diff options
Diffstat (limited to 'contrib/llvm/lib/Target/X86/X86InstrCompiler.td')
| -rw-r--r-- | contrib/llvm/lib/Target/X86/X86InstrCompiler.td | 44 |
1 files changed, 26 insertions, 18 deletions
diff --git a/contrib/llvm/lib/Target/X86/X86InstrCompiler.td b/contrib/llvm/lib/Target/X86/X86InstrCompiler.td index 3c27eb8..d003d02 100644 --- a/contrib/llvm/lib/Target/X86/X86InstrCompiler.td +++ b/contrib/llvm/lib/Target/X86/X86InstrCompiler.td @@ -43,7 +43,8 @@ let hasSideEffects = 0, isNotDuplicable = 1, Uses = [ESP] in // Pessimistically assume ADJCALLSTACKDOWN / ADJCALLSTACKUP will become // sub / add which can clobber EFLAGS. let Defs = [ESP, EFLAGS], Uses = [ESP] in { -def ADJCALLSTACKDOWN32 : I<0, Pseudo, (outs), (ins i32imm:$amt1, i32imm:$amt2), +def ADJCALLSTACKDOWN32 : I<0, Pseudo, (outs), + (ins i32imm:$amt1, i32imm:$amt2, i32imm:$amt3), "#ADJCALLSTACKDOWN", []>, Requires<[NotLP64]>; @@ -52,8 +53,8 @@ def ADJCALLSTACKUP32 : I<0, Pseudo, (outs), (ins i32imm:$amt1, i32imm:$amt2), [(X86callseq_end timm:$amt1, timm:$amt2)]>, Requires<[NotLP64]>; } -def : Pat<(X86callseq_start timm:$amt1), - (ADJCALLSTACKDOWN32 i32imm:$amt1, 0)>, Requires<[NotLP64]>; +def : Pat<(X86callseq_start timm:$amt1, timm:$amt2), + (ADJCALLSTACKDOWN32 i32imm:$amt1, i32imm:$amt2, 0)>, Requires<[NotLP64]>; // ADJCALLSTACKDOWN/UP implicitly use/def RSP because they may be expanded into @@ -62,7 +63,8 @@ def : Pat<(X86callseq_start timm:$amt1), // Pessimistically assume ADJCALLSTACKDOWN / ADJCALLSTACKUP will become // sub / add which can clobber EFLAGS. let Defs = [RSP, EFLAGS], Uses = [RSP] in { -def ADJCALLSTACKDOWN64 : I<0, Pseudo, (outs), (ins i32imm:$amt1, i32imm:$amt2), +def ADJCALLSTACKDOWN64 : I<0, Pseudo, (outs), + (ins i32imm:$amt1, i32imm:$amt2, i32imm:$amt3), "#ADJCALLSTACKDOWN", []>, Requires<[IsLP64]>; @@ -71,8 +73,8 @@ def ADJCALLSTACKUP64 : I<0, Pseudo, (outs), (ins i32imm:$amt1, i32imm:$amt2), [(X86callseq_end timm:$amt1, timm:$amt2)]>, Requires<[IsLP64]>; } -def : Pat<(X86callseq_start timm:$amt1), - (ADJCALLSTACKDOWN64 i32imm:$amt1, 0)>, Requires<[IsLP64]>; +def : Pat<(X86callseq_start timm:$amt1, timm:$amt2), + (ADJCALLSTACKDOWN64 i32imm:$amt1, i32imm:$amt2, 0)>, Requires<[IsLP64]>; // x86-64 va_start lowering magic. @@ -259,20 +261,20 @@ def MORESTACK_RET_RESTORE_R10 : I<0, Pseudo, (outs), (ins), // Alias instruction mapping movr0 to xor. // FIXME: remove when we can teach regalloc that xor reg, reg is ok. let Defs = [EFLAGS], isReMaterializable = 1, isAsCheapAsAMove = 1, - isPseudo = 1, AddedComplexity = 20 in + isPseudo = 1, AddedComplexity = 10 in def MOV32r0 : I<0, Pseudo, (outs GR32:$dst), (ins), "", [(set GR32:$dst, 0)], IIC_ALU_NONMEM>, Sched<[WriteZero]>; // Other widths can also make use of the 32-bit xor, which may have a smaller // encoding and avoid partial register updates. +let AddedComplexity = 10 in { def : Pat<(i8 0), (EXTRACT_SUBREG (MOV32r0), sub_8bit)>; def : Pat<(i16 0), (EXTRACT_SUBREG (MOV32r0), sub_16bit)>; -def : Pat<(i64 0), (SUBREG_TO_REG (i64 0), (MOV32r0), sub_32bit)> { - let AddedComplexity = 20; +def : Pat<(i64 0), (SUBREG_TO_REG (i64 0), (MOV32r0), sub_32bit)>; } let Predicates = [OptForSize, NotSlowIncDec, Not64BitMode], - AddedComplexity = 15 in { + AddedComplexity = 10 in { // Pseudo instructions for materializing 1 and -1 using XOR+INC/DEC, // which only require 3 bytes compared to MOV32ri which requires 5. let Defs = [EFLAGS], isReMaterializable = 1, isPseudo = 1 in { @@ -287,7 +289,7 @@ let Predicates = [OptForSize, NotSlowIncDec, Not64BitMode], def : Pat<(i16 -1), (EXTRACT_SUBREG (MOV32r_1), sub_16bit)>; } -let isReMaterializable = 1, isPseudo = 1, AddedComplexity = 10 in { +let isReMaterializable = 1, isPseudo = 1, AddedComplexity = 5 in { // AddedComplexity higher than MOV64ri but lower than MOV32r0 and MOV32r1. // FIXME: Add itinerary class and Schedule. def MOV32ImmSExti8 : I<0, Pseudo, (outs GR32:$dst), (ins i32i8imm:$src), "", @@ -772,11 +774,11 @@ defm LCMPXCHG8B : LCMPXCHG_UnOp<0xC7, MRM1m, "cmpxchg8b", // the pseudo. The argument feeding EBX is ebx_input. // // The additional argument, $ebx_save, is a temporary register used to -// save the value of RBX accross the actual instruction. +// save the value of RBX across the actual instruction. // // To make sure the register assigned to $ebx_save does not interfere with // the definition of the actual instruction, we use a definition $dst which -// is tied to $rbx_save. That way, the live-range of $rbx_save spans accross +// is tied to $rbx_save. That way, the live-range of $rbx_save spans across // the instruction and we are sure we will have a valid register to restore // the value of RBX. let Defs = [EAX, EDX, EBX, EFLAGS], Uses = [EAX, ECX, EDX], @@ -1271,11 +1273,11 @@ def or_is_add : PatFrag<(ops node:$lhs, node:$rhs), (or node:$lhs, node:$rhs),[{ if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N->getOperand(1))) return CurDAG->MaskedValueIsZero(N->getOperand(0), CN->getAPIntValue()); - APInt KnownZero0, KnownOne0; - CurDAG->computeKnownBits(N->getOperand(0), KnownZero0, KnownOne0, 0); - APInt KnownZero1, KnownOne1; - CurDAG->computeKnownBits(N->getOperand(1), KnownZero1, KnownOne1, 0); - return (~KnownZero0 & ~KnownZero1) == 0; + KnownBits Known0; + CurDAG->computeKnownBits(N->getOperand(0), Known0, 0); + KnownBits Known1; + CurDAG->computeKnownBits(N->getOperand(1), Known1, 0); + return (~Known0.Zero & ~Known1.Zero) == 0; }]>; @@ -1743,6 +1745,12 @@ def : Pat<(X86sub_flag 0, GR16:$src), (NEG16r GR16:$src)>; def : Pat<(X86sub_flag 0, GR32:$src), (NEG32r GR32:$src)>; def : Pat<(X86sub_flag 0, GR64:$src), (NEG64r GR64:$src)>; +// sub reg, relocImm +def : Pat<(X86sub_flag GR64:$src1, i64relocImmSExt8_su:$src2), + (SUB64ri8 GR64:$src1, i64relocImmSExt8_su:$src2)>; +def : Pat<(X86sub_flag GR64:$src1, i64relocImmSExt32_su:$src2), + (SUB64ri32 GR64:$src1, i64relocImmSExt32_su:$src2)>; + // mul reg, reg def : Pat<(mul GR16:$src1, GR16:$src2), (IMUL16rr GR16:$src1, GR16:$src2)>; |
