From f7de577a08a705970f0fd8f3c1bb40f7040e4476 Mon Sep 17 00:00:00 2001 From: Craig Topper Date: Tue, 22 Nov 2011 01:57:35 +0000 Subject: Fix shuffle decoding logic to handle UNPCKLPS/UNPCKLPD on 256-bit vectors correctly. Add support for decoding UNPCKHPS/UNPCKHPD for AVX 128-bit and 256-bit forms. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@145055 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/Utils/X86ShuffleDecode.cpp | 45 +++++++++++++++++++++---------- 1 file changed, 31 insertions(+), 14 deletions(-) (limited to 'lib/Target/X86/Utils/X86ShuffleDecode.cpp') diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp index aeb3309..8acd3c3 100644 --- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp +++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp @@ -142,11 +142,32 @@ void DecodeSHUFPSMask(unsigned NElts, unsigned Imm, } } -void DecodeUNPCKHPMask(unsigned NElts, - SmallVectorImpl &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 +void DecodeUNPCKHPSMask(unsigned NElts, + SmallVectorImpl &ShuffleMask) { + DecodeUNPCKHPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask); +} + +void DecodeUNPCKHPDMask(unsigned NElts, + SmallVectorImpl &ShuffleMask) { + DecodeUNPCKHPMask(MVT::getVectorVT(MVT::i64, NElts), ShuffleMask); +} + +void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl &ShuffleMask) { + unsigned NumElts = VT.getVectorNumElements(); + + // 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 s = 0; s < NumLanes; ++s) { + unsigned Start = s * NumLaneElts + NumLaneElts/2; + unsigned End = s * NumLaneElts + NumLaneElts; + for (unsigned i = Start; i != End; ++i) { + ShuffleMask.push_back(i); // Reads from dest/src1 + ShuffleMask.push_back(i+NumElts); // Reads from src/src2 + } } } @@ -163,8 +184,7 @@ void DecodeUNPCKLPDMask(unsigned NElts, /// 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 &ShuffleMask) { +void DecodeUNPCKLPMask(EVT VT, SmallVectorImpl &ShuffleMask) { unsigned NumElts = VT.getVectorNumElements(); // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate @@ -173,16 +193,13 @@ 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) { + unsigned Start = s * NumLaneElts; + unsigned End = s * NumLaneElts + NumLaneElts/2; for (unsigned i = Start; i != End; ++i) { - ShuffleMask.push_back(i); // Reads from dest/src1 - ShuffleMask.push_back(i+NumLaneElts); // Reads from src/src2 + 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; } } -- cgit v1.1 From 796c193768547459cd6cbd667c8a43fedd601022 Mon Sep 17 00:00:00 2001 From: Craig Topper Date: Tue, 22 Nov 2011 14:27:57 +0000 Subject: More fixes to the X86InstComments for shuffle instructions. In particular add AVX flavors of many instructions and fix the destination operand for some of the existing AVX entries. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@145063 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/Utils/X86ShuffleDecode.cpp | 20 -------------------- 1 file changed, 20 deletions(-) (limited to 'lib/Target/X86/Utils/X86ShuffleDecode.cpp') diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp index 8acd3c3..f6c9d7b 100644 --- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp +++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp @@ -142,16 +142,6 @@ void DecodeSHUFPSMask(unsigned NElts, unsigned Imm, } } -void DecodeUNPCKHPSMask(unsigned NElts, - SmallVectorImpl &ShuffleMask) { - DecodeUNPCKHPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask); -} - -void DecodeUNPCKHPDMask(unsigned NElts, - SmallVectorImpl &ShuffleMask) { - DecodeUNPCKHPMask(MVT::getVectorVT(MVT::i64, NElts), ShuffleMask); -} - void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl &ShuffleMask) { unsigned NumElts = VT.getVectorNumElements(); @@ -171,16 +161,6 @@ void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl &ShuffleMask) { } } -void DecodeUNPCKLPSMask(unsigned NElts, - SmallVectorImpl &ShuffleMask) { - DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask); -} - -void DecodeUNPCKLPDMask(unsigned NElts, - SmallVectorImpl &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. -- cgit v1.1