1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 04:02:41 +01:00
llvm-mirror/lib/Target/X86/X86InstrXOP.td
Simon Pilgrim ac4c196247 [X86][XOP] Add support for the matching of the VPCMOV bit select instruction
XOP has the VPCMOV instruction that performs the common vector bit select operation OR( AND( SRC1, SRC3 ), AND( SRC2, ~SRC3 ) )

This patch adds tablegen pattern matching for this instruction.

Differential Revision: http://reviews.llvm.org/D8841

llvm-svn: 251975
2015-11-03 20:27:01 +00:00

345 lines
16 KiB
TableGen

//===-- X86InstrXOP.td - XOP Instruction Set ---------------*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file describes XOP (eXtended OPerations)
//
//===----------------------------------------------------------------------===//
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:$src))]>, XOP;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP;
}
let ExeDomain = SSEPackedInt in {
defm VPHSUBWD : xop2op<0xE2, "vphsubwd", int_x86_xop_vphsubwd, loadv2i64>;
defm VPHSUBDQ : xop2op<0xE3, "vphsubdq", int_x86_xop_vphsubdq, loadv2i64>;
defm VPHSUBBW : xop2op<0xE1, "vphsubbw", int_x86_xop_vphsubbw, loadv2i64>;
defm VPHADDWQ : xop2op<0xC7, "vphaddwq", int_x86_xop_vphaddwq, loadv2i64>;
defm VPHADDWD : xop2op<0xC6, "vphaddwd", int_x86_xop_vphaddwd, loadv2i64>;
defm VPHADDUWQ : xop2op<0xD7, "vphadduwq", int_x86_xop_vphadduwq, loadv2i64>;
defm VPHADDUWD : xop2op<0xD6, "vphadduwd", int_x86_xop_vphadduwd, loadv2i64>;
defm VPHADDUDQ : xop2op<0xDB, "vphaddudq", int_x86_xop_vphaddudq, loadv2i64>;
defm VPHADDUBW : xop2op<0xD1, "vphaddubw", int_x86_xop_vphaddubw, loadv2i64>;
defm VPHADDUBQ : xop2op<0xD3, "vphaddubq", int_x86_xop_vphaddubq, loadv2i64>;
defm VPHADDUBD : xop2op<0xD2, "vphaddubd", int_x86_xop_vphaddubd, loadv2i64>;
defm VPHADDDQ : xop2op<0xCB, "vphadddq", int_x86_xop_vphadddq, loadv2i64>;
defm VPHADDBW : xop2op<0xC1, "vphaddbw", int_x86_xop_vphaddbw, loadv2i64>;
defm VPHADDBQ : xop2op<0xC3, "vphaddbq", int_x86_xop_vphaddbq, loadv2i64>;
defm VPHADDBD : xop2op<0xC2, "vphaddbd", int_x86_xop_vphaddbd, loadv2i64>;
}
// Scalar load 2 addr operand instructions
multiclass xop2opsld<bits<8> opc, string OpcodeStr, Intrinsic Int,
Operand memop, ComplexPattern mem_cpat> {
def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (Int VR128:$src))]>, XOP;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst), (ins memop:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (Int (bitconvert mem_cpat:$src)))]>, XOP;
}
multiclass xop2op128<bits<8> opc, string OpcodeStr, Intrinsic Int,
PatFrag memop> {
def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (Int VR128:$src))]>, XOP;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP;
}
multiclass xop2op256<bits<8> opc, string OpcodeStr, Intrinsic Int,
PatFrag memop> {
def rrY : IXOP<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR256:$dst, (Int VR256:$src))]>, XOP, VEX_L;
def rmY : IXOP<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR256:$dst, (Int (bitconvert (memop addr:$src))))]>, XOP, VEX_L;
}
let ExeDomain = SSEPackedSingle in {
defm VFRCZSS : xop2opsld<0x82, "vfrczss", int_x86_xop_vfrcz_ss,
ssmem, sse_load_f32>;
defm VFRCZPS : xop2op128<0x80, "vfrczps", int_x86_xop_vfrcz_ps, loadv4f32>;
defm VFRCZPS : xop2op256<0x80, "vfrczps", int_x86_xop_vfrcz_ps_256, loadv8f32>;
}
let ExeDomain = SSEPackedDouble in {
defm VFRCZSD : xop2opsld<0x83, "vfrczsd", int_x86_xop_vfrcz_sd,
sdmem, sse_load_f64>;
defm VFRCZPD : xop2op128<0x81, "vfrczpd", int_x86_xop_vfrcz_pd, loadv2f64>;
defm VFRCZPD : xop2op256<0x81, "vfrczpd", int_x86_xop_vfrcz_pd_256, loadv4f64>;
}
multiclass xop3op<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType vt128> {
def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1), (vt128 VR128:$src2))))]>,
XOP_4VOp3, Sched<[WriteVarVecShift]>;
def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1),
(vt128 (bitconvert (loadv2i64 addr:$src2))))))]>,
XOP_4V, VEX_W, Sched<[WriteVarVecShift, ReadAfterLd]>;
def mr : IXOP<opc, MRMSrcMem, (outs VR128:$dst),
(ins i128mem:$src1, VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (vt128 (bitconvert (loadv2i64 addr:$src1))),
(vt128 VR128:$src2))))]>,
XOP_4VOp3, Sched<[WriteVarVecShift, ReadAfterLd]>;
}
let ExeDomain = SSEPackedInt in {
defm VPROTB : xop3op<0x90, "vprotb", X86vprot, v16i8>;
defm VPROTD : xop3op<0x92, "vprotd", X86vprot, v4i32>;
defm VPROTQ : xop3op<0x93, "vprotq", X86vprot, v2i64>;
defm VPROTW : xop3op<0x91, "vprotw", X86vprot, v8i16>;
defm VPSHAB : xop3op<0x98, "vpshab", X86vpsha, v16i8>;
defm VPSHAD : xop3op<0x9A, "vpshad", X86vpsha, v4i32>;
defm VPSHAQ : xop3op<0x9B, "vpshaq", X86vpsha, v2i64>;
defm VPSHAW : xop3op<0x99, "vpshaw", X86vpsha, v8i16>;
defm VPSHLB : xop3op<0x94, "vpshlb", X86vpshl, v16i8>;
defm VPSHLD : xop3op<0x96, "vpshld", X86vpshl, v4i32>;
defm VPSHLQ : xop3op<0x97, "vpshlq", X86vpshl, v2i64>;
defm VPSHLW : xop3op<0x95, "vpshlw", X86vpshl, v8i16>;
}
multiclass xop3opimm<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType vt128> {
def ri : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, u8imm:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1), imm:$src2)))]>, XOP;
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;
}
let ExeDomain = SSEPackedInt in {
defm VPROTB : xop3opimm<0xC0, "vprotb", X86vproti, v16i8>;
defm VPROTD : xop3opimm<0xC2, "vprotd", X86vproti, v4i32>;
defm VPROTQ : xop3opimm<0xC3, "vprotq", X86vproti, v2i64>;
defm VPROTW : xop3opimm<0xC1, "vprotw", X86vproti, v8i16>;
}
// Instruction where second source can be memory, but third must be register
multiclass xop4opm2<bits<8> opc, string OpcodeStr, Intrinsic Int> {
let isCommutable = 1 in
def rr : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int VR128:$src1, VR128:$src2, VR128:$src3))]>, XOP_4V, VEX_I8IMM;
def rm : IXOPi8<opc, MRMSrcMem, (outs VR128:$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:$src3))]>, XOP_4V, VEX_I8IMM;
}
let ExeDomain = SSEPackedInt in {
defm VPMADCSWD : xop4opm2<0xB6, "vpmadcswd", int_x86_xop_vpmadcswd>;
defm VPMADCSSWD : xop4opm2<0xA6, "vpmadcsswd", int_x86_xop_vpmadcsswd>;
defm VPMACSWW : xop4opm2<0x95, "vpmacsww", int_x86_xop_vpmacsww>;
defm VPMACSWD : xop4opm2<0x96, "vpmacswd", int_x86_xop_vpmacswd>;
defm VPMACSSWW : xop4opm2<0x85, "vpmacssww", int_x86_xop_vpmacssww>;
defm VPMACSSWD : xop4opm2<0x86, "vpmacsswd", int_x86_xop_vpmacsswd>;
defm VPMACSSDQL : xop4opm2<0x87, "vpmacssdql", int_x86_xop_vpmacssdql>;
defm VPMACSSDQH : xop4opm2<0x8F, "vpmacssdqh", int_x86_xop_vpmacssdqh>;
defm VPMACSSDD : xop4opm2<0x8E, "vpmacssdd", int_x86_xop_vpmacssdd>;
defm VPMACSDQL : xop4opm2<0x97, "vpmacsdql", int_x86_xop_vpmacsdql>;
defm VPMACSDQH : xop4opm2<0x9F, "vpmacsdqh", int_x86_xop_vpmacsdqh>;
defm VPMACSDD : xop4opm2<0x9E, "vpmacsdd", int_x86_xop_vpmacsdd>;
}
// Instruction where second source can be memory, third must be imm8
multiclass xopvpcom<bits<8> opc, string Suffix, SDNode OpNode, ValueType vt128> {
let isCommutable = 1 in
def ri : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, XOPCC:$cc),
!strconcat("vpcom${cc}", Suffix,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1), (vt128 VR128:$src2),
i8immZExt3:$cc)))]>,
XOP_4V;
def mi : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, XOPCC:$cc),
!strconcat("vpcom${cc}", Suffix,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (vt128 VR128:$src1),
(vt128 (bitconvert (loadv2i64 addr:$src2))),
i8immZExt3:$cc)))]>,
XOP_4V;
let isAsmParserOnly = 1, hasSideEffects = 0 in {
def ri_alt : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, u8imm:$src3),
!strconcat("vpcom", Suffix,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[]>, XOP_4V;
let mayLoad = 1 in
def mi_alt : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, u8imm:$src3),
!strconcat("vpcom", Suffix,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[]>, XOP_4V;
}
}
let ExeDomain = SSEPackedInt in { // SSE integer instructions
defm VPCOMB : xopvpcom<0xCC, "b", X86vpcom, v16i8>;
defm VPCOMW : xopvpcom<0xCD, "w", X86vpcom, v8i16>;
defm VPCOMD : xopvpcom<0xCE, "d", X86vpcom, v4i32>;
defm VPCOMQ : xopvpcom<0xCF, "q", X86vpcom, v2i64>;
defm VPCOMUB : xopvpcom<0xEC, "ub", X86vpcomu, v16i8>;
defm VPCOMUW : xopvpcom<0xED, "uw", X86vpcomu, v8i16>;
defm VPCOMUD : xopvpcom<0xEE, "ud", X86vpcomu, v4i32>;
defm VPCOMUQ : xopvpcom<0xEF, "uq", X86vpcomu, v2i64>;
}
// Instruction where either second or third source can be memory
multiclass xop4op<bits<8> opc, string OpcodeStr, Intrinsic Int> {
def rr : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst, (Int VR128:$src1, VR128:$src2, VR128:$src3))]>,
XOP_4V, VEX_I8IMM;
def rm : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i128mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(Int VR128:$src1, VR128:$src2,
(bitconvert (loadv2i64 addr:$src3))))]>,
XOP_4V, VEX_I8IMM, VEX_W, MemOp4;
def mr : IXOPi8<opc, MRMSrcMem, (outs VR128:$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:$src3))]>,
XOP_4V, VEX_I8IMM;
}
let ExeDomain = SSEPackedInt in {
defm VPPERM : xop4op<0xA3, "vpperm", int_x86_xop_vpperm>;
defm VPCMOV : xop4op<0xA2, "vpcmov", int_x86_xop_vpcmov>;
}
multiclass xop4op256<bits<8> opc, string OpcodeStr, Intrinsic Int> {
def rrY : IXOPi8<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst, (Int VR256:$src1, VR256:$src2, VR256:$src3))]>,
XOP_4V, VEX_I8IMM, VEX_L;
def rmY : IXOPi8<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, i256mem:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst,
(Int VR256:$src1, VR256:$src2,
(bitconvert (loadv4i64 addr:$src3))))]>,
XOP_4V, VEX_I8IMM, VEX_W, MemOp4, VEX_L;
def mrY : IXOPi8<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, VR256:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR256:$dst,
(Int VR256:$src1, (bitconvert (loadv4i64 addr:$src2)),
VR256:$src3))]>,
XOP_4V, VEX_I8IMM, VEX_L;
}
let ExeDomain = SSEPackedInt in
defm VPCMOV : xop4op256<0xA2, "vpcmov", int_x86_xop_vpcmov_256>;
let Predicates = [HasXOP] in {
def : Pat<(v2i64 (or (and VR128:$src3, VR128:$src1),
(X86andnp VR128:$src3, VR128:$src2))),
(VPCMOVrr VR128:$src1, VR128:$src2, VR128:$src3)>;
def : Pat<(v4i64 (or (and VR256:$src3, VR256:$src1),
(X86andnp VR256:$src3, VR256:$src2))),
(VPCMOVrrY VR256:$src1, VR256:$src2, VR256:$src3)>;
}
multiclass xop5op<bits<8> opc, string OpcodeStr, Intrinsic Int128,
Intrinsic Int256, PatFrag ld_128, PatFrag ld_256> {
def rr : IXOP5<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, VR128:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set VR128:$dst,
(Int128 VR128:$src1, VR128:$src2, VR128:$src3, imm:$src4))]>;
def rm : IXOP5<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, f128mem:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set VR128:$dst,
(Int128 VR128:$src1, VR128:$src2, (ld_128 addr:$src3), imm:$src4))]>,
VEX_W, MemOp4;
def mr : IXOP5<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, f128mem:$src2, VR128:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set VR128:$dst,
(Int128 VR128:$src1, (ld_128 addr:$src2), VR128:$src3, imm:$src4))]>;
def rrY : IXOP5<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, VR256:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set VR256:$dst,
(Int256 VR256:$src1, VR256:$src2, VR256:$src3, imm:$src4))]>, VEX_L;
def rmY : IXOP5<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, f256mem:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set VR256:$dst,
(Int256 VR256:$src1, VR256:$src2, (ld_256 addr:$src3), imm:$src4))]>,
VEX_W, MemOp4, VEX_L;
def mrY : IXOP5<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, VR256:$src3, u8imm:$src4),
!strconcat(OpcodeStr,
"\t{$src4, $src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3, $src4}"),
[(set VR256:$dst,
(Int256 VR256:$src1, (ld_256 addr:$src2), VR256:$src3, imm:$src4))]>,
VEX_L;
}
let ExeDomain = SSEPackedDouble in
defm VPERMIL2PD : xop5op<0x49, "vpermil2pd", int_x86_xop_vpermil2pd,
int_x86_xop_vpermil2pd_256, loadv2f64, loadv4f64>;
let ExeDomain = SSEPackedSingle in
defm VPERMIL2PS : xop5op<0x48, "vpermil2ps", int_x86_xop_vpermil2ps,
int_x86_xop_vpermil2ps_256, loadv4f32, loadv8f32>;