replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[X86] Cleanup SchedWriteFMA classes and use X86SchedWriteWidths directly. 

Rename scalar and XMM versions, this is to match/simplify an upcoming change to split MUL/DIV/SQRT scalar/xmm/ymm/zmm classes.

llvm-svn: 331531
This commit is contained in:
Simon Pilgrim 2018-05-04 15:20:18 +00:00
parent 12e42f0f81
commit 67cc246dca
12 changed files with 80 additions and 82 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
llvm/lib/Target/X86/X86InstrAVX512.td
View File

@ -6430,29 +6430,29 @@ let Constraints = "$src1 = $dst", hasSideEffects = 0 in {
defm r_Int: AVX512_maskable_3src_scalar<opc, MRMSrcReg, _, (outs _.RC:$dst),
(ins _.RC:$src2, _.RC:$src3), OpcodeStr,
"$src3, $src2", "$src2, $src3", RHS_VEC_r, 1, 1>,
AVX512FMA3Base, Sched<[WriteFMAS]>;
AVX512FMA3Base, Sched<[SchedWriteFMA.Scl]>;
defm m_Int: AVX512_maskable_3src_scalar<opc, MRMSrcMem, _, (outs _.RC:$dst),
(ins _.RC:$src2, _.IntScalarMemOp:$src3), OpcodeStr,
"$src3, $src2", "$src2, $src3", RHS_VEC_m, 1, 1>,
AVX512FMA3Base, Sched<[WriteFMASLd, ReadAfterLd]>;
AVX512FMA3Base, Sched<[SchedWriteFMA.Scl.Folded, ReadAfterLd]>;
defm rb_Int: AVX512_maskable_3src_scalar<opc, MRMSrcReg, _, (outs _.RC:$dst),
(ins _.RC:$src2, _.RC:$src3, AVX512RC:$rc),
OpcodeStr, "$rc, $src3, $src2", "$src2, $src3, $rc", RHS_VEC_rb, 1, 1>,
AVX512FMA3Base, EVEX_B, EVEX_RC, Sched<[WriteFMAS]>;
AVX512FMA3Base, EVEX_B, EVEX_RC, Sched<[SchedWriteFMA.Scl]>;
let isCodeGenOnly = 1, isCommutable = 1 in {
def r : AVX512FMA3S<opc, MRMSrcReg, (outs _.FRC:$dst),
(ins _.FRC:$src1, _.FRC:$src2, _.FRC:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!if(MaskOnlyReg, [], [RHS_r])>, Sched<[WriteFMAS]>;
!if(MaskOnlyReg, [], [RHS_r])>, Sched<[SchedWriteFMA.Scl]>;
def m : AVX512FMA3S<opc, MRMSrcMem, (outs _.FRC:$dst),
(ins _.FRC:$src1, _.FRC:$src2, _.ScalarMemOp:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[RHS_m]>, Sched<[WriteFMASLd, ReadAfterLd]>;
[RHS_m]>, Sched<[SchedWriteFMA.Scl.Folded, ReadAfterLd]>;
}// isCodeGenOnly = 1
}// Constraints = "$src1 = $dst"
}

122
llvm/lib/Target/X86/X86InstrFMA.td
View File

@ -101,23 +101,22 @@ multiclass fma3p_forms<bits<8> opc132, bits<8> opc213, bits<8> opc231,
string OpcodeStr, string PackTy, string Suff,
PatFrag MemFrag128, PatFrag MemFrag256,
SDNode Op, ValueType OpTy128, ValueType OpTy256,
X86FoldableSchedWrite sched128,
X86FoldableSchedWrite sched256> {
X86SchedWriteWidths sched> {
defm NAME#213#Suff : fma3p_rm_213<opc213, !strconcat(OpcodeStr, "213", PackTy),
VR128, OpTy128, f128mem, MemFrag128, Op, sched128>;
VR128, OpTy128, f128mem, MemFrag128, Op, sched.XMM>;
defm NAME#231#Suff : fma3p_rm_231<opc231, !strconcat(OpcodeStr, "231", PackTy),
VR128, OpTy128, f128mem, MemFrag128, Op, sched128>;
VR128, OpTy128, f128mem, MemFrag128, Op, sched.XMM>;
defm NAME#132#Suff : fma3p_rm_132<opc132, !strconcat(OpcodeStr, "132", PackTy),
VR128, OpTy128, f128mem, MemFrag128, Op, sched128>;
VR128, OpTy128, f128mem, MemFrag128, Op, sched.XMM>;
defm NAME#213#Suff#Y : fma3p_rm_213<opc213, !strconcat(OpcodeStr, "213", PackTy),
VR256, OpTy256, f256mem, MemFrag256, Op, sched256>,
VR256, OpTy256, f256mem, MemFrag256, Op, sched.YMM>,
VEX_L;
defm NAME#231#Suff#Y : fma3p_rm_231<opc231, !strconcat(OpcodeStr, "231", PackTy),
VR256, OpTy256, f256mem, MemFrag256, Op, sched256>,
VR256, OpTy256, f256mem, MemFrag256, Op, sched.YMM>,
VEX_L;
defm NAME#132#Suff#Y : fma3p_rm_132<opc132, !strconcat(OpcodeStr, "132", PackTy),
VR256, OpTy256, f256mem, MemFrag256, Op, sched256>,
VR256, OpTy256, f256mem, MemFrag256, Op, sched.YMM>,
VEX_L;
}
@ -125,45 +124,45 @@ multiclass fma3p_forms<bits<8> opc132, bits<8> opc213, bits<8> opc231,
let ExeDomain = SSEPackedSingle in {
defm VFMADD : fma3p_forms<0x98, 0xA8, 0xB8, "vfmadd", "ps", "PS",
loadv4f32, loadv8f32, X86Fmadd, v4f32, v8f32,
WriteFMA, WriteFMAY>;
SchedWriteFMA>;
defm VFMSUB : fma3p_forms<0x9A, 0xAA, 0xBA, "vfmsub", "ps", "PS",
loadv4f32, loadv8f32, X86Fmsub, v4f32, v8f32,
WriteFMA, WriteFMAY>;
SchedWriteFMA>;
defm VFMADDSUB : fma3p_forms<0x96, 0xA6, 0xB6, "vfmaddsub", "ps", "PS",
loadv4f32, loadv8f32, X86Fmaddsub, v4f32, v8f32,
WriteFMA, WriteFMAY>;
SchedWriteFMA>;
defm VFMSUBADD : fma3p_forms<0x97, 0xA7, 0xB7, "vfmsubadd", "ps", "PS",
loadv4f32, loadv8f32, X86Fmsubadd, v4f32, v8f32,
WriteFMA, WriteFMAY>;
SchedWriteFMA>;
}
let ExeDomain = SSEPackedDouble in {
defm VFMADD : fma3p_forms<0x98, 0xA8, 0xB8, "vfmadd", "pd", "PD",
loadv2f64, loadv4f64, X86Fmadd, v2f64,
v4f64, WriteFMA, WriteFMAY>, VEX_W;
v4f64, SchedWriteFMA>, VEX_W;
defm VFMSUB : fma3p_forms<0x9A, 0xAA, 0xBA, "vfmsub", "pd", "PD",
loadv2f64, loadv4f64, X86Fmsub, v2f64,
v4f64, WriteFMA, WriteFMAY>, VEX_W;
v4f64, SchedWriteFMA>, VEX_W;
defm VFMADDSUB : fma3p_forms<0x96, 0xA6, 0xB6, "vfmaddsub", "pd", "PD",
loadv2f64, loadv4f64, X86Fmaddsub,
v2f64, v4f64, WriteFMA, WriteFMAY>, VEX_W;
v2f64, v4f64, SchedWriteFMA>, VEX_W;
defm VFMSUBADD : fma3p_forms<0x97, 0xA7, 0xB7, "vfmsubadd", "pd", "PD",
loadv2f64, loadv4f64, X86Fmsubadd,
v2f64, v4f64, WriteFMA, WriteFMAY>, VEX_W;
v2f64, v4f64, SchedWriteFMA>, VEX_W;
}
// Fused Negative Multiply-Add
let ExeDomain = SSEPackedSingle in {
defm VFNMADD : fma3p_forms<0x9C, 0xAC, 0xBC, "vfnmadd", "ps", "PS", loadv4f32,
loadv8f32, X86Fnmadd, v4f32, v8f32, WriteFMA, WriteFMAY>;
loadv8f32, X86Fnmadd, v4f32, v8f32, SchedWriteFMA>;
defm VFNMSUB : fma3p_forms<0x9E, 0xAE, 0xBE, "vfnmsub", "ps", "PS", loadv4f32,
loadv8f32, X86Fnmsub, v4f32, v8f32, WriteFMA, WriteFMAY>;
loadv8f32, X86Fnmsub, v4f32, v8f32, SchedWriteFMA>;
}
let ExeDomain = SSEPackedDouble in {
defm VFNMADD : fma3p_forms<0x9C, 0xAC, 0xBC, "vfnmadd", "pd", "PD", loadv2f64,
loadv4f64, X86Fnmadd, v2f64, v4f64, WriteFMA, WriteFMAY>, VEX_W;
loadv4f64, X86Fnmadd, v2f64, v4f64, SchedWriteFMA>, VEX_W;
defm VFNMSUB : fma3p_forms<0x9E, 0xAE, 0xBE, "vfnmsub", "pd", "PD", loadv2f64,
loadv4f64, X86Fnmsub, v2f64, v4f64, WriteFMA, WriteFMAY>, VEX_W;
loadv4f64, X86Fnmsub, v2f64, v4f64, SchedWriteFMA>, VEX_W;
}
// All source register operands of FMA opcodes defined in fma3s_rm multiclass
@ -356,14 +355,14 @@ multiclass fma3s<bits<8> opc132, bits<8> opc213, bits<8> opc231,
}
defm VFMADD : fma3s<0x99, 0xA9, 0xB9, "vfmadd", X86Fmadds1, X86Fmadd,
WriteFMAS>, VEX_LIG;
SchedWriteFMA.Scl>, VEX_LIG;
defm VFMSUB : fma3s<0x9B, 0xAB, 0xBB, "vfmsub", X86Fmsubs1, X86Fmsub,
WriteFMAS>, VEX_LIG;
SchedWriteFMA.Scl>, VEX_LIG;
defm VFNMADD : fma3s<0x9D, 0xAD, 0xBD, "vfnmadd", X86Fnmadds1, X86Fnmadd,
WriteFMAS>, VEX_LIG;
SchedWriteFMA.Scl>, VEX_LIG;
defm VFNMSUB : fma3s<0x9F, 0xAF, 0xBF, "vfnmsub", X86Fnmsubs1, X86Fnmsub,
WriteFMAS>, VEX_LIG;
SchedWriteFMA.Scl>, VEX_LIG;
//===----------------------------------------------------------------------===//
// FMA4 - AMD 4 operand Fused Multiply-Add instructions
@ -450,8 +449,7 @@ let hasSideEffects = 0 in
multiclass fma4p<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType OpVT128, ValueType OpVT256,
PatFrag ld_frag128, PatFrag ld_frag256,
X86FoldableSchedWrite sched128,
X86FoldableSchedWrite sched256> {
X86SchedWriteWidths sched> {
let isCommutable = 1 in
def rr : FMA4<opc, MRMSrcRegOp4, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
@ -459,21 +457,21 @@ multiclass fma4p<bits<8> opc, string OpcodeStr, SDNode OpNode,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(OpVT128 (OpNode VR128:$src1, VR128:$src2, VR128:$src3)))]>,
VEX_W, Sched<[sched128]>;
VEX_W, Sched<[sched.XMM]>;
def rm : FMA4<opc, MRMSrcMemOp4, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, f128mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst, (OpNode VR128:$src1, VR128:$src2,
(ld_frag128 addr:$src3)))]>, VEX_W,
Sched<[sched128.Folded, ReadAfterLd, ReadAfterLd]>;
Sched<[sched.XMM.Folded, ReadAfterLd, ReadAfterLd]>;
def mr : FMA4<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, f128mem:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(OpNode VR128:$src1, (ld_frag128 addr:$src2), VR128:$src3))]>,
Sched<[sched128.Folded, ReadAfterLd,
Sched<[sched.XMM.Folded, ReadAfterLd,
// f128mem:$src2
ReadDefault, ReadDefault, ReadDefault, ReadDefault,
ReadDefault,
@ -486,21 +484,21 @@ multiclass fma4p<bits<8> opc, string OpcodeStr, SDNode OpNode,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst,
(OpVT256 (OpNode VR256:$src1, VR256:$src2, VR256:$src3)))]>,
VEX_W, VEX_L, Sched<[sched256]>;
VEX_W, VEX_L, Sched<[sched.YMM]>;
def Yrm : FMA4<opc, MRMSrcMemOp4, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, f256mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst, (OpNode VR256:$src1, VR256:$src2,
(ld_frag256 addr:$src3)))]>, VEX_W, VEX_L,
Sched<[sched256.Folded, ReadAfterLd, ReadAfterLd]>;
Sched<[sched.YMM.Folded, ReadAfterLd, ReadAfterLd]>;
def Ymr : FMA4<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst, (OpNode VR256:$src1,
(ld_frag256 addr:$src2), VR256:$src3))]>, VEX_L,
Sched<[sched256.Folded, ReadAfterLd,
Sched<[sched.YMM.Folded, ReadAfterLd,
// f256mem:$src2
ReadDefault, ReadDefault, ReadDefault, ReadDefault,
ReadDefault,
@ -512,78 +510,78 @@ let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in {
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), []>,
Sched<[sched128]>, FoldGenData<NAME#rr>;
Sched<[sched.XMM]>, FoldGenData<NAME#rr>;
def Yrr_REV : FMA4<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"), []>,
VEX_L, Sched<[sched256]>, FoldGenData<NAME#Yrr>;
VEX_L, Sched<[sched.YMM]>, FoldGenData<NAME#Yrr>;
} // isCodeGenOnly = 1
}
let ExeDomain = SSEPackedSingle in {
// Scalar Instructions
defm VFMADDSS4 : fma4s<0x6A, "vfmaddss", FR32, f32mem, f32, X86Fmadd, loadf32,
WriteFMAS>,
SchedWriteFMA.Scl>,
fma4s_int<0x6A, "vfmaddss", ssmem, v4f32, sse_load_f32,
X86Fmadd4s, WriteFMAS>;
X86Fmadd4s, SchedWriteFMA.Scl>;
defm VFMSUBSS4 : fma4s<0x6E, "vfmsubss", FR32, f32mem, f32, X86Fmsub, loadf32,
WriteFMAS>,
SchedWriteFMA.Scl>,
fma4s_int<0x6E, "vfmsubss", ssmem, v4f32, sse_load_f32,
X86Fmsub4s, WriteFMAS>;
X86Fmsub4s, SchedWriteFMA.Scl>;
defm VFNMADDSS4 : fma4s<0x7A, "vfnmaddss", FR32, f32mem, f32,
X86Fnmadd, loadf32, WriteFMAS>,
X86Fnmadd, loadf32, SchedWriteFMA.Scl>,
fma4s_int<0x7A, "vfnmaddss", ssmem, v4f32, sse_load_f32,
X86Fnmadd4s, WriteFMAS>;
X86Fnmadd4s, SchedWriteFMA.Scl>;
defm VFNMSUBSS4 : fma4s<0x7E, "vfnmsubss", FR32, f32mem, f32,
X86Fnmsub, loadf32, WriteFMAS>,
X86Fnmsub, loadf32, SchedWriteFMA.Scl>,
fma4s_int<0x7E, "vfnmsubss", ssmem, v4f32, sse_load_f32,
X86Fnmsub4s, WriteFMAS>;
X86Fnmsub4s, SchedWriteFMA.Scl>;
// Packed Instructions
defm VFMADDPS4 : fma4p<0x68, "vfmaddps", X86Fmadd, v4f32, v8f32,
loadv4f32, loadv8f32, WriteFMA, WriteFMAY>;
loadv4f32, loadv8f32, SchedWriteFMA>;
defm VFMSUBPS4 : fma4p<0x6C, "vfmsubps", X86Fmsub, v4f32, v8f32,
loadv4f32, loadv8f32, WriteFMA, WriteFMAY>;
loadv4f32, loadv8f32, SchedWriteFMA>;
defm VFNMADDPS4 : fma4p<0x78, "vfnmaddps", X86Fnmadd, v4f32, v8f32,
loadv4f32, loadv8f32, WriteFMA, WriteFMAY>;
loadv4f32, loadv8f32, SchedWriteFMA>;
defm VFNMSUBPS4 : fma4p<0x7C, "vfnmsubps", X86Fnmsub, v4f32, v8f32,
loadv4f32, loadv8f32, WriteFMA, WriteFMAY>;
loadv4f32, loadv8f32, SchedWriteFMA>;
defm VFMADDSUBPS4 : fma4p<0x5C, "vfmaddsubps", X86Fmaddsub, v4f32, v8f32,
loadv4f32, loadv8f32, WriteFMA, WriteFMAY>;
loadv4f32, loadv8f32, SchedWriteFMA>;
defm VFMSUBADDPS4 : fma4p<0x5E, "vfmsubaddps", X86Fmsubadd, v4f32, v8f32,
loadv4f32, loadv8f32, WriteFMA, WriteFMAY>;
loadv4f32, loadv8f32, SchedWriteFMA>;
}
let ExeDomain = SSEPackedDouble in {
// Scalar Instructions
defm VFMADDSD4 : fma4s<0x6B, "vfmaddsd", FR64, f64mem, f64, X86Fmadd, loadf64,
WriteFMAS>,
SchedWriteFMA.Scl>,
fma4s_int<0x6B, "vfmaddsd", sdmem, v2f64, sse_load_f64,
X86Fmadd4s, WriteFMAS>;
X86Fmadd4s, SchedWriteFMA.Scl>;
defm VFMSUBSD4 : fma4s<0x6F, "vfmsubsd", FR64, f64mem, f64, X86Fmsub, loadf64,
WriteFMAS>,
SchedWriteFMA.Scl>,
fma4s_int<0x6F, "vfmsubsd", sdmem, v2f64, sse_load_f64,
X86Fmsub4s, WriteFMAS>;
X86Fmsub4s, SchedWriteFMA.Scl>;
defm VFNMADDSD4 : fma4s<0x7B, "vfnmaddsd", FR64, f64mem, f64,
X86Fnmadd, loadf64, WriteFMAS>,
X86Fnmadd, loadf64, SchedWriteFMA.Scl>,
fma4s_int<0x7B, "vfnmaddsd", sdmem, v2f64, sse_load_f64,
X86Fnmadd4s, WriteFMAS>;
X86Fnmadd4s, SchedWriteFMA.Scl>;
defm VFNMSUBSD4 : fma4s<0x7F, "vfnmsubsd", FR64, f64mem, f64,
X86Fnmsub, loadf64, WriteFMAS>,
X86Fnmsub, loadf64, SchedWriteFMA.Scl>,
fma4s_int<0x7F, "vfnmsubsd", sdmem, v2f64, sse_load_f64,
X86Fnmsub4s, WriteFMAS>;
X86Fnmsub4s, SchedWriteFMA.Scl>;
// Packed Instructions
defm VFMADDPD4 : fma4p<0x69, "vfmaddpd", X86Fmadd, v2f64, v4f64,
loadv2f64, loadv4f64, WriteFMA, WriteFMAY>;
loadv2f64, loadv4f64, SchedWriteFMA>;
defm VFMSUBPD4 : fma4p<0x6D, "vfmsubpd", X86Fmsub, v2f64, v4f64,
loadv2f64, loadv4f64, WriteFMA, WriteFMAY>;
loadv2f64, loadv4f64, SchedWriteFMA>;
defm VFNMADDPD4 : fma4p<0x79, "vfnmaddpd", X86Fnmadd, v2f64, v4f64,
loadv2f64, loadv4f64, WriteFMA, WriteFMAY>;
loadv2f64, loadv4f64, SchedWriteFMA>;
defm VFNMSUBPD4 : fma4p<0x7D, "vfnmsubpd", X86Fnmsub, v2f64, v4f64,
loadv2f64, loadv4f64, WriteFMA, WriteFMAY>;
loadv2f64, loadv4f64, SchedWriteFMA>;
defm VFMADDSUBPD4 : fma4p<0x5D, "vfmaddsubpd", X86Fmaddsub, v2f64, v4f64,
loadv2f64, loadv4f64, WriteFMA, WriteFMAY>;
loadv2f64, loadv4f64, SchedWriteFMA>;
defm VFMSUBADDPD4 : fma4p<0x5F, "vfmsubaddpd", X86Fmsubadd, v2f64, v4f64,
loadv2f64, loadv4f64, WriteFMA, WriteFMAY>;
loadv2f64, loadv4f64, SchedWriteFMA>;
}

2
llvm/lib/Target/X86/X86SchedBroadwell.td
View File

@ -171,7 +171,7 @@ defm : BWWriteResPair<WriteFRcpY, [BWPort0], 5, [1], 1, 7>; // Floating point
defm : BWWriteResPair<WriteFRsqrt, [BWPort0], 5, [1], 1, 5>; // Floating point reciprocal square root estimate.
defm : BWWriteResPair<WriteFRsqrtY,[BWPort0], 5, [1], 1, 7>; // Floating point reciprocal square root estimate (YMM/ZMM).
defm : BWWriteResPair<WriteFMA, [BWPort01], 5, [1], 1, 5>; // Fused Multiply Add.
defm : BWWriteResPair<WriteFMAS, [BWPort01], 5, [1], 1, 5>; // Fused Multiply Add (Scalar).
defm : BWWriteResPair<WriteFMAX, [BWPort01], 5, [1], 1, 5>; // Fused Multiply Add (XMM).
defm : BWWriteResPair<WriteFMAY, [BWPort01], 5, [1], 1, 6>; // Fused Multiply Add (YMM/ZMM).
defm : BWWriteResPair<WriteDPPD, [BWPort0,BWPort1,BWPort5], 9, [1,1,1], 3, 5>; // Floating point double dot product.
defm : BWWriteResPair<WriteDPPS, [BWPort0,BWPort1,BWPort5], 14, [2,1,1], 4, 5>; // Floating point single dot product.

4
llvm/lib/Target/X86/X86SchedHaswell.td
View File

@ -166,8 +166,8 @@ defm : HWWriteResPair<WriteFSqrtY, [HWPort0], 15, [1], 1, 7>;
defm : HWWriteResPair<WriteCvtF2I, [HWPort1], 3>;
defm : HWWriteResPair<WriteCvtI2F, [HWPort1], 4>;
defm : HWWriteResPair<WriteCvtF2F, [HWPort1], 3>;
defm : HWWriteResPair<WriteFMA, [HWPort01], 5, [1], 1, 6>;
defm : HWWriteResPair<WriteFMAS, [HWPort01], 5, [1], 1, 5>;
defm : HWWriteResPair<WriteFMA, [HWPort01], 5, [1], 1, 5>;
defm : HWWriteResPair<WriteFMAX, [HWPort01], 5, [1], 1, 6>;
defm : HWWriteResPair<WriteFMAY, [HWPort01], 5, [1], 1, 7>;
defm : HWWriteResPair<WriteDPPD, [HWPort0,HWPort1,HWPort5], 9, [1,1,1], 3, 6>;
defm : HWWriteResPair<WriteDPPS, [HWPort0,HWPort1,HWPort5], 14, [2,1,1], 4, 6>;

2
llvm/lib/Target/X86/X86SchedSandyBridge.td
View File

@ -351,7 +351,7 @@ defm : SBWriteResPair<WriteFVarShuffle256, [SBPort5], 1, [1], 1, 7>;
defm : SBWriteResPair<WriteShuffle256, [SBPort5], 1, [1], 1, 7>;
defm : SBWriteResPair<WriteVarShuffle256, [SBPort5], 1, [1], 1, 7>;
defm : SBWriteResPair<WriteFMA, [SBPort01], 5>;
defm : SBWriteResPair<WriteFMAS, [SBPort01], 5>;
defm : SBWriteResPair<WriteFMAX, [SBPort01], 5>;
defm : SBWriteResPair<WriteFMAY, [SBPort01], 5>;
// Remaining SNB instrs.

4
llvm/lib/Target/X86/X86SchedSkylakeClient.td
View File

@ -166,8 +166,8 @@ defm : SKLWriteResPair<WriteFRcp, [SKLPort0], 4, [1], 1, 5>; // Floating point
defm : SKLWriteResPair<WriteFRcpY, [SKLPort0], 4, [1], 1, 5>; // Floating point reciprocal estimate (YMM/ZMM).
defm : SKLWriteResPair<WriteFRsqrt, [SKLPort0], 4, [1], 1, 5>; // Floating point reciprocal square root estimate.
defm : SKLWriteResPair<WriteFRsqrtY,[SKLPort0], 4, [1], 1, 5>; // Floating point reciprocal square root estimate (YMM/ZMM).
defm : SKLWriteResPair<WriteFMA, [SKLPort01], 4, [1], 1, 6>; // Fused Multiply Add.
defm : SKLWriteResPair<WriteFMAS, [SKLPort01], 4, [1], 1, 5>; // Fused Multiply Add (Scalar).
defm : SKLWriteResPair<WriteFMA, [SKLPort01], 4, [1], 1, 5>; // Fused Multiply Add.
defm : SKLWriteResPair<WriteFMAX, [SKLPort01], 4, [1], 1, 6>; // Fused Multiply Add (XMM).
defm : SKLWriteResPair<WriteFMAY, [SKLPort01], 4, [1], 1, 7>; // Fused Multiply Add (YMM/ZMM).
defm : SKLWriteResPair<WriteDPPD, [SKLPort5,SKLPort01], 9, [1,2], 3, 6>; // Floating point double dot product.
defm : SKLWriteResPair<WriteDPPS, [SKLPort5,SKLPort01], 13, [1,3], 4, 6>; // Floating point single dot product.

4
llvm/lib/Target/X86/X86SchedSkylakeServer.td
View File

@ -166,8 +166,8 @@ defm : SKXWriteResPair<WriteFRcp, [SKXPort0], 4, [1], 1, 6>; // Floating poin
defm : SKXWriteResPair<WriteFRcpY, [SKXPort0], 4, [1], 1, 7>; // Floating point reciprocal estimate (YMM/ZMM).
defm : SKXWriteResPair<WriteFRsqrt, [SKXPort0], 4, [1], 1, 6>; // Floating point reciprocal square root estimate.
defm : SKXWriteResPair<WriteFRsqrtY,[SKXPort0], 4, [1], 1, 7>; // Floating point reciprocal square root estimate (YMM/ZMM).
defm : SKXWriteResPair<WriteFMA, [SKXPort015], 4, [1], 1, 6>; // Fused Multiply Add.
defm : SKXWriteResPair<WriteFMAS, [SKXPort015], 4, [1], 1, 5>; // Fused Multiply Add (Scalar).
defm : SKXWriteResPair<WriteFMA, [SKXPort015], 4, [1], 1, 5>; // Fused Multiply Add.
defm : SKXWriteResPair<WriteFMAX, [SKXPort015], 4, [1], 1, 6>; // Fused Multiply Add (XMM).
defm : SKXWriteResPair<WriteFMAY, [SKXPort015], 4, [1], 1, 7>; // Fused Multiply Add (YMM/ZMM).
defm : SKXWriteResPair<WriteDPPD, [SKXPort5,SKXPort015], 9, [1,2], 3, 6>; // Floating point double dot product.
defm : SKXWriteResPair<WriteDPPS, [SKXPort5,SKXPort015], 13, [1,3], 4, 6>; // Floating point single dot product.

4
llvm/lib/Target/X86/X86Schedule.td
View File

@ -117,7 +117,7 @@ defm WriteFRcpY : X86SchedWritePair; // Floating point reciprocal estimate (YMM
defm WriteFRsqrt : X86SchedWritePair; // Floating point reciprocal square root estimate.
defm WriteFRsqrtY: X86SchedWritePair; // Floating point reciprocal square root estimate (YMM/ZMM).
defm WriteFMA : X86SchedWritePair; // Fused Multiply Add.
defm WriteFMAS : X86SchedWritePair; // Fused Multiply Add (Scalar).
defm WriteFMAX : X86SchedWritePair; // Fused Multiply Add (XMM).
defm WriteFMAY : X86SchedWritePair; // Fused Multiply Add (YMM/ZMM).
defm WriteDPPD : X86SchedWritePair; // Floating point double dot product.
defm WriteDPPS : X86SchedWritePair; // Floating point single dot product.
@ -248,7 +248,7 @@ def SchedWriteFCmp
def SchedWriteFMul
: X86SchedWriteWidths<WriteFMul, WriteFMul, WriteFMulY, WriteFMulY>;
def SchedWriteFMA
: X86SchedWriteWidths<WriteFMAS, WriteFMA, WriteFMAY, WriteFMAY>;
: X86SchedWriteWidths<WriteFMA, WriteFMAX, WriteFMAY, WriteFMAY>;
def SchedWriteDPPD
: X86SchedWriteWidths<WriteDPPD, WriteDPPD, WriteDPPD, WriteDPPD>;
def SchedWriteDPPS

2
llvm/lib/Target/X86/X86ScheduleAtom.td
View File

@ -227,7 +227,7 @@ defm : AtomWriteResPair<WriteFShuffleY, [AtomPort0], [AtomPort0]>; // NOTE
defm : AtomWriteResPair<WriteFVarShuffle, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteFVarShuffleY, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteFMA, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteFMAS, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteFMAX, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteFMAY, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteDPPD, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.
defm : AtomWriteResPair<WriteDPPS, [AtomPort0], [AtomPort0]>; // NOTE: Doesn't exist on Atom.

2
llvm/lib/Target/X86/X86ScheduleBtVer2.td
View File

@ -323,7 +323,7 @@ defm : JWriteResFpuPair<WriteFCom, [JFPU0, JFPA, JALU0], 3>;
defm : JWriteResFpuPair<WriteFMul, [JFPU1, JFPM], 2>;
defm : JWriteResYMMPair<WriteFMulY, [JFPU1, JFPM], 2, [2,2], 2>;
defm : JWriteResFpuPair<WriteFMA, [JFPU1, JFPM], 2>; // NOTE: Doesn't exist on Jaguar.
defm : JWriteResFpuPair<WriteFMAS, [JFPU1, JFPM], 2>; // NOTE: Doesn't exist on Jaguar.
defm : JWriteResFpuPair<WriteFMAX, [JFPU1, JFPM], 2>; // NOTE: Doesn't exist on Jaguar.
defm : JWriteResFpuPair<WriteFMAY, [JFPU1, JFPM], 2>; // NOTE: Doesn't exist on Jaguar.
defm : JWriteResFpuPair<WriteDPPD, [JFPU1, JFPM, JFPA], 9, [1, 3, 3], 3>;
defm : JWriteResFpuPair<WriteDPPS, [JFPU1, JFPM, JFPA], 11, [1, 3, 3], 5>;

2
llvm/lib/Target/X86/X86ScheduleSLM.td
View File

@ -318,7 +318,7 @@ defm : SLMWriteResPair<WriteVarShuffle256, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVarVecShift, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVarVecShiftY, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteFMA, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteFMAS, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteFMAX, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteFMAY, [SLM_FPC_RSV0], 1>;
// Instruction overrides

2
llvm/lib/Target/X86/X86ScheduleZnver1.td
View File

@ -220,7 +220,7 @@ defm : ZnWriteResFpuPair<WriteFVarShuffleY,[ZnFPU12], 1>;
defm : ZnWriteResFpuPair<WriteFMul, [ZnFPU0], 5>;
defm : ZnWriteResFpuPair<WriteFMulY, [ZnFPU0], 5>;
defm : ZnWriteResFpuPair<WriteFMA, [ZnFPU03], 5>;
defm : ZnWriteResFpuPair<WriteFMAS, [ZnFPU03], 5>;
defm : ZnWriteResFpuPair<WriteFMAX, [ZnFPU03], 5>;
defm : ZnWriteResFpuPair<WriteFMAY, [ZnFPU03], 5>;
defm : ZnWriteResFpuPair<WriteFRcp, [ZnFPU01], 5>;
defm : ZnWriteResFpuPair<WriteFRcpY, [ZnFPU01], 5>;