//
// The LLVM Compiler Infrastructure
//
-// This file was developed by Evan Cheng and is distributed under the
-// University of Illinois Open Source License. See LICENSE.TXT for details.
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
//
//===----------------------------------------------------------------------===//
-// Some 'special' instructions
-def IMPLICIT_DEF_VR64 : I<0, Pseudo, (outs VR64:$dst), (ins),
- "#IMPLICIT_DEF $dst",
- [(set VR64:$dst, (v8i8 (undef)))]>,
- Requires<[HasMMX]>;
-
-// 64-bit vector undef's.
-def : Pat<(v8i8 (undef)), (IMPLICIT_DEF_VR64)>;
-def : Pat<(v4i16 (undef)), (IMPLICIT_DEF_VR64)>;
-def : Pat<(v2i32 (undef)), (IMPLICIT_DEF_VR64)>;
-def : Pat<(v1i64 (undef)), (IMPLICIT_DEF_VR64)>;
-
//===----------------------------------------------------------------------===//
// MMX Pattern Fragments
//===----------------------------------------------------------------------===//
return X86::isPSHUFDMask(N);
}], MMX_SHUFFLE_get_shuf_imm>;
-// Patterns for: vector_shuffle v1, v2, <4, 5, 2, 3>; etc.
-def MMX_MOVL_shuffle_mask : PatLeaf<(build_vector), [{
- return X86::isMOVLMask(N);
-}]>;
-
//===----------------------------------------------------------------------===//
// MMX Multiclasses
//===----------------------------------------------------------------------===//
//
multiclass MMXI_binop_rm_v1i64<bits<8> opc, string OpcodeStr, SDNode OpNode,
bit Commutable = 0> {
- def rr : MMXI<opc, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src1, VR64:$src2),
+ def rr : MMXI<opc, MRMSrcReg, (outs VR64:$dst),
+ (ins VR64:$src1, VR64:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst, (v1i64 (OpNode VR64:$src1, VR64:$src2)))]> {
let isCommutable = Commutable;
}
- def rm : MMXI<opc, MRMSrcMem, (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2),
+ def rm : MMXI<opc, MRMSrcMem, (outs VR64:$dst),
+ (ins VR64:$src1, i64mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst,
(OpNode VR64:$src1,(load_mmx addr:$src2)))]>;
}
multiclass MMXI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
- string OpcodeStr, Intrinsic IntId> {
- def rr : MMXI<opc, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src1, VR64:$src2),
+ string OpcodeStr, Intrinsic IntId,
+ Intrinsic IntId2> {
+ def rr : MMXI<opc, MRMSrcReg, (outs VR64:$dst),
+ (ins VR64:$src1, VR64:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst, (IntId VR64:$src1, VR64:$src2))]>;
- def rm : MMXI<opc, MRMSrcMem, (outs VR64:$dst), (ins VR64:$src1, i64mem:$src2),
+ def rm : MMXI<opc, MRMSrcMem, (outs VR64:$dst),
+ (ins VR64:$src1, i64mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst, (IntId VR64:$src1,
(bitconvert (load_mmx addr:$src2))))]>;
- def ri : MMXIi8<opc2, ImmForm, (outs VR64:$dst), (ins VR64:$src1, i32i8imm:$src2),
+ def ri : MMXIi8<opc2, ImmForm, (outs VR64:$dst),
+ (ins VR64:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
- [(set VR64:$dst, (IntId VR64:$src1,
- (scalar_to_vector (i32 imm:$src2))))]>;
+ [(set VR64:$dst, (IntId2 VR64:$src1, (i32 imm:$src2)))]>;
}
}
// Data Transfer Instructions
def MMX_MOVD64rr : MMXI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src),
- "movd\t{$src, $dst|$dst, $src}", []>;
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set VR64:$dst, (v2i32 (scalar_to_vector GR32:$src)))]>;
+let canFoldAsLoad = 1, isReMaterializable = 1 in
def MMX_MOVD64rm : MMXI<0x6E, MRMSrcMem, (outs VR64:$dst), (ins i32mem:$src),
- "movd\t{$src, $dst|$dst, $src}", []>;
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set VR64:$dst, (v2i32 (scalar_to_vector (loadi32 addr:$src))))]>;
+let mayStore = 1 in
def MMX_MOVD64mr : MMXI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR64:$src),
"movd\t{$src, $dst|$dst, $src}", []>;
+let neverHasSideEffects = 1 in
def MMX_MOVD64to64rr : MMXRI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR64:$src),
"movd\t{$src, $dst|$dst, $src}", []>;
+let neverHasSideEffects = 1 in
+def MMX_MOVD64from64rr : MMXRI<0x7E, MRMSrcReg, (outs GR64:$dst), (ins VR64:$src),
+ "movd\t{$src, $dst|$dst, $src}", []>;
+
+let neverHasSideEffects = 1 in
def MMX_MOVQ64rr : MMXI<0x6F, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src),
"movq\t{$src, $dst|$dst, $src}", []>;
+let canFoldAsLoad = 1, isReMaterializable = 1, mayHaveSideEffects = 1 in
def MMX_MOVQ64rm : MMXI<0x6F, MRMSrcMem, (outs VR64:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (load_mmx addr:$src))]>;
"movq\t{$src, $dst|$dst, $src}",
[(store (v1i64 VR64:$src), addr:$dst)]>;
-def MMX_MOVDQ2Qrr : MMXID<0xD6, MRMDestMem, (outs VR64:$dst), (ins VR128:$src),
+def MMX_MOVDQ2Qrr : SDIi8<0xD6, MRMDestMem, (outs VR64:$dst), (ins VR128:$src),
"movdq2q\t{$src, $dst|$dst, $src}",
[(set VR64:$dst,
- (v1i64 (vector_extract (v2i64 VR128:$src),
- (iPTR 0))))]>;
+ (v1i64 (bitconvert
+ (i64 (vector_extract (v2i64 VR128:$src),
+ (iPTR 0))))))]>;
-def MMX_MOVQ2DQrr : MMXIS<0xD6, MRMDestMem, (outs VR128:$dst), (ins VR64:$src),
- "movq2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (bitconvert (v1i64 VR64:$src)))]>;
+def MMX_MOVQ2DQrr : SSDIi8<0xD6, MRMDestMem, (outs VR128:$dst), (ins VR64:$src),
+ "movq2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (v2i64 (vector_shuffle immAllZerosV,
+ (v2i64 (scalar_to_vector (i64 (bitconvert VR64:$src)))),
+ MOVL_shuffle_mask)))]>;
def MMX_MOVNTQmr : MMXI<0xE7, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src),
"movntq\t{$src, $dst|$dst, $src}",
let AddedComplexity = 15 in
// movd to MMX register zero-extends
-def MMX_MOVZDI2PDIrr : MMX2I<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src),
+def MMX_MOVZDI2PDIrr : MMXI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst,
- (v2i32 (vector_shuffle immAllZerosV,
- (v2i32 (scalar_to_vector GR32:$src)),
- MMX_MOVL_shuffle_mask)))]>;
+ [(set VR64:$dst,
+ (v2i32 (X86vzmovl (v2i32 (scalar_to_vector GR32:$src)))))]>;
let AddedComplexity = 20 in
-def MMX_MOVZDI2PDIrm : MMX2I<0x6E, MRMSrcMem, (outs VR64:$dst), (ins i32mem:$src),
+def MMX_MOVZDI2PDIrm : MMXI<0x6E, MRMSrcMem, (outs VR64:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst,
- (v2i32 (vector_shuffle immAllZerosV,
- (v2i32 (scalar_to_vector
- (loadi32 addr:$src))),
- MMX_MOVL_shuffle_mask)))]>;
+ [(set VR64:$dst,
+ (v2i32 (X86vzmovl (v2i32
+ (scalar_to_vector (loadi32 addr:$src))))))]>;
// Arithmetic Instructions
// Shift Instructions
defm MMX_PSRLW : MMXI_binop_rmi_int<0xD1, 0x71, MRM2r, "psrlw",
- int_x86_mmx_psrl_w>;
+ int_x86_mmx_psrl_w, int_x86_mmx_psrli_w>;
defm MMX_PSRLD : MMXI_binop_rmi_int<0xD2, 0x72, MRM2r, "psrld",
- int_x86_mmx_psrl_d>;
+ int_x86_mmx_psrl_d, int_x86_mmx_psrli_d>;
defm MMX_PSRLQ : MMXI_binop_rmi_int<0xD3, 0x73, MRM2r, "psrlq",
- int_x86_mmx_psrl_q>;
+ int_x86_mmx_psrl_q, int_x86_mmx_psrli_q>;
defm MMX_PSLLW : MMXI_binop_rmi_int<0xF1, 0x71, MRM6r, "psllw",
- int_x86_mmx_psll_w>;
+ int_x86_mmx_psll_w, int_x86_mmx_pslli_w>;
defm MMX_PSLLD : MMXI_binop_rmi_int<0xF2, 0x72, MRM6r, "pslld",
- int_x86_mmx_psll_d>;
+ int_x86_mmx_psll_d, int_x86_mmx_pslli_d>;
defm MMX_PSLLQ : MMXI_binop_rmi_int<0xF3, 0x73, MRM6r, "psllq",
- int_x86_mmx_psll_q>;
+ int_x86_mmx_psll_q, int_x86_mmx_pslli_q>;
defm MMX_PSRAW : MMXI_binop_rmi_int<0xE1, 0x71, MRM4r, "psraw",
- int_x86_mmx_psra_w>;
+ int_x86_mmx_psra_w, int_x86_mmx_psrai_w>;
defm MMX_PSRAD : MMXI_binop_rmi_int<0xE2, 0x72, MRM4r, "psrad",
- int_x86_mmx_psra_d>;
+ int_x86_mmx_psra_d, int_x86_mmx_psrai_d>;
+
+// Shift up / down and insert zero's.
+def : Pat<(v1i64 (X86vshl VR64:$src, (i8 imm:$amt))),
+ (v1i64 (MMX_PSLLQri VR64:$src, imm:$amt))>;
+def : Pat<(v1i64 (X86vshr VR64:$src, (i8 imm:$amt))),
+ (v1i64 (MMX_PSRLQri VR64:$src, imm:$amt))>;
// Comparison Instructions
defm MMX_PCMPEQB : MMXI_binop_rm_int<0x74, "pcmpeqb", int_x86_mmx_pcmpeq_b>;
MMX_PSHUFW_shuffle_mask:$src2)))]>;
// -- Conversion Instructions
+let neverHasSideEffects = 1 in {
def MMX_CVTPD2PIrr : MMX2I<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvtpd2pi\t{$src, $dst|$dst, $src}", []>;
+let mayLoad = 1 in
def MMX_CVTPD2PIrm : MMX2I<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
"cvtpd2pi\t{$src, $dst|$dst, $src}", []>;
def MMX_CVTPI2PDrr : MMX2I<0x2A, MRMSrcReg, (outs VR128:$dst), (ins VR64:$src),
"cvtpi2pd\t{$src, $dst|$dst, $src}", []>;
+let mayLoad = 1 in
def MMX_CVTPI2PDrm : MMX2I<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"cvtpi2pd\t{$src, $dst|$dst, $src}", []>;
def MMX_CVTPI2PSrr : MMXI<0x2A, MRMSrcReg, (outs VR128:$dst), (ins VR64:$src),
"cvtpi2ps\t{$src, $dst|$dst, $src}", []>;
+let mayLoad = 1 in
def MMX_CVTPI2PSrm : MMXI<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"cvtpi2ps\t{$src, $dst|$dst, $src}", []>;
def MMX_CVTPS2PIrr : MMXI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvtps2pi\t{$src, $dst|$dst, $src}", []>;
+let mayLoad = 1 in
def MMX_CVTPS2PIrm : MMXI<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f64mem:$src),
"cvtps2pi\t{$src, $dst|$dst, $src}", []>;
def MMX_CVTTPD2PIrr : MMX2I<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvttpd2pi\t{$src, $dst|$dst, $src}", []>;
+let mayLoad = 1 in
def MMX_CVTTPD2PIrm : MMX2I<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
"cvttpd2pi\t{$src, $dst|$dst, $src}", []>;
def MMX_CVTTPS2PIrr : MMXI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvttps2pi\t{$src, $dst|$dst, $src}", []>;
+let mayLoad = 1 in
def MMX_CVTTPS2PIrm : MMXI<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f64mem:$src),
"cvttps2pi\t{$src, $dst|$dst, $src}", []>;
+} // end neverHasSideEffects
+
// Extract / Insert
def MMX_X86pextrw : SDNode<"X86ISD::PEXTRW", SDTypeProfile<1, 2, []>, []>;
[(set GR32:$dst, (int_x86_mmx_pmovmskb VR64:$src))]>;
// Misc.
+let Uses = [EDI] in
def MMX_MASKMOVQ : MMXI<0xF7, MRMDestMem, (outs), (ins VR64:$src, VR64:$mask),
"maskmovq\t{$mask, $src|$src, $mask}",
- [(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, EDI)]>,
- Imp<[EDI],[]>;
+ [(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, EDI)]>;
+let Uses = [RDI] in
+def MMX_MASKMOVQ64: MMXI64<0xF7, MRMDestMem, (outs), (ins VR64:$src, VR64:$mask),
+ "maskmovq\t{$mask, $src|$src, $mask}",
+ [(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, RDI)]>;
//===----------------------------------------------------------------------===//
// Alias Instructions
//===----------------------------------------------------------------------===//
// Alias instructions that map zero vector to pxor.
-// FIXME: remove when we can teach regalloc that xor reg, reg is ok.
let isReMaterializable = 1 in {
def MMX_V_SET0 : MMXI<0xEF, MRMInitReg, (outs VR64:$dst), (ins),
"pxor\t$dst, $dst",
- [(set VR64:$dst, (v1i64 immAllZerosV))]>;
+ [(set VR64:$dst, (v2i32 immAllZerosV))]>;
def MMX_V_SETALLONES : MMXI<0x76, MRMInitReg, (outs VR64:$dst), (ins),
"pcmpeqd\t$dst, $dst",
- [(set VR64:$dst, (v1i64 immAllOnesV))]>;
+ [(set VR64:$dst, (v2i32 immAllOnesV))]>;
+}
+
+let Predicates = [HasMMX] in {
+ def : Pat<(v1i64 immAllZerosV), (MMX_V_SET0)>;
+ def : Pat<(v4i16 immAllZerosV), (MMX_V_SET0)>;
+ def : Pat<(v8i8 immAllZerosV), (MMX_V_SET0)>;
}
//===----------------------------------------------------------------------===//
(MMX_MOVQ64mr addr:$dst, VR64:$src)>;
def : Pat<(store (v2i32 VR64:$src), addr:$dst),
(MMX_MOVQ64mr addr:$dst, VR64:$src)>;
+def : Pat<(store (v2f32 VR64:$src), addr:$dst),
+ (MMX_MOVQ64mr addr:$dst, VR64:$src)>;
def : Pat<(store (v1i64 VR64:$src), addr:$dst),
(MMX_MOVQ64mr addr:$dst, VR64:$src)>;
-// 64-bit vector all zero's.
-def : Pat<(v8i8 immAllZerosV), (MMX_V_SET0)>;
-def : Pat<(v4i16 immAllZerosV), (MMX_V_SET0)>;
-def : Pat<(v2i32 immAllZerosV), (MMX_V_SET0)>;
-def : Pat<(v1i64 immAllZerosV), (MMX_V_SET0)>;
-
-// 64-bit vector all one's.
-def : Pat<(v8i8 immAllOnesV), (MMX_V_SETALLONES)>;
-def : Pat<(v4i16 immAllOnesV), (MMX_V_SETALLONES)>;
-def : Pat<(v2i32 immAllOnesV), (MMX_V_SETALLONES)>;
-def : Pat<(v1i64 immAllOnesV), (MMX_V_SETALLONES)>;
-
// Bit convert.
def : Pat<(v8i8 (bitconvert (v1i64 VR64:$src))), (v8i8 VR64:$src)>;
def : Pat<(v8i8 (bitconvert (v2i32 VR64:$src))), (v8i8 VR64:$src)>;
+def : Pat<(v8i8 (bitconvert (v2f32 VR64:$src))), (v8i8 VR64:$src)>;
def : Pat<(v8i8 (bitconvert (v4i16 VR64:$src))), (v8i8 VR64:$src)>;
def : Pat<(v4i16 (bitconvert (v1i64 VR64:$src))), (v4i16 VR64:$src)>;
def : Pat<(v4i16 (bitconvert (v2i32 VR64:$src))), (v4i16 VR64:$src)>;
+def : Pat<(v4i16 (bitconvert (v2f32 VR64:$src))), (v4i16 VR64:$src)>;
def : Pat<(v4i16 (bitconvert (v8i8 VR64:$src))), (v4i16 VR64:$src)>;
def : Pat<(v2i32 (bitconvert (v1i64 VR64:$src))), (v2i32 VR64:$src)>;
+def : Pat<(v2i32 (bitconvert (v2f32 VR64:$src))), (v2i32 VR64:$src)>;
def : Pat<(v2i32 (bitconvert (v4i16 VR64:$src))), (v2i32 VR64:$src)>;
def : Pat<(v2i32 (bitconvert (v8i8 VR64:$src))), (v2i32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v1i64 VR64:$src))), (v2f32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v2i32 VR64:$src))), (v2f32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v4i16 VR64:$src))), (v2f32 VR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v8i8 VR64:$src))), (v2f32 VR64:$src)>;
def : Pat<(v1i64 (bitconvert (v2i32 VR64:$src))), (v1i64 VR64:$src)>;
+def : Pat<(v1i64 (bitconvert (v2f32 VR64:$src))), (v1i64 VR64:$src)>;
def : Pat<(v1i64 (bitconvert (v4i16 VR64:$src))), (v1i64 VR64:$src)>;
def : Pat<(v1i64 (bitconvert (v8i8 VR64:$src))), (v1i64 VR64:$src)>;
(MMX_MOVD64to64rr GR64:$src)>;
def : Pat<(v2i32 (bitconvert (i64 GR64:$src))),
(MMX_MOVD64to64rr GR64:$src)>;
+def : Pat<(v2f32 (bitconvert (i64 GR64:$src))),
+ (MMX_MOVD64to64rr GR64:$src)>;
def : Pat<(v4i16 (bitconvert (i64 GR64:$src))),
(MMX_MOVD64to64rr GR64:$src)>;
def : Pat<(v8i8 (bitconvert (i64 GR64:$src))),
(MMX_MOVD64to64rr GR64:$src)>;
-
-def MMX_X86s2vec : SDNode<"X86ISD::S2VEC", SDTypeProfile<1, 1, []>, []>;
+def : Pat<(i64 (bitconvert (v1i64 VR64:$src))),
+ (MMX_MOVD64from64rr VR64:$src)>;
+def : Pat<(i64 (bitconvert (v2i32 VR64:$src))),
+ (MMX_MOVD64from64rr VR64:$src)>;
+def : Pat<(i64 (bitconvert (v2f32 VR64:$src))),
+ (MMX_MOVD64from64rr VR64:$src)>;
+def : Pat<(i64 (bitconvert (v4i16 VR64:$src))),
+ (MMX_MOVD64from64rr VR64:$src)>;
+def : Pat<(i64 (bitconvert (v8i8 VR64:$src))),
+ (MMX_MOVD64from64rr VR64:$src)>;
// Move scalar to XMM zero-extended
// movd to XMM register zero-extends
let AddedComplexity = 15 in {
- def : Pat<(v8i8 (vector_shuffle immAllZerosV,
- (v8i8 (MMX_X86s2vec GR32:$src)), MMX_MOVL_shuffle_mask)),
- (MMX_MOVZDI2PDIrr GR32:$src)>;
- def : Pat<(v4i16 (vector_shuffle immAllZerosV,
- (v4i16 (MMX_X86s2vec GR32:$src)), MMX_MOVL_shuffle_mask)),
- (MMX_MOVZDI2PDIrr GR32:$src)>;
- def : Pat<(v2i32 (vector_shuffle immAllZerosV,
- (v2i32 (MMX_X86s2vec GR32:$src)), MMX_MOVL_shuffle_mask)),
- (MMX_MOVZDI2PDIrr GR32:$src)>;
+ def : Pat<(v8i8 (X86vzmovl (bc_v8i8 (v2i32 (scalar_to_vector GR32:$src))))),
+ (MMX_MOVZDI2PDIrr GR32:$src)>;
+ def : Pat<(v4i16 (X86vzmovl (bc_v4i16 (v2i32 (scalar_to_vector GR32:$src))))),
+ (MMX_MOVZDI2PDIrr GR32:$src)>;
}
-// Scalar to v2i32 / v4i16 / v8i8. The source may be a GR32, but only the lower
+let AddedComplexity = 20 in {
+ def : Pat<(v8i8 (X86vzmovl (bc_v8i8 (load_mmx addr:$src)))),
+ (MMX_MOVZDI2PDIrm addr:$src)>;
+ def : Pat<(v4i16 (X86vzmovl (bc_v4i16 (load_mmx addr:$src)))),
+ (MMX_MOVZDI2PDIrm addr:$src)>;
+ def : Pat<(v2i32 (X86vzmovl (bc_v2i32 (load_mmx addr:$src)))),
+ (MMX_MOVZDI2PDIrm addr:$src)>;
+}
+
+// Clear top half.
+let AddedComplexity = 15 in {
+ def : Pat<(v8i8 (X86vzmovl VR64:$src)),
+ (MMX_PUNPCKLDQrr VR64:$src, (MMX_V_SET0))>;
+ def : Pat<(v4i16 (X86vzmovl VR64:$src)),
+ (MMX_PUNPCKLDQrr VR64:$src, (MMX_V_SET0))>;
+ def : Pat<(v2i32 (X86vzmovl VR64:$src)),
+ (MMX_PUNPCKLDQrr VR64:$src, (MMX_V_SET0))>;
+}
+
+// Scalar to v4i16 / v8i8. The source may be a GR32, but only the lower
// 8 or 16-bits matter.
-def : Pat<(v8i8 (MMX_X86s2vec GR32:$src)), (MMX_MOVD64rr GR32:$src)>;
-def : Pat<(v4i16 (MMX_X86s2vec GR32:$src)), (MMX_MOVD64rr GR32:$src)>;
-def : Pat<(v2i32 (MMX_X86s2vec GR32:$src)), (MMX_MOVD64rr GR32:$src)>;
+def : Pat<(bc_v8i8 (v2i32 (scalar_to_vector GR32:$src))),
+ (MMX_MOVD64rr GR32:$src)>;
+def : Pat<(bc_v4i16 (v2i32 (scalar_to_vector GR32:$src))),
+ (MMX_MOVD64rr GR32:$src)>;
// Patterns to perform canonical versions of vector shuffling.
let AddedComplexity = 10 in {
def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v2i32 immAllOnesV))),
VR64:$src2)),
(MMX_PANDNrr VR64:$src1, VR64:$src2)>;
-def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v4i16 immAllOnesV))),
+def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v4i16 immAllOnesV_bc))),
VR64:$src2)),
(MMX_PANDNrr VR64:$src1, VR64:$src2)>;
-def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v8i8 immAllOnesV))),
+def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v8i8 immAllOnesV_bc))),
VR64:$src2)),
(MMX_PANDNrr VR64:$src1, VR64:$src2)>;
def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v2i32 immAllOnesV))),
(load addr:$src2))),
(MMX_PANDNrm VR64:$src1, addr:$src2)>;
-def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v4i16 immAllOnesV))),
+def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v4i16 immAllOnesV_bc))),
(load addr:$src2))),
(MMX_PANDNrm VR64:$src1, addr:$src2)>;
-def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v8i8 immAllOnesV))),
+def : Pat<(v1i64 (and (xor VR64:$src1, (bc_v1i64 (v8i8 immAllOnesV_bc))),
(load addr:$src2))),
(MMX_PANDNrm VR64:$src1, addr:$src2)>;
+
+// Move MMX to lower 64-bit of XMM
+def : Pat<(v2i64 (scalar_to_vector (i64 (bitconvert VR64:$src)))),
+ (v2i64 (MMX_MOVQ2DQrr VR64:$src))>;
+
+// Move lower 64-bit of XMM to MMX.
+def : Pat<(v2i32 (bitconvert (i64 (vector_extract (v2i64 VR128:$src),
+ (iPTR 0))))),
+ (v2i32 (MMX_MOVDQ2Qrr VR128:$src))>;
+def : Pat<(v4i16 (bitconvert (i64 (vector_extract (v2i64 VR128:$src),
+ (iPTR 0))))),
+ (v4i16 (MMX_MOVDQ2Qrr VR128:$src))>;
+def : Pat<(v8i8 (bitconvert (i64 (vector_extract (v2i64 VR128:$src),
+ (iPTR 0))))),
+ (v8i8 (MMX_MOVDQ2Qrr VR128:$src))>;
+
projects / oota-llvm.git / blobdiff