; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	25
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	26 ; SI-DENORM-SLOWFMAF-NOT: v_fma
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	27 ; SI-DENORM-SLOWFMAF-NOT: v_mad
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	28
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	29 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	30 ; SI-DENORM-SLOWFMAF: v_add_f32_e32 [[RESULT:v[0-9]+]], [[TMP]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	31
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	32 ; SI-DENORM: buffer_store_dword [[RESULT]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	33 ; SI-STD: buffer_store_dword [[C]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	34 define amdgpu_kernel void @combine_to_mad_f32_0(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	35 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	36 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	37 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	38 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	39 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	40
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	41 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	42 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	43 %c = load volatile float, float addrspace(1)* %gep.2
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	44
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	45 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	46 %fma = fadd float %mul, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	47 store float %fma, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	48 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	49 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	50
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	51 ; (fadd (fmul x, y), z) -> (fma x, y, z)
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	52 ; FUNC-LABEL: {{^}}combine_to_mad_f32_0_2use:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	53 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	54 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	55 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	56 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	57
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	58 ; SI-STD-DAG: v_mac_f32_e32 [[C]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	59 ; SI-STD-DAG: v_mac_f32_e32 [[D]], [[A]], [[B]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	60
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	61 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	62 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], [[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	63
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	64 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	65 ; SI-DENORM-SLOWFMAF-DAG: v_add_f32_e32 [[RESULT0:v[0-9]+]], [[TMP]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	66 ; SI-DENORM-SLOWFMAF-DAG: v_add_f32_e32 [[RESULT1:v[0-9]+]], [[TMP]], [[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	67
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	68 ; SI-DENORM-DAG: buffer_store_dword [[RESULT0]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	69 ; SI-DENORM-DAG: buffer_store_dword [[RESULT1]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	70 ; SI-STD-DAG: buffer_store_dword [[C]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	71 ; SI-STD-DAG: buffer_store_dword [[D]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	72 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	73 define amdgpu_kernel void @combine_to_mad_f32_0_2use(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	74 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	75 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	76 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	77 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	78 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	79 %gep.out.0 = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	80 %gep.out.1 = getelementptr float, float addrspace(1)* %gep.out.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	81
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	82 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	83 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	84 %c = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	85 %d = load volatile float, float addrspace(1)* %gep.3
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	86
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	87 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	88 %fma0 = fadd float %mul, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	89 %fma1 = fadd float %mul, %d
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	90
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	91 store volatile float %fma0, float addrspace(1)* %gep.out.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	92 store volatile float %fma1, float addrspace(1)* %gep.out.1
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	93 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	94 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	95
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	96 ; (fadd x, (fmul y, z)) -> (fma y, z, x)
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	97 ; FUNC-LABEL: {{^}}combine_to_mad_f32_1:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	98 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	99 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	100 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	101
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	102 ; SI-STD: v_mac_f32_e32 [[C]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	103 ; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	104
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	105 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	106 ; SI-DENORM-SLOWFMAF: v_add_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	107
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	108 ; SI-DENORM: buffer_store_dword [[RESULT]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	109 ; SI-STD: buffer_store_dword [[C]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	110 define amdgpu_kernel void @combine_to_mad_f32_1(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	111 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	112 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	113 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	114 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	115 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	116
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	117 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	118 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	119 %c = load volatile float, float addrspace(1)* %gep.2
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	120
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	121 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	122 %fma = fadd float %c, %mul
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	123 store float %fma, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	124 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	125 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	126
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	127 ; (fsub (fmul x, y), z) -> (fma x, y, (fneg z))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	128 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_0_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	129 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	130 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	131 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	132
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	133 ; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], -[[C]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	134 ; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], [[A]], [[B]], -[[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	135
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	136 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	137 ; SI-DENORM-SLOWFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[TMP]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	138
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	139 ; SI: buffer_store_dword [[RESULT]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	140 define amdgpu_kernel void @combine_to_mad_fsub_0_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	141 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	142 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	143 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	144 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	145 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	146
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	147 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	148 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	149 %c = load volatile float, float addrspace(1)* %gep.2
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	150
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	151 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	152 %fma = fsub float %mul, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	153 store float %fma, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	154 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	155 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	156
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	157 ; (fsub (fmul x, y), z) -> (fma x, y, (fneg z))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	158 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_0_f32_2use:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	159 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	160 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	161 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	162 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	163
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	164 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], -[[C]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	165 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	166
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	167 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], [[A]], [[B]], -[[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	168 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	169
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	170 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	171 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT0:v[0-9]+]], [[TMP]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	172 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT1:v[0-9]+]], [[TMP]], [[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	173
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	174 ; SI-DAG: buffer_store_dword [[RESULT0]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	175 ; SI-DAG: buffer_store_dword [[RESULT1]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	176 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	177 define amdgpu_kernel void @combine_to_mad_fsub_0_f32_2use(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	178 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	179 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	180 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	181 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	182 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	183 %gep.out.0 = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	184 %gep.out.1 = getelementptr float, float addrspace(1)* %gep.out.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	185
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	186 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	187 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	188 %c = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	189 %d = load volatile float, float addrspace(1)* %gep.3
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	190
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	191 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	192 %fma0 = fsub float %mul, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	193 %fma1 = fsub float %mul, %d
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	194 store volatile float %fma0, float addrspace(1)* %gep.out.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	195 store volatile float %fma1, float addrspace(1)* %gep.out.1
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	196 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	197 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	198
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	199 ; (fsub x, (fmul y, z)) -> (fma (fneg y), z, x)
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	200 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_1_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	201 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	202 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	203 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	204
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	205 ; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], [[C]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	206 ; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	207
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	208 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	209 ; SI-DENORM-SLOWFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[C]], [[TMP]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	210
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	211 ; SI: buffer_store_dword [[RESULT]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	212 define amdgpu_kernel void @combine_to_mad_fsub_1_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	213 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	214 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	215 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	216 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	217 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	218
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	219 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	220 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	221 %c = load volatile float, float addrspace(1)* %gep.2
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	222
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	223 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	224 %fma = fsub float %c, %mul
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	225 store float %fma, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	226 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	227 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	228
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	229 ; (fsub x, (fmul y, z)) -> (fma (fneg y), z, x)
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	230 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_1_f32_2use:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	231 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	232 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	233 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	234
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	235 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], [[C]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	236 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], [[D]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	237
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	238 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	239 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], [[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	240
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	241 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	242 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT0:v[0-9]+]], [[C]], [[TMP]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	243 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT1:v[0-9]+]], [[D]], [[TMP]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	244
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	245 ; SI-DAG: buffer_store_dword [[RESULT0]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	246 ; SI-DAG: buffer_store_dword [[RESULT1]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	247 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	248 define amdgpu_kernel void @combine_to_mad_fsub_1_f32_2use(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	249 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	250 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	251 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	252 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	253 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	254 %gep.out.0 = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	255 %gep.out.1 = getelementptr float, float addrspace(1)* %gep.out.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	256
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	257 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	258 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	259 %c = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	260 %d = load volatile float, float addrspace(1)* %gep.3
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	261
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	262 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	263 %fma0 = fsub float %c, %mul
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	264 %fma1 = fsub float %d, %mul
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	265 store volatile float %fma0, float addrspace(1)* %gep.out.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	266 store volatile float %fma1, float addrspace(1)* %gep.out.1
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	267 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	268 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	269
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	270 ; (fsub (fneg (fmul x, y)), z) -> (fma (fneg x), y, (fneg z))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	271 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_2_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	272 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	273 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	274 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	275
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	276 ; SI-STD: v_mad_f32 [[RESULT:v[0-9]+]], [[A]], -[[B]], -[[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	277
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	278 ; SI-DENORM-FASTFMAF: v_fma_f32 [[RESULT:v[0-9]+]], -[[A]], [[B]], -[[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	279
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	280 ; SI-DENORM-SLOWFMAF: v_mul_f32_e64 [[TMP:v[0-9]+]], [[A]], -[[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	281 ; SI-DENORM-SLOWFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[TMP]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	282
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	283 ; SI: buffer_store_dword [[RESULT]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	284 define amdgpu_kernel void @combine_to_mad_fsub_2_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	285 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	286 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	287 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	288 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	289 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	290
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	291 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	292 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	293 %c = load volatile float, float addrspace(1)* %gep.2
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	294
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	295 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	296 %mul.neg = fsub float -0.0, %mul
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	297 %fma = fsub float %mul.neg, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	298
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	299 store float %fma, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	300 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	301 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	302
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	303 ; (fsub (fneg (fmul x, y)), z) -> (fma (fneg x), y, (fneg z))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	304 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_2_f32_2uses_neg:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	305 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	306 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	307 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	308
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	309 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], [[A]], -[[B]], -[[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	310 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], [[A]], -[[B]], -[[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	311
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	312 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	313 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], -[[A]], [[B]], -[[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	314
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	315 ; SI-DENORM-SLOWFMAF: v_mul_f32_e64 [[TMP:v[0-9]+]], [[A]], -[[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	316 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT0:v[0-9]+]], [[TMP]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	317 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT1:v[0-9]+]], [[TMP]], [[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	318
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	319 ; SI-DAG: buffer_store_dword [[RESULT0]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	320 ; SI-DAG: buffer_store_dword [[RESULT1]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	321 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	322 define amdgpu_kernel void @combine_to_mad_fsub_2_f32_2uses_neg(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	323 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	324 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	325 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	326 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	327 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	328 %gep.out.0 = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	329 %gep.out.1 = getelementptr float, float addrspace(1)* %gep.out.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	330
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	331 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	332 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	333 %c = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	334 %d = load volatile float, float addrspace(1)* %gep.3
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	335
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	336 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	337 %mul.neg = fsub float -0.0, %mul
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	338 %fma0 = fsub float %mul.neg, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	339 %fma1 = fsub float %mul.neg, %d
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	340
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	341 store volatile float %fma0, float addrspace(1)* %gep.out.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	342 store volatile float %fma1, float addrspace(1)* %gep.out.1
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	343 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	344 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	345
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	346 ; (fsub (fneg (fmul x, y)), z) -> (fma (fneg x), y, (fneg z))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	347 ; FUNC-LABEL: {{^}}combine_to_mad_fsub_2_f32_2uses_mul:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	348 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	349 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	350 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	351
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	352 ; SI-STD-DAG: v_mad_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	353 ; SI-STD-DAG: v_mad_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	354
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	355 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT0:v[0-9]+]], -[[A]], [[B]], -[[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	356 ; SI-DENORM-FASTFMAF-DAG: v_fma_f32 [[RESULT1:v[0-9]+]], [[A]], [[B]], -[[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	357
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	358 ; SI-DENORM-SLOWFMAF: v_mul_f32_e32 [[TMP:v[0-9]+]], [[A]], [[B]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	359 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e64 [[RESULT0:v[0-9]+]], -[[TMP]], [[C]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	360 ; SI-DENORM-SLOWFMAF-DAG: v_sub_f32_e32 [[RESULT1:v[0-9]+]], [[TMP]], [[D]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	361
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	362 ; SI-DAG: buffer_store_dword [[RESULT0]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	363 ; SI-DAG: buffer_store_dword [[RESULT1]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	364 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	365 define amdgpu_kernel void @combine_to_mad_fsub_2_f32_2uses_mul(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	366 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	367 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	368 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	369 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	370 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	371 %gep.out.0 = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	372 %gep.out.1 = getelementptr float, float addrspace(1)* %gep.out.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	373
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	374 %a = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	375 %b = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	376 %c = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	377 %d = load volatile float, float addrspace(1)* %gep.3
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	378
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	379 %mul = fmul float %a, %b
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	380 %mul.neg = fsub float -0.0, %mul
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	381 %fma0 = fsub float %mul.neg, %c
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	382 %fma1 = fsub float %mul, %d
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	383
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	384 store volatile float %fma0, float addrspace(1)* %gep.out.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	385 store volatile float %fma1, float addrspace(1)* %gep.out.1
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	386 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	387 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	388
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	389 ; fold (fsub (fma x, y, (fmul u, v)), z) -> (fma x, y (fma u, v, (fneg z)))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	390
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	391 ; FUNC-LABEL: {{^}}aggressive_combine_to_mad_fsub_0_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	392 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	393 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	394 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	395 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	396 ; SI-DAG: buffer_load_dword [[E:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	397
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	398 ; SI-STD: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	399 ; SI-STD: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	400 ; SI-STD: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[TMP1]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	401
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	402 ; SI-DENORM: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	403 ; SI-DENORM: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	404 ; SI-DENORM: v_sub_f32_e32 [[RESULT1:v[0-9]+]], [[TMP1]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	405
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	406 ; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	407 define amdgpu_kernel void @aggressive_combine_to_mad_fsub_0_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	408 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	409 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	410 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	411 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	412 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	413 %gep.4 = getelementptr float, float addrspace(1)* %gep.0, i32 4
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	414 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	415
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	416 %x = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	417 %y = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	418 %z = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	419 %u = load volatile float, float addrspace(1)* %gep.3
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	420 %v = load volatile float, float addrspace(1)* %gep.4
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	421
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	422 %tmp0 = fmul float %u, %v
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	423 %tmp1 = call float @llvm.fma.f32(float %x, float %y, float %tmp0) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	424 %tmp2 = fsub float %tmp1, %z
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	425
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	426 store float %tmp2, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	427 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	428 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	429
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	430 ; fold (fsub x, (fma y, z, (fmul u, v)))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	431 ; -> (fma (fneg y), z, (fma (fneg u), v, x))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	432
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	433 ; FUNC-LABEL: {{^}}aggressive_combine_to_mad_fsub_1_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	434 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	435 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	436 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	437 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	438 ; SI-DAG: buffer_load_dword [[E:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	439
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	440 ; SI-STD: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	441 ; SI-STD: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	442 ; SI-STD: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[A]], [[TMP1]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	443
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	444 ; SI-DENORM: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	445 ; SI-DENORM: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	446 ; SI-DENORM: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[A]], [[TMP1]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	447
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	448 ; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	449 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	450 define amdgpu_kernel void @aggressive_combine_to_mad_fsub_1_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	451 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	452 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	453 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	454 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	455 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	456 %gep.4 = getelementptr float, float addrspace(1)* %gep.0, i32 4
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	457 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	458
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	459 %x = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	460 %y = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	461 %z = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	462 %u = load volatile float, float addrspace(1)* %gep.3
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	463 %v = load volatile float, float addrspace(1)* %gep.4
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	464
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	465 %tmp0 = fmul float %u, %v
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	466 %tmp1 = call float @llvm.fma.f32(float %y, float %z, float %tmp0) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	467 %tmp2 = fsub float %x, %tmp1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	468
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	469 store float %tmp2, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	470 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	471 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	472
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	473 ; fold (fsub (fma x, y, (fmul u, v)), z) -> (fma x, y (fma u, v, (fneg z)))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	474
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	475 ; FUNC-LABEL: {{^}}aggressive_combine_to_mad_fsub_2_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	476 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	477 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	478 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	479 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	480 ; SI-DAG: buffer_load_dword [[E:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	481
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	482 ; SI-STD-SAFE: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	483 ; SI-STD-SAFE: v_mac_f32_e32 [[TMP0]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	484 ; SI-STD-SAFE: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[TMP0]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	485
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	486 ; SI-STD-UNSAFE: v_mad_f32 [[RESULT:v[0-9]+]], [[D]], [[E]], -[[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	487 ; SI-STD-UNSAFE: v_mac_f32_e32 [[RESULT]], [[A]], [[B]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	488
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	489 ; SI-DENORM-FASTFMAF: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	490 ; SI-DENORM-FASTFMAF: v_fma_f32 [[TMP1:v[0-9]+]], [[A]], [[B]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	491 ; SI-DENORM-FASTFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[TMP1]], [[C]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	492
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	493 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	494 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP1:v[0-9]+]], [[A]], [[B]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	495 ; SI-DENORM-SLOWFMAF: v_add_f32_e32 [[TMP2:v[0-9]+]], [[TMP1]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	496 ; SI-DENORM-SLOWFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[TMP2]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	497
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	498 ; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	499 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	500 define amdgpu_kernel void @aggressive_combine_to_mad_fsub_2_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	501 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	502 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	503 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	504 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	505 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	506 %gep.4 = getelementptr float, float addrspace(1)* %gep.0, i32 4
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	507 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	508
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	509 %x = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	510 %y = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	511 %z = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	512 %u = load volatile float, float addrspace(1)* %gep.3
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	513 %v = load volatile float, float addrspace(1)* %gep.4
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	514
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	515 %tmp0 = fmul float %u, %v
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	516 %tmp1 = call float @llvm.fmuladd.f32(float %x, float %y, float %tmp0) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	517 %tmp2 = fsub float %tmp1, %z
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	518
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	519 store float %tmp2, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	520 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	521 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	522
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	523 ; fold (fsub x, (fmuladd y, z, (fmul u, v)))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	524 ; -> (fmuladd (fneg y), z, (fmuladd (fneg u), v, x))
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	525
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	526 ; FUNC-LABEL: {{^}}aggressive_combine_to_mad_fsub_3_f32:
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	527 ; SI-DAG: buffer_load_dword [[A:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	528 ; SI-DAG: buffer_load_dword [[B:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	529 ; SI-DAG: buffer_load_dword [[C:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	530 ; SI-DAG: buffer_load_dword [[D:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:12{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	531 ; SI-DAG: buffer_load_dword [[E:v[0-9]+]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:16{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	532
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	533 ; SI-STD-SAFE: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	534 ; SI-STD-SAFE: v_mac_f32_e32 [[TMP0]], [[B]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	535 ; SI-STD-SAFE: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[A]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	536
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	537 ; SI-STD-UNSAFE: v_mad_f32 [[TMP:v[0-9]+]], -[[D]], [[E]], [[A]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	538 ; SI-STD-UNSAFE: v_mad_f32 [[RESULT:v[0-9]+]], -[[B]], [[C]], [[TMP]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	539
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	540 ; SI-DENORM-FASTFMAF: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	541 ; SI-DENORM-FASTFMAF: v_fma_f32 [[TMP1:v[0-9]+]], [[B]], [[C]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	542 ; SI-DENORM-FASTFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[A]], [[TMP1]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	543
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	544 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP0:v[0-9]+]], [[D]], [[E]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	545 ; SI-DENORM-SLOWFMAF-DAG: v_mul_f32_e32 [[TMP1:v[0-9]+]], [[B]], [[C]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	546 ; SI-DENORM-SLOWFMAF: v_add_f32_e32 [[TMP2:v[0-9]+]], [[TMP1]], [[TMP0]]
803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	547 ; SI-DENORM-SLOWFMAF: v_sub_f32_e32 [[RESULT:v[0-9]+]], [[A]], [[TMP2]]
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	548
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	549 ; SI: buffer_store_dword [[RESULT]], v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	550 ; SI: s_endpgm
121 803732b1fca8 LLVM 5.0 kono parents: 120 diff changeset	551 define amdgpu_kernel void @aggressive_combine_to_mad_fsub_3_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) #1 {
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	552 %tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	553 %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	554 %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	555 %gep.2 = getelementptr float, float addrspace(1)* %gep.0, i32 2
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	556 %gep.3 = getelementptr float, float addrspace(1)* %gep.0, i32 3
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	557 %gep.4 = getelementptr float, float addrspace(1)* %gep.0, i32 4
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	558 %gep.out = getelementptr float, float addrspace(1)* %out, i32 %tid
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	559
120 1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	560 %x = load volatile float, float addrspace(1)* %gep.0
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	561 %y = load volatile float, float addrspace(1)* %gep.1
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	562 %z = load volatile float, float addrspace(1)* %gep.2
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	563 %u = load volatile float, float addrspace(1)* %gep.3
1172e4bd9c6f update 4.0.0 mir3636 parents: 95 diff changeset	564 %v = load volatile float, float addrspace(1)* %gep.4
95 afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	565
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	566 %tmp0 = fmul float %u, %v
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	567 %tmp1 = call float @llvm.fmuladd.f32(float %y, float %z, float %tmp0) #0
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	568 %tmp2 = fsub float %x, %tmp1
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	569
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	570 store float %tmp2, float addrspace(1)* %gep.out
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	571 ret void
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	572 }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	573
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	574 attributes #0 = { nounwind readnone }
afa8332a0e37 LLVM 3.8 Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp> parents: diff changeset	575 attributes #1 = { nounwind }

Mercurial > hg > Members > tobaru > cbc > CbC_llvm

annotate test/CodeGen/AMDGPU/mad-combine.ll @ 128:c347d3398279 default tip