From: Simon Pilgrim Date: Thu, 4 Aug 2016 10:51:41 +0000 (+0000) Subject: [X86][SSE] Add initial costs for vector CTTZ/CTLZ X-Git-Tag: llvmorg-4.0.0-rc1~13340 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=5d5ca9c0cb784737c64a536fa6b173bf30a93bc1;p=platform%2Fupstream%2Fllvm.git [X86][SSE] Add initial costs for vector CTTZ/CTLZ llvm-svn: 277716 --- diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp index 50e6308..6ea4d42 100644 --- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp +++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp @@ -948,7 +948,9 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy, // Costs should match the codegen from: // BITREVERSE: llvm\test\CodeGen\X86\vector-bitreverse.ll // BSWAP: llvm\test\CodeGen\X86\bswap-vector.ll + // CTLZ: llvm\test\CodeGen\X86\vector-lzcnt-*.ll // CTPOP: llvm\test\CodeGen\X86\vector-popcnt-*.ll + // CTTZ: llvm\test\CodeGen\X86\vector-tzcnt-*.ll static const CostTblEntry XOPCostTbl[] = { { ISD::BITREVERSE, MVT::v4i64, 4 }, { ISD::BITREVERSE, MVT::v8i32, 4 }, @@ -971,10 +973,18 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy, { ISD::BSWAP, MVT::v4i64, 1 }, { ISD::BSWAP, MVT::v8i32, 1 }, { ISD::BSWAP, MVT::v16i16, 1 }, + { ISD::CTLZ, MVT::v4i64, 23 }, + { ISD::CTLZ, MVT::v8i32, 18 }, + { ISD::CTLZ, MVT::v16i16, 14 }, + { ISD::CTLZ, MVT::v32i8, 9 }, { ISD::CTPOP, MVT::v4i64, 7 }, { ISD::CTPOP, MVT::v8i32, 11 }, { ISD::CTPOP, MVT::v16i16, 9 }, - { ISD::CTPOP, MVT::v32i8, 6 } + { ISD::CTPOP, MVT::v32i8, 6 }, + { ISD::CTTZ, MVT::v4i64, 10 }, + { ISD::CTTZ, MVT::v8i32, 14 }, + { ISD::CTTZ, MVT::v16i16, 12 }, + { ISD::CTTZ, MVT::v32i8, 9 } }; static const CostTblEntry AVX1CostTbl[] = { { ISD::BITREVERSE, MVT::v4i64, 10 }, @@ -984,10 +994,18 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy, { ISD::BSWAP, MVT::v4i64, 4 }, { ISD::BSWAP, MVT::v8i32, 4 }, { ISD::BSWAP, MVT::v16i16, 4 }, + { ISD::CTLZ, MVT::v4i64, 46 }, + { ISD::CTLZ, MVT::v8i32, 36 }, + { ISD::CTLZ, MVT::v16i16, 28 }, + { ISD::CTLZ, MVT::v32i8, 18 }, { ISD::CTPOP, MVT::v4i64, 14 }, { ISD::CTPOP, MVT::v8i32, 22 }, { ISD::CTPOP, MVT::v16i16, 18 }, - { ISD::CTPOP, MVT::v32i8, 12 } + { ISD::CTPOP, MVT::v32i8, 12 }, + { ISD::CTTZ, MVT::v4i64, 20 }, + { ISD::CTTZ, MVT::v8i32, 28 }, + { ISD::CTTZ, MVT::v16i16, 24 }, + { ISD::CTTZ, MVT::v32i8, 18 }, }; static const CostTblEntry SSSE3CostTbl[] = { { ISD::BITREVERSE, MVT::v2i64, 5 }, @@ -997,19 +1015,32 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy, { ISD::BSWAP, MVT::v2i64, 1 }, { ISD::BSWAP, MVT::v4i32, 1 }, { ISD::BSWAP, MVT::v8i16, 1 }, + { ISD::CTLZ, MVT::v2i64, 23 }, + { ISD::CTLZ, MVT::v4i32, 18 }, + { ISD::CTLZ, MVT::v8i16, 14 }, + { ISD::CTLZ, MVT::v16i8, 9 }, { ISD::CTPOP, MVT::v2i64, 7 }, { ISD::CTPOP, MVT::v4i32, 11 }, { ISD::CTPOP, MVT::v8i16, 9 }, - { ISD::CTPOP, MVT::v16i8, 6 } + { ISD::CTPOP, MVT::v16i8, 6 }, + { ISD::CTTZ, MVT::v2i64, 10 }, + { ISD::CTTZ, MVT::v4i32, 14 }, + { ISD::CTTZ, MVT::v8i16, 12 }, + { ISD::CTTZ, MVT::v16i8, 9 } }; static const CostTblEntry SSE2CostTbl[] = { { ISD::BSWAP, MVT::v2i64, 7 }, { ISD::BSWAP, MVT::v4i32, 7 }, { ISD::BSWAP, MVT::v8i16, 7 }, + /* ISD::CTLZ - currently scalarized pre-SSSE3 */ { ISD::CTPOP, MVT::v2i64, 12 }, { ISD::CTPOP, MVT::v4i32, 15 }, { ISD::CTPOP, MVT::v8i16, 13 }, - { ISD::CTPOP, MVT::v16i8, 10 } + { ISD::CTPOP, MVT::v16i8, 10 }, + { ISD::CTTZ, MVT::v2i64, 14 }, + { ISD::CTTZ, MVT::v4i32, 18 }, + { ISD::CTTZ, MVT::v8i16, 16 }, + { ISD::CTTZ, MVT::v16i8, 13 } }; unsigned ISD = ISD::DELETED_NODE; @@ -1022,9 +1053,15 @@ int X86TTIImpl::getIntrinsicInstrCost(Intrinsic::ID IID, Type *RetTy, case Intrinsic::bswap: ISD = ISD::BSWAP; break; + case Intrinsic::ctlz: + ISD = ISD::CTLZ; + break; case Intrinsic::ctpop: ISD = ISD::CTPOP; break; + case Intrinsic::cttz: + ISD = ISD::CTTZ; + break; } // Legalize the type. diff --git a/llvm/test/Analysis/CostModel/X86/ctbits-cost.ll b/llvm/test/Analysis/CostModel/X86/ctbits-cost.ll index 7ec5de4..b3b9a8c 100644 --- a/llvm/test/Analysis/CostModel/X86/ctbits-cost.ll +++ b/llvm/test/Analysis/CostModel/X86/ctbits-cost.ll @@ -221,144 +221,176 @@ declare <32 x i8> @llvm.ctlz.v32i8(<32 x i8>, i1) define <2 x i64> @var_ctlz_v2i64(<2 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v2i64': -; SSE: Found an estimated cost of 6 for instruction: %ctlz -; AVX: Found an estimated cost of 6 for instruction: %ctlz -; XOP: Found an estimated cost of 6 for instruction: %ctlz +; SSE2: Found an estimated cost of 6 for instruction: %ctlz +; SSE42: Found an estimated cost of 23 for instruction: %ctlz +; AVX: Found an estimated cost of 23 for instruction: %ctlz +; XOP: Found an estimated cost of 23 for instruction: %ctlz %ctlz = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %a, i1 0) ret <2 x i64> %ctlz } define <2 x i64> @var_ctlz_v2i64u(<2 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v2i64u': -; SSE: Found an estimated cost of 6 for instruction: %ctlz -; AVX: Found an estimated cost of 6 for instruction: %ctlz -; XOP: Found an estimated cost of 6 for instruction: %ctlz +; SSE2: Found an estimated cost of 6 for instruction: %ctlz +; SSE42: Found an estimated cost of 23 for instruction: %ctlz +; AVX: Found an estimated cost of 23 for instruction: %ctlz +; XOP: Found an estimated cost of 23 for instruction: %ctlz %ctlz = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %a, i1 1) ret <2 x i64> %ctlz } define <4 x i64> @var_ctlz_v4i64(<4 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v4i64': -; SSE: Found an estimated cost of 12 for instruction: %ctlz -; AVX: Found an estimated cost of 12 for instruction: %ctlz -; XOP: Found an estimated cost of 12 for instruction: %ctlz +; SSE2: Found an estimated cost of 12 for instruction: %ctlz +; SSE42: Found an estimated cost of 46 for instruction: %ctlz +; AVX1: Found an estimated cost of 46 for instruction: %ctlz +; AVX2: Found an estimated cost of 23 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 46 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 23 for instruction: %ctlz %ctlz = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %a, i1 0) ret <4 x i64> %ctlz } define <4 x i64> @var_ctlz_v4i64u(<4 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v4i64u': -; SSE: Found an estimated cost of 12 for instruction: %ctlz -; AVX: Found an estimated cost of 12 for instruction: %ctlz -; XOP: Found an estimated cost of 12 for instruction: %ctlz +; SSE2: Found an estimated cost of 12 for instruction: %ctlz +; SSE42: Found an estimated cost of 46 for instruction: %ctlz +; AVX1: Found an estimated cost of 46 for instruction: %ctlz +; AVX2: Found an estimated cost of 23 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 46 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 23 for instruction: %ctlz %ctlz = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %a, i1 1) ret <4 x i64> %ctlz } define <4 x i32> @var_ctlz_v4i32(<4 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v4i32': -; SSE: Found an estimated cost of 12 for instruction: %ctlz -; AVX: Found an estimated cost of 12 for instruction: %ctlz -; XOP: Found an estimated cost of 12 for instruction: %ctlz +; SSE2: Found an estimated cost of 12 for instruction: %ctlz +; SSE42: Found an estimated cost of 18 for instruction: %ctlz +; AVX: Found an estimated cost of 18 for instruction: %ctlz +; XOP: Found an estimated cost of 18 for instruction: %ctlz %ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %a, i1 0) ret <4 x i32> %ctlz } define <4 x i32> @var_ctlz_v4i32u(<4 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v4i32u': -; SSE: Found an estimated cost of 12 for instruction: %ctlz -; AVX: Found an estimated cost of 12 for instruction: %ctlz -; XOP: Found an estimated cost of 12 for instruction: %ctlz +; SSE2: Found an estimated cost of 12 for instruction: %ctlz +; SSE42: Found an estimated cost of 18 for instruction: %ctlz +; AVX: Found an estimated cost of 18 for instruction: %ctlz +; XOP: Found an estimated cost of 18 for instruction: %ctlz %ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %a, i1 1) ret <4 x i32> %ctlz } define <8 x i32> @var_ctlz_v8i32(<8 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v8i32': -; SSE: Found an estimated cost of 24 for instruction: %ctlz -; AVX: Found an estimated cost of 24 for instruction: %ctlz -; XOP: Found an estimated cost of 24 for instruction: %ctlz +; SSE2: Found an estimated cost of 24 for instruction: %ctlz +; SSE42: Found an estimated cost of 36 for instruction: %ctlz +; AVX1: Found an estimated cost of 36 for instruction: %ctlz +; AVX2: Found an estimated cost of 18 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 36 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 18 for instruction: %ctlz %ctlz = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %a, i1 0) ret <8 x i32> %ctlz } define <8 x i32> @var_ctlz_v8i32u(<8 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v8i32u': -; SSE: Found an estimated cost of 24 for instruction: %ctlz -; AVX: Found an estimated cost of 24 for instruction: %ctlz -; XOP: Found an estimated cost of 24 for instruction: %ctlz +; SSE2: Found an estimated cost of 24 for instruction: %ctlz +; SSE42: Found an estimated cost of 36 for instruction: %ctlz +; AVX1: Found an estimated cost of 36 for instruction: %ctlz +; AVX2: Found an estimated cost of 18 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 36 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 18 for instruction: %ctlz %ctlz = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %a, i1 1) ret <8 x i32> %ctlz } define <8 x i16> @var_ctlz_v8i16(<8 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v8i16': -; SSE: Found an estimated cost of 24 for instruction: %ctlz -; AVX: Found an estimated cost of 24 for instruction: %ctlz -; XOP: Found an estimated cost of 24 for instruction: %ctlz +; SSE2: Found an estimated cost of 24 for instruction: %ctlz +; SSE42: Found an estimated cost of 14 for instruction: %ctlz +; AVX: Found an estimated cost of 14 for instruction: %ctlz +; XOP: Found an estimated cost of 14 for instruction: %ctlz %ctlz = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %a, i1 0) ret <8 x i16> %ctlz } define <8 x i16> @var_ctlz_v8i16u(<8 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v8i16u': -; SSE: Found an estimated cost of 24 for instruction: %ctlz -; AVX: Found an estimated cost of 24 for instruction: %ctlz -; XOP: Found an estimated cost of 24 for instruction: %ctlz +; SSE2: Found an estimated cost of 24 for instruction: %ctlz +; SSE42: Found an estimated cost of 14 for instruction: %ctlz +; AVX: Found an estimated cost of 14 for instruction: %ctlz +; XOP: Found an estimated cost of 14 for instruction: %ctlz %ctlz = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %a, i1 1) ret <8 x i16> %ctlz } define <16 x i16> @var_ctlz_v16i16(<16 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v16i16': -; SSE: Found an estimated cost of 48 for instruction: %ctlz -; AVX: Found an estimated cost of 48 for instruction: %ctlz -; XOP: Found an estimated cost of 48 for instruction: %ctlz +; SSE2: Found an estimated cost of 48 for instruction: %ctlz +; SSE42: Found an estimated cost of 28 for instruction: %ctlz +; AVX1: Found an estimated cost of 28 for instruction: %ctlz +; AVX2: Found an estimated cost of 14 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 28 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 14 for instruction: %ctlz %ctlz = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %a, i1 0) ret <16 x i16> %ctlz } define <16 x i16> @var_ctlz_v16i16u(<16 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v16i16u': -; SSE: Found an estimated cost of 48 for instruction: %ctlz -; AVX: Found an estimated cost of 48 for instruction: %ctlz -; XOP: Found an estimated cost of 48 for instruction: %ctlz +; SSE2: Found an estimated cost of 48 for instruction: %ctlz +; SSE42: Found an estimated cost of 28 for instruction: %ctlz +; AVX1: Found an estimated cost of 28 for instruction: %ctlz +; AVX2: Found an estimated cost of 14 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 28 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 14 for instruction: %ctlz %ctlz = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %a, i1 1) ret <16 x i16> %ctlz } define <16 x i8> @var_ctlz_v16i8(<16 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v16i8': -; SSE: Found an estimated cost of 48 for instruction: %ctlz -; AVX: Found an estimated cost of 48 for instruction: %ctlz -; XOP: Found an estimated cost of 48 for instruction: %ctlz +; SSE2: Found an estimated cost of 48 for instruction: %ctlz +; SSE42: Found an estimated cost of 9 for instruction: %ctlz +; AVX: Found an estimated cost of 9 for instruction: %ctlz +; XOP: Found an estimated cost of 9 for instruction: %ctlz %ctlz = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %a, i1 0) ret <16 x i8> %ctlz } define <16 x i8> @var_ctlz_v16i8u(<16 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v16i8u': -; SSE: Found an estimated cost of 48 for instruction: %ctlz -; AVX: Found an estimated cost of 48 for instruction: %ctlz -; XOP: Found an estimated cost of 48 for instruction: %ctlz +; SSE2: Found an estimated cost of 48 for instruction: %ctlz +; SSE42: Found an estimated cost of 9 for instruction: %ctlz +; AVX: Found an estimated cost of 9 for instruction: %ctlz +; XOP: Found an estimated cost of 9 for instruction: %ctlz %ctlz = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %a, i1 1) ret <16 x i8> %ctlz } define <32 x i8> @var_ctlz_v32i8(<32 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v32i8': -; SSE: Found an estimated cost of 96 for instruction: %ctlz -; AVX: Found an estimated cost of 96 for instruction: %ctlz -; XOP: Found an estimated cost of 96 for instruction: %ctlz +; SSE2: Found an estimated cost of 96 for instruction: %ctlz +; SSE42: Found an estimated cost of 18 for instruction: %ctlz +; AVX1: Found an estimated cost of 18 for instruction: %ctlz +; AVX2: Found an estimated cost of 9 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 18 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 9 for instruction: %ctlz %ctlz = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> %a, i1 0) ret <32 x i8> %ctlz } define <32 x i8> @var_ctlz_v32i8u(<32 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_ctlz_v32i8u': -; SSE: Found an estimated cost of 96 for instruction: %ctlz -; AVX: Found an estimated cost of 96 for instruction: %ctlz -; XOP: Found an estimated cost of 96 for instruction: %ctlz +; SSE2: Found an estimated cost of 96 for instruction: %ctlz +; SSE42: Found an estimated cost of 18 for instruction: %ctlz +; AVX1: Found an estimated cost of 18 for instruction: %ctlz +; AVX2: Found an estimated cost of 9 for instruction: %ctlz +; XOPAVX1: Found an estimated cost of 18 for instruction: %ctlz +; XOPAVX2: Found an estimated cost of 9 for instruction: %ctlz %ctlz = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> %a, i1 1) ret <32 x i8> %ctlz } @@ -440,144 +472,176 @@ declare <32 x i8> @llvm.cttz.v32i8(<32 x i8>, i1) define <2 x i64> @var_cttz_v2i64(<2 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v2i64': -; SSE: Found an estimated cost of 6 for instruction: %cttz -; AVX: Found an estimated cost of 6 for instruction: %cttz -; XOP: Found an estimated cost of 6 for instruction: %cttz +; SSE2: Found an estimated cost of 14 for instruction: %cttz +; SSE42: Found an estimated cost of 10 for instruction: %cttz +; AVX: Found an estimated cost of 10 for instruction: %cttz +; XOP: Found an estimated cost of 10 for instruction: %cttz %cttz = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %a, i1 0) ret <2 x i64> %cttz } define <2 x i64> @var_cttz_v2i64u(<2 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v2i64u': -; SSE: Found an estimated cost of 6 for instruction: %cttz -; AVX: Found an estimated cost of 6 for instruction: %cttz -; XOP: Found an estimated cost of 6 for instruction: %cttz +; SSE2: Found an estimated cost of 14 for instruction: %cttz +; SSE42: Found an estimated cost of 10 for instruction: %cttz +; AVX: Found an estimated cost of 10 for instruction: %cttz +; XOP: Found an estimated cost of 10 for instruction: %cttz %cttz = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %a, i1 1) ret <2 x i64> %cttz } define <4 x i64> @var_cttz_v4i64(<4 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v4i64': -; SSE: Found an estimated cost of 12 for instruction: %cttz -; AVX: Found an estimated cost of 12 for instruction: %cttz -; XOP: Found an estimated cost of 12 for instruction: %cttz +; SSE2: Found an estimated cost of 28 for instruction: %cttz +; SSE42: Found an estimated cost of 20 for instruction: %cttz +; AVX1: Found an estimated cost of 20 for instruction: %cttz +; AVX2: Found an estimated cost of 10 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 20 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 10 for instruction: %cttz %cttz = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> %a, i1 0) ret <4 x i64> %cttz } define <4 x i64> @var_cttz_v4i64u(<4 x i64> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v4i64u': -; SSE: Found an estimated cost of 12 for instruction: %cttz -; AVX: Found an estimated cost of 12 for instruction: %cttz -; XOP: Found an estimated cost of 12 for instruction: %cttz +; SSE2: Found an estimated cost of 28 for instruction: %cttz +; SSE42: Found an estimated cost of 20 for instruction: %cttz +; AVX1: Found an estimated cost of 20 for instruction: %cttz +; AVX2: Found an estimated cost of 10 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 20 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 10 for instruction: %cttz %cttz = call <4 x i64> @llvm.cttz.v4i64(<4 x i64> %a, i1 1) ret <4 x i64> %cttz } define <4 x i32> @var_cttz_v4i32(<4 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v4i32': -; SSE: Found an estimated cost of 12 for instruction: %cttz -; AVX: Found an estimated cost of 12 for instruction: %cttz -; XOP: Found an estimated cost of 12 for instruction: %cttz +; SSE2: Found an estimated cost of 18 for instruction: %cttz +; SSE42: Found an estimated cost of 14 for instruction: %cttz +; AVX: Found an estimated cost of 14 for instruction: %cttz +; XOP: Found an estimated cost of 14 for instruction: %cttz %cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %a, i1 0) ret <4 x i32> %cttz } define <4 x i32> @var_cttz_v4i32u(<4 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v4i32u': -; SSE: Found an estimated cost of 12 for instruction: %cttz -; AVX: Found an estimated cost of 12 for instruction: %cttz -; XOP: Found an estimated cost of 12 for instruction: %cttz +; SSE2: Found an estimated cost of 18 for instruction: %cttz +; SSE42: Found an estimated cost of 14 for instruction: %cttz +; AVX: Found an estimated cost of 14 for instruction: %cttz +; XOP: Found an estimated cost of 14 for instruction: %cttz %cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %a, i1 1) ret <4 x i32> %cttz } define <8 x i32> @var_cttz_v8i32(<8 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v8i32': -; SSE: Found an estimated cost of 24 for instruction: %cttz -; AVX: Found an estimated cost of 24 for instruction: %cttz -; XOP: Found an estimated cost of 24 for instruction: %cttz +; SSE2: Found an estimated cost of 36 for instruction: %cttz +; SSE42: Found an estimated cost of 28 for instruction: %cttz +; AVX1: Found an estimated cost of 28 for instruction: %cttz +; AVX2: Found an estimated cost of 14 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 28 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 14 for instruction: %cttz %cttz = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> %a, i1 0) ret <8 x i32> %cttz } define <8 x i32> @var_cttz_v8i32u(<8 x i32> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v8i32u': -; SSE: Found an estimated cost of 24 for instruction: %cttz -; AVX: Found an estimated cost of 24 for instruction: %cttz -; XOP: Found an estimated cost of 24 for instruction: %cttz +; SSE2: Found an estimated cost of 36 for instruction: %cttz +; SSE42: Found an estimated cost of 28 for instruction: %cttz +; AVX1: Found an estimated cost of 28 for instruction: %cttz +; AVX2: Found an estimated cost of 14 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 28 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 14 for instruction: %cttz %cttz = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> %a, i1 1) ret <8 x i32> %cttz } define <8 x i16> @var_cttz_v8i16(<8 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v8i16': -; SSE: Found an estimated cost of 24 for instruction: %cttz -; AVX: Found an estimated cost of 24 for instruction: %cttz -; XOP: Found an estimated cost of 24 for instruction: %cttz +; SSE2: Found an estimated cost of 16 for instruction: %cttz +; SSE42: Found an estimated cost of 12 for instruction: %cttz +; AVX: Found an estimated cost of 12 for instruction: %cttz +; XOP: Found an estimated cost of 12 for instruction: %cttz %cttz = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %a, i1 0) ret <8 x i16> %cttz } define <8 x i16> @var_cttz_v8i16u(<8 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v8i16u': -; SSE: Found an estimated cost of 24 for instruction: %cttz -; AVX: Found an estimated cost of 24 for instruction: %cttz -; XOP: Found an estimated cost of 24 for instruction: %cttz +; SSE2: Found an estimated cost of 16 for instruction: %cttz +; SSE42: Found an estimated cost of 12 for instruction: %cttz +; AVX: Found an estimated cost of 12 for instruction: %cttz +; XOP: Found an estimated cost of 12 for instruction: %cttz %cttz = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %a, i1 1) ret <8 x i16> %cttz } define <16 x i16> @var_cttz_v16i16(<16 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v16i16': -; SSE: Found an estimated cost of 48 for instruction: %cttz -; AVX: Found an estimated cost of 48 for instruction: %cttz -; XOP: Found an estimated cost of 48 for instruction: %cttz +; SSE2: Found an estimated cost of 32 for instruction: %cttz +; SSE42: Found an estimated cost of 24 for instruction: %cttz +; AVX1: Found an estimated cost of 24 for instruction: %cttz +; AVX2: Found an estimated cost of 12 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 24 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 12 for instruction: %cttz %cttz = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> %a, i1 0) ret <16 x i16> %cttz } define <16 x i16> @var_cttz_v16i16u(<16 x i16> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v16i16u': -; SSE: Found an estimated cost of 48 for instruction: %cttz -; AVX: Found an estimated cost of 48 for instruction: %cttz -; XOP: Found an estimated cost of 48 for instruction: %cttz +; SSE2: Found an estimated cost of 32 for instruction: %cttz +; SSE42: Found an estimated cost of 24 for instruction: %cttz +; AVX1: Found an estimated cost of 24 for instruction: %cttz +; AVX2: Found an estimated cost of 12 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 24 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 12 for instruction: %cttz %cttz = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> %a, i1 1) ret <16 x i16> %cttz } define <16 x i8> @var_cttz_v16i8(<16 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v16i8': -; SSE: Found an estimated cost of 48 for instruction: %cttz -; AVX: Found an estimated cost of 48 for instruction: %cttz -; XOP: Found an estimated cost of 48 for instruction: %cttz +; SSE2: Found an estimated cost of 13 for instruction: %cttz +; SSE42: Found an estimated cost of 9 for instruction: %cttz +; AVX: Found an estimated cost of 9 for instruction: %cttz +; XOP: Found an estimated cost of 9 for instruction: %cttz %cttz = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %a, i1 0) ret <16 x i8> %cttz } define <16 x i8> @var_cttz_v16i8u(<16 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v16i8u': -; SSE: Found an estimated cost of 48 for instruction: %cttz -; AVX: Found an estimated cost of 48 for instruction: %cttz -; XOP: Found an estimated cost of 48 for instruction: %cttz +; SSE2: Found an estimated cost of 13 for instruction: %cttz +; SSE42: Found an estimated cost of 9 for instruction: %cttz +; AVX: Found an estimated cost of 9 for instruction: %cttz +; XOP: Found an estimated cost of 9 for instruction: %cttz %cttz = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %a, i1 1) ret <16 x i8> %cttz } define <32 x i8> @var_cttz_v32i8(<32 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v32i8': -; SSE: Found an estimated cost of 96 for instruction: %cttz -; AVX: Found an estimated cost of 96 for instruction: %cttz -; XOP: Found an estimated cost of 96 for instruction: %cttz +; SSE2: Found an estimated cost of 26 for instruction: %cttz +; SSE42: Found an estimated cost of 18 for instruction: %cttz +; AVX1: Found an estimated cost of 18 for instruction: %cttz +; AVX2: Found an estimated cost of 9 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 18 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 9 for instruction: %cttz %cttz = call <32 x i8> @llvm.cttz.v32i8(<32 x i8> %a, i1 0) ret <32 x i8> %cttz } define <32 x i8> @var_cttz_v32i8u(<32 x i8> %a) { ; CHECK: 'Cost Model Analysis' for function 'var_cttz_v32i8u': -; SSE: Found an estimated cost of 96 for instruction: %cttz -; AVX: Found an estimated cost of 96 for instruction: %cttz -; XOP: Found an estimated cost of 96 for instruction: %cttz +; SSE2: Found an estimated cost of 26 for instruction: %cttz +; SSE42: Found an estimated cost of 18 for instruction: %cttz +; AVX1: Found an estimated cost of 18 for instruction: %cttz +; AVX2: Found an estimated cost of 9 for instruction: %cttz +; XOPAVX1: Found an estimated cost of 18 for instruction: %cttz +; XOPAVX2: Found an estimated cost of 9 for instruction: %cttz %cttz = call <32 x i8> @llvm.cttz.v32i8(<32 x i8> %a, i1 1) ret <32 x i8> %cttz } diff --git a/llvm/test/Analysis/CostModel/X86/scalarize.ll b/llvm/test/Analysis/CostModel/X86/scalarize.ll index fd4e300..5380887 100644 --- a/llvm/test/Analysis/CostModel/X86/scalarize.ll +++ b/llvm/test/Analysis/CostModel/X86/scalarize.ll @@ -28,11 +28,11 @@ define void @test_scalarized_intrinsics() { ; CHECK64: cost of 1 {{.*}}bswap.v2i64 %r3 = call %i8 @llvm.bswap.v2i64(%i8 undef) -; CHECK32: cost of 12 {{.*}}cttz.v4i32 -; CHECK64: cost of 12 {{.*}}cttz.v4i32 +; CHECK32: cost of 14 {{.*}}cttz.v4i32 +; CHECK64: cost of 14 {{.*}}cttz.v4i32 %r4 = call %i4 @llvm.cttz.v4i32(%i4 undef) ; CHECK32: cost of 10 {{.*}}cttz.v2i64 -; CHECK64: cost of 6 {{.*}}cttz.v2i64 +; CHECK64: cost of 10 {{.*}}cttz.v2i64 %r5 = call %i8 @llvm.cttz.v2i64(%i8 undef) ; CHECK32: ret diff --git a/llvm/test/Transforms/SLPVectorizer/X86/ctlz.ll b/llvm/test/Transforms/SLPVectorizer/X86/ctlz.ll index 8e28197..f9bca44 100644 --- a/llvm/test/Transforms/SLPVectorizer/X86/ctlz.ll +++ b/llvm/test/Transforms/SLPVectorizer/X86/ctlz.ll @@ -1,7 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py -; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX +; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE2 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE42 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" @@ -105,32 +106,65 @@ define void @ctlz_4i32() #0 { } define void @ctlz_8i32() #0 { -; CHECK-LABEL: @ctlz_8i32( -; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false) -; CHECK-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 -; CHECK-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 -; CHECK-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 -; CHECK-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 -; CHECK-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 -; CHECK-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 -; CHECK-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 -; CHECK-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 -; CHECK-NEXT: ret void +; SSE-LABEL: @ctlz_8i32( +; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; SSE-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false) +; SSE-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false) +; SSE-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false) +; SSE-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false) +; SSE-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false) +; SSE-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false) +; SSE-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false) +; SSE-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false) +; SSE-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; SSE-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; SSE-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; SSE-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; SSE-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; SSE-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; SSE-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; SSE-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; SSE-NEXT: ret void +; +; AVX1-LABEL: @ctlz_8i32( +; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; AVX1-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false) +; AVX1-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false) +; AVX1-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false) +; AVX1-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false) +; AVX1-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false) +; AVX1-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false) +; AVX1-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false) +; AVX1-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false) +; AVX1-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; AVX1-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; AVX1-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; AVX1-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; AVX1-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; AVX1-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; AVX1-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; AVX1-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; AVX1-NEXT: ret void +; +; AVX2-LABEL: @ctlz_8i32( +; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2 +; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> [[TMP1]], i1 false) +; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2 +; AVX2-NEXT: ret void ; %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 @@ -160,32 +194,44 @@ define void @ctlz_8i32() #0 { } define void @ctlz_8i16() #0 { -; CHECK-LABEL: @ctlz_8i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 false) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 false) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 false) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 false) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 false) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 false) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 false) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 false) -; CHECK-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_8i16( +; SSE2-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 +; SSE2-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 +; SSE2-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 +; SSE2-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 +; SSE2-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 +; SSE2-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 +; SSE2-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 +; SSE2-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 false) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 false) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 false) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 false) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 false) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 false) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 false) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 false) +; SSE2-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 +; SSE2-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 +; SSE2-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 +; SSE2-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 +; SSE2-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 +; SSE2-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 +; SSE2-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 +; SSE2-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_8i16( +; SSE42-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; SSE42-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false) +; SSE42-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_8i16( +; AVX-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; AVX-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false) +; AVX-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; AVX-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 @@ -215,56 +261,71 @@ define void @ctlz_8i16() #0 { } define void @ctlz_16i16() #0 { -; CHECK-LABEL: @ctlz_16i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 -; CHECK-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 -; CHECK-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 -; CHECK-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 -; CHECK-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 -; CHECK-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 -; CHECK-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 -; CHECK-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 false) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 false) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 false) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 false) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 false) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 false) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 false) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 false) -; CHECK-NEXT: [[CTLZ8:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD8]], i1 false) -; CHECK-NEXT: [[CTLZ9:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD9]], i1 false) -; CHECK-NEXT: [[CTLZ10:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD10]], i1 false) -; CHECK-NEXT: [[CTLZ11:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD11]], i1 false) -; CHECK-NEXT: [[CTLZ12:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD12]], i1 false) -; CHECK-NEXT: [[CTLZ13:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD13]], i1 false) -; CHECK-NEXT: [[CTLZ14:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD14]], i1 false) -; CHECK-NEXT: [[CTLZ15:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD15]], i1 false) -; CHECK-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 -; CHECK-NEXT: store i16 [[CTLZ8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 -; CHECK-NEXT: store i16 [[CTLZ9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 -; CHECK-NEXT: store i16 [[CTLZ10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 -; CHECK-NEXT: store i16 [[CTLZ11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 -; CHECK-NEXT: store i16 [[CTLZ12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 -; CHECK-NEXT: store i16 [[CTLZ13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 -; CHECK-NEXT: store i16 [[CTLZ14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 -; CHECK-NEXT: store i16 [[CTLZ15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_16i16( +; SSE2-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 +; SSE2-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 +; SSE2-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 +; SSE2-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 +; SSE2-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 +; SSE2-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 +; SSE2-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 +; SSE2-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 +; SSE2-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 +; SSE2-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 +; SSE2-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 +; SSE2-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 +; SSE2-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 +; SSE2-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 +; SSE2-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 +; SSE2-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 false) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 false) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 false) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 false) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 false) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 false) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 false) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 false) +; SSE2-NEXT: [[CTLZ8:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD8]], i1 false) +; SSE2-NEXT: [[CTLZ9:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD9]], i1 false) +; SSE2-NEXT: [[CTLZ10:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD10]], i1 false) +; SSE2-NEXT: [[CTLZ11:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD11]], i1 false) +; SSE2-NEXT: [[CTLZ12:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD12]], i1 false) +; SSE2-NEXT: [[CTLZ13:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD13]], i1 false) +; SSE2-NEXT: [[CTLZ14:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD14]], i1 false) +; SSE2-NEXT: [[CTLZ15:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD15]], i1 false) +; SSE2-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 +; SSE2-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 +; SSE2-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 +; SSE2-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 +; SSE2-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 +; SSE2-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 +; SSE2-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 +; SSE2-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 +; SSE2-NEXT: store i16 [[CTLZ8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 +; SSE2-NEXT: store i16 [[CTLZ9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 +; SSE2-NEXT: store i16 [[CTLZ10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 +; SSE2-NEXT: store i16 [[CTLZ11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 +; SSE2-NEXT: store i16 [[CTLZ12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 +; SSE2-NEXT: store i16 [[CTLZ13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 +; SSE2-NEXT: store i16 [[CTLZ14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 +; SSE2-NEXT: store i16 [[CTLZ15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_16i16( +; SSE42-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; SSE42-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE42-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false) +; SSE42-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP2]], i1 false) +; SSE42-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; SSE42-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_16i16( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2 +; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> [[TMP1]], i1 false) +; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2 +; AVX-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 @@ -318,56 +379,68 @@ define void @ctlz_16i16() #0 { } define void @ctlz_16i8() #0 { -; CHECK-LABEL: @ctlz_16i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 false) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 false) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 false) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 false) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 false) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 false) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 false) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 false) -; CHECK-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 false) -; CHECK-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 false) -; CHECK-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 false) -; CHECK-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 false) -; CHECK-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 false) -; CHECK-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 false) -; CHECK-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 false) -; CHECK-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 false) -; CHECK-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_16i8( +; SSE2-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 +; SSE2-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 +; SSE2-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 +; SSE2-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 +; SSE2-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 +; SSE2-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 +; SSE2-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 +; SSE2-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 +; SSE2-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 +; SSE2-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 +; SSE2-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 +; SSE2-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 +; SSE2-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 +; SSE2-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 +; SSE2-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 +; SSE2-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 false) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 false) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 false) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 false) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 false) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 false) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 false) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 false) +; SSE2-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 false) +; SSE2-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 false) +; SSE2-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 false) +; SSE2-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 false) +; SSE2-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 false) +; SSE2-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 false) +; SSE2-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 false) +; SSE2-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 false) +; SSE2-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 +; SSE2-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 +; SSE2-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 +; SSE2-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 +; SSE2-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 +; SSE2-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 +; SSE2-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 +; SSE2-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 +; SSE2-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 +; SSE2-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 +; SSE2-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 +; SSE2-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 +; SSE2-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 +; SSE2-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 +; SSE2-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 +; SSE2-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_16i8( +; SSE42-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; SSE42-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false) +; SSE42-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_16i8( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; AVX-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false) +; AVX-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; AVX-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 @@ -421,104 +494,122 @@ define void @ctlz_16i8() #0 { } define void @ctlz_32i8() #0 { -; CHECK-LABEL: @ctlz_32i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 -; CHECK-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 -; CHECK-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 -; CHECK-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 -; CHECK-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 -; CHECK-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 -; CHECK-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 -; CHECK-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 -; CHECK-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 -; CHECK-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 -; CHECK-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 -; CHECK-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 -; CHECK-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 -; CHECK-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 -; CHECK-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 -; CHECK-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 false) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 false) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 false) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 false) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 false) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 false) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 false) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 false) -; CHECK-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 false) -; CHECK-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 false) -; CHECK-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 false) -; CHECK-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 false) -; CHECK-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 false) -; CHECK-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 false) -; CHECK-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 false) -; CHECK-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 false) -; CHECK-NEXT: [[CTLZ16:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD16]], i1 false) -; CHECK-NEXT: [[CTLZ17:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD17]], i1 false) -; CHECK-NEXT: [[CTLZ18:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD18]], i1 false) -; CHECK-NEXT: [[CTLZ19:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD19]], i1 false) -; CHECK-NEXT: [[CTLZ20:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD20]], i1 false) -; CHECK-NEXT: [[CTLZ21:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD21]], i1 false) -; CHECK-NEXT: [[CTLZ22:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD22]], i1 false) -; CHECK-NEXT: [[CTLZ23:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD23]], i1 false) -; CHECK-NEXT: [[CTLZ24:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD24]], i1 false) -; CHECK-NEXT: [[CTLZ25:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD25]], i1 false) -; CHECK-NEXT: [[CTLZ26:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD26]], i1 false) -; CHECK-NEXT: [[CTLZ27:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD27]], i1 false) -; CHECK-NEXT: [[CTLZ28:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD28]], i1 false) -; CHECK-NEXT: [[CTLZ29:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD29]], i1 false) -; CHECK-NEXT: [[CTLZ30:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD30]], i1 false) -; CHECK-NEXT: [[CTLZ31:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD31]], i1 false) -; CHECK-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 -; CHECK-NEXT: store i8 [[CTLZ16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 -; CHECK-NEXT: store i8 [[CTLZ17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 -; CHECK-NEXT: store i8 [[CTLZ18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 -; CHECK-NEXT: store i8 [[CTLZ19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 -; CHECK-NEXT: store i8 [[CTLZ20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 -; CHECK-NEXT: store i8 [[CTLZ21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 -; CHECK-NEXT: store i8 [[CTLZ22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 -; CHECK-NEXT: store i8 [[CTLZ23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 -; CHECK-NEXT: store i8 [[CTLZ24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 -; CHECK-NEXT: store i8 [[CTLZ25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 -; CHECK-NEXT: store i8 [[CTLZ26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 -; CHECK-NEXT: store i8 [[CTLZ27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 -; CHECK-NEXT: store i8 [[CTLZ28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 -; CHECK-NEXT: store i8 [[CTLZ29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 -; CHECK-NEXT: store i8 [[CTLZ30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 -; CHECK-NEXT: store i8 [[CTLZ31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_32i8( +; SSE2-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 +; SSE2-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 +; SSE2-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 +; SSE2-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 +; SSE2-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 +; SSE2-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 +; SSE2-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 +; SSE2-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 +; SSE2-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 +; SSE2-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 +; SSE2-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 +; SSE2-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 +; SSE2-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 +; SSE2-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 +; SSE2-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 +; SSE2-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 +; SSE2-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 +; SSE2-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 +; SSE2-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 +; SSE2-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 +; SSE2-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 +; SSE2-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 +; SSE2-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 +; SSE2-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 +; SSE2-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 +; SSE2-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 +; SSE2-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 +; SSE2-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 +; SSE2-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 +; SSE2-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 +; SSE2-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 +; SSE2-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 false) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 false) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 false) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 false) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 false) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 false) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 false) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 false) +; SSE2-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 false) +; SSE2-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 false) +; SSE2-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 false) +; SSE2-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 false) +; SSE2-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 false) +; SSE2-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 false) +; SSE2-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 false) +; SSE2-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 false) +; SSE2-NEXT: [[CTLZ16:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD16]], i1 false) +; SSE2-NEXT: [[CTLZ17:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD17]], i1 false) +; SSE2-NEXT: [[CTLZ18:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD18]], i1 false) +; SSE2-NEXT: [[CTLZ19:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD19]], i1 false) +; SSE2-NEXT: [[CTLZ20:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD20]], i1 false) +; SSE2-NEXT: [[CTLZ21:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD21]], i1 false) +; SSE2-NEXT: [[CTLZ22:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD22]], i1 false) +; SSE2-NEXT: [[CTLZ23:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD23]], i1 false) +; SSE2-NEXT: [[CTLZ24:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD24]], i1 false) +; SSE2-NEXT: [[CTLZ25:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD25]], i1 false) +; SSE2-NEXT: [[CTLZ26:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD26]], i1 false) +; SSE2-NEXT: [[CTLZ27:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD27]], i1 false) +; SSE2-NEXT: [[CTLZ28:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD28]], i1 false) +; SSE2-NEXT: [[CTLZ29:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD29]], i1 false) +; SSE2-NEXT: [[CTLZ30:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD30]], i1 false) +; SSE2-NEXT: [[CTLZ31:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD31]], i1 false) +; SSE2-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 +; SSE2-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 +; SSE2-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 +; SSE2-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 +; SSE2-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 +; SSE2-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 +; SSE2-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 +; SSE2-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 +; SSE2-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 +; SSE2-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 +; SSE2-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 +; SSE2-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 +; SSE2-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 +; SSE2-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 +; SSE2-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 +; SSE2-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 +; SSE2-NEXT: store i8 [[CTLZ16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 +; SSE2-NEXT: store i8 [[CTLZ17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 +; SSE2-NEXT: store i8 [[CTLZ18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 +; SSE2-NEXT: store i8 [[CTLZ19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 +; SSE2-NEXT: store i8 [[CTLZ20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 +; SSE2-NEXT: store i8 [[CTLZ21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 +; SSE2-NEXT: store i8 [[CTLZ22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 +; SSE2-NEXT: store i8 [[CTLZ23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 +; SSE2-NEXT: store i8 [[CTLZ24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 +; SSE2-NEXT: store i8 [[CTLZ25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 +; SSE2-NEXT: store i8 [[CTLZ26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 +; SSE2-NEXT: store i8 [[CTLZ27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 +; SSE2-NEXT: store i8 [[CTLZ28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 +; SSE2-NEXT: store i8 [[CTLZ29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 +; SSE2-NEXT: store i8 [[CTLZ30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 +; SSE2-NEXT: store i8 [[CTLZ31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_32i8( +; SSE42-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; SSE42-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 +; SSE42-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false) +; SSE42-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 false) +; SSE42-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; SSE42-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_32i8( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; AVX-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 +; AVX-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false) +; AVX-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 false) +; AVX-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; AVX-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 +; AVX-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 @@ -705,32 +796,65 @@ define void @ctlz_undef_4i32() #0 { } define void @ctlz_undef_8i32() #0 { -; CHECK-LABEL: @ctlz_undef_8i32( -; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true) -; CHECK-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 -; CHECK-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 -; CHECK-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 -; CHECK-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 -; CHECK-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 -; CHECK-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 -; CHECK-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 -; CHECK-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 -; CHECK-NEXT: ret void +; SSE-LABEL: @ctlz_undef_8i32( +; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; SSE-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true) +; SSE-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true) +; SSE-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true) +; SSE-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true) +; SSE-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true) +; SSE-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true) +; SSE-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true) +; SSE-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true) +; SSE-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; SSE-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; SSE-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; SSE-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; SSE-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; SSE-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; SSE-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; SSE-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; SSE-NEXT: ret void +; +; AVX1-LABEL: @ctlz_undef_8i32( +; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; AVX1-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true) +; AVX1-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true) +; AVX1-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true) +; AVX1-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true) +; AVX1-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true) +; AVX1-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true) +; AVX1-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true) +; AVX1-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true) +; AVX1-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; AVX1-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; AVX1-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; AVX1-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; AVX1-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; AVX1-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; AVX1-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; AVX1-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; AVX1-NEXT: ret void +; +; AVX2-LABEL: @ctlz_undef_8i32( +; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2 +; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> [[TMP1]], i1 true) +; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2 +; AVX2-NEXT: ret void ; %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 @@ -760,32 +884,44 @@ define void @ctlz_undef_8i32() #0 { } define void @ctlz_undef_8i16() #0 { -; CHECK-LABEL: @ctlz_undef_8i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 true) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 true) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 true) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 true) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 true) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 true) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 true) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 true) -; CHECK-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_undef_8i16( +; SSE2-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 +; SSE2-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 +; SSE2-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 +; SSE2-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 +; SSE2-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 +; SSE2-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 +; SSE2-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 +; SSE2-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 true) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 true) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 true) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 true) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 true) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 true) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 true) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 true) +; SSE2-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 +; SSE2-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 +; SSE2-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 +; SSE2-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 +; SSE2-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 +; SSE2-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 +; SSE2-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 +; SSE2-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_undef_8i16( +; SSE42-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; SSE42-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true) +; SSE42-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_undef_8i16( +; AVX-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; AVX-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true) +; AVX-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; AVX-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 @@ -815,56 +951,71 @@ define void @ctlz_undef_8i16() #0 { } define void @ctlz_undef_16i16() #0 { -; CHECK-LABEL: @ctlz_undef_16i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 -; CHECK-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 -; CHECK-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 -; CHECK-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 -; CHECK-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 -; CHECK-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 -; CHECK-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 -; CHECK-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 true) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 true) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 true) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 true) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 true) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 true) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 true) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 true) -; CHECK-NEXT: [[CTLZ8:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD8]], i1 true) -; CHECK-NEXT: [[CTLZ9:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD9]], i1 true) -; CHECK-NEXT: [[CTLZ10:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD10]], i1 true) -; CHECK-NEXT: [[CTLZ11:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD11]], i1 true) -; CHECK-NEXT: [[CTLZ12:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD12]], i1 true) -; CHECK-NEXT: [[CTLZ13:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD13]], i1 true) -; CHECK-NEXT: [[CTLZ14:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD14]], i1 true) -; CHECK-NEXT: [[CTLZ15:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD15]], i1 true) -; CHECK-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 -; CHECK-NEXT: store i16 [[CTLZ8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 -; CHECK-NEXT: store i16 [[CTLZ9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 -; CHECK-NEXT: store i16 [[CTLZ10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 -; CHECK-NEXT: store i16 [[CTLZ11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 -; CHECK-NEXT: store i16 [[CTLZ12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 -; CHECK-NEXT: store i16 [[CTLZ13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 -; CHECK-NEXT: store i16 [[CTLZ14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 -; CHECK-NEXT: store i16 [[CTLZ15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_undef_16i16( +; SSE2-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 +; SSE2-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 +; SSE2-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 +; SSE2-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 +; SSE2-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 +; SSE2-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 +; SSE2-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 +; SSE2-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 +; SSE2-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 +; SSE2-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 +; SSE2-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 +; SSE2-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 +; SSE2-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 +; SSE2-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 +; SSE2-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 +; SSE2-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD0]], i1 true) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD1]], i1 true) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD2]], i1 true) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD3]], i1 true) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD4]], i1 true) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD5]], i1 true) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD6]], i1 true) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD7]], i1 true) +; SSE2-NEXT: [[CTLZ8:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD8]], i1 true) +; SSE2-NEXT: [[CTLZ9:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD9]], i1 true) +; SSE2-NEXT: [[CTLZ10:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD10]], i1 true) +; SSE2-NEXT: [[CTLZ11:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD11]], i1 true) +; SSE2-NEXT: [[CTLZ12:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD12]], i1 true) +; SSE2-NEXT: [[CTLZ13:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD13]], i1 true) +; SSE2-NEXT: [[CTLZ14:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD14]], i1 true) +; SSE2-NEXT: [[CTLZ15:%.*]] = call i16 @llvm.ctlz.i16(i16 [[LD15]], i1 true) +; SSE2-NEXT: store i16 [[CTLZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 +; SSE2-NEXT: store i16 [[CTLZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 +; SSE2-NEXT: store i16 [[CTLZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 +; SSE2-NEXT: store i16 [[CTLZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 +; SSE2-NEXT: store i16 [[CTLZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 +; SSE2-NEXT: store i16 [[CTLZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 +; SSE2-NEXT: store i16 [[CTLZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 +; SSE2-NEXT: store i16 [[CTLZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 +; SSE2-NEXT: store i16 [[CTLZ8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 +; SSE2-NEXT: store i16 [[CTLZ9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 +; SSE2-NEXT: store i16 [[CTLZ10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 +; SSE2-NEXT: store i16 [[CTLZ11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 +; SSE2-NEXT: store i16 [[CTLZ12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 +; SSE2-NEXT: store i16 [[CTLZ13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 +; SSE2-NEXT: store i16 [[CTLZ14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 +; SSE2-NEXT: store i16 [[CTLZ15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_undef_16i16( +; SSE42-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; SSE42-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE42-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true) +; SSE42-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP2]], i1 true) +; SSE42-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; SSE42-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_undef_16i16( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2 +; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> [[TMP1]], i1 true) +; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2 +; AVX-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 @@ -918,56 +1069,68 @@ define void @ctlz_undef_16i16() #0 { } define void @ctlz_undef_16i8() #0 { -; CHECK-LABEL: @ctlz_undef_16i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 true) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 true) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 true) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 true) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 true) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 true) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 true) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 true) -; CHECK-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 true) -; CHECK-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 true) -; CHECK-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 true) -; CHECK-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 true) -; CHECK-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 true) -; CHECK-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 true) -; CHECK-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 true) -; CHECK-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 true) -; CHECK-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_undef_16i8( +; SSE2-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 +; SSE2-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 +; SSE2-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 +; SSE2-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 +; SSE2-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 +; SSE2-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 +; SSE2-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 +; SSE2-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 +; SSE2-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 +; SSE2-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 +; SSE2-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 +; SSE2-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 +; SSE2-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 +; SSE2-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 +; SSE2-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 +; SSE2-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 true) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 true) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 true) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 true) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 true) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 true) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 true) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 true) +; SSE2-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 true) +; SSE2-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 true) +; SSE2-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 true) +; SSE2-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 true) +; SSE2-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 true) +; SSE2-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 true) +; SSE2-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 true) +; SSE2-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 true) +; SSE2-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 +; SSE2-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 +; SSE2-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 +; SSE2-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 +; SSE2-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 +; SSE2-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 +; SSE2-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 +; SSE2-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 +; SSE2-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 +; SSE2-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 +; SSE2-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 +; SSE2-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 +; SSE2-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 +; SSE2-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 +; SSE2-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 +; SSE2-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_undef_16i8( +; SSE42-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; SSE42-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true) +; SSE42-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_undef_16i8( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; AVX-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true) +; AVX-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; AVX-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 @@ -1021,104 +1184,122 @@ define void @ctlz_undef_16i8() #0 { } define void @ctlz_undef_32i8() #0 { -; CHECK-LABEL: @ctlz_undef_32i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 -; CHECK-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 -; CHECK-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 -; CHECK-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 -; CHECK-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 -; CHECK-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 -; CHECK-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 -; CHECK-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 -; CHECK-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 -; CHECK-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 -; CHECK-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 -; CHECK-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 -; CHECK-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 -; CHECK-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 -; CHECK-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 -; CHECK-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 -; CHECK-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 true) -; CHECK-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 true) -; CHECK-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 true) -; CHECK-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 true) -; CHECK-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 true) -; CHECK-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 true) -; CHECK-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 true) -; CHECK-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 true) -; CHECK-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 true) -; CHECK-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 true) -; CHECK-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 true) -; CHECK-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 true) -; CHECK-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 true) -; CHECK-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 true) -; CHECK-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 true) -; CHECK-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 true) -; CHECK-NEXT: [[CTLZ16:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD16]], i1 true) -; CHECK-NEXT: [[CTLZ17:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD17]], i1 true) -; CHECK-NEXT: [[CTLZ18:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD18]], i1 true) -; CHECK-NEXT: [[CTLZ19:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD19]], i1 true) -; CHECK-NEXT: [[CTLZ20:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD20]], i1 true) -; CHECK-NEXT: [[CTLZ21:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD21]], i1 true) -; CHECK-NEXT: [[CTLZ22:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD22]], i1 true) -; CHECK-NEXT: [[CTLZ23:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD23]], i1 true) -; CHECK-NEXT: [[CTLZ24:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD24]], i1 true) -; CHECK-NEXT: [[CTLZ25:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD25]], i1 true) -; CHECK-NEXT: [[CTLZ26:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD26]], i1 true) -; CHECK-NEXT: [[CTLZ27:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD27]], i1 true) -; CHECK-NEXT: [[CTLZ28:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD28]], i1 true) -; CHECK-NEXT: [[CTLZ29:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD29]], i1 true) -; CHECK-NEXT: [[CTLZ30:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD30]], i1 true) -; CHECK-NEXT: [[CTLZ31:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD31]], i1 true) -; CHECK-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 -; CHECK-NEXT: store i8 [[CTLZ16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 -; CHECK-NEXT: store i8 [[CTLZ17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 -; CHECK-NEXT: store i8 [[CTLZ18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 -; CHECK-NEXT: store i8 [[CTLZ19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 -; CHECK-NEXT: store i8 [[CTLZ20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 -; CHECK-NEXT: store i8 [[CTLZ21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 -; CHECK-NEXT: store i8 [[CTLZ22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 -; CHECK-NEXT: store i8 [[CTLZ23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 -; CHECK-NEXT: store i8 [[CTLZ24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 -; CHECK-NEXT: store i8 [[CTLZ25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 -; CHECK-NEXT: store i8 [[CTLZ26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 -; CHECK-NEXT: store i8 [[CTLZ27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 -; CHECK-NEXT: store i8 [[CTLZ28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 -; CHECK-NEXT: store i8 [[CTLZ29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 -; CHECK-NEXT: store i8 [[CTLZ30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 -; CHECK-NEXT: store i8 [[CTLZ31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 -; CHECK-NEXT: ret void +; SSE2-LABEL: @ctlz_undef_32i8( +; SSE2-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 +; SSE2-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 +; SSE2-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 +; SSE2-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 +; SSE2-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 +; SSE2-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 +; SSE2-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 +; SSE2-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 +; SSE2-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 +; SSE2-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 +; SSE2-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 +; SSE2-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 +; SSE2-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 +; SSE2-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 +; SSE2-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 +; SSE2-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 +; SSE2-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 +; SSE2-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 +; SSE2-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 +; SSE2-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 +; SSE2-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 +; SSE2-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 +; SSE2-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 +; SSE2-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 +; SSE2-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 +; SSE2-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 +; SSE2-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 +; SSE2-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 +; SSE2-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 +; SSE2-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 +; SSE2-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 +; SSE2-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 +; SSE2-NEXT: [[CTLZ0:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD0]], i1 true) +; SSE2-NEXT: [[CTLZ1:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD1]], i1 true) +; SSE2-NEXT: [[CTLZ2:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD2]], i1 true) +; SSE2-NEXT: [[CTLZ3:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD3]], i1 true) +; SSE2-NEXT: [[CTLZ4:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD4]], i1 true) +; SSE2-NEXT: [[CTLZ5:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD5]], i1 true) +; SSE2-NEXT: [[CTLZ6:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD6]], i1 true) +; SSE2-NEXT: [[CTLZ7:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD7]], i1 true) +; SSE2-NEXT: [[CTLZ8:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD8]], i1 true) +; SSE2-NEXT: [[CTLZ9:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD9]], i1 true) +; SSE2-NEXT: [[CTLZ10:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD10]], i1 true) +; SSE2-NEXT: [[CTLZ11:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD11]], i1 true) +; SSE2-NEXT: [[CTLZ12:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD12]], i1 true) +; SSE2-NEXT: [[CTLZ13:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD13]], i1 true) +; SSE2-NEXT: [[CTLZ14:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD14]], i1 true) +; SSE2-NEXT: [[CTLZ15:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD15]], i1 true) +; SSE2-NEXT: [[CTLZ16:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD16]], i1 true) +; SSE2-NEXT: [[CTLZ17:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD17]], i1 true) +; SSE2-NEXT: [[CTLZ18:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD18]], i1 true) +; SSE2-NEXT: [[CTLZ19:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD19]], i1 true) +; SSE2-NEXT: [[CTLZ20:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD20]], i1 true) +; SSE2-NEXT: [[CTLZ21:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD21]], i1 true) +; SSE2-NEXT: [[CTLZ22:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD22]], i1 true) +; SSE2-NEXT: [[CTLZ23:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD23]], i1 true) +; SSE2-NEXT: [[CTLZ24:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD24]], i1 true) +; SSE2-NEXT: [[CTLZ25:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD25]], i1 true) +; SSE2-NEXT: [[CTLZ26:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD26]], i1 true) +; SSE2-NEXT: [[CTLZ27:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD27]], i1 true) +; SSE2-NEXT: [[CTLZ28:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD28]], i1 true) +; SSE2-NEXT: [[CTLZ29:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD29]], i1 true) +; SSE2-NEXT: [[CTLZ30:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD30]], i1 true) +; SSE2-NEXT: [[CTLZ31:%.*]] = call i8 @llvm.ctlz.i8(i8 [[LD31]], i1 true) +; SSE2-NEXT: store i8 [[CTLZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 +; SSE2-NEXT: store i8 [[CTLZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 +; SSE2-NEXT: store i8 [[CTLZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 +; SSE2-NEXT: store i8 [[CTLZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 +; SSE2-NEXT: store i8 [[CTLZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 +; SSE2-NEXT: store i8 [[CTLZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 +; SSE2-NEXT: store i8 [[CTLZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 +; SSE2-NEXT: store i8 [[CTLZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 +; SSE2-NEXT: store i8 [[CTLZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 +; SSE2-NEXT: store i8 [[CTLZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 +; SSE2-NEXT: store i8 [[CTLZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 +; SSE2-NEXT: store i8 [[CTLZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 +; SSE2-NEXT: store i8 [[CTLZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 +; SSE2-NEXT: store i8 [[CTLZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 +; SSE2-NEXT: store i8 [[CTLZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 +; SSE2-NEXT: store i8 [[CTLZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 +; SSE2-NEXT: store i8 [[CTLZ16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 +; SSE2-NEXT: store i8 [[CTLZ17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 +; SSE2-NEXT: store i8 [[CTLZ18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 +; SSE2-NEXT: store i8 [[CTLZ19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 +; SSE2-NEXT: store i8 [[CTLZ20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 +; SSE2-NEXT: store i8 [[CTLZ21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 +; SSE2-NEXT: store i8 [[CTLZ22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 +; SSE2-NEXT: store i8 [[CTLZ23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 +; SSE2-NEXT: store i8 [[CTLZ24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 +; SSE2-NEXT: store i8 [[CTLZ25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 +; SSE2-NEXT: store i8 [[CTLZ26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 +; SSE2-NEXT: store i8 [[CTLZ27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 +; SSE2-NEXT: store i8 [[CTLZ28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 +; SSE2-NEXT: store i8 [[CTLZ29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 +; SSE2-NEXT: store i8 [[CTLZ30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 +; SSE2-NEXT: store i8 [[CTLZ31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 +; SSE2-NEXT: ret void +; +; SSE42-LABEL: @ctlz_undef_32i8( +; SSE42-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; SSE42-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 +; SSE42-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true) +; SSE42-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 true) +; SSE42-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; SSE42-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 +; SSE42-NEXT: ret void +; +; AVX-LABEL: @ctlz_undef_32i8( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; AVX-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 +; AVX-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true) +; AVX-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 true) +; AVX-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; AVX-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 +; AVX-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 diff --git a/llvm/test/Transforms/SLPVectorizer/X86/cttz.ll b/llvm/test/Transforms/SLPVectorizer/X86/cttz.ll index 1ede065..751fea7 100644 --- a/llvm/test/Transforms/SLPVectorizer/X86/cttz.ll +++ b/llvm/test/Transforms/SLPVectorizer/X86/cttz.ll @@ -1,7 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py -; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX +; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE2 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE42 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" @@ -105,32 +106,65 @@ define void @cttz_4i32() #0 { } define void @cttz_8i32() #0 { -; CHECK-LABEL: @cttz_8i32( -; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 false) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 false) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 false) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 false) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 false) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 false) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 false) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 false) -; CHECK-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 -; CHECK-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 -; CHECK-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 -; CHECK-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 -; CHECK-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 -; CHECK-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 -; CHECK-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 -; CHECK-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 -; CHECK-NEXT: ret void +; SSE-LABEL: @cttz_8i32( +; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; SSE-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 false) +; SSE-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 false) +; SSE-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 false) +; SSE-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 false) +; SSE-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 false) +; SSE-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 false) +; SSE-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 false) +; SSE-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 false) +; SSE-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; SSE-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; SSE-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; SSE-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; SSE-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; SSE-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; SSE-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; SSE-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; SSE-NEXT: ret void +; +; AVX1-LABEL: @cttz_8i32( +; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; AVX1-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 false) +; AVX1-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 false) +; AVX1-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 false) +; AVX1-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 false) +; AVX1-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 false) +; AVX1-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 false) +; AVX1-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 false) +; AVX1-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 false) +; AVX1-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; AVX1-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; AVX1-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; AVX1-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; AVX1-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; AVX1-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; AVX1-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; AVX1-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; AVX1-NEXT: ret void +; +; AVX2-LABEL: @cttz_8i32( +; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2 +; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> [[TMP1]], i1 false) +; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2 +; AVX2-NEXT: ret void ; %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 @@ -161,30 +195,9 @@ define void @cttz_8i32() #0 { define void @cttz_8i16() #0 { ; CHECK-LABEL: @cttz_8i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD0]], i1 false) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD1]], i1 false) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD2]], i1 false) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD3]], i1 false) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD4]], i1 false) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD5]], i1 false) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD6]], i1 false) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD7]], i1 false) -; CHECK-NEXT: store i16 [[CTTZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTTZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTTZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTTZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTTZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTTZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTTZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTTZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 +; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 false) +; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 ; CHECK-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 @@ -215,56 +228,20 @@ define void @cttz_8i16() #0 { } define void @cttz_16i16() #0 { -; CHECK-LABEL: @cttz_16i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 -; CHECK-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 -; CHECK-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 -; CHECK-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 -; CHECK-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 -; CHECK-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 -; CHECK-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 -; CHECK-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD0]], i1 false) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD1]], i1 false) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD2]], i1 false) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD3]], i1 false) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD4]], i1 false) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD5]], i1 false) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD6]], i1 false) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD7]], i1 false) -; CHECK-NEXT: [[CTTZ8:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD8]], i1 false) -; CHECK-NEXT: [[CTTZ9:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD9]], i1 false) -; CHECK-NEXT: [[CTTZ10:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD10]], i1 false) -; CHECK-NEXT: [[CTTZ11:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD11]], i1 false) -; CHECK-NEXT: [[CTTZ12:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD12]], i1 false) -; CHECK-NEXT: [[CTTZ13:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD13]], i1 false) -; CHECK-NEXT: [[CTTZ14:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD14]], i1 false) -; CHECK-NEXT: [[CTTZ15:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD15]], i1 false) -; CHECK-NEXT: store i16 [[CTTZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTTZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTTZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTTZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTTZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTTZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTTZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTTZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 -; CHECK-NEXT: store i16 [[CTTZ8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 -; CHECK-NEXT: store i16 [[CTTZ9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 -; CHECK-NEXT: store i16 [[CTTZ10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 -; CHECK-NEXT: store i16 [[CTTZ11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 -; CHECK-NEXT: store i16 [[CTTZ12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 -; CHECK-NEXT: store i16 [[CTTZ13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 -; CHECK-NEXT: store i16 [[CTTZ14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 -; CHECK-NEXT: store i16 [[CTTZ15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 -; CHECK-NEXT: ret void +; SSE-LABEL: @cttz_16i16( +; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 false) +; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP2]], i1 false) +; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE-NEXT: ret void +; +; AVX-LABEL: @cttz_16i16( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2 +; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> [[TMP1]], i1 false) +; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2 +; AVX-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 @@ -319,54 +296,9 @@ define void @cttz_16i16() #0 { define void @cttz_16i8() #0 { ; CHECK-LABEL: @cttz_16i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD0]], i1 false) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD1]], i1 false) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD2]], i1 false) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD3]], i1 false) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD4]], i1 false) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD5]], i1 false) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD6]], i1 false) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD7]], i1 false) -; CHECK-NEXT: [[CTTZ8:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD8]], i1 false) -; CHECK-NEXT: [[CTTZ9:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD9]], i1 false) -; CHECK-NEXT: [[CTTZ10:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD10]], i1 false) -; CHECK-NEXT: [[CTTZ11:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD11]], i1 false) -; CHECK-NEXT: [[CTTZ12:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD12]], i1 false) -; CHECK-NEXT: [[CTTZ13:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD13]], i1 false) -; CHECK-NEXT: [[CTTZ14:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD14]], i1 false) -; CHECK-NEXT: [[CTTZ15:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD15]], i1 false) -; CHECK-NEXT: store i8 [[CTTZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTTZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTTZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTTZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTTZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTTZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTTZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTTZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTTZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTTZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTTZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTTZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTTZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTTZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTTZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTTZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 +; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 false) +; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 ; CHECK-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 @@ -422,102 +354,12 @@ define void @cttz_16i8() #0 { define void @cttz_32i8() #0 { ; CHECK-LABEL: @cttz_32i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 -; CHECK-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 -; CHECK-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 -; CHECK-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 -; CHECK-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 -; CHECK-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 -; CHECK-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 -; CHECK-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 -; CHECK-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 -; CHECK-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 -; CHECK-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 -; CHECK-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 -; CHECK-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 -; CHECK-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 -; CHECK-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 -; CHECK-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD0]], i1 false) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD1]], i1 false) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD2]], i1 false) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD3]], i1 false) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD4]], i1 false) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD5]], i1 false) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD6]], i1 false) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD7]], i1 false) -; CHECK-NEXT: [[CTTZ8:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD8]], i1 false) -; CHECK-NEXT: [[CTTZ9:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD9]], i1 false) -; CHECK-NEXT: [[CTTZ10:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD10]], i1 false) -; CHECK-NEXT: [[CTTZ11:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD11]], i1 false) -; CHECK-NEXT: [[CTTZ12:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD12]], i1 false) -; CHECK-NEXT: [[CTTZ13:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD13]], i1 false) -; CHECK-NEXT: [[CTTZ14:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD14]], i1 false) -; CHECK-NEXT: [[CTTZ15:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD15]], i1 false) -; CHECK-NEXT: [[CTTZ16:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD16]], i1 false) -; CHECK-NEXT: [[CTTZ17:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD17]], i1 false) -; CHECK-NEXT: [[CTTZ18:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD18]], i1 false) -; CHECK-NEXT: [[CTTZ19:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD19]], i1 false) -; CHECK-NEXT: [[CTTZ20:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD20]], i1 false) -; CHECK-NEXT: [[CTTZ21:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD21]], i1 false) -; CHECK-NEXT: [[CTTZ22:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD22]], i1 false) -; CHECK-NEXT: [[CTTZ23:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD23]], i1 false) -; CHECK-NEXT: [[CTTZ24:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD24]], i1 false) -; CHECK-NEXT: [[CTTZ25:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD25]], i1 false) -; CHECK-NEXT: [[CTTZ26:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD26]], i1 false) -; CHECK-NEXT: [[CTTZ27:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD27]], i1 false) -; CHECK-NEXT: [[CTTZ28:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD28]], i1 false) -; CHECK-NEXT: [[CTTZ29:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD29]], i1 false) -; CHECK-NEXT: [[CTTZ30:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD30]], i1 false) -; CHECK-NEXT: [[CTTZ31:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD31]], i1 false) -; CHECK-NEXT: store i8 [[CTTZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTTZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTTZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTTZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTTZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTTZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTTZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTTZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTTZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTTZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTTZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTTZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTTZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTTZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTTZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTTZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 -; CHECK-NEXT: store i8 [[CTTZ16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 -; CHECK-NEXT: store i8 [[CTTZ17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 -; CHECK-NEXT: store i8 [[CTTZ18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 -; CHECK-NEXT: store i8 [[CTTZ19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 -; CHECK-NEXT: store i8 [[CTTZ20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 -; CHECK-NEXT: store i8 [[CTTZ21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 -; CHECK-NEXT: store i8 [[CTTZ22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 -; CHECK-NEXT: store i8 [[CTTZ23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 -; CHECK-NEXT: store i8 [[CTTZ24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 -; CHECK-NEXT: store i8 [[CTTZ25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 -; CHECK-NEXT: store i8 [[CTTZ26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 -; CHECK-NEXT: store i8 [[CTTZ27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 -; CHECK-NEXT: store i8 [[CTTZ28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 -; CHECK-NEXT: store i8 [[CTTZ29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 -; CHECK-NEXT: store i8 [[CTTZ30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 -; CHECK-NEXT: store i8 [[CTTZ31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 +; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 +; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 false) +; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP2]], i1 false) +; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 ; CHECK-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 @@ -705,32 +547,65 @@ define void @cttz_undef_4i32() #0 { } define void @cttz_undef_8i32() #0 { -; CHECK-LABEL: @cttz_undef_8i32( -; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 true) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 true) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 true) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 true) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 true) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 true) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 true) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 true) -; CHECK-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 -; CHECK-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 -; CHECK-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 -; CHECK-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 -; CHECK-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 -; CHECK-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 -; CHECK-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 -; CHECK-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 -; CHECK-NEXT: ret void +; SSE-LABEL: @cttz_undef_8i32( +; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; SSE-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 true) +; SSE-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 true) +; SSE-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 true) +; SSE-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 true) +; SSE-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 true) +; SSE-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 true) +; SSE-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 true) +; SSE-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 true) +; SSE-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; SSE-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; SSE-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; SSE-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; SSE-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; SSE-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; SSE-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; SSE-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; SSE-NEXT: ret void +; +; AVX1-LABEL: @cttz_undef_8i32( +; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 +; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 +; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 +; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 +; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 +; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 +; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 +; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 +; AVX1-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 true) +; AVX1-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 true) +; AVX1-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 true) +; AVX1-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 true) +; AVX1-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 true) +; AVX1-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 true) +; AVX1-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 true) +; AVX1-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 true) +; AVX1-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 +; AVX1-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 +; AVX1-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 +; AVX1-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 +; AVX1-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 +; AVX1-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 +; AVX1-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 +; AVX1-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 +; AVX1-NEXT: ret void +; +; AVX2-LABEL: @cttz_undef_8i32( +; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2 +; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> [[TMP1]], i1 true) +; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2 +; AVX2-NEXT: ret void ; %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 @@ -761,30 +636,9 @@ define void @cttz_undef_8i32() #0 { define void @cttz_undef_8i16() #0 { ; CHECK-LABEL: @cttz_undef_8i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD0]], i1 true) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD1]], i1 true) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD2]], i1 true) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD3]], i1 true) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD4]], i1 true) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD5]], i1 true) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD6]], i1 true) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD7]], i1 true) -; CHECK-NEXT: store i16 [[CTTZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTTZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTTZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTTZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTTZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTTZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTTZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTTZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 +; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 true) +; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 ; CHECK-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 @@ -815,56 +669,20 @@ define void @cttz_undef_8i16() #0 { } define void @cttz_undef_16i16() #0 { -; CHECK-LABEL: @cttz_undef_16i16( -; CHECK-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 -; CHECK-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 -; CHECK-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 -; CHECK-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 -; CHECK-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 -; CHECK-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 -; CHECK-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 -; CHECK-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 -; CHECK-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 -; CHECK-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 -; CHECK-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 -; CHECK-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 -; CHECK-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 -; CHECK-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 -; CHECK-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 -; CHECK-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD0]], i1 true) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD1]], i1 true) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD2]], i1 true) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD3]], i1 true) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD4]], i1 true) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD5]], i1 true) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD6]], i1 true) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD7]], i1 true) -; CHECK-NEXT: [[CTTZ8:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD8]], i1 true) -; CHECK-NEXT: [[CTTZ9:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD9]], i1 true) -; CHECK-NEXT: [[CTTZ10:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD10]], i1 true) -; CHECK-NEXT: [[CTTZ11:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD11]], i1 true) -; CHECK-NEXT: [[CTTZ12:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD12]], i1 true) -; CHECK-NEXT: [[CTTZ13:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD13]], i1 true) -; CHECK-NEXT: [[CTTZ14:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD14]], i1 true) -; CHECK-NEXT: [[CTTZ15:%.*]] = call i16 @llvm.cttz.i16(i16 [[LD15]], i1 true) -; CHECK-NEXT: store i16 [[CTTZ0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 -; CHECK-NEXT: store i16 [[CTTZ1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 -; CHECK-NEXT: store i16 [[CTTZ2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 -; CHECK-NEXT: store i16 [[CTTZ3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 -; CHECK-NEXT: store i16 [[CTTZ4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 -; CHECK-NEXT: store i16 [[CTTZ5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 -; CHECK-NEXT: store i16 [[CTTZ6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 -; CHECK-NEXT: store i16 [[CTTZ7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 -; CHECK-NEXT: store i16 [[CTTZ8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 -; CHECK-NEXT: store i16 [[CTTZ9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 -; CHECK-NEXT: store i16 [[CTTZ10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 -; CHECK-NEXT: store i16 [[CTTZ11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 -; CHECK-NEXT: store i16 [[CTTZ12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 -; CHECK-NEXT: store i16 [[CTTZ13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 -; CHECK-NEXT: store i16 [[CTTZ14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 -; CHECK-NEXT: store i16 [[CTTZ15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 -; CHECK-NEXT: ret void +; SSE-LABEL: @cttz_undef_16i16( +; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 +; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 true) +; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP2]], i1 true) +; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 +; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2 +; SSE-NEXT: ret void +; +; AVX-LABEL: @cttz_undef_16i16( +; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2 +; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> [[TMP1]], i1 true) +; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2 +; AVX-NEXT: ret void ; %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 @@ -919,54 +737,9 @@ define void @cttz_undef_16i16() #0 { define void @cttz_undef_16i8() #0 { ; CHECK-LABEL: @cttz_undef_16i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD0]], i1 true) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD1]], i1 true) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD2]], i1 true) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD3]], i1 true) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD4]], i1 true) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD5]], i1 true) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD6]], i1 true) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD7]], i1 true) -; CHECK-NEXT: [[CTTZ8:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD8]], i1 true) -; CHECK-NEXT: [[CTTZ9:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD9]], i1 true) -; CHECK-NEXT: [[CTTZ10:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD10]], i1 true) -; CHECK-NEXT: [[CTTZ11:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD11]], i1 true) -; CHECK-NEXT: [[CTTZ12:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD12]], i1 true) -; CHECK-NEXT: [[CTTZ13:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD13]], i1 true) -; CHECK-NEXT: [[CTTZ14:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD14]], i1 true) -; CHECK-NEXT: [[CTTZ15:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD15]], i1 true) -; CHECK-NEXT: store i8 [[CTTZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTTZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTTZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTTZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTTZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTTZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTTZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTTZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTTZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTTZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTTZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTTZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTTZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTTZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTTZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTTZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 +; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 true) +; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 ; CHECK-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 @@ -1022,102 +795,12 @@ define void @cttz_undef_16i8() #0 { define void @cttz_undef_32i8() #0 { ; CHECK-LABEL: @cttz_undef_32i8( -; CHECK-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 -; CHECK-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 -; CHECK-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 -; CHECK-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 -; CHECK-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 -; CHECK-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 -; CHECK-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 -; CHECK-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 -; CHECK-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 -; CHECK-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 -; CHECK-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 -; CHECK-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 -; CHECK-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 -; CHECK-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 -; CHECK-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 -; CHECK-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 -; CHECK-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 -; CHECK-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 -; CHECK-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 -; CHECK-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 -; CHECK-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 -; CHECK-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 -; CHECK-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 -; CHECK-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 -; CHECK-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 -; CHECK-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 -; CHECK-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 -; CHECK-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 -; CHECK-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 -; CHECK-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 -; CHECK-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 -; CHECK-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 -; CHECK-NEXT: [[CTTZ0:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD0]], i1 true) -; CHECK-NEXT: [[CTTZ1:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD1]], i1 true) -; CHECK-NEXT: [[CTTZ2:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD2]], i1 true) -; CHECK-NEXT: [[CTTZ3:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD3]], i1 true) -; CHECK-NEXT: [[CTTZ4:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD4]], i1 true) -; CHECK-NEXT: [[CTTZ5:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD5]], i1 true) -; CHECK-NEXT: [[CTTZ6:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD6]], i1 true) -; CHECK-NEXT: [[CTTZ7:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD7]], i1 true) -; CHECK-NEXT: [[CTTZ8:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD8]], i1 true) -; CHECK-NEXT: [[CTTZ9:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD9]], i1 true) -; CHECK-NEXT: [[CTTZ10:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD10]], i1 true) -; CHECK-NEXT: [[CTTZ11:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD11]], i1 true) -; CHECK-NEXT: [[CTTZ12:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD12]], i1 true) -; CHECK-NEXT: [[CTTZ13:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD13]], i1 true) -; CHECK-NEXT: [[CTTZ14:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD14]], i1 true) -; CHECK-NEXT: [[CTTZ15:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD15]], i1 true) -; CHECK-NEXT: [[CTTZ16:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD16]], i1 true) -; CHECK-NEXT: [[CTTZ17:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD17]], i1 true) -; CHECK-NEXT: [[CTTZ18:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD18]], i1 true) -; CHECK-NEXT: [[CTTZ19:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD19]], i1 true) -; CHECK-NEXT: [[CTTZ20:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD20]], i1 true) -; CHECK-NEXT: [[CTTZ21:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD21]], i1 true) -; CHECK-NEXT: [[CTTZ22:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD22]], i1 true) -; CHECK-NEXT: [[CTTZ23:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD23]], i1 true) -; CHECK-NEXT: [[CTTZ24:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD24]], i1 true) -; CHECK-NEXT: [[CTTZ25:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD25]], i1 true) -; CHECK-NEXT: [[CTTZ26:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD26]], i1 true) -; CHECK-NEXT: [[CTTZ27:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD27]], i1 true) -; CHECK-NEXT: [[CTTZ28:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD28]], i1 true) -; CHECK-NEXT: [[CTTZ29:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD29]], i1 true) -; CHECK-NEXT: [[CTTZ30:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD30]], i1 true) -; CHECK-NEXT: [[CTTZ31:%.*]] = call i8 @llvm.cttz.i8(i8 [[LD31]], i1 true) -; CHECK-NEXT: store i8 [[CTTZ0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 -; CHECK-NEXT: store i8 [[CTTZ1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 -; CHECK-NEXT: store i8 [[CTTZ2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 -; CHECK-NEXT: store i8 [[CTTZ3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 -; CHECK-NEXT: store i8 [[CTTZ4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 -; CHECK-NEXT: store i8 [[CTTZ5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 -; CHECK-NEXT: store i8 [[CTTZ6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 -; CHECK-NEXT: store i8 [[CTTZ7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 -; CHECK-NEXT: store i8 [[CTTZ8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 -; CHECK-NEXT: store i8 [[CTTZ9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 -; CHECK-NEXT: store i8 [[CTTZ10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 -; CHECK-NEXT: store i8 [[CTTZ11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 -; CHECK-NEXT: store i8 [[CTTZ12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 -; CHECK-NEXT: store i8 [[CTTZ13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 -; CHECK-NEXT: store i8 [[CTTZ14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 -; CHECK-NEXT: store i8 [[CTTZ15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 -; CHECK-NEXT: store i8 [[CTTZ16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 -; CHECK-NEXT: store i8 [[CTTZ17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 -; CHECK-NEXT: store i8 [[CTTZ18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 -; CHECK-NEXT: store i8 [[CTTZ19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 -; CHECK-NEXT: store i8 [[CTTZ20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 -; CHECK-NEXT: store i8 [[CTTZ21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 -; CHECK-NEXT: store i8 [[CTTZ22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 -; CHECK-NEXT: store i8 [[CTTZ23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 -; CHECK-NEXT: store i8 [[CTTZ24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 -; CHECK-NEXT: store i8 [[CTTZ25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 -; CHECK-NEXT: store i8 [[CTTZ26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 -; CHECK-NEXT: store i8 [[CTTZ27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 -; CHECK-NEXT: store i8 [[CTTZ28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 -; CHECK-NEXT: store i8 [[CTTZ29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 -; CHECK-NEXT: store i8 [[CTTZ30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 -; CHECK-NEXT: store i8 [[CTTZ31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 +; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 +; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 +; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 true) +; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP2]], i1 true) +; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 +; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 ; CHECK-NEXT: ret void ; %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1