def bc_v4i32 : PatFrag<(ops node:$in), (v4i32 (bitconvert node:$in))>;
def bc_v2i64 : PatFrag<(ops node:$in), (v2i64 (bitconvert node:$in))>;
+def vzmovl_v2i64 : PatFrag<(ops node:$src),
+ (bitconvert (v2i64 (X86vzmovl
+ (v2i64 (scalar_to_vector (loadi64 node:$src))))))>;
+def vzmovl_v4i32 : PatFrag<(ops node:$src),
+ (bitconvert (v4i32 (X86vzmovl
+ (v4i32 (scalar_to_vector (loadi32 node:$src))))))>;
+
+def vzload_v2i64 : PatFrag<(ops node:$src),
+ (bitconvert (v2i64 (X86vzload node:$src)))>;
+
+
def fp32imm0 : PatLeaf<(f32 fpimm), [{
return N->isExactlyValue(+0.0);
}]>;
def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst,
- (IntId (bitconvert (v4i32 (load addr:$src)))))]>, OpSize;
+ [(set VR128:$dst,
+ (IntId (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))]>,
+ OpSize;
}
defm PMOVSXBW : SS41I_binop_rm_int8<0x20, "pmovsxbw", int_x86_sse41_pmovsxbw>;
defm PMOVZXWD : SS41I_binop_rm_int8<0x33, "pmovzxwd", int_x86_sse41_pmovzxwd>;
defm PMOVZXDQ : SS41I_binop_rm_int8<0x35, "pmovzxdq", int_x86_sse41_pmovzxdq>;
+// Common patterns involving scalar load.
+def : Pat<(int_x86_sse41_pmovsxbw (vzmovl_v2i64 addr:$src)),
+ (PMOVSXBWrm addr:$src)>, Requires<[HasSSE41]>;
+def : Pat<(int_x86_sse41_pmovsxbw (vzload_v2i64 addr:$src)),
+ (PMOVSXBWrm addr:$src)>, Requires<[HasSSE41]>;
+
+def : Pat<(int_x86_sse41_pmovsxwd (vzmovl_v2i64 addr:$src)),
+ (PMOVSXWDrm addr:$src)>, Requires<[HasSSE41]>;
+def : Pat<(int_x86_sse41_pmovsxwd (vzload_v2i64 addr:$src)),
+ (PMOVSXWDrm addr:$src)>, Requires<[HasSSE41]>;
+
+def : Pat<(int_x86_sse41_pmovsxdq (vzmovl_v2i64 addr:$src)),
+ (PMOVSXDQrm addr:$src)>, Requires<[HasSSE41]>;
+def : Pat<(int_x86_sse41_pmovsxdq (vzload_v2i64 addr:$src)),
+ (PMOVSXDQrm addr:$src)>, Requires<[HasSSE41]>;
+
+def : Pat<(int_x86_sse41_pmovzxbw (vzmovl_v2i64 addr:$src)),
+ (PMOVZXBWrm addr:$src)>, Requires<[HasSSE41]>;
+def : Pat<(int_x86_sse41_pmovzxbw (vzload_v2i64 addr:$src)),
+ (PMOVZXBWrm addr:$src)>, Requires<[HasSSE41]>;
+
+def : Pat<(int_x86_sse41_pmovzxwd (vzmovl_v2i64 addr:$src)),
+ (PMOVZXWDrm addr:$src)>, Requires<[HasSSE41]>;
+def : Pat<(int_x86_sse41_pmovzxwd (vzload_v2i64 addr:$src)),
+ (PMOVZXWDrm addr:$src)>, Requires<[HasSSE41]>;
+
+def : Pat<(int_x86_sse41_pmovzxdq (vzmovl_v2i64 addr:$src)),
+ (PMOVZXDQrm addr:$src)>, Requires<[HasSSE41]>;
+def : Pat<(int_x86_sse41_pmovzxdq (vzload_v2i64 addr:$src)),
+ (PMOVZXDQrm addr:$src)>, Requires<[HasSSE41]>;
+
+
multiclass SS41I_binop_rm_int4<bits<8> opc, string OpcodeStr, Intrinsic IntId> {
def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst,
- (IntId (bitconvert (v4i32 (load addr:$src)))))]>, OpSize;
+ [(set VR128:$dst,
+ (IntId (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))]>,
+ OpSize;
}
defm PMOVSXBD : SS41I_binop_rm_int4<0x21, "pmovsxbd", int_x86_sse41_pmovsxbd>;
defm PMOVZXBD : SS41I_binop_rm_int4<0x31, "pmovzxbd", int_x86_sse41_pmovzxbd>;
defm PMOVZXWQ : SS41I_binop_rm_int4<0x34, "pmovzxwq", int_x86_sse41_pmovzxwq>;
+// Common patterns involving scalar load
+def : Pat<(int_x86_sse41_pmovsxbd (vzmovl_v4i32 addr:$src)),
+ (PMOVSXBDrm addr:$src)>;
+def : Pat<(int_x86_sse41_pmovsxwq (vzmovl_v4i32 addr:$src)),
+ (PMOVSXWQrm addr:$src)>;
+
+def : Pat<(int_x86_sse41_pmovzxbd (vzmovl_v4i32 addr:$src)),
+ (PMOVZXBDrm addr:$src)>;
+def : Pat<(int_x86_sse41_pmovzxwq (vzmovl_v4i32 addr:$src)),
+ (PMOVZXWQrm addr:$src)>;
+
+
multiclass SS41I_binop_rm_int2<bits<8> opc, string OpcodeStr, Intrinsic IntId> {
def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (IntId VR128:$src))]>, OpSize;
+ // Expecting a i16 load any extended to i32 value.
def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i16mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst,
- (IntId (bitconvert (v4i32 (load addr:$src)))))]>, OpSize;
+ [(set VR128:$dst, (IntId (bitconvert
+ (v4i32 (scalar_to_vector (loadi16_anyext addr:$src))))))]>,
+ OpSize;
}
defm PMOVSXBQ : SS41I_binop_rm_int2<0x22, "pmovsxbq", int_x86_sse41_pmovsxbq>;
defm PMOVZXBQ : SS41I_binop_rm_int2<0x32, "pmovsxbq", int_x86_sse41_pmovzxbq>;
+// Common patterns involving scalar load
+def : Pat<(int_x86_sse41_pmovsxbq
+ (bitconvert (v4i32 (X86vzmovl
+ (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
+ (PMOVSXBQrm addr:$src)>;
+
+def : Pat<(int_x86_sse41_pmovzxbq
+ (bitconvert (v4i32 (X86vzmovl
+ (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
+ (PMOVZXBQrm addr:$src)>;
+
/// SS41I_binop_ext8 - SSE 4.1 extract 8 bits to 32 bit reg or 8 bit mem
multiclass SS41I_extract8<bits<8> opc, string OpcodeStr> {
--- /dev/null
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse41 | not grep movd
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse41 | not grep movq
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse41 | grep pmovsxbd
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse41 | grep pmovsxwd
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse41 | grep pmovsxbq
+; RUN: llvm-as < %s | llc -march=x86-64 -mattr=sse41 | grep movq | count 1
+
+define <2 x i64> @t1(i32* %p) nounwind {
+entry:
+ %0 = load i32* %p, align 4 ; <i32> [#uses=1]
+ %1 = insertelement <4 x i32> undef, i32 %0, i32 0 ; <<4 x i32>> [#uses=1]
+ %2 = insertelement <4 x i32> %1, i32 0, i32 1 ; <<4 x i32>> [#uses=1]
+ %3 = insertelement <4 x i32> %2, i32 0, i32 2 ; <<4 x i32>> [#uses=1]
+ %4 = insertelement <4 x i32> %3, i32 0, i32 3 ; <<4 x i32>> [#uses=1]
+ %5 = bitcast <4 x i32> %4 to <16 x i8> ; <<16 x i8>> [#uses=1]
+ %6 = tail call <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8> %5) nounwind readnone ; <<4 x i32>> [#uses=1]
+ %7 = bitcast <4 x i32> %6 to <2 x i64> ; <<2 x i64>> [#uses=1]
+ ret <2 x i64> %7
+}
+
+declare <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8>) nounwind readnone
+
+define <2 x i64> @t2(i64* %p) nounwind readonly {
+entry:
+ %0 = load i64* %p ; <i64> [#uses=1]
+ %tmp2 = insertelement <2 x i64> zeroinitializer, i64 %0, i32 0 ; <<2 x i64>> [#uses=1]
+ %1 = bitcast <2 x i64> %tmp2 to <8 x i16> ; <<8 x i16>> [#uses=1]
+ %2 = tail call <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16> %1) nounwind readnone ; <<4 x i32>> [#uses=1]
+ %3 = bitcast <4 x i32> %2 to <2 x i64> ; <<2 x i64>> [#uses=1]
+ ret <2 x i64> %3
+}
+
+declare <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16>) nounwind readnone
+
+
+@gv = external global i16 ; <i16*> [#uses=1]
+
+define <2 x i64> @t3() nounwind {
+entry:
+ %0 = load i16* @gv, align 2 ; <i16> [#uses=1]
+ %1 = insertelement <8 x i16> undef, i16 %0, i32 0 ; <<8 x i16>> [#uses=1]
+ %2 = bitcast <8 x i16> %1 to <16 x i8> ; <<16 x i8>> [#uses=1]
+ %3 = tail call <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8> %2) nounwind readnone ; <<2 x i64>> [#uses=1]
+ ret <2 x i64> %3
+}
+
+declare <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8>) nounwind readnone