llvm-6502/lib/Target/R600/AMDGPUInstructions.td
Tom Stellard 83f0a5a5e8 R600: Add pattern for SHA-256 Ma function
This can be optimized using the BFI_INT instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181033 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-03 17:21:20 +00:00

291 lines
8.7 KiB
TableGen

//===-- AMDGPUInstructions.td - Common instruction defs ---*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains instruction defs that are common to all hw codegen
// targets.
//
//===----------------------------------------------------------------------===//
class AMDGPUInst <dag outs, dag ins, string asm, list<dag> pattern> : Instruction {
field bit isRegisterLoad = 0;
field bit isRegisterStore = 0;
let Namespace = "AMDGPU";
let OutOperandList = outs;
let InOperandList = ins;
let AsmString = asm;
let Pattern = pattern;
let Itinerary = NullALU;
let TSFlags{63} = isRegisterLoad;
let TSFlags{62} = isRegisterStore;
}
class AMDGPUShaderInst <dag outs, dag ins, string asm, list<dag> pattern>
: AMDGPUInst<outs, ins, asm, pattern> {
field bits<32> Inst = 0xffffffff;
}
def InstFlag : OperandWithDefaultOps <i32, (ops (i32 0))>;
def COND_EQ : PatLeaf <
(cond),
[{switch(N->get()){{default: return false;
case ISD::SETOEQ: case ISD::SETUEQ:
case ISD::SETEQ: return true;}}}]
>;
def COND_NE : PatLeaf <
(cond),
[{switch(N->get()){{default: return false;
case ISD::SETONE: case ISD::SETUNE:
case ISD::SETNE: return true;}}}]
>;
def COND_GT : PatLeaf <
(cond),
[{switch(N->get()){{default: return false;
case ISD::SETOGT: case ISD::SETUGT:
case ISD::SETGT: return true;}}}]
>;
def COND_GE : PatLeaf <
(cond),
[{switch(N->get()){{default: return false;
case ISD::SETOGE: case ISD::SETUGE:
case ISD::SETGE: return true;}}}]
>;
def COND_LT : PatLeaf <
(cond),
[{switch(N->get()){{default: return false;
case ISD::SETOLT: case ISD::SETULT:
case ISD::SETLT: return true;}}}]
>;
def COND_LE : PatLeaf <
(cond),
[{switch(N->get()){{default: return false;
case ISD::SETOLE: case ISD::SETULE:
case ISD::SETLE: return true;}}}]
>;
def COND_NULL : PatLeaf <
(cond),
[{return false;}]
>;
//===----------------------------------------------------------------------===//
// Load/Store Pattern Fragments
//===----------------------------------------------------------------------===//
def zextloadi8_global : PatFrag<(ops node:$ptr), (zextloadi8 node:$ptr), [{
return isGlobalLoad(dyn_cast<LoadSDNode>(N));
}]>;
class Constants {
int TWO_PI = 0x40c90fdb;
int PI = 0x40490fdb;
int TWO_PI_INV = 0x3e22f983;
int FP_UINT_MAX_PLUS_1 = 0x4f800000; // 1 << 32 in floating point encoding
}
def CONST : Constants;
def FP_ZERO : PatLeaf <
(fpimm),
[{return N->getValueAPF().isZero();}]
>;
def FP_ONE : PatLeaf <
(fpimm),
[{return N->isExactlyValue(1.0);}]
>;
let isCodeGenOnly = 1, isPseudo = 1 in {
let usesCustomInserter = 1 in {
class CLAMP <RegisterClass rc> : AMDGPUShaderInst <
(outs rc:$dst),
(ins rc:$src0),
"CLAMP $dst, $src0",
[(set f32:$dst, (int_AMDIL_clamp f32:$src0, (f32 FP_ZERO), (f32 FP_ONE)))]
>;
class FABS <RegisterClass rc> : AMDGPUShaderInst <
(outs rc:$dst),
(ins rc:$src0),
"FABS $dst, $src0",
[(set f32:$dst, (fabs f32:$src0))]
>;
class FNEG <RegisterClass rc> : AMDGPUShaderInst <
(outs rc:$dst),
(ins rc:$src0),
"FNEG $dst, $src0",
[(set f32:$dst, (fneg f32:$src0))]
>;
} // usesCustomInserter = 1
multiclass RegisterLoadStore <RegisterClass dstClass, Operand addrClass,
ComplexPattern addrPat> {
def RegisterLoad : AMDGPUShaderInst <
(outs dstClass:$dst),
(ins addrClass:$addr, i32imm:$chan),
"RegisterLoad $dst, $addr",
[(set i32:$dst, (AMDGPUregister_load addrPat:$addr, (i32 timm:$chan)))]
> {
let isRegisterLoad = 1;
}
def RegisterStore : AMDGPUShaderInst <
(outs),
(ins dstClass:$val, addrClass:$addr, i32imm:$chan),
"RegisterStore $val, $addr",
[(AMDGPUregister_store i32:$val, addrPat:$addr, (i32 timm:$chan))]
> {
let isRegisterStore = 1;
}
}
} // End isCodeGenOnly = 1, isPseudo = 1
/* Generic helper patterns for intrinsics */
/* -------------------------------------- */
class POW_Common <AMDGPUInst log_ieee, AMDGPUInst exp_ieee, AMDGPUInst mul>
: Pat <
(fpow f32:$src0, f32:$src1),
(exp_ieee (mul f32:$src1, (log_ieee f32:$src0)))
>;
/* Other helper patterns */
/* --------------------- */
/* Extract element pattern */
class Extract_Element <ValueType sub_type, ValueType vec_type, int sub_idx,
SubRegIndex sub_reg>
: Pat<
(sub_type (vector_extract vec_type:$src, sub_idx)),
(EXTRACT_SUBREG $src, sub_reg)
>;
/* Insert element pattern */
class Insert_Element <ValueType elem_type, ValueType vec_type,
int sub_idx, SubRegIndex sub_reg>
: Pat <
(vector_insert vec_type:$vec, elem_type:$elem, sub_idx),
(INSERT_SUBREG $vec, $elem, sub_reg)
>;
// Vector Build pattern
class Vector1_Build <ValueType vecType, ValueType elemType,
RegisterClass rc> : Pat <
(vecType (build_vector elemType:$src)),
(vecType (COPY_TO_REGCLASS $src, rc))
>;
class Vector2_Build <ValueType vecType, ValueType elemType> : Pat <
(vecType (build_vector elemType:$sub0, elemType:$sub1)),
(INSERT_SUBREG (INSERT_SUBREG
(vecType (IMPLICIT_DEF)), $sub0, sub0), $sub1, sub1)
>;
class Vector4_Build <ValueType vecType, ValueType elemType> : Pat <
(vecType (build_vector elemType:$x, elemType:$y, elemType:$z, elemType:$w)),
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(vecType (IMPLICIT_DEF)), $x, sub0), $y, sub1), $z, sub2), $w, sub3)
>;
class Vector8_Build <ValueType vecType, ValueType elemType> : Pat <
(vecType (build_vector elemType:$sub0, elemType:$sub1,
elemType:$sub2, elemType:$sub3,
elemType:$sub4, elemType:$sub5,
elemType:$sub6, elemType:$sub7)),
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(vecType (IMPLICIT_DEF)), $sub0, sub0), $sub1, sub1),
$sub2, sub2), $sub3, sub3),
$sub4, sub4), $sub5, sub5),
$sub6, sub6), $sub7, sub7)
>;
class Vector16_Build <ValueType vecType, ValueType elemType> : Pat <
(vecType (build_vector elemType:$sub0, elemType:$sub1,
elemType:$sub2, elemType:$sub3,
elemType:$sub4, elemType:$sub5,
elemType:$sub6, elemType:$sub7,
elemType:$sub8, elemType:$sub9,
elemType:$sub10, elemType:$sub11,
elemType:$sub12, elemType:$sub13,
elemType:$sub14, elemType:$sub15)),
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG
(vecType (IMPLICIT_DEF)), $sub0, sub0), $sub1, sub1),
$sub2, sub2), $sub3, sub3),
$sub4, sub4), $sub5, sub5),
$sub6, sub6), $sub7, sub7),
$sub8, sub8), $sub9, sub9),
$sub10, sub10), $sub11, sub11),
$sub12, sub12), $sub13, sub13),
$sub14, sub14), $sub15, sub15)
>;
// XXX: Convert to new syntax and use COPY_TO_REG, once the DFAPacketizer
// can handle COPY instructions.
// bitconvert pattern
class BitConvert <ValueType dt, ValueType st, RegisterClass rc> : Pat <
(dt (bitconvert (st rc:$src0))),
(dt rc:$src0)
>;
// XXX: Convert to new syntax and use COPY_TO_REG, once the DFAPacketizer
// can handle COPY instructions.
class DwordAddrPat<ValueType vt, RegisterClass rc> : Pat <
(vt (AMDGPUdwordaddr (vt rc:$addr))),
(vt rc:$addr)
>;
// BFI_INT patterns
multiclass BFIPatterns <Instruction BFI_INT> {
// Definition from ISA doc:
// (y & x) | (z & ~x)
def : Pat <
(or (and i32:$y, i32:$x), (and i32:$z, (not i32:$x))),
(BFI_INT $x, $y, $z)
>;
// SHA-256 Ch function
// z ^ (x & (y ^ z))
def : Pat <
(xor i32:$z, (and i32:$x, (xor i32:$y, i32:$z))),
(BFI_INT $x, $y, $z)
>;
}
// SHA-256 Ma patterns
// ((x & z) | (y & (x | z))) -> BFI_INT (XOR x, y), z, y
class SHA256MaPattern <Instruction BFI_INT, Instruction XOR> : Pat <
(or (and i32:$x, i32:$z), (and i32:$y, (or i32:$x, i32:$z))),
(BFI_INT (XOR i32:$x, i32:$y), i32:$z, i32:$y)
>;
include "R600Instructions.td"
include "SIInstrInfo.td"