summaryrefslogtreecommitdiffstats
path: root/lib/Target/X86/X86InstrSSE.td
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Target/X86/X86InstrSSE.td')
-rw-r--r--lib/Target/X86/X86InstrSSE.td560
1 files changed, 396 insertions, 164 deletions
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 5d6ef36..96fc932 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -1,10 +1,10 @@
//====- X86InstrSSE.td - Describe the X86 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 the X86 SSE instruction set, defining the instructions,
@@ -36,22 +36,22 @@ def X86frcp : SDNode<"X86ISD::FRCP", SDTFPUnaryOp>;
def X86fsrl : SDNode<"X86ISD::FSRL", SDTX86FPShiftOp>;
def X86comi : SDNode<"X86ISD::COMI", SDTX86CmpTest>;
def X86ucomi : SDNode<"X86ISD::UCOMI", SDTX86CmpTest>;
-def X86pshufb : SDNode<"X86ISD::PSHUFB",
+def X86pshufb : SDNode<"X86ISD::PSHUFB",
SDTypeProfile<1, 2, [SDTCisVT<0, v16i8>, SDTCisSameAs<0,1>,
SDTCisSameAs<0,2>]>>;
def X86pextrb : SDNode<"X86ISD::PEXTRB",
SDTypeProfile<1, 2, [SDTCisVT<0, i32>, SDTCisPtrTy<2>]>>;
def X86pextrw : SDNode<"X86ISD::PEXTRW",
SDTypeProfile<1, 2, [SDTCisVT<0, i32>, SDTCisPtrTy<2>]>>;
-def X86pinsrb : SDNode<"X86ISD::PINSRB",
+def X86pinsrb : SDNode<"X86ISD::PINSRB",
SDTypeProfile<1, 3, [SDTCisVT<0, v16i8>, SDTCisSameAs<0,1>,
SDTCisVT<2, i32>, SDTCisPtrTy<3>]>>;
-def X86pinsrw : SDNode<"X86ISD::PINSRW",
+def X86pinsrw : SDNode<"X86ISD::PINSRW",
SDTypeProfile<1, 3, [SDTCisVT<0, v8i16>, SDTCisSameAs<0,1>,
SDTCisVT<2, i32>, SDTCisPtrTy<3>]>>;
-def X86insrtps : SDNode<"X86ISD::INSERTPS",
+def X86insrtps : SDNode<"X86ISD::INSERTPS",
SDTypeProfile<1, 3, [SDTCisVT<0, v4f32>, SDTCisSameAs<0,1>,
- SDTCisVT<2, f32>, SDTCisPtrTy<3>]>>;
+ SDTCisVT<2, v4f32>, SDTCisPtrTy<3>]>>;
def X86vzmovl : SDNode<"X86ISD::VZEXT_MOVL",
SDTypeProfile<1, 1, [SDTCisSameAs<0,1>]>>;
def X86vzload : SDNode<"X86ISD::VZEXT_LOAD", SDTLoad,
@@ -69,6 +69,10 @@ def X86pcmpgtw : SDNode<"X86ISD::PCMPGTW", SDTIntBinOp>;
def X86pcmpgtd : SDNode<"X86ISD::PCMPGTD", SDTIntBinOp>;
def X86pcmpgtq : SDNode<"X86ISD::PCMPGTQ", SDTIntBinOp>;
+def SDTX86CmpPTest : SDTypeProfile<0, 2, [SDTCisVT<0, v4f32>,
+ SDTCisVT<1, v4f32>]>;
+def X86ptest : SDNode<"X86ISD::PTEST", SDTX86CmpPTest>;
+
//===----------------------------------------------------------------------===//
// SSE Complex Patterns
//===----------------------------------------------------------------------===//
@@ -83,11 +87,13 @@ def sse_load_f64 : ComplexPattern<v2f64, 5, "SelectScalarSSELoad", [],
def ssmem : Operand<v4f32> {
let PrintMethod = "printf32mem";
- let MIOperandInfo = (ops ptr_rc, i8imm, ptr_rc, i32imm, i8imm);
+ let MIOperandInfo = (ops ptr_rc, i8imm, ptr_rc_nosp, i32imm, i8imm);
+ let ParserMatchClass = X86MemAsmOperand;
}
def sdmem : Operand<v2f64> {
let PrintMethod = "printf64mem";
- let MIOperandInfo = (ops ptr_rc, i8imm, ptr_rc, i32imm, i8imm);
+ let MIOperandInfo = (ops ptr_rc, i8imm, ptr_rc_nosp, i32imm, i8imm);
+ let ParserMatchClass = X86MemAsmOperand;
}
//===----------------------------------------------------------------------===//
@@ -179,13 +185,13 @@ def SHUFFLE_get_shuf_imm : SDNodeXForm<vector_shuffle, [{
return getI8Imm(X86::getShuffleSHUFImmediate(N));
}]>;
-// SHUFFLE_get_pshufhw_imm xform function: convert vector_shuffle mask to
+// SHUFFLE_get_pshufhw_imm xform function: convert vector_shuffle mask to
// PSHUFHW imm.
def SHUFFLE_get_pshufhw_imm : SDNodeXForm<vector_shuffle, [{
return getI8Imm(X86::getShufflePSHUFHWImmediate(N));
}]>;
-// SHUFFLE_get_pshuflw_imm xform function: convert vector_shuffle mask to
+// SHUFFLE_get_pshuflw_imm xform function: convert vector_shuffle mask to
// PSHUFLW imm.
def SHUFFLE_get_pshuflw_imm : SDNodeXForm<vector_shuffle, [{
return getI8Imm(X86::getShufflePSHUFLWImmediate(N));
@@ -360,25 +366,25 @@ def Int_CVTPS2PIrr : PSI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
[(set VR64:$dst, (int_x86_sse_cvtps2pi VR128:$src))]>;
def Int_CVTPS2PIrm : PSI<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f64mem:$src),
"cvtps2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvtps2pi
+ [(set VR64:$dst, (int_x86_sse_cvtps2pi
(load addr:$src)))]>;
def Int_CVTTPS2PIrr: PSI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvttps2pi\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (int_x86_sse_cvttps2pi VR128:$src))]>;
def Int_CVTTPS2PIrm: PSI<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f64mem:$src),
"cvttps2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvttps2pi
+ [(set VR64:$dst, (int_x86_sse_cvttps2pi
(load addr:$src)))]>;
let Constraints = "$src1 = $dst" in {
- def Int_CVTPI2PSrr : PSI<0x2A, MRMSrcReg,
+ def Int_CVTPI2PSrr : PSI<0x2A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR64:$src2),
"cvtpi2ps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst, (int_x86_sse_cvtpi2ps VR128:$src1,
VR64:$src2))]>;
- def Int_CVTPI2PSrm : PSI<0x2A, MRMSrcMem,
+ def Int_CVTPI2PSrm : PSI<0x2A, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i64mem:$src2),
"cvtpi2ps\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst, (int_x86_sse_cvtpi2ps VR128:$src1,
+ [(set VR128:$dst, (int_x86_sse_cvtpi2ps VR128:$src1,
(load addr:$src2)))]>;
}
@@ -407,11 +413,11 @@ let Constraints = "$src1 = $dst" in {
// Comparison instructions
let Constraints = "$src1 = $dst", neverHasSideEffects = 1 in {
- def CMPSSrr : SSIi8<0xC2, MRMSrcReg,
+ def CMPSSrr : SSIi8<0xC2, MRMSrcReg,
(outs FR32:$dst), (ins FR32:$src1, FR32:$src, SSECC:$cc),
"cmp${cc}ss\t{$src, $dst|$dst, $src}", []>;
let mayLoad = 1 in
- def CMPSSrm : SSIi8<0xC2, MRMSrcMem,
+ def CMPSSrm : SSIi8<0xC2, MRMSrcMem,
(outs FR32:$dst), (ins FR32:$src1, f32mem:$src, SSECC:$cc),
"cmp${cc}ss\t{$src, $dst|$dst, $src}", []>;
}
@@ -428,13 +434,15 @@ def UCOMISSrm: PSI<0x2E, MRMSrcMem, (outs), (ins FR32:$src1, f32mem:$src2),
// Aliases to match intrinsics which expect XMM operand(s).
let Constraints = "$src1 = $dst" in {
- def Int_CMPSSrr : SSIi8<0xC2, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc),
+ def Int_CMPSSrr : SSIi8<0xC2, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src,
+ SSECC:$cc),
"cmp${cc}ss\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cmp_ss VR128:$src1,
- VR128:$src, imm:$cc))]>;
- def Int_CMPSSrm : SSIi8<0xC2, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f32mem:$src, SSECC:$cc),
+ VR128:$src, imm:$cc))]>;
+ def Int_CMPSSrm : SSIi8<0xC2, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, f32mem:$src,
+ SSECC:$cc),
"cmp${cc}ss\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cmp_ss VR128:$src1,
(load addr:$src), imm:$cc))]>;
@@ -460,18 +468,19 @@ def Int_COMISSrm: PSI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
(implicit EFLAGS)]>;
} // Defs = [EFLAGS]
-// Aliases of packed SSE1 instructions for scalar use. These all have names that
-// start with 'Fs'.
+// Aliases of packed SSE1 instructions for scalar use. These all have names
+// that start with 'Fs'.
// Alias instructions that map fld0 to pxor for sse.
-let isReMaterializable = 1, isAsCheapAsAMove = 1 in
+let isReMaterializable = 1, isAsCheapAsAMove = 1, isCodeGenOnly = 1,
+ canFoldAsLoad = 1 in
def FsFLD0SS : I<0xEF, MRMInitReg, (outs FR32:$dst), (ins),
"pxor\t$dst, $dst", [(set FR32:$dst, fp32imm0)]>,
Requires<[HasSSE1]>, TB, OpSize;
// Alias instruction to do FR32 reg-to-reg copy using movaps. Upper bits are
// disregarded.
-let neverHasSideEffects = 1 in
+let neverHasSideEffects = 1 in
def FsMOVAPSrr : PSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
"movaps\t{$src, $dst|$dst, $src}", []>;
@@ -552,7 +561,7 @@ multiclass basic_sse1_fp_binop_rm<bits<8> opc, string OpcodeStr,
(ins FR32:$src1, f32mem:$src2),
!strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
[(set FR32:$dst, (OpNode FR32:$src1, (load addr:$src2)))]>;
-
+
// Vector operation, reg+reg.
def PSrr : PSI<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
@@ -616,7 +625,7 @@ multiclass sse1_fp_binop_rm<bits<8> opc, string OpcodeStr,
(ins FR32:$src1, f32mem:$src2),
!strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
[(set FR32:$dst, (OpNode FR32:$src1, (load addr:$src2)))]>;
-
+
// Vector operation, reg+reg.
def PSrr : PSI<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
@@ -671,7 +680,7 @@ defm MIN : sse1_fp_binop_rm<0x5D, "min", X86fmin,
// SSE packed FP Instructions
// Move Instructions
-let neverHasSideEffects = 1 in
+let neverHasSideEffects = 1 in
def MOVAPSrr : PSI<0x28, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movaps\t{$src, $dst|$dst, $src}", []>;
let canFoldAsLoad = 1, isReMaterializable = 1, mayHaveSideEffects = 1 in
@@ -708,13 +717,13 @@ let Constraints = "$src1 = $dst" in {
def MOVLPSrm : PSI<0x12, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
"movlps\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
+ [(set VR128:$dst,
(movlp VR128:$src1,
(bc_v4f32 (v2f64 (scalar_to_vector (loadf64 addr:$src2))))))]>;
def MOVHPSrm : PSI<0x16, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
"movhps\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
+ [(set VR128:$dst,
(movhp VR128:$src1,
(bc_v4f32 (v2f64 (scalar_to_vector (loadf64 addr:$src2))))))]>;
} // AddedComplexity
@@ -789,7 +798,7 @@ multiclass sse1_fp_unop_rm<bits<8> opc, string OpcodeStr,
def SSm : SSI<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
[(set FR32:$dst, (OpNode (load addr:$src)))]>;
-
+
// Vector operation, reg.
def PSr : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
@@ -890,12 +899,12 @@ let Constraints = "$src1 = $dst" in {
}
let Constraints = "$src1 = $dst" in {
- def CMPPSrri : PSIi8<0xC2, MRMSrcReg,
+ def CMPPSrri : PSIi8<0xC2, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc),
"cmp${cc}ps\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cmp_ps VR128:$src1,
VR128:$src, imm:$cc))]>;
- def CMPPSrmi : PSIi8<0xC2, MRMSrcMem,
+ def CMPPSrmi : PSIi8<0xC2, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src, SSECC:$cc),
"cmp${cc}ps\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cmp_ps VR128:$src1,
@@ -909,13 +918,13 @@ def : Pat<(v4i32 (X86cmpps (v4f32 VR128:$src1), (memop addr:$src2), imm:$cc)),
// Shuffle and unpack instructions
let Constraints = "$src1 = $dst" in {
let isConvertibleToThreeAddress = 1 in // Convert to pshufd
- def SHUFPSrri : PSIi8<0xC6, MRMSrcReg,
+ def SHUFPSrri : PSIi8<0xC6, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1,
VR128:$src2, i8imm:$src3),
"shufps\t{$src3, $src2, $dst|$dst, $src2, $src3}",
[(set VR128:$dst,
(v4f32 (shufp:$src3 VR128:$src1, VR128:$src2)))]>;
- def SHUFPSrmi : PSIi8<0xC6, MRMSrcMem,
+ def SHUFPSrmi : PSIi8<0xC6, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1,
f128mem:$src2, i8imm:$src3),
"shufps\t{$src3, $src2, $dst|$dst, $src2, $src3}",
@@ -924,24 +933,24 @@ let Constraints = "$src1 = $dst" in {
VR128:$src1, (memopv4f32 addr:$src2))))]>;
let AddedComplexity = 10 in {
- def UNPCKHPSrr : PSI<0x15, MRMSrcReg,
+ def UNPCKHPSrr : PSI<0x15, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"unpckhps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v4f32 (unpckh VR128:$src1, VR128:$src2)))]>;
- def UNPCKHPSrm : PSI<0x15, MRMSrcMem,
+ def UNPCKHPSrm : PSI<0x15, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
"unpckhps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v4f32 (unpckh VR128:$src1,
(memopv4f32 addr:$src2))))]>;
- def UNPCKLPSrr : PSI<0x14, MRMSrcReg,
+ def UNPCKLPSrr : PSI<0x14, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"unpcklps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v4f32 (unpckl VR128:$src1, VR128:$src2)))]>;
- def UNPCKLPSrm : PSI<0x14, MRMSrcMem,
+ def UNPCKLPSrm : PSI<0x14, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
"unpcklps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
@@ -984,7 +993,8 @@ def STMXCSR : PSI<0xAE, MRM3m, (outs), (ins i32mem:$dst),
// Alias instructions that map zero vector to pxor / xorp* for sse.
// We set canFoldAsLoad because this can be converted to a constant-pool
// load of an all-zeros value if folding it would be beneficial.
-let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1 in
+let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
+ isCodeGenOnly = 1 in
def V_SET0 : PSI<0x57, MRMInitReg, (outs VR128:$dst), (ins),
"xorps\t$dst, $dst",
[(set VR128:$dst, (v4i32 immAllZerosV))]>;
@@ -1046,14 +1056,14 @@ let AddedComplexity = 20 in
def MOVZSS2PSrm : SSI<0x10, MRMSrcMem, (outs VR128:$dst), (ins f32mem:$src),
"movss\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v4f32 (X86vzmovl (v4f32 (scalar_to_vector
- (loadf32 addr:$src))))))]>;
+ (loadf32 addr:$src))))))]>;
def : Pat<(v4f32 (X86vzmovl (loadv4f32 addr:$src))),
(MOVZSS2PSrm addr:$src)>;
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// SSE2 Instructions
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// Move Instructions
let neverHasSideEffects = 1 in
@@ -1077,7 +1087,7 @@ def CVTTSD2SIrm : SDI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f64mem:$src),
def CVTSD2SSrr : SDI<0x5A, MRMSrcReg, (outs FR32:$dst), (ins FR64:$src),
"cvtsd2ss\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (fround FR64:$src))]>;
-def CVTSD2SSrm : SDI<0x5A, MRMSrcMem, (outs FR32:$dst), (ins f64mem:$src),
+def CVTSD2SSrm : SDI<0x5A, MRMSrcMem, (outs FR32:$dst), (ins f64mem:$src),
"cvtsd2ss\t{$src, $dst|$dst, $src}",
[(set FR32:$dst, (fround (loadf64 addr:$src)))]>;
def CVTSI2SDrr : SDI<0x2A, MRMSrcReg, (outs FR64:$dst), (ins GR32:$src),
@@ -1087,6 +1097,27 @@ def CVTSI2SDrm : SDI<0x2A, MRMSrcMem, (outs FR64:$dst), (ins i32mem:$src),
"cvtsi2sd\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (sint_to_fp (loadi32 addr:$src)))]>;
+def CVTPD2DQrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
+def CVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
+def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtps2dq\t{$src, $dst|$dst, $src}", []>;
+def CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtps2dq\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PSrr : PSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2ps\t{$src, $dst|$dst, $src}", []>;
+def CVTDQ2PSrm : PSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtdq2ps\t{$src, $dst|$dst, $src}", []>;
+def COMISDrr: PDI<0x2F, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
+ "comisd\t{$src2, $src1|$src1, $src2}", []>;
+def COMISDrm: PDI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
+ "comisd\t{$src2, $src1|$src1, $src2}", []>;
+
// SSE2 instructions with XS prefix
def CVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst), (ins FR32:$src),
"cvtss2sd\t{$src, $dst|$dst, $src}",
@@ -1112,21 +1143,21 @@ def Int_CVTPD2PIrr : PDI<0x2D, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
[(set VR64:$dst, (int_x86_sse_cvtpd2pi VR128:$src))]>;
def Int_CVTPD2PIrm : PDI<0x2D, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
"cvtpd2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvtpd2pi
+ [(set VR64:$dst, (int_x86_sse_cvtpd2pi
(memop addr:$src)))]>;
def Int_CVTTPD2PIrr: PDI<0x2C, MRMSrcReg, (outs VR64:$dst), (ins VR128:$src),
"cvttpd2pi\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (int_x86_sse_cvttpd2pi VR128:$src))]>;
def Int_CVTTPD2PIrm: PDI<0x2C, MRMSrcMem, (outs VR64:$dst), (ins f128mem:$src),
"cvttpd2pi\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst, (int_x86_sse_cvttpd2pi
+ [(set VR64:$dst, (int_x86_sse_cvttpd2pi
(memop addr:$src)))]>;
def Int_CVTPI2PDrr : PDI<0x2A, MRMSrcReg, (outs VR128:$dst), (ins VR64:$src),
"cvtpi2pd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse_cvtpi2pd VR64:$src))]>;
def Int_CVTPI2PDrm : PDI<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"cvtpi2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse_cvtpi2pd
+ [(set VR128:$dst, (int_x86_sse_cvtpi2pd
(load addr:$src)))]>;
// Aliases for intrinsics
@@ -1141,11 +1172,11 @@ def Int_CVTTSD2SIrm : SDI<0x2C, MRMSrcMem, (outs GR32:$dst), (ins f128mem:$src),
// Comparison instructions
let Constraints = "$src1 = $dst", neverHasSideEffects = 1 in {
- def CMPSDrr : SDIi8<0xC2, MRMSrcReg,
+ def CMPSDrr : SDIi8<0xC2, MRMSrcReg,
(outs FR64:$dst), (ins FR64:$src1, FR64:$src, SSECC:$cc),
"cmp${cc}sd\t{$src, $dst|$dst, $src}", []>;
let mayLoad = 1 in
- def CMPSDrm : SDIi8<0xC2, MRMSrcMem,
+ def CMPSDrm : SDIi8<0xC2, MRMSrcMem,
(outs FR64:$dst), (ins FR64:$src1, f64mem:$src, SSECC:$cc),
"cmp${cc}sd\t{$src, $dst|$dst, $src}", []>;
}
@@ -1162,13 +1193,15 @@ def UCOMISDrm: PDI<0x2E, MRMSrcMem, (outs), (ins FR64:$src1, f64mem:$src2),
// Aliases to match intrinsics which expect XMM operand(s).
let Constraints = "$src1 = $dst" in {
- def Int_CMPSDrr : SDIi8<0xC2, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc),
+ def Int_CMPSDrr : SDIi8<0xC2, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src,
+ SSECC:$cc),
"cmp${cc}sd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
VR128:$src, imm:$cc))]>;
- def Int_CMPSDrm : SDIi8<0xC2, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f64mem:$src, SSECC:$cc),
+ def Int_CMPSDrm : SDIi8<0xC2, MRMSrcMem,
+ (outs VR128:$dst), (ins VR128:$src1, f64mem:$src,
+ SSECC:$cc),
"cmp${cc}sd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cmp_sd VR128:$src1,
(load addr:$src), imm:$cc))]>;
@@ -1194,11 +1227,12 @@ def Int_COMISDrm: PDI<0x2F, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
(implicit EFLAGS)]>;
} // Defs = [EFLAGS]
-// Aliases of packed SSE2 instructions for scalar use. These all have names that
-// start with 'Fs'.
+// Aliases of packed SSE2 instructions for scalar use. These all have names
+// that start with 'Fs'.
// Alias instructions that map fld0 to pxor for sse.
-let isReMaterializable = 1, isAsCheapAsAMove = 1 in
+let isReMaterializable = 1, isAsCheapAsAMove = 1, isCodeGenOnly = 1,
+ canFoldAsLoad = 1 in
def FsFLD0SD : I<0xEF, MRMInitReg, (outs FR64:$dst), (ins),
"pxor\t$dst, $dst", [(set FR64:$dst, fpimm0)]>,
Requires<[HasSSE2]>, TB, OpSize;
@@ -1286,7 +1320,7 @@ multiclass basic_sse2_fp_binop_rm<bits<8> opc, string OpcodeStr,
(ins FR64:$src1, f64mem:$src2),
!strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
[(set FR64:$dst, (OpNode FR64:$src1, (load addr:$src2)))]>;
-
+
// Vector operation, reg+reg.
def PDrr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
@@ -1350,7 +1384,7 @@ multiclass sse2_fp_binop_rm<bits<8> opc, string OpcodeStr,
(ins FR64:$src1, f64mem:$src2),
!strconcat(OpcodeStr, "sd\t{$src2, $dst|$dst, $src2}"),
[(set FR64:$dst, (OpNode FR64:$src1, (load addr:$src2)))]>;
-
+
// Vector operation, reg+reg.
def PDrr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
@@ -1402,7 +1436,7 @@ defm MAX : sse2_fp_binop_rm<0x5F, "max", X86fmax,
defm MIN : sse2_fp_binop_rm<0x5D, "min", X86fmin,
int_x86_sse2_min_sd, int_x86_sse2_min_pd>;
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// SSE packed FP Instructions
// Move Instructions
@@ -1442,13 +1476,13 @@ let Constraints = "$src1 = $dst" in {
def MOVLPDrm : PDI<0x12, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
"movlpd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
+ [(set VR128:$dst,
(v2f64 (movlp VR128:$src1,
(scalar_to_vector (loadf64 addr:$src2)))))]>;
def MOVHPDrm : PDI<0x16, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
"movhpd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
+ [(set VR128:$dst,
(v2f64 (movhp VR128:$src1,
(scalar_to_vector (loadf64 addr:$src2)))))]>;
} // AddedComplexity
@@ -1564,7 +1598,7 @@ def Int_CVTSD2SSrr: SDI<0x5A, MRMSrcReg,
[(set VR128:$dst, (int_x86_sse2_cvtsd2ss VR128:$src1,
VR128:$src2))]>;
def Int_CVTSD2SSrm: SDI<0x5A, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
+ (outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
"cvtsd2ss\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst, (int_x86_sse2_cvtsd2ss VR128:$src1,
(load addr:$src2)))]>;
@@ -1612,7 +1646,7 @@ multiclass sse2_fp_unop_rm<bits<8> opc, string OpcodeStr,
def SDm : SDI<opc, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
[(set FR64:$dst, (OpNode (load addr:$src)))]>;
-
+
// Vector operation, reg.
def PDr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
@@ -1712,12 +1746,12 @@ let Constraints = "$src1 = $dst" in {
}
let Constraints = "$src1 = $dst" in {
- def CMPPDrri : PDIi8<0xC2, MRMSrcReg,
+ def CMPPDrri : PDIi8<0xC2, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src, SSECC:$cc),
"cmp${cc}pd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cmp_pd VR128:$src1,
VR128:$src, imm:$cc))]>;
- def CMPPDrmi : PDIi8<0xC2, MRMSrcMem,
+ def CMPPDrmi : PDIi8<0xC2, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src, SSECC:$cc),
"cmp${cc}pd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cmp_pd VR128:$src1,
@@ -1730,12 +1764,12 @@ def : Pat<(v2i64 (X86cmppd (v2f64 VR128:$src1), (memop addr:$src2), imm:$cc)),
// Shuffle and unpack instructions
let Constraints = "$src1 = $dst" in {
- def SHUFPDrri : PDIi8<0xC6, MRMSrcReg,
+ def SHUFPDrri : PDIi8<0xC6, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2, i8imm:$src3),
"shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
[(set VR128:$dst,
(v2f64 (shufp:$src3 VR128:$src1, VR128:$src2)))]>;
- def SHUFPDrmi : PDIi8<0xC6, MRMSrcMem,
+ def SHUFPDrmi : PDIi8<0xC6, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1,
f128mem:$src2, i8imm:$src3),
"shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
@@ -1744,24 +1778,24 @@ let Constraints = "$src1 = $dst" in {
VR128:$src1, (memopv2f64 addr:$src2))))]>;
let AddedComplexity = 10 in {
- def UNPCKHPDrr : PDI<0x15, MRMSrcReg,
+ def UNPCKHPDrr : PDI<0x15, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"unpckhpd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v2f64 (unpckh VR128:$src1, VR128:$src2)))]>;
- def UNPCKHPDrm : PDI<0x15, MRMSrcMem,
+ def UNPCKHPDrm : PDI<0x15, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
"unpckhpd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v2f64 (unpckh VR128:$src1,
(memopv2f64 addr:$src2))))]>;
- def UNPCKLPDrr : PDI<0x14, MRMSrcReg,
+ def UNPCKLPDrr : PDI<0x14, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"unpcklpd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v2f64 (unpckl VR128:$src1, VR128:$src2)))]>;
- def UNPCKLPDrm : PDI<0x14, MRMSrcMem,
+ def UNPCKLPDrm : PDI<0x14, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f128mem:$src2),
"unpcklpd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
@@ -1770,7 +1804,7 @@ let Constraints = "$src1 = $dst" in {
} // Constraints = "$src1 = $dst"
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// SSE integer instructions
// Move Instructions
@@ -1825,14 +1859,17 @@ multiclass PDI_binop_rm_int<bits<8> opc, string OpcodeStr, Intrinsic IntId,
multiclass PDI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
string OpcodeStr,
Intrinsic IntId, Intrinsic IntId2> {
- def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
+ def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1,
+ VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst, (IntId VR128:$src1, VR128:$src2))]>;
- def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
+ def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1,
+ i128mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst, (IntId VR128:$src1,
- (bitconvert (memopv2i64 addr:$src2))))]>;
- def ri : PDIi8<opc2, ImmForm, (outs VR128:$dst), (ins VR128:$src1, i32i8imm:$src2),
+ (bitconvert (memopv2i64 addr:$src2))))]>;
+ def ri : PDIi8<opc2, ImmForm, (outs VR128:$dst), (ins VR128:$src1,
+ i32i8imm:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst, (IntId2 VR128:$src1, (i32 imm:$src2)))]>;
}
@@ -1840,15 +1877,17 @@ multiclass PDI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
/// PDI_binop_rm - Simple SSE2 binary operator.
multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType OpVT, bit Commutable = 0> {
- def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
+ def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1,
+ VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst, (OpVT (OpNode VR128:$src1, VR128:$src2)))]> {
let isCommutable = Commutable;
}
- def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
+ def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1,
+ i128mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst, (OpVT (OpNode VR128:$src1,
- (bitconvert (memopv2i64 addr:$src2)))))]>;
+ (bitconvert (memopv2i64 addr:$src2)))))]>;
}
/// PDI_binop_rm_v2i64 - Simple SSE2 binary operator whose type is v2i64.
@@ -1858,14 +1897,17 @@ multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
///
multiclass PDI_binop_rm_v2i64<bits<8> opc, string OpcodeStr, SDNode OpNode,
bit Commutable = 0> {
- def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
+ def rr : PDI<opc, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst, (v2i64 (OpNode VR128:$src1, VR128:$src2)))]> {
let isCommutable = Commutable;
}
- def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
+ def rm : PDI<opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, i128mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (OpNode VR128:$src1,(memopv2i64 addr:$src2)))]>;
+ [(set VR128:$dst, (OpNode VR128:$src1,
+ (memopv2i64 addr:$src2)))]>;
}
} // Constraints = "$src1 = $dst"
@@ -2029,8 +2071,8 @@ def PSHUFDmi : PDIi8<0x70, MRMSrcMem,
(outs VR128:$dst), (ins i128mem:$src1, i8imm:$src2),
"pshufd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst, (v4i32 (pshufd:$src2
- (bc_v4i32(memopv2i64 addr:$src1)),
- (undef))))]>;
+ (bc_v4i32(memopv2i64 addr:$src1)),
+ (undef))))]>;
// SSE2 with ImmT == Imm8 and XS prefix.
def PSHUFHWri : Ii8<0x70, MRMSrcReg,
@@ -2043,8 +2085,8 @@ def PSHUFHWmi : Ii8<0x70, MRMSrcMem,
(outs VR128:$dst), (ins i128mem:$src1, i8imm:$src2),
"pshufhw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst, (v8i16 (pshufhw:$src2
- (bc_v8i16 (memopv2i64 addr:$src1)),
- (undef))))]>,
+ (bc_v8i16 (memopv2i64 addr:$src1)),
+ (undef))))]>,
XS, Requires<[HasSSE2]>;
// SSE2 with ImmT == Imm8 and XD prefix.
@@ -2064,90 +2106,90 @@ def PSHUFLWmi : Ii8<0x70, MRMSrcMem,
let Constraints = "$src1 = $dst" in {
- def PUNPCKLBWrr : PDI<0x60, MRMSrcReg,
+ def PUNPCKLBWrr : PDI<0x60, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpcklbw\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v16i8 (unpckl VR128:$src1, VR128:$src2)))]>;
- def PUNPCKLBWrm : PDI<0x60, MRMSrcMem,
+ def PUNPCKLBWrm : PDI<0x60, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpcklbw\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(unpckl VR128:$src1,
(bc_v16i8 (memopv2i64 addr:$src2))))]>;
- def PUNPCKLWDrr : PDI<0x61, MRMSrcReg,
+ def PUNPCKLWDrr : PDI<0x61, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpcklwd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v8i16 (unpckl VR128:$src1, VR128:$src2)))]>;
- def PUNPCKLWDrm : PDI<0x61, MRMSrcMem,
+ def PUNPCKLWDrm : PDI<0x61, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpcklwd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(unpckl VR128:$src1,
(bc_v8i16 (memopv2i64 addr:$src2))))]>;
- def PUNPCKLDQrr : PDI<0x62, MRMSrcReg,
+ def PUNPCKLDQrr : PDI<0x62, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpckldq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v4i32 (unpckl VR128:$src1, VR128:$src2)))]>;
- def PUNPCKLDQrm : PDI<0x62, MRMSrcMem,
+ def PUNPCKLDQrm : PDI<0x62, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpckldq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(unpckl VR128:$src1,
(bc_v4i32 (memopv2i64 addr:$src2))))]>;
- def PUNPCKLQDQrr : PDI<0x6C, MRMSrcReg,
+ def PUNPCKLQDQrr : PDI<0x6C, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpcklqdq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v2i64 (unpckl VR128:$src1, VR128:$src2)))]>;
- def PUNPCKLQDQrm : PDI<0x6C, MRMSrcMem,
+ def PUNPCKLQDQrm : PDI<0x6C, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpcklqdq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v2i64 (unpckl VR128:$src1,
(memopv2i64 addr:$src2))))]>;
-
- def PUNPCKHBWrr : PDI<0x68, MRMSrcReg,
+
+ def PUNPCKHBWrr : PDI<0x68, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpckhbw\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v16i8 (unpckh VR128:$src1, VR128:$src2)))]>;
- def PUNPCKHBWrm : PDI<0x68, MRMSrcMem,
+ def PUNPCKHBWrm : PDI<0x68, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpckhbw\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst,
- (unpckh VR128:$src1,
+ [(set VR128:$dst,
+ (unpckh VR128:$src1,
(bc_v16i8 (memopv2i64 addr:$src2))))]>;
- def PUNPCKHWDrr : PDI<0x69, MRMSrcReg,
+ def PUNPCKHWDrr : PDI<0x69, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpckhwd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v8i16 (unpckh VR128:$src1, VR128:$src2)))]>;
- def PUNPCKHWDrm : PDI<0x69, MRMSrcMem,
+ def PUNPCKHWDrm : PDI<0x69, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpckhwd\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(unpckh VR128:$src1,
(bc_v8i16 (memopv2i64 addr:$src2))))]>;
- def PUNPCKHDQrr : PDI<0x6A, MRMSrcReg,
+ def PUNPCKHDQrr : PDI<0x6A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpckhdq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v4i32 (unpckh VR128:$src1, VR128:$src2)))]>;
- def PUNPCKHDQrm : PDI<0x6A, MRMSrcMem,
+ def PUNPCKHDQrm : PDI<0x6A, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpckhdq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(unpckh VR128:$src1,
(bc_v4i32 (memopv2i64 addr:$src2))))]>;
- def PUNPCKHQDQrr : PDI<0x6D, MRMSrcReg,
+ def PUNPCKHQDQrr : PDI<0x6D, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"punpckhqdq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(v2i64 (unpckh VR128:$src1, VR128:$src2)))]>;
- def PUNPCKHQDQrm : PDI<0x6D, MRMSrcMem,
+ def PUNPCKHQDQrm : PDI<0x6D, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
"punpckhqdq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
@@ -2172,7 +2214,7 @@ let Constraints = "$src1 = $dst" in {
(outs VR128:$dst), (ins VR128:$src1,
i16mem:$src2, i32i8imm:$src3),
"pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- [(set VR128:$dst,
+ [(set VR128:$dst,
(X86pinsrw VR128:$src1, (extloadi16 addr:$src2),
imm:$src3))]>;
}
@@ -2202,7 +2244,7 @@ def MOVNTDQmr : PDI<0xE7, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
[(int_x86_sse2_movnt_dq addr:$dst, VR128:$src)]>;
def MOVNTImr : I<0xC3, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src),
"movnti\t{$src, $dst|$dst, $src}",
- [(int_x86_sse2_movnt_i addr:$dst, GR32:$src)]>,
+ [(int_x86_sse2_movnt_i addr:$dst, GR32:$src)]>,
TB, Requires<[HasSSE2]>;
// Flush cache
@@ -2217,17 +2259,18 @@ def MFENCE : I<0xAE, MRM6r, (outs), (ins),
"mfence", [(int_x86_sse2_mfence)]>, TB, Requires<[HasSSE2]>;
//TODO: custom lower this so as to never even generate the noop
-def : Pat<(membarrier (i8 imm:$ll), (i8 imm:$ls), (i8 imm:$sl), (i8 imm:$ss),
+def : Pat<(membarrier (i8 imm:$ll), (i8 imm:$ls), (i8 imm:$sl), (i8 imm:$ss),
(i8 0)), (NOOP)>;
def : Pat<(membarrier (i8 0), (i8 0), (i8 0), (i8 1), (i8 1)), (SFENCE)>;
def : Pat<(membarrier (i8 1), (i8 0), (i8 0), (i8 0), (i8 1)), (LFENCE)>;
-def : Pat<(membarrier (i8 imm:$ll), (i8 imm:$ls), (i8 imm:$sl), (i8 imm:$ss),
+def : Pat<(membarrier (i8 imm:$ll), (i8 imm:$ls), (i8 imm:$sl), (i8 imm:$ss),
(i8 1)), (MFENCE)>;
// Alias instructions that map zero vector to pxor / xorp* for sse.
// We set canFoldAsLoad because this can be converted to a constant-pool
// load of an all-ones value if folding it would be beneficial.
-let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1 in
+let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
+ isCodeGenOnly = 1 in
def V_SETALLONES : PDI<0x76, MRMInitReg, (outs VR128:$dst), (ins),
"pcmpeqd\t$dst, $dst",
[(set VR128:$dst, (v4i32 immAllOnesV))]>;
@@ -2240,7 +2283,7 @@ def MOVSD2PDrr : SDI<0x10, MRMSrcReg, (outs VR128:$dst), (ins FR64:$src),
(v2f64 (scalar_to_vector FR64:$src)))]>;
def MOVSD2PDrm : SDI<0x10, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
"movsd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
+ [(set VR128:$dst,
(v2f64 (scalar_to_vector (loadf64 addr:$src))))]>;
def MOVDI2PDIrr : PDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
@@ -2399,9 +2442,9 @@ def : Pat<(v2i64 (X86vzmovl (bc_v2i64 (loadv4i32 addr:$src)))),
(MOVZPQILo2PQIrm addr:$src)>;
}
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// SSE3 Instructions
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// Move Instructions
def MOVSHDUPrr : S3SI<0x16, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
@@ -2525,9 +2568,9 @@ let AddedComplexity = 20 in
def : Pat<(v4i32 (movsldup (bc_v4i32 (memopv2i64 addr:$src)), (undef))),
(MOVSLDUPrm addr:$src)>, Requires<[HasSSE3]>;
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// SSSE3 Instructions
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
/// SS3I_unop_rm_int_8 - Simple SSSE3 unary operator whose type is v*i8.
multiclass SS3I_unop_rm_int_8<bits<8> opc, string OpcodeStr,
@@ -2801,12 +2844,13 @@ def : Pat<(X86pshufb VR128:$src, VR128:$mask),
def : Pat<(X86pshufb VR128:$src, (bc_v16i8 (memopv2i64 addr:$mask))),
(PSHUFBrm128 VR128:$src, addr:$mask)>, Requires<[HasSSSE3]>;
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
// Non-Instruction Patterns
-//===----------------------------------------------------------------------===//
+//===---------------------------------------------------------------------===//
-// extload f32 -> f64. This matches load+fextend because we have a hack in
-// the isel (PreprocessForFPConvert) that can introduce loads after dag combine.
+// extload f32 -> f64. This matches load+fextend because we have a hack in
+// the isel (PreprocessForFPConvert) that can introduce loads after dag
+// combine.
// Since these loads aren't folded into the fextend, we have to match it
// explicitly here.
let Predicates = [HasSSE2] in
@@ -2884,12 +2928,12 @@ def : Pat<(v4f32 (pshufd:$src2 VR128:$src1, (undef))),
Requires<[HasSSE2]>;
// Special unary SHUFPDrri case.
def : Pat<(v2i64 (pshufd:$src3 VR128:$src1, (undef))),
- (SHUFPDrri VR128:$src1, VR128:$src1,
+ (SHUFPDrri VR128:$src1, VR128:$src1,
(SHUFFLE_get_shuf_imm VR128:$src3))>,
Requires<[HasSSE2]>;
// Special unary SHUFPDrri case.
def : Pat<(v2f64 (pshufd:$src3 VR128:$src1, (undef))),
- (SHUFPDrri VR128:$src1, VR128:$src1,
+ (SHUFPDrri VR128:$src1, VR128:$src1,
(SHUFFLE_get_shuf_imm VR128:$src3))>,
Requires<[HasSSE2]>;
// Unary v4f32 shuffle with PSHUF* in order to fold a load.
@@ -2899,16 +2943,16 @@ def : Pat<(pshufd:$src2 (bc_v4i32 (memopv4f32 addr:$src1)), (undef)),
// Special binary v4i32 shuffle cases with SHUFPS.
def : Pat<(v4i32 (shufp:$src3 VR128:$src1, (v4i32 VR128:$src2))),
- (SHUFPSrri VR128:$src1, VR128:$src2,
+ (SHUFPSrri VR128:$src1, VR128:$src2,
(SHUFFLE_get_shuf_imm VR128:$src3))>,
Requires<[HasSSE2]>;
def : Pat<(v4i32 (shufp:$src3 VR128:$src1, (bc_v4i32 (memopv2i64 addr:$src2)))),
- (SHUFPSrmi VR128:$src1, addr:$src2,
+ (SHUFPSrmi VR128:$src1, addr:$src2,
(SHUFFLE_get_shuf_imm VR128:$src3))>,
Requires<[HasSSE2]>;
// Special binary v2i64 shuffle cases using SHUFPDrri.
def : Pat<(v2i64 (shufp:$src3 VR128:$src1, VR128:$src2)),
- (SHUFPDrri VR128:$src1, VR128:$src2,
+ (SHUFPDrri VR128:$src1, VR128:$src2,
(SHUFFLE_get_shuf_imm VR128:$src3))>,
Requires<[HasSSE2]>;
@@ -3030,7 +3074,7 @@ def : Pat<(v4i32 (movlp VR128:$src1, VR128:$src2)),
// vector_shuffle v1, v2 <4, 5, 2, 3> using SHUFPSrri (we prefer movsd, but
// fall back to this for SSE1)
def : Pat<(v4f32 (movlp:$src3 VR128:$src1, (v4f32 VR128:$src2))),
- (SHUFPSrri VR128:$src2, VR128:$src1,
+ (SHUFPSrri VR128:$src2, VR128:$src1,
(SHUFFLE_get_shuf_imm VR128:$src3))>, Requires<[HasSSE1]>;
// Set lowest element and zero upper elements.
@@ -3097,7 +3141,7 @@ def : Pat<(store (v8i16 VR128:$src), addr:$dst),
(MOVUPSmr addr:$dst, VR128:$src)>, Requires<[HasSSE2]>;
def : Pat<(store (v16i8 VR128:$src), addr:$dst),
(MOVUPSmr addr:$dst, VR128:$src)>, Requires<[HasSSE2]>;
-
+
//===----------------------------------------------------------------------===//
// SSE4.1 Instructions
//===----------------------------------------------------------------------===//
@@ -3108,7 +3152,7 @@ multiclass sse41_fp_unop_rm<bits<8> opcps, bits<8> opcpd,
Intrinsic V2F64Int> {
// Intrinsic operation, reg.
// Vector intrinsic operation, reg
- def PSr_Int : SS4AIi8<opcps, MRMSrcReg,
+ def PSr_Int : SS4AIi8<opcps, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
"ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
@@ -3149,41 +3193,41 @@ multiclass sse41_fp_binop_rm<bits<8> opcss, bits<8> opcsd,
Intrinsic F64Int> {
// Intrinsic operation, reg.
def SSr_Int : SS4AIi8<opcss, MRMSrcReg,
- (outs VR128:$dst),
+ (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i32i8imm:$src3),
!strconcat(OpcodeStr,
"ss\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(F32Int VR128:$src1, VR128:$src2, imm:$src3))]>,
OpSize;
// Intrinsic operation, mem.
- def SSm_Int : SS4AIi8<opcss, MRMSrcMem,
- (outs VR128:$dst),
+ def SSm_Int : SS4AIi8<opcss, MRMSrcMem,
+ (outs VR128:$dst),
(ins VR128:$src1, ssmem:$src2, i32i8imm:$src3),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"ss\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(F32Int VR128:$src1, sse_load_f32:$src2, imm:$src3))]>,
OpSize;
// Intrinsic operation, reg.
def SDr_Int : SS4AIi8<opcsd, MRMSrcReg,
- (outs VR128:$dst),
+ (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i32i8imm:$src3),
!strconcat(OpcodeStr,
"sd\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(F64Int VR128:$src1, VR128:$src2, imm:$src3))]>,
OpSize;
// Intrinsic operation, mem.
def SDm_Int : SS4AIi8<opcsd, MRMSrcMem,
- (outs VR128:$dst),
+ (outs VR128:$dst),
(ins VR128:$src1, sdmem:$src2, i32i8imm:$src3),
!strconcat(OpcodeStr,
"sd\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(F64Int VR128:$src1, sse_load_f64:$src2, imm:$src3))]>,
OpSize;
}
@@ -3302,9 +3346,9 @@ let Constraints = "$src1 = $dst" in {
Intrinsic IntId128, bit Commutable = 0> {
def rri : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i32i8imm:$src3),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(IntId128 VR128:$src1, VR128:$src2, imm:$src3))]>,
OpSize {
let isCommutable = Commutable;
@@ -3339,7 +3383,7 @@ let Uses = [XMM0], Constraints = "$src1 = $dst" in {
multiclass SS41I_ternary_int<bits<8> opc, string OpcodeStr, Intrinsic IntId> {
def rr0 : SS48I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}"),
[(set VR128:$dst, (IntId VR128:$src1, VR128:$src2, XMM0))]>,
OpSize;
@@ -3471,13 +3515,13 @@ def : Pat<(int_x86_sse41_pmovzxbq
multiclass SS41I_extract8<bits<8> opc, string OpcodeStr> {
def rr : SS4AIi8<opc, MRMDestReg, (outs GR32:$dst),
(ins VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set GR32:$dst, (X86pextrb (v16i8 VR128:$src1), imm:$src2))]>,
OpSize;
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i8mem:$dst, VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[]>, OpSize;
// FIXME:
@@ -3492,7 +3536,7 @@ defm PEXTRB : SS41I_extract8<0x14, "pextrb">;
multiclass SS41I_extract16<bits<8> opc, string OpcodeStr> {
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i16mem:$dst, VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[]>, OpSize;
// FIXME:
@@ -3507,13 +3551,13 @@ defm PEXTRW : SS41I_extract16<0x15, "pextrw">;
multiclass SS41I_extract32<bits<8> opc, string OpcodeStr> {
def rr : SS4AIi8<opc, MRMDestReg, (outs GR32:$dst),
(ins VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set GR32:$dst,
(extractelt (v4i32 VR128:$src1), imm:$src2))]>, OpSize;
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i32mem:$dst, VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(store (extractelt (v4i32 VR128:$src1), imm:$src2),
addr:$dst)]>, OpSize;
@@ -3527,14 +3571,14 @@ defm PEXTRD : SS41I_extract32<0x16, "pextrd">;
multiclass SS41I_extractf32<bits<8> opc, string OpcodeStr> {
def rr : SS4AIi8<opc, MRMDestReg, (outs GR32:$dst),
(ins VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set GR32:$dst,
(extractelt (bc_v4i32 (v4f32 VR128:$src1)), imm:$src2))]>,
OpSize;
- def mr : SS4AIi8<opc, MRMDestMem, (outs),
+ def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins f32mem:$dst, VR128:$src1, i32i8imm:$src2),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(store (extractelt (bc_v4i32 (v4f32 VR128:$src1)), imm:$src2),
addr:$dst)]>, OpSize;
@@ -3553,15 +3597,15 @@ let Constraints = "$src1 = $dst" in {
multiclass SS41I_insert8<bits<8> opc, string OpcodeStr> {
def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, GR32:$src2, i32i8imm:$src3),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(X86pinsrb VR128:$src1, GR32:$src2, imm:$src3))]>, OpSize;
def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i8mem:$src2, i32i8imm:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(X86pinsrb VR128:$src1, (extloadi8 addr:$src2),
imm:$src3))]>, OpSize;
}
@@ -3573,16 +3617,16 @@ let Constraints = "$src1 = $dst" in {
multiclass SS41I_insert32<bits<8> opc, string OpcodeStr> {
def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, GR32:$src2, i32i8imm:$src3),
- !strconcat(OpcodeStr,
+ !strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(v4i32 (insertelt VR128:$src1, GR32:$src2, imm:$src3)))]>,
OpSize;
def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i32mem:$src2, i32i8imm:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
+ [(set VR128:$dst,
(v4i32 (insertelt VR128:$src1, (loadi32 addr:$src2),
imm:$src3)))]>, OpSize;
}
@@ -3590,37 +3634,57 @@ let Constraints = "$src1 = $dst" in {
defm PINSRD : SS41I_insert32<0x22, "pinsrd">;
+// insertps has a few different modes, there's the first two here below which
+// are optimized inserts that won't zero arbitrary elements in the destination
+// vector. The next one matches the intrinsic and could zero arbitrary elements
+// in the target vector.
let Constraints = "$src1 = $dst" in {
multiclass SS41I_insertf32<bits<8> opc, string OpcodeStr> {
def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, FR32:$src2, i32i8imm:$src3),
- !strconcat(OpcodeStr,
+ (ins VR128:$src1, VR128:$src2, i32i8imm:$src3),
+ !strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
- (X86insrtps VR128:$src1, FR32:$src2, imm:$src3))]>, OpSize;
+ [(set VR128:$dst,
+ (X86insrtps VR128:$src1, VR128:$src2, imm:$src3))]>,
+ OpSize;
def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, f32mem:$src2, i32i8imm:$src3),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR128:$dst,
- (X86insrtps VR128:$src1, (loadf32 addr:$src2),
+ [(set VR128:$dst,
+ (X86insrtps VR128:$src1,
+ (v4f32 (scalar_to_vector (loadf32 addr:$src2))),
imm:$src3))]>, OpSize;
}
}
defm INSERTPS : SS41I_insertf32<0x21, "insertps">;
+def : Pat<(int_x86_sse41_insertps VR128:$src1, VR128:$src2, imm:$src3),
+ (INSERTPSrr VR128:$src1, VR128:$src2, imm:$src3)>;
+
+// ptest instruction we'll lower to this in X86ISelLowering primarily from
+// the intel intrinsic that corresponds to this.
let Defs = [EFLAGS] in {
def PTESTrr : SS48I<0x17, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
- "ptest \t{$src2, $src1|$src1, $src2}", []>, OpSize;
+ "ptest \t{$src2, $src1|$src1, $src2}",
+ [(X86ptest VR128:$src1, VR128:$src2),
+ (implicit EFLAGS)]>, OpSize;
def PTESTrm : SS48I<0x17, MRMSrcMem, (outs), (ins VR128:$src1, i128mem:$src2),
- "ptest \t{$src2, $src1|$src1, $src2}", []>, OpSize;
+ "ptest \t{$src2, $src1|$src1, $src2}",
+ [(X86ptest VR128:$src1, (load addr:$src2)),
+ (implicit EFLAGS)]>, OpSize;
}
def MOVNTDQArm : SS48I<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movntdqa\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse41_movntdqa addr:$src))]>;
+
+//===----------------------------------------------------------------------===//
+// SSE4.2 Instructions
+//===----------------------------------------------------------------------===//
+
/// SS42I_binop_rm_int - Simple SSE 4.2 binary operator
let Constraints = "$src1 = $dst" in {
multiclass SS42I_binop_rm_int<bits<8> opc, string OpcodeStr,
@@ -3647,3 +3711,171 @@ def : Pat<(v2i64 (X86pcmpgtq VR128:$src1, VR128:$src2)),
(PCMPGTQrr VR128:$src1, VR128:$src2)>;
def : Pat<(v2i64 (X86pcmpgtq VR128:$src1, (memop addr:$src2))),
(PCMPGTQrm VR128:$src1, addr:$src2)>;
+
+// crc intrinsic instruction
+// This set of instructions are only rm, the only difference is the size
+// of r and m.
+let Constraints = "$src1 = $dst" in {
+ def CRC32m8 : SS42FI<0xF0, MRMSrcMem, (outs GR32:$dst),
+ (ins GR32:$src1, i8mem:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR32:$dst,
+ (int_x86_sse42_crc32_8 GR32:$src1,
+ (load addr:$src2)))]>, OpSize;
+ def CRC32r8 : SS42FI<0xF0, MRMSrcReg, (outs GR32:$dst),
+ (ins GR32:$src1, GR8:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR32:$dst,
+ (int_x86_sse42_crc32_8 GR32:$src1, GR8:$src2))]>,
+ OpSize;
+ def CRC32m16 : SS42FI<0xF1, MRMSrcMem, (outs GR32:$dst),
+ (ins GR32:$src1, i16mem:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR32:$dst,
+ (int_x86_sse42_crc32_16 GR32:$src1,
+ (load addr:$src2)))]>,
+ OpSize;
+ def CRC32r16 : SS42FI<0xF1, MRMSrcReg, (outs GR32:$dst),
+ (ins GR32:$src1, GR16:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR32:$dst,
+ (int_x86_sse42_crc32_16 GR32:$src1, GR16:$src2))]>,
+ OpSize;
+ def CRC32m32 : SS42FI<0xF1, MRMSrcMem, (outs GR32:$dst),
+ (ins GR32:$src1, i32mem:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR32:$dst,
+ (int_x86_sse42_crc32_32 GR32:$src1,
+ (load addr:$src2)))]>, OpSize;
+ def CRC32r32 : SS42FI<0xF1, MRMSrcReg, (outs GR32:$dst),
+ (ins GR32:$src1, GR32:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR32:$dst,
+ (int_x86_sse42_crc32_32 GR32:$src1, GR32:$src2))]>,
+ OpSize;
+ def CRC64m64 : SS42FI<0xF0, MRMSrcMem, (outs GR64:$dst),
+ (ins GR64:$src1, i64mem:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR64:$dst,
+ (int_x86_sse42_crc32_64 GR64:$src1,
+ (load addr:$src2)))]>,
+ OpSize, REX_W;
+ def CRC64r64 : SS42FI<0xF0, MRMSrcReg, (outs GR64:$dst),
+ (ins GR64:$src1, GR64:$src2),
+ "crc32 \t{$src2, $src1|$src1, $src2}",
+ [(set GR64:$dst,
+ (int_x86_sse42_crc32_64 GR64:$src1, GR64:$src2))]>,
+ OpSize, REX_W;
+}
+
+// String/text processing instructions.
+let Defs = [EFLAGS], usesCustomDAGSchedInserter = 1 in {
+def PCMPISTRM128REG : SS42AI<0, Pseudo, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ "#PCMPISTRM128rr PSEUDO!",
+ [(set VR128:$dst,
+ (int_x86_sse42_pcmpistrm128 VR128:$src1, VR128:$src2,
+ imm:$src3))]>, OpSize;
+def PCMPISTRM128MEM : SS42AI<0, Pseudo, (outs VR128:$dst),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ "#PCMPISTRM128rm PSEUDO!",
+ [(set VR128:$dst,
+ (int_x86_sse42_pcmpistrm128 VR128:$src1,
+ (load addr:$src2),
+ imm:$src3))]>, OpSize;
+}
+
+let Defs = [XMM0, EFLAGS] in {
+def PCMPISTRM128rr : SS42AI<0x62, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ "pcmpistrm\t{$src3, $src2, $src1|$src1, $src2, $src3}",
+ []>, OpSize;
+def PCMPISTRM128rm : SS42AI<0x62, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ "pcmpistrm\t{$src3, $src2, $src1|$src1, $src2, $src3}",
+ []>, OpSize;
+}
+
+let Defs = [EFLAGS], Uses = [EAX, EDX],
+ usesCustomDAGSchedInserter = 1 in {
+def PCMPESTRM128REG : SS42AI<0, Pseudo, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src3, i8imm:$src5),
+ "#PCMPESTRM128rr PSEUDO!",
+ [(set VR128:$dst,
+ (int_x86_sse42_pcmpestrm128 VR128:$src1, EAX,
+ VR128:$src3,
+ EDX, imm:$src5))]>, OpSize;
+def PCMPESTRM128MEM : SS42AI<0, Pseudo, (outs VR128:$dst),
+ (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
+ "#PCMPESTRM128rm PSEUDO!",
+ [(set VR128:$dst,
+ (int_x86_sse42_pcmpestrm128 VR128:$src1, EAX,
+ (load addr:$src3),
+ EDX, imm:$src5))]>, OpSize;
+}
+
+let Defs = [XMM0, EFLAGS], Uses = [EAX, EDX] in {
+def PCMPESTRM128rr : SS42AI<0x60, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src3, i8imm:$src5),
+ "pcmpestrm\t{$src5, $src3, $src1|$src1, $src3, $src5}",
+ []>, OpSize;
+def PCMPESTRM128rm : SS42AI<0x60, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
+ "pcmpestrm\t{$src5, $src3, $src1|$src1, $src3, $src5}",
+ []>, OpSize;
+}
+
+let Defs = [ECX, EFLAGS] in {
+ multiclass SS42AI_pcmpistri<Intrinsic IntId128> {
+ def rr : SS42AI<0x63, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ "pcmpistri\t{$src3, $src2, $src1|$src1, $src2, $src3}",
+ [(set ECX,
+ (IntId128 VR128:$src1, VR128:$src2, imm:$src3)),
+ (implicit EFLAGS)]>,
+ OpSize;
+ def rm : SS42AI<0x63, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ "pcmpistri\t{$src3, $src2, $src1|$src1, $src2, $src3}",
+ [(set ECX,
+ (IntId128 VR128:$src1, (load addr:$src2), imm:$src3)),
+ (implicit EFLAGS)]>,
+ OpSize;
+ }
+}
+
+defm PCMPISTRI : SS42AI_pcmpistri<int_x86_sse42_pcmpistri128>;
+defm PCMPISTRIA : SS42AI_pcmpistri<int_x86_sse42_pcmpistria128>;
+defm PCMPISTRIC : SS42AI_pcmpistri<int_x86_sse42_pcmpistric128>;
+defm PCMPISTRIO : SS42AI_pcmpistri<int_x86_sse42_pcmpistrio128>;
+defm PCMPISTRIS : SS42AI_pcmpistri<int_x86_sse42_pcmpistris128>;
+defm PCMPISTRIZ : SS42AI_pcmpistri<int_x86_sse42_pcmpistriz128>;
+
+let Defs = [ECX, EFLAGS] in {
+let Uses = [EAX, EDX] in {
+ multiclass SS42AI_pcmpestri<Intrinsic IntId128> {
+ def rr : SS42AI<0x61, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src3, i8imm:$src5),
+ "pcmpestri\t{$src5, $src3, $src1|$src1, $src3, $src5}",
+ [(set ECX,
+ (IntId128 VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5)),
+ (implicit EFLAGS)]>,
+ OpSize;
+ def rm : SS42AI<0x61, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
+ "pcmpestri\t{$src5, $src3, $src1|$src1, $src3, $src5}",
+ [(set ECX,
+ (IntId128 VR128:$src1, EAX, (load addr:$src3),
+ EDX, imm:$src5)),
+ (implicit EFLAGS)]>,
+ OpSize;
+ }
+}
+}
+
+defm PCMPESTRI : SS42AI_pcmpestri<int_x86_sse42_pcmpestri128>;
+defm PCMPESTRIA : SS42AI_pcmpestri<int_x86_sse42_pcmpestria128>;
+defm PCMPESTRIC : SS42AI_pcmpestri<int_x86_sse42_pcmpestric128>;
+defm PCMPESTRIO : SS42AI_pcmpestri<int_x86_sse42_pcmpestrio128>;
+defm PCMPESTRIS : SS42AI_pcmpestri<int_x86_sse42_pcmpestris128>;
+defm PCMPESTRIZ : SS42AI_pcmpestri<int_x86_sse42_pcmpestriz128>;
OpenPOWER on IntegriCloud