[X86] Remove X86ISD::FMADD in favor ISD::FMA

There's no reason to have a target specific node with the same semantics as a target independent opcode.

This should simplify D36335 so that it doesn't need to touch X86ISelDAGToDAG.cpp

Differential Revision: https://reviews.llvm.org/D36983

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@311568 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Craig Topper 2017-08-23 16:28:04 +00:00
parent d0ef4bebe9
commit 47b719dfa8
6 changed files with 43 additions and 45 deletions

View File

@ -24646,7 +24646,6 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
case X86ISD::VPCOM: return "X86ISD::VPCOM";
case X86ISD::VPCOMU: return "X86ISD::VPCOMU";
case X86ISD::VPERMIL2: return "X86ISD::VPERMIL2";
case X86ISD::FMADD: return "X86ISD::FMADD";
case X86ISD::FMSUB: return "X86ISD::FMSUB";
case X86ISD::FNMADD: return "X86ISD::FNMADD";
case X86ISD::FNMSUB: return "X86ISD::FNMSUB";
@ -34011,10 +34010,10 @@ static SDValue combineFneg(SDNode *N, SelectionDAG &DAG,
unsigned NewOpcode = 0;
if (Arg.hasOneUse()) {
switch (Arg.getOpcode()) {
case X86ISD::FMADD: NewOpcode = X86ISD::FNMSUB; break;
case ISD::FMA: NewOpcode = X86ISD::FNMSUB; break;
case X86ISD::FMSUB: NewOpcode = X86ISD::FNMADD; break;
case X86ISD::FNMADD: NewOpcode = X86ISD::FMSUB; break;
case X86ISD::FNMSUB: NewOpcode = X86ISD::FMADD; break;
case X86ISD::FNMSUB: NewOpcode = ISD::FMA; break;
case X86ISD::FMADD_RND: NewOpcode = X86ISD::FNMSUB_RND; break;
case X86ISD::FMSUB_RND: NewOpcode = X86ISD::FNMADD_RND; break;
case X86ISD::FNMADD_RND: NewOpcode = X86ISD::FMSUB_RND; break;
@ -34608,14 +34607,14 @@ static SDValue combineFMA(SDNode *N, SelectionDAG &DAG,
unsigned NewOpcode;
if (!NegMul)
NewOpcode = (!NegC) ? X86ISD::FMADD : X86ISD::FMSUB;
NewOpcode = (!NegC) ? ISD::FMA : X86ISD::FMSUB;
else
NewOpcode = (!NegC) ? X86ISD::FNMADD : X86ISD::FNMSUB;
// For FMA and FMADD, we risk reconstructing the node we started with.
// For FMA, we risk reconstructing the node we started with.
// In order to avoid this, we check for negation or opcode change. If
// one of the two happened, then it is a new node and we return it.
if (N->getOpcode() == X86ISD::FMADD || N->getOpcode() == ISD::FMA) {
if (N->getOpcode() == ISD::FMA) {
if (HasNeg || NewOpcode != N->getOpcode())
return DAG.getNode(NewOpcode, dl, VT, A, B, C);
return SDValue();
@ -34623,21 +34622,21 @@ static SDValue combineFMA(SDNode *N, SelectionDAG &DAG,
if (N->getOpcode() == X86ISD::FMADD_RND) {
switch (NewOpcode) {
case X86ISD::FMADD: NewOpcode = X86ISD::FMADD_RND; break;
case ISD::FMA: NewOpcode = X86ISD::FMADD_RND; break;
case X86ISD::FMSUB: NewOpcode = X86ISD::FMSUB_RND; break;
case X86ISD::FNMADD: NewOpcode = X86ISD::FNMADD_RND; break;
case X86ISD::FNMSUB: NewOpcode = X86ISD::FNMSUB_RND; break;
}
} else if (N->getOpcode() == X86ISD::FMADDS1_RND) {
switch (NewOpcode) {
case X86ISD::FMADD: NewOpcode = X86ISD::FMADDS1_RND; break;
case ISD::FMA: NewOpcode = X86ISD::FMADDS1_RND; break;
case X86ISD::FMSUB: NewOpcode = X86ISD::FMSUBS1_RND; break;
case X86ISD::FNMADD: NewOpcode = X86ISD::FNMADDS1_RND; break;
case X86ISD::FNMSUB: NewOpcode = X86ISD::FNMSUBS1_RND; break;
}
} else if (N->getOpcode() == X86ISD::FMADDS3_RND) {
switch (NewOpcode) {
case X86ISD::FMADD: NewOpcode = X86ISD::FMADDS3_RND; break;
case ISD::FMA: NewOpcode = X86ISD::FMADDS3_RND; break;
case X86ISD::FMSUB: NewOpcode = X86ISD::FMSUBS3_RND; break;
case X86ISD::FNMADD: NewOpcode = X86ISD::FNMADDS3_RND; break;
case X86ISD::FNMSUB: NewOpcode = X86ISD::FNMSUBS3_RND; break;
@ -35789,7 +35788,6 @@ SDValue X86TargetLowering::PerformDAGCombine(SDNode *N,
case X86ISD::VPERM2X128:
case X86ISD::VZEXT_MOVL:
case ISD::VECTOR_SHUFFLE: return combineShuffle(N, DAG, DCI,Subtarget);
case X86ISD::FMADD:
case X86ISD::FMADD_RND:
case X86ISD::FMADDS1_RND:
case X86ISD::FMADDS3_RND:

View File

@ -474,7 +474,7 @@ namespace llvm {
VPMADD52L, VPMADD52H,
// FMA nodes.
FMADD,
// We use the target independent ISD::FMA for the non-inverted case.
FNMADD,
FMSUB,
FNMSUB,

View File

@ -6261,7 +6261,7 @@ multiclass avx512_fma3p_213_f<bits<8> opc, string OpcodeStr, SDNode OpNode,
avx512vl_f64_info, "PD">, VEX_W;
}
defm VFMADD213 : avx512_fma3p_213_f<0xA8, "vfmadd213", X86Fmadd, X86FmaddRnd>;
defm VFMADD213 : avx512_fma3p_213_f<0xA8, "vfmadd213", fma, X86FmaddRnd>;
defm VFMSUB213 : avx512_fma3p_213_f<0xAA, "vfmsub213", X86Fmsub, X86FmsubRnd>;
defm VFMADDSUB213 : avx512_fma3p_213_f<0xA6, "vfmaddsub213", X86Fmaddsub, X86FmaddsubRnd>;
defm VFMSUBADD213 : avx512_fma3p_213_f<0xA7, "vfmsubadd213", X86Fmsubadd, X86FmsubaddRnd>;
@ -6346,7 +6346,7 @@ multiclass avx512_fma3p_231_f<bits<8> opc, string OpcodeStr, SDNode OpNode,
avx512vl_f64_info, "PD">, VEX_W;
}
defm VFMADD231 : avx512_fma3p_231_f<0xB8, "vfmadd231", X86Fmadd, X86FmaddRnd>;
defm VFMADD231 : avx512_fma3p_231_f<0xB8, "vfmadd231", fma, X86FmaddRnd>;
defm VFMSUB231 : avx512_fma3p_231_f<0xBA, "vfmsub231", X86Fmsub, X86FmsubRnd>;
defm VFMADDSUB231 : avx512_fma3p_231_f<0xB6, "vfmaddsub231", X86Fmaddsub, X86FmaddsubRnd>;
defm VFMSUBADD231 : avx512_fma3p_231_f<0xB7, "vfmsubadd231", X86Fmsubadd, X86FmsubaddRnd>;
@ -6420,7 +6420,7 @@ multiclass avx512_fma3p_132_f<bits<8> opc, string OpcodeStr, SDNode OpNode,
avx512vl_f64_info, "PD">, VEX_W;
}
defm VFMADD132 : avx512_fma3p_132_f<0x98, "vfmadd132", X86Fmadd, X86FmaddRnd>;
defm VFMADD132 : avx512_fma3p_132_f<0x98, "vfmadd132", fma, X86FmaddRnd>;
defm VFMSUB132 : avx512_fma3p_132_f<0x9A, "vfmsub132", X86Fmsub, X86FmsubRnd>;
defm VFMADDSUB132 : avx512_fma3p_132_f<0x96, "vfmaddsub132", X86Fmaddsub, X86FmaddsubRnd>;
defm VFMSUBADD132 : avx512_fma3p_132_f<0x97, "vfmsubadd132", X86Fmsubadd, X86FmsubaddRnd>;
@ -6514,7 +6514,7 @@ multiclass avx512_fma3s<bits<8> opc213, bits<8> opc231, bits<8> opc132,
}
}
defm VFMADD : avx512_fma3s<0xA9, 0xB9, 0x99, "vfmadd", X86Fmadd, X86FmaddRnds1,
defm VFMADD : avx512_fma3s<0xA9, 0xB9, 0x99, "vfmadd", fma, X86FmaddRnds1,
X86FmaddRnds3>;
defm VFMSUB : avx512_fma3s<0xAB, 0xBB, 0x9B, "vfmsub", X86Fmsub, X86FmsubRnds1,
X86FmsubRnds3>;

View File

@ -89,7 +89,7 @@ multiclass fma3p_forms<bits<8> opc132, bits<8> opc213, bits<8> opc231,
// Fused Multiply-Add
let ExeDomain = SSEPackedSingle in {
defm VFMADD : fma3p_forms<0x98, 0xA8, 0xB8, "vfmadd", "ps", "PS",
loadv4f32, loadv8f32, X86Fmadd, v4f32, v8f32>;
loadv4f32, loadv8f32, fma, v4f32, v8f32>;
defm VFMSUB : fma3p_forms<0x9A, 0xAA, 0xBA, "vfmsub", "ps", "PS",
loadv4f32, loadv8f32, X86Fmsub, v4f32, v8f32>;
defm VFMADDSUB : fma3p_forms<0x96, 0xA6, 0xB6, "vfmaddsub", "ps", "PS",
@ -102,7 +102,7 @@ let ExeDomain = SSEPackedSingle in {
let ExeDomain = SSEPackedDouble in {
defm VFMADD : fma3p_forms<0x98, 0xA8, 0xB8, "vfmadd", "pd", "PD",
loadv2f64, loadv4f64, X86Fmadd, v2f64,
loadv2f64, loadv4f64, fma, v2f64,
v4f64>, VEX_W;
defm VFMSUB : fma3p_forms<0x9A, 0xAA, 0xBA, "vfmsub", "pd", "PD",
loadv2f64, loadv4f64, X86Fmsub, v2f64,
@ -254,7 +254,7 @@ multiclass fma3s<bits<8> opc132, bits<8> opc213, bits<8> opc231,
}
defm VFMADD : fma3s<0x99, 0xA9, 0xB9, "vfmadd", int_x86_fma_vfmadd_ss,
int_x86_fma_vfmadd_sd, X86Fmadd>, VEX_LIG;
int_x86_fma_vfmadd_sd, fma>, VEX_LIG;
defm VFMSUB : fma3s<0x9B, 0xAB, 0xBB, "vfmsub", int_x86_fma_vfmsub_ss,
int_x86_fma_vfmsub_sd, X86Fmsub>, VEX_LIG;
@ -390,7 +390,7 @@ let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in {
let ExeDomain = SSEPackedSingle in {
// Scalar Instructions
defm VFMADDSS4 : fma4s<0x6A, "vfmaddss", FR32, f32mem, f32, X86Fmadd, loadf32>,
defm VFMADDSS4 : fma4s<0x6A, "vfmaddss", FR32, f32mem, f32, fma, loadf32>,
fma4s_int<0x6A, "vfmaddss", ssmem, sse_load_f32,
int_x86_fma_vfmadd_ss>;
defm VFMSUBSS4 : fma4s<0x6E, "vfmsubss", FR32, f32mem, f32, X86Fmsub, loadf32>,
@ -405,7 +405,7 @@ let ExeDomain = SSEPackedSingle in {
fma4s_int<0x7E, "vfnmsubss", ssmem, sse_load_f32,
int_x86_fma_vfnmsub_ss>;
// Packed Instructions
defm VFMADDPS4 : fma4p<0x68, "vfmaddps", X86Fmadd, v4f32, v8f32,
defm VFMADDPS4 : fma4p<0x68, "vfmaddps", fma, v4f32, v8f32,
loadv4f32, loadv8f32>;
defm VFMSUBPS4 : fma4p<0x6C, "vfmsubps", X86Fmsub, v4f32, v8f32,
loadv4f32, loadv8f32>;
@ -421,7 +421,7 @@ let ExeDomain = SSEPackedSingle in {
let ExeDomain = SSEPackedDouble in {
// Scalar Instructions
defm VFMADDSD4 : fma4s<0x6B, "vfmaddsd", FR64, f64mem, f64, X86Fmadd, loadf64>,
defm VFMADDSD4 : fma4s<0x6B, "vfmaddsd", FR64, f64mem, f64, fma, loadf64>,
fma4s_int<0x6B, "vfmaddsd", sdmem, sse_load_f64,
int_x86_fma_vfmadd_sd>;
defm VFMSUBSD4 : fma4s<0x6F, "vfmsubsd", FR64, f64mem, f64, X86Fmsub, loadf64>,
@ -436,7 +436,7 @@ let ExeDomain = SSEPackedDouble in {
fma4s_int<0x7F, "vfnmsubsd", sdmem, sse_load_f64,
int_x86_fma_vfnmsub_sd>;
// Packed Instructions
defm VFMADDPD4 : fma4p<0x69, "vfmaddpd", X86Fmadd, v2f64, v4f64,
defm VFMADDPD4 : fma4p<0x69, "vfmaddpd", fma, v2f64, v4f64,
loadv2f64, loadv4f64>;
defm VFMSUBPD4 : fma4p<0x6D, "vfmsubpd", X86Fmsub, v2f64, v4f64,
loadv2f64, loadv4f64>;

View File

@ -481,7 +481,7 @@ def X86fsqrtRnds : SDNode<"X86ISD::FSQRTS_RND", SDTFPBinOpRound>;
def X86fgetexpRnd : SDNode<"X86ISD::FGETEXP_RND", SDTFPUnaryOpRound>;
def X86fgetexpRnds : SDNode<"X86ISD::FGETEXPS_RND", SDTFPBinOpRound>;
def X86Fmadd : SDNode<"X86ISD::FMADD", SDTFPTernaryOp>;
// No need for FMADD because we use ISD::FMA.
def X86Fnmadd : SDNode<"X86ISD::FNMADD", SDTFPTernaryOp>;
def X86Fmsub : SDNode<"X86ISD::FMSUB", SDTFPTernaryOp>;
def X86Fnmsub : SDNode<"X86ISD::FNMSUB", SDTFPTernaryOp>;

View File

@ -1149,13 +1149,13 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
X86ISD::CVTPS2PH, 0),
X86_INTRINSIC_DATA(avx512_mask_vcvtps2ph_512, INTR_TYPE_2OP_MASK,
X86ISD::CVTPS2PH, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_pd_128, FMA_OP_MASK, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_pd_256, FMA_OP_MASK, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_pd_512, FMA_OP_MASK, X86ISD::FMADD,
X86_INTRINSIC_DATA(avx512_mask_vfmadd_pd_128, FMA_OP_MASK, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_pd_256, FMA_OP_MASK, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_pd_512, FMA_OP_MASK, ISD::FMA,
X86ISD::FMADD_RND),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_128, FMA_OP_MASK, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_256, FMA_OP_MASK, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_512, FMA_OP_MASK, X86ISD::FMADD,
X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_128, FMA_OP_MASK, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_256, FMA_OP_MASK, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_ps_512, FMA_OP_MASK, ISD::FMA,
X86ISD::FMADD_RND),
X86_INTRINSIC_DATA(avx512_mask_vfmadd_sd, FMA_OP_SCALAR_MASK, X86ISD::FMADDS1_RND, 0),
@ -1271,13 +1271,13 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
X86ISD::VPMADD52L, 0),
X86_INTRINSIC_DATA(avx512_mask_vpmadd52l_uq_512 , FMA_OP_MASK,
X86ISD::VPMADD52L, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_128, FMA_OP_MASK3, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_256, FMA_OP_MASK3, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_512, FMA_OP_MASK3, X86ISD::FMADD,
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_128, FMA_OP_MASK3, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_256, FMA_OP_MASK3, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_pd_512, FMA_OP_MASK3, ISD::FMA,
X86ISD::FMADD_RND),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_128, FMA_OP_MASK3, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_256, FMA_OP_MASK3, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_512, FMA_OP_MASK3, X86ISD::FMADD,
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_128, FMA_OP_MASK3, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_256, FMA_OP_MASK3, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_ps_512, FMA_OP_MASK3, ISD::FMA,
X86ISD::FMADD_RND),
X86_INTRINSIC_DATA(avx512_mask3_vfmadd_sd, FMA_OP_SCALAR_MASK3, X86ISD::FMADDS3_RND, 0),
@ -1349,13 +1349,13 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
X86ISD::VPTERNLOG, 0),
X86_INTRINSIC_DATA(avx512_maskz_pternlog_q_512, TERLOG_OP_MASKZ,
X86ISD::VPTERNLOG, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_pd_128, FMA_OP_MASKZ, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_pd_256, FMA_OP_MASKZ, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_pd_512, FMA_OP_MASKZ, X86ISD::FMADD,
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_pd_128, FMA_OP_MASKZ, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_pd_256, FMA_OP_MASKZ, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_pd_512, FMA_OP_MASKZ, ISD::FMA,
X86ISD::FMADD_RND),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_128, FMA_OP_MASKZ, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_256, FMA_OP_MASKZ, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_512, FMA_OP_MASKZ, X86ISD::FMADD,
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_128, FMA_OP_MASKZ, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_256, FMA_OP_MASKZ, ISD::FMA, 0),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_ps_512, FMA_OP_MASKZ, ISD::FMA,
X86ISD::FMADD_RND),
X86_INTRINSIC_DATA(avx512_maskz_vfmadd_sd, FMA_OP_SCALAR_MASKZ, X86ISD::FMADDS1_RND, 0),
@ -1524,10 +1524,10 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
X86_INTRINSIC_DATA(avx512_vcvtss2usi64, INTR_TYPE_2OP, X86ISD::CVTS2UI_RND, 0),
X86_INTRINSIC_DATA(avx512_vpermilvar_pd_512, INTR_TYPE_2OP, X86ISD::VPERMILPV, 0),
X86_INTRINSIC_DATA(avx512_vpermilvar_ps_512, INTR_TYPE_2OP, X86ISD::VPERMILPV, 0),
X86_INTRINSIC_DATA(fma_vfmadd_pd, INTR_TYPE_3OP, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(fma_vfmadd_pd_256, INTR_TYPE_3OP, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(fma_vfmadd_ps, INTR_TYPE_3OP, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(fma_vfmadd_ps_256, INTR_TYPE_3OP, X86ISD::FMADD, 0),
X86_INTRINSIC_DATA(fma_vfmadd_pd, INTR_TYPE_3OP, ISD::FMA, 0),
X86_INTRINSIC_DATA(fma_vfmadd_pd_256, INTR_TYPE_3OP, ISD::FMA, 0),
X86_INTRINSIC_DATA(fma_vfmadd_ps, INTR_TYPE_3OP, ISD::FMA, 0),
X86_INTRINSIC_DATA(fma_vfmadd_ps_256, INTR_TYPE_3OP, ISD::FMA, 0),
X86_INTRINSIC_DATA(fma_vfmaddsub_pd, INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
X86_INTRINSIC_DATA(fma_vfmaddsub_pd_256, INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
X86_INTRINSIC_DATA(fma_vfmaddsub_ps, INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),