[llvm] r318758 - [X86][XOP] Add missing scheduler classes to XOP instructions
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 21 04:02:18 PST 2017
Author: rksimon
Date: Tue Nov 21 04:02:18 2017
New Revision: 318758
URL: http://llvm.org/viewvc/llvm-project?rev=318758&view=rev
Log:
[X86][XOP] Add missing scheduler classes to XOP instructions
All match equivalent basic classes (WritePHAdd, WriteFAdd etc.) according to both the AMD 15h SOG and Agner's tables.
Modified:
llvm/trunk/lib/Target/X86/X86InstrXOP.td
Modified: llvm/trunk/lib/Target/X86/X86InstrXOP.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrXOP.td?rev=318758&r1=318757&r2=318758&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrXOP.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrXOP.td Tue Nov 21 04:02:18 2017
@@ -14,10 +14,11 @@
multiclass xop2op<bits<8> opc, string OpcodeStr, Intrinsic Int, PatFrag memop> {
def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int VR128:$src))]>, XOP;
+ [(set VR128:$dst, (Int VR128:$src))]>, XOP, Sched<[WritePHAdd]>;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP;
+ [(set VR128:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP,
+ Sched<[WritePHAdd, ReadAfterLd]>;
}
let ExeDomain = SSEPackedInt in {
@@ -43,30 +44,33 @@ multiclass xop2opsld<bits<8> opc, string
Operand memop, ComplexPattern mem_cpat> {
def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int VR128:$src))]>, XOP;
+ [(set VR128:$dst, (Int VR128:$src))]>, XOP, Sched<[WriteFAdd]>;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst), (ins memop:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int (bitconvert mem_cpat:$src)))]>, XOP;
+ [(set VR128:$dst, (Int (bitconvert mem_cpat:$src)))]>, XOP,
+ Sched<[WriteFAdd, ReadAfterLd]>;
}
multiclass xop2op128<bits<8> opc, string OpcodeStr, Intrinsic Int,
PatFrag memop> {
def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int VR128:$src))]>, XOP;
+ [(set VR128:$dst, (Int VR128:$src))]>, XOP, Sched<[WriteFAdd]>;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP;
+ [(set VR128:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP,
+ Sched<[WriteFAdd, ReadAfterLd]>;
}
multiclass xop2op256<bits<8> opc, string OpcodeStr, Intrinsic Int,
PatFrag memop> {
def rrY : IXOP<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (Int VR256:$src))]>, XOP, VEX_L;
+ [(set VR256:$dst, (Int VR256:$src))]>, XOP, VEX_L, Sched<[WriteFAdd]>;
def rmY : IXOP<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP, VEX_L;
+ [(set VR256:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP, VEX_L,
+ Sched<[WriteFAdd, ReadAfterLd]>;
}
let ExeDomain = SSEPackedSingle in {
@@ -135,12 +139,14 @@ multiclass xop3opimm<bits<8> opc, string
(ins VR128:$src1, u8imm:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
- (vt128 (OpNode (vt128 VR128:$src1), imm:$src2)))]>, XOP;
+ (vt128 (OpNode (vt128 VR128:$src1), imm:$src2)))]>,
+ XOP, Sched<[WriteVecShift]>;
def mi : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins i128mem:$src1, u8imm:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
- (vt128 (OpNode (vt128 (bitconvert (loadv2i64 addr:$src1))), imm:$src2)))]>, XOP;
+ (vt128 (OpNode (vt128 (bitconvert (loadv2i64 addr:$src1))), imm:$src2)))]>,
+ XOP, Sched<[WriteVecShift, ReadAfterLd]>;
}
let ExeDomain = SSEPackedInt in {
@@ -158,14 +164,15 @@ multiclass xop4opm2<bits<8> opc, string
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
- (Int VR128:$src1, VR128:$src2, VR128:$src3))]>, XOP_4V;
+ (Int VR128:$src1, VR128:$src2, VR128:$src3))]>, XOP_4V,
+ Sched<[WriteVecIMul]>;
def rm : IXOPi8Reg<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int VR128:$src1, (bitconvert (loadv2i64 addr:$src2)),
- VR128:$src3))]>, XOP_4V;
+ VR128:$src3))]>, XOP_4V, Sched<[WriteVecIMul, ReadAfterLd]>;
}
let ExeDomain = SSEPackedInt in {
@@ -214,7 +221,7 @@ multiclass xopvpcom<bits<8> opc, string
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1), (vt128 VR128:$src2),
imm:$cc)))]>,
- XOP_4V;
+ XOP_4V, Sched<[WriteVecALU, ReadAfterLd]>;
def mi : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, XOPCC:$cc),
!strconcat("vpcom${cc}", Suffix,
@@ -223,19 +230,19 @@ multiclass xopvpcom<bits<8> opc, string
(vt128 (OpNode (vt128 VR128:$src1),
(vt128 (bitconvert (loadv2i64 addr:$src2))),
imm:$cc)))]>,
- XOP_4V;
+ XOP_4V, Sched<[WriteVecALU, ReadAfterLd]>;
let isAsmParserOnly = 1, hasSideEffects = 0 in {
def ri_alt : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, u8imm:$src3),
!strconcat("vpcom", Suffix,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
- []>, XOP_4V;
+ []>, XOP_4V, Sched<[WriteVecALU, ReadAfterLd]>;
let mayLoad = 1 in
def mi_alt : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, u8imm:$src3),
!strconcat("vpcom", Suffix,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
- []>, XOP_4V;
+ []>, XOP_4V, Sched<[WriteVecALU, ReadAfterLd]>;
}
}
@@ -259,7 +266,7 @@ multiclass xop4op<bits<8> opc, string Op
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1), (vt128 VR128:$src2),
(vt128 VR128:$src3))))]>,
- XOP_4V;
+ XOP_4V, Sched<[WriteShuffle]>;
def rrm : IXOPi8Reg<opc, MRMSrcMemOp4, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i128mem:$src3),
!strconcat(OpcodeStr,
@@ -267,7 +274,7 @@ multiclass xop4op<bits<8> opc, string Op
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1), (vt128 VR128:$src2),
(vt128 (bitconvert (loadv2i64 addr:$src3))))))]>,
- XOP_4V, VEX_W;
+ XOP_4V, VEX_W, Sched<[WriteShuffle, ReadAfterLd]>;
def rmr : IXOPi8Reg<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, VR128:$src3),
!strconcat(OpcodeStr,
@@ -275,14 +282,14 @@ multiclass xop4op<bits<8> opc, string Op
[(set VR128:$dst,
(v16i8 (OpNode (vt128 VR128:$src1), (vt128 (bitconvert (loadv2i64 addr:$src2))),
(vt128 VR128:$src3))))]>,
- XOP_4V;
+ XOP_4V, Sched<[WriteShuffle, ReadAfterLd]>;
// For disassembler
let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in
def rrr_REV : IXOPi8Reg<opc, MRMSrcRegOp4, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
- []>, XOP_4V, VEX_W, FoldGenData<NAME#rrr>;
+ []>, XOP_4V, VEX_W, Sched<[WriteShuffle]>, FoldGenData<NAME#rrr>;
}
let ExeDomain = SSEPackedInt in {
@@ -297,28 +304,29 @@ multiclass xop4op_int<bits<8> opc, strin
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set RC:$dst, (VT (or (and RC:$src3, RC:$src1),
- (X86andnp RC:$src3, RC:$src2))))]>, XOP_4V;
+ (X86andnp RC:$src3, RC:$src2))))]>, XOP_4V,
+ Sched<[WriteShuffle]>;
def rrm : IXOPi8Reg<opc, MRMSrcMemOp4, (outs RC:$dst),
(ins RC:$src1, RC:$src2, x86memop:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set RC:$dst, (VT (or (and (load addr:$src3), RC:$src1),
(X86andnp (load addr:$src3), RC:$src2))))]>,
- XOP_4V, VEX_W;
+ XOP_4V, VEX_W, Sched<[WriteShuffle, ReadAfterLd]>;
def rmr : IXOPi8Reg<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2, RC:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set RC:$dst, (VT (or (and RC:$src3, RC:$src1),
(X86andnp RC:$src3, (load addr:$src2)))))]>,
- XOP_4V;
+ XOP_4V, Sched<[WriteShuffle, ReadAfterLd]>;
// For disassembler
let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in
def rrr_REV : IXOPi8Reg<opc, MRMSrcRegOp4, (outs RC:$dst),
(ins RC:$src1, RC:$src2, RC:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
- []>, XOP_4V, VEX_W, FoldGenData<NAME#rrr>;
+ []>, XOP_4V, VEX_W, Sched<[WriteShuffle]>, FoldGenData<NAME#rrr>;
}
let ExeDomain = SSEPackedInt in {
@@ -335,7 +343,8 @@ multiclass xop_vpermil2<bits<8> Opc, str
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set RC:$dst,
- (VT (X86vpermil2 RC:$src1, RC:$src2, RC:$src3, (i8 imm:$src4))))]>;
+ (VT (X86vpermil2 RC:$src1, RC:$src2, RC:$src3, (i8 imm:$src4))))]>,
+ Sched<[WriteFShuffle]>;
def rm : IXOP5<Opc, MRMSrcMemOp4, (outs RC:$dst),
(ins RC:$src1, RC:$src2, intmemop:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
@@ -343,21 +352,23 @@ multiclass xop_vpermil2<bits<8> Opc, str
[(set RC:$dst,
(VT (X86vpermil2 RC:$src1, RC:$src2,
(bitconvert (IntLdFrag addr:$src3)),
- (i8 imm:$src4))))]>, VEX_W;
+ (i8 imm:$src4))))]>, VEX_W,
+ Sched<[WriteFShuffle, ReadAfterLd]>;
def mr : IXOP5<Opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, fpmemop:$src2, RC:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set RC:$dst,
(VT (X86vpermil2 RC:$src1, (FPLdFrag addr:$src2),
- RC:$src3, (i8 imm:$src4))))]>;
+ RC:$src3, (i8 imm:$src4))))]>,
+ Sched<[WriteFShuffle, ReadAfterLd]>;
// For disassembler
let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in
def rr_REV : IXOP5<Opc, MRMSrcRegOp4, (outs RC:$dst),
(ins RC:$src1, RC:$src2, RC:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
- []>, VEX_W, FoldGenData<NAME#rr>;
+ []>, VEX_W, Sched<[WriteFShuffle]>, FoldGenData<NAME#rr>;
}
let ExeDomain = SSEPackedDouble in {
More information about the llvm-commits
mailing list