aboutsummaryrefslogtreecommitdiff
path: root/contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp')
-rw-r--r--contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp193
1 files changed, 66 insertions, 127 deletions
diff --git a/contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index aeb3309d09aa..32c722acc437 100644
--- a/contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/contrib/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -20,7 +20,7 @@
namespace llvm {
-void DecodeINSERTPSMask(unsigned Imm, SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeINSERTPSMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
// Defaults the copying the dest value.
ShuffleMask.push_back(0);
ShuffleMask.push_back(1);
@@ -44,8 +44,7 @@ void DecodeINSERTPSMask(unsigned Imm, SmallVectorImpl<unsigned> &ShuffleMask) {
}
// <3,1> or <6,7,2,3>
-void DecodeMOVHLPSMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeMOVHLPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask) {
for (unsigned i = NElts/2; i != NElts; ++i)
ShuffleMask.push_back(NElts+i);
@@ -54,8 +53,7 @@ void DecodeMOVHLPSMask(unsigned NElts,
}
// <0,2> or <0,1,4,5>
-void DecodeMOVLHPSMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeMOVLHPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask) {
for (unsigned i = 0; i != NElts/2; ++i)
ShuffleMask.push_back(i);
@@ -63,16 +61,26 @@ void DecodeMOVLHPSMask(unsigned NElts,
ShuffleMask.push_back(NElts+i);
}
-void DecodePSHUFMask(unsigned NElts, unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- for (unsigned i = 0; i != NElts; ++i) {
- ShuffleMask.push_back(Imm % NElts);
- Imm /= NElts;
+/// DecodePSHUFMask - This decodes the shuffle masks for pshufd, and vpermilp*.
+/// VT indicates the type of the vector allowing it to handle different
+/// datatypes and vector widths.
+void DecodePSHUFMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
+ unsigned NumElts = VT.getVectorNumElements();
+
+ unsigned NumLanes = VT.getSizeInBits() / 128;
+ unsigned NumLaneElts = NumElts / NumLanes;
+
+ int NewImm = Imm;
+ for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
+ for (unsigned i = 0; i != NumLaneElts; ++i) {
+ ShuffleMask.push_back(NewImm % NumLaneElts + l);
+ NewImm /= NumLaneElts;
+ }
+ if (NumLaneElts == 4) NewImm = Imm; // reload imm
}
}
-void DecodePSHUFHWMask(unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodePSHUFHWMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
ShuffleMask.push_back(0);
ShuffleMask.push_back(1);
ShuffleMask.push_back(2);
@@ -83,8 +91,7 @@ void DecodePSHUFHWMask(unsigned Imm,
}
}
-void DecodePSHUFLWMask(unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodePSHUFLWMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
for (unsigned i = 0; i != 4; ++i) {
ShuffleMask.push_back((Imm & 3));
Imm >>= 2;
@@ -95,76 +102,35 @@ void DecodePSHUFLWMask(unsigned Imm,
ShuffleMask.push_back(7);
}
-void DecodePUNPCKLBWMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i8, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLWDMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i16, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLDQMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLQDQMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i64, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLMask(EVT VT,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(VT, ShuffleMask);
-}
-
-void DecodePUNPCKHMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- for (unsigned i = 0; i != NElts/2; ++i) {
- ShuffleMask.push_back(i+NElts/2);
- ShuffleMask.push_back(i+NElts+NElts/2);
- }
-}
+/// DecodeSHUFPMask - This decodes the shuffle masks for shufp*. VT indicates
+/// the type of the vector allowing it to handle different datatypes and vector
+/// widths.
+void DecodeSHUFPMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
+ unsigned NumElts = VT.getVectorNumElements();
-void DecodeSHUFPSMask(unsigned NElts, unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- // Part that reads from dest.
- for (unsigned i = 0; i != NElts/2; ++i) {
- ShuffleMask.push_back(Imm % NElts);
- Imm /= NElts;
- }
- // Part that reads from src.
- for (unsigned i = 0; i != NElts/2; ++i) {
- ShuffleMask.push_back(Imm % NElts + NElts);
- Imm /= NElts;
- }
-}
+ unsigned NumLanes = VT.getSizeInBits() / 128;
+ unsigned NumLaneElts = NumElts / NumLanes;
-void DecodeUNPCKHPMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- for (unsigned i = 0; i != NElts/2; ++i) {
- ShuffleMask.push_back(i+NElts/2); // Reads from dest
- ShuffleMask.push_back(i+NElts+NElts/2); // Reads from src
+ int NewImm = Imm;
+ for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
+ // Part that reads from dest.
+ for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+ ShuffleMask.push_back(NewImm % NumLaneElts + l);
+ NewImm /= NumLaneElts;
+ }
+ // Part that reads from src.
+ for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+ ShuffleMask.push_back(NewImm % NumLaneElts + NumElts + l);
+ NewImm /= NumLaneElts;
+ }
+ if (NumLaneElts == 4) NewImm = Imm; // reload imm
}
}
-void DecodeUNPCKLPSMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask);
-}
-
-void DecodeUNPCKLPDMask(unsigned NElts,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i64, NElts), ShuffleMask);
-}
-
-/// DecodeUNPCKLPMask - This decodes the shuffle masks for unpcklps/unpcklpd
-/// etc. VT indicates the type of the vector allowing it to handle different
-/// datatypes and vector widths.
-void DecodeUNPCKLPMask(EVT VT,
- SmallVectorImpl<unsigned> &ShuffleMask) {
+/// DecodeUNPCKHMask - This decodes the shuffle masks for unpckhps/unpckhpd
+/// and punpckh*. VT indicates the type of the vector allowing it to handle
+/// different datatypes and vector widths.
+void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<int> &ShuffleMask) {
unsigned NumElts = VT.getVectorNumElements();
// Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
@@ -173,55 +139,36 @@ void DecodeUNPCKLPMask(EVT VT,
if (NumLanes == 0 ) NumLanes = 1; // Handle MMX
unsigned NumLaneElts = NumElts / NumLanes;
- unsigned Start = 0;
- unsigned End = NumLaneElts / 2;
- for (unsigned s = 0; s < NumLanes; ++s) {
- for (unsigned i = Start; i != End; ++i) {
- ShuffleMask.push_back(i); // Reads from dest/src1
- ShuffleMask.push_back(i+NumLaneElts); // Reads from src/src2
+ for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
+ for (unsigned i = l + NumLaneElts/2, e = l + NumLaneElts; i != e; ++i) {
+ ShuffleMask.push_back(i); // Reads from dest/src1
+ ShuffleMask.push_back(i+NumElts); // Reads from src/src2
}
- // Process the next 128 bits.
- Start += NumLaneElts;
- End += NumLaneElts;
}
}
-// DecodeVPERMILPSMask - Decodes VPERMILPS permutes for any 128-bit 32-bit
-// elements. For 256-bit vectors, it's considered as two 128 lanes, the
-// referenced elements can't cross lanes and the mask of the first lane must
-// be the same of the second.
-void DecodeVPERMILPSMask(unsigned NumElts, unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- unsigned NumLanes = (NumElts*32)/128;
- unsigned LaneSize = NumElts/NumLanes;
-
- for (unsigned l = 0; l != NumLanes; ++l) {
- for (unsigned i = 0; i != LaneSize; ++i) {
- unsigned Idx = (Imm >> (i*2)) & 0x3 ;
- ShuffleMask.push_back(Idx+(l*LaneSize));
- }
- }
-}
+/// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
+/// and punpckl*. VT indicates the type of the vector allowing it to handle
+/// different datatypes and vector widths.
+void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<int> &ShuffleMask) {
+ unsigned NumElts = VT.getVectorNumElements();
-// DecodeVPERMILPDMask - Decodes VPERMILPD permutes for any 128-bit 64-bit
-// elements. For 256-bit vectors, it's considered as two 128 lanes, the
-// referenced elements can't cross lanes but the mask of the first lane can
-// be the different of the second (not like VPERMILPS).
-void DecodeVPERMILPDMask(unsigned NumElts, unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- unsigned NumLanes = (NumElts*64)/128;
- unsigned LaneSize = NumElts/NumLanes;
-
- for (unsigned l = 0; l < NumLanes; ++l) {
- for (unsigned i = l*LaneSize; i < LaneSize*(l+1); ++i) {
- unsigned Idx = (Imm >> i) & 0x1;
- ShuffleMask.push_back(Idx+(l*LaneSize));
+ // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
+ // independently on 128-bit lanes.
+ unsigned NumLanes = VT.getSizeInBits() / 128;
+ if (NumLanes == 0 ) NumLanes = 1; // Handle MMX
+ unsigned NumLaneElts = NumElts / NumLanes;
+
+ for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
+ for (unsigned i = l, e = l + NumLaneElts/2; i != e; ++i) {
+ ShuffleMask.push_back(i); // Reads from dest/src1
+ ShuffleMask.push_back(i+NumElts); // Reads from src/src2
}
}
}
-void DecodeVPERM2F128Mask(EVT VT, unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeVPERM2X128Mask(EVT VT, unsigned Imm,
+ SmallVectorImpl<int> &ShuffleMask) {
unsigned HalfSize = VT.getVectorNumElements()/2;
unsigned FstHalfBegin = (Imm & 0x3) * HalfSize;
unsigned SndHalfBegin = ((Imm >> 4) & 0x3) * HalfSize;
@@ -232,12 +179,4 @@ void DecodeVPERM2F128Mask(EVT VT, unsigned Imm,
ShuffleMask.push_back(i);
}
-void DecodeVPERM2F128Mask(unsigned Imm,
- SmallVectorImpl<unsigned> &ShuffleMask) {
- // VPERM2F128 is used by any 256-bit EVT, but X86InstComments only
- // has information about the instruction and not the types. So for
- // instruction comments purpose, assume the 256-bit vector is v4i64.
- return DecodeVPERM2F128Mask(MVT::v4i64, Imm, ShuffleMask);
-}
-
} // llvm namespace
generated by cgit v1.2.3 (git 2.25.1) at 2025-03-14 17:44:11 +0000