[LV][NFC] Remove instcombine pass from RUN lines in target tests (#205848)
There is still one test remaining: LoopVectorize/X86/x86-interleaved-store-accesses-with-gaps.ll but this looks more like a phase-ordering test and should probably be handled separately.
This commit is contained in:
David Sherwood
GitHub
parent
b2e725548b
commit
10755f42bb
7 changed files with 623 additions and 597 deletions
|
|
@ -1,7 +1,7 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --filter-out-after "^scalar.ph:" --version 2
|
||||
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -tail-folding-policy=dont-fold-tail %s 2>&1 | FileCheck %s -check-prefix=SCALAR_EPILOGUE
|
||||
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -tail-folding-policy=prefer-fold-tail -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA
|
||||
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -tail-folding-policy=prefer-fold-tail %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL
|
||||
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize -mattr=+v -tail-folding-policy=dont-fold-tail %s 2>&1 | FileCheck %s -check-prefix=SCALAR_EPILOGUE
|
||||
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize -mattr=+v -tail-folding-policy=prefer-fold-tail -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA
|
||||
; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize -mattr=+v -tail-folding-policy=prefer-fold-tail %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL
|
||||
|
||||
target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
|
||||
|
||||
|
|
@ -11,84 +11,85 @@ define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noali
|
|||
; SCALAR_EPILOGUE-NEXT: entry:
|
||||
; SCALAR_EPILOGUE-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP0]], 64
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
|
||||
; SCALAR_EPILOGUE-NEXT: [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[TMP1]], i32 32)
|
||||
; SCALAR_EPILOGUE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 1024, [[UMAX]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
|
||||
; SCALAR_EPILOGUE: vector.ph:
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP1:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP2:%.*]] = shl nuw i32 [[TMP1]], 4
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP2]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP3:%.*]] = shl nuw i32 [[TMP2]], 4
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_VEC:%.*]] = sub i32 1024, [[N_MOD_VF]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP2]], i64 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP3]], i64 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; SCALAR_EPILOGUE-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; SCALAR_EPILOGUE: vector.body:
|
||||
; SCALAR_EPILOGUE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SCALAR_EPILOGUE-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP4:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP5:%.*]] = shl i32 [[INDEX]], 1
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP5:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP6:%.*]] = shl i32 [[INDEX]], 1
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr align 1 [[TMP7]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
|
||||
; SCALAR_EPILOGUE-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP10:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP11:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP11]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP10]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP13]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP10:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP11:%.*]] = select <vscale x 16 x i1> [[TMP10]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP11]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP13]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; SCALAR_EPILOGUE-NEXT: call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP12]], <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP2]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
|
||||
; SCALAR_EPILOGUE: middle.block:
|
||||
; SCALAR_EPILOGUE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[CMP_N:%.*]] = icmp eq i32 1024, [[N_VEC]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
|
||||
; SCALAR_EPILOGUE: scalar.ph:
|
||||
;
|
||||
; PREDICATED_DATA-LABEL: define void @masked_strided_factor2
|
||||
; PREDICATED_DATA-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
|
||||
; PREDICATED_DATA-NEXT: entry:
|
||||
; PREDICATED_DATA-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-NEXT: br label [[VECTOR_PH:%.*]]
|
||||
; PREDICATED_DATA: vector.ph:
|
||||
; PREDICATED_DATA-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; PREDICATED_DATA-NEXT: [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
|
||||
; PREDICATED_DATA-NEXT: [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1023
|
||||
; PREDICATED_DATA-NEXT: [[TMP2:%.*]] = sub i32 [[TMP1]], 1
|
||||
; PREDICATED_DATA-NEXT: [[N_RND_UP:%.*]] = add i32 1024, [[TMP2]]
|
||||
; PREDICATED_DATA-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP1]]
|
||||
; PREDICATED_DATA-NEXT: [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: [[TMP2:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; PREDICATED_DATA-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; PREDICATED_DATA: vector.body:
|
||||
; PREDICATED_DATA-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; PREDICATED_DATA-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP2]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; PREDICATED_DATA-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; PREDICATED_DATA-NEXT: [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 1024)
|
||||
; PREDICATED_DATA-NEXT: [[TMP3:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP4:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP3]], <vscale x 16 x i1> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: [[TMP5:%.*]] = shl i32 [[INDEX]], 1
|
||||
; PREDICATED_DATA-NEXT: [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; PREDICATED_DATA-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP4:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP5:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: [[TMP6:%.*]] = shl i32 [[INDEX]], 1
|
||||
; PREDICATED_DATA-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; PREDICATED_DATA-NEXT: [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr align 1 [[TMP7]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
|
||||
; PREDICATED_DATA-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; PREDICATED_DATA-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; PREDICATED_DATA-NEXT: [[TMP10:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP11:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; PREDICATED_DATA-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP11]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP10]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP13]])
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP10:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP11:%.*]] = select <vscale x 16 x i1> [[TMP10]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP11]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP13]])
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; PREDICATED_DATA-NEXT: call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP12]], <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
|
||||
; PREDICATED_DATA-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP1]]
|
||||
; PREDICATED_DATA-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
|
||||
|
|
@ -102,9 +103,9 @@ define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noali
|
|||
; PREDICATED_DATA-WITH-EVL-LABEL: define void @masked_strided_factor2
|
||||
; PREDICATED_DATA-WITH-EVL-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0:[0-9]+]] {
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: entry:
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: br label [[VECTOR_PH:%.*]]
|
||||
; PREDICATED_DATA-WITH-EVL: vector.ph:
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP0:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
|
|
@ -118,27 +119,26 @@ define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noali
|
|||
; PREDICATED_DATA-WITH-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP2:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP3:%.*]] = shl i32 [[INDEX]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP4]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL:%.*]] = shl nuw nsw i32 [[TMP1]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP3]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL:%.*]] = mul nuw nsw i32 [[TMP1]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[WIDE_VP_LOAD:%.*]] = call <vscale x 32 x i8> @llvm.vp.load.nxv32i8.p0(ptr align 1 [[TMP5]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], i32 [[INTERLEAVE_EVL]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[WIDE_VP_LOAD:%.*]] = call <vscale x 32 x i8> @llvm.vp.load.nxv32i8.p0(ptr align 1 [[TMP4]], <vscale x 32 x i1> [[INTERLEAVED_MASK]], i32 [[INTERLEAVE_EVL]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_VP_LOAD]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP8:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP9:%.*]] = sext i32 [[TMP3]] to i64
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP10:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP9]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP11:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP8]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL3:%.*]] = shl nuw nsw i32 [[TMP1]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP7:%.*]] = icmp slt <vscale x 16 x i8> [[TMP5]], [[TMP6]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP8:%.*]] = select <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP5]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP3]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP10:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP8]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL3:%.*]] = mul nuw nsw i32 [[TMP1]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_MASK4:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP11]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: call void @llvm.vp.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP10]], <vscale x 32 x i1> [[INTERLEAVED_MASK4]], i32 [[INTERLEAVE_EVL3]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP10]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: call void @llvm.vp.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP9]], <vscale x 32 x i1> [[INTERLEAVED_MASK4]], i32 [[INTERLEAVE_EVL3]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[CURRENT_ITERATION_NEXT]] = add nuw i32 [[TMP1]], [[INDEX]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[AVL_NEXT]] = sub nuw i32 [[AVL]], [[TMP1]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP12:%.*]] = icmp eq i32 [[AVL_NEXT]], 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP11:%.*]] = icmp eq i32 [[AVL_NEXT]], 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
|
||||
; PREDICATED_DATA-WITH-EVL: middle.block:
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: br label [[FOR_END:%.*]]
|
||||
; PREDICATED_DATA-WITH-EVL: for.end:
|
||||
|
|
@ -185,97 +185,100 @@ define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noali
|
|||
; SCALAR_EPILOGUE-NEXT: entry:
|
||||
; SCALAR_EPILOGUE-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i32 [[TMP0]], 64
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
|
||||
; SCALAR_EPILOGUE-NEXT: [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[TMP1]], i32 32)
|
||||
; SCALAR_EPILOGUE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 1024, [[UMAX]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
|
||||
; SCALAR_EPILOGUE: vector.ph:
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP1:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP2:%.*]] = shl nuw i32 [[TMP1]], 4
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP2]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_VEC:%.*]] = sub nuw nsw i32 1024, [[N_MOD_VF]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP3:%.*]] = shl nuw i32 [[TMP2]], 4
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_MOD_VF:%.*]] = urem i32 1024, [[TMP3]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[N_VEC:%.*]] = sub i32 1024, [[N_MOD_VF]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP2]], i64 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP3]], i64 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; SCALAR_EPILOGUE-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; SCALAR_EPILOGUE: vector.body:
|
||||
; SCALAR_EPILOGUE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SCALAR_EPILOGUE-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP4:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP5:%.*]] = shl i32 [[INDEX]], 2
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP5:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP6:%.*]] = shl i32 [[INDEX]], 2
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr align 1 [[TMP7]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
|
||||
; SCALAR_EPILOGUE-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP12:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP12]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP14:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP14]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP16:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP16]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; SCALAR_EPILOGUE-NEXT: call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP17]], <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP2]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP12:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP13:%.*]] = select <vscale x 16 x i1> [[TMP12]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP14:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP15:%.*]] = icmp slt <vscale x 16 x i8> [[TMP10]], [[TMP11]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP16:%.*]] = select <vscale x 16 x i1> [[TMP15]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP10]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP17:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP16]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; SCALAR_EPILOGUE-NEXT: call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP18]], <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
|
||||
; SCALAR_EPILOGUE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP18:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
|
||||
; SCALAR_EPILOGUE-NEXT: [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
|
||||
; SCALAR_EPILOGUE: middle.block:
|
||||
; SCALAR_EPILOGUE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
|
||||
; SCALAR_EPILOGUE-NEXT: [[CMP_N:%.*]] = icmp eq i32 1024, [[N_VEC]]
|
||||
; SCALAR_EPILOGUE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
|
||||
; SCALAR_EPILOGUE: scalar.ph:
|
||||
;
|
||||
; PREDICATED_DATA-LABEL: define void @masked_strided_factor4
|
||||
; PREDICATED_DATA-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
|
||||
; PREDICATED_DATA-NEXT: entry:
|
||||
; PREDICATED_DATA-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-NEXT: br label [[VECTOR_PH:%.*]]
|
||||
; PREDICATED_DATA: vector.ph:
|
||||
; PREDICATED_DATA-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
|
||||
; PREDICATED_DATA-NEXT: [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 4
|
||||
; PREDICATED_DATA-NEXT: [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1023
|
||||
; PREDICATED_DATA-NEXT: [[TMP2:%.*]] = sub i32 [[TMP1]], 1
|
||||
; PREDICATED_DATA-NEXT: [[N_RND_UP:%.*]] = add i32 1024, [[TMP2]]
|
||||
; PREDICATED_DATA-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP1]]
|
||||
; PREDICATED_DATA-NEXT: [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: [[TMP2:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; PREDICATED_DATA-NEXT: [[TMP3:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP1]], i64 0
|
||||
; PREDICATED_DATA-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; PREDICATED_DATA: vector.body:
|
||||
; PREDICATED_DATA-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; PREDICATED_DATA-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP2]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; PREDICATED_DATA-NEXT: [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP3]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; PREDICATED_DATA-NEXT: [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 1024)
|
||||
; PREDICATED_DATA-NEXT: [[TMP3:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP4:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP3]], <vscale x 16 x i1> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: [[TMP5:%.*]] = shl i32 [[INDEX]], 2
|
||||
; PREDICATED_DATA-NEXT: [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; PREDICATED_DATA-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP4:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP5:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> zeroinitializer
|
||||
; PREDICATED_DATA-NEXT: [[TMP6:%.*]] = shl i32 [[INDEX]], 2
|
||||
; PREDICATED_DATA-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP6]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; PREDICATED_DATA-NEXT: [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr align 1 [[TMP7]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
|
||||
; PREDICATED_DATA-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; PREDICATED_DATA-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; PREDICATED_DATA-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
|
||||
; PREDICATED_DATA-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
|
||||
; PREDICATED_DATA-NEXT: [[TMP12:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP12]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP14:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP14]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP16:%.*]] = sext i32 [[TMP5]] to i64
|
||||
; PREDICATED_DATA-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP16]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]])
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]], <vscale x 16 x i1> [[TMP4]])
|
||||
; PREDICATED_DATA-NEXT: call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP17]], <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
|
||||
; PREDICATED_DATA-NEXT: [[TMP12:%.*]] = icmp slt <vscale x 16 x i8> [[TMP8]], [[TMP9]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP13:%.*]] = select <vscale x 16 x i1> [[TMP12]], <vscale x 16 x i8> [[TMP9]], <vscale x 16 x i8> [[TMP8]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP14:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP15:%.*]] = icmp slt <vscale x 16 x i8> [[TMP10]], [[TMP11]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP16:%.*]] = select <vscale x 16 x i1> [[TMP15]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP10]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP17:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP16]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP6]]
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]])
|
||||
; PREDICATED_DATA-NEXT: [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP5]])
|
||||
; PREDICATED_DATA-NEXT: call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP18]], <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
|
||||
; PREDICATED_DATA-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP1]]
|
||||
; PREDICATED_DATA-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP18:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
|
||||
; PREDICATED_DATA-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
|
||||
; PREDICATED_DATA-NEXT: [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
|
||||
; PREDICATED_DATA-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
|
||||
; PREDICATED_DATA: middle.block:
|
||||
; PREDICATED_DATA-NEXT: br label [[FOR_END:%.*]]
|
||||
; PREDICATED_DATA: for.end:
|
||||
|
|
@ -284,9 +287,9 @@ define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noali
|
|||
; PREDICATED_DATA-WITH-EVL-LABEL: define void @masked_strided_factor4
|
||||
; PREDICATED_DATA-WITH-EVL-SAME: (ptr noalias readonly captures(none) [[P:%.*]], ptr noalias captures(none) [[Q:%.*]], i8 zeroext [[GUARD:%.*]]) #[[ATTR0]] {
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: entry:
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: br label [[VECTOR_PH:%.*]]
|
||||
; PREDICATED_DATA-WITH-EVL: vector.ph:
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[CONV:%.*]] = zext i8 [[GUARD]] to i32
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP0:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
|
||||
|
|
@ -300,25 +303,25 @@ define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noali
|
|||
; PREDICATED_DATA-WITH-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP2:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP3:%.*]] = shl i32 [[INDEX]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP4]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL:%.*]] = shl nuw nsw i32 [[TMP1]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[P]], i32 [[TMP3]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL:%.*]] = mul nuw nsw i32 [[TMP1]], 4
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[WIDE_VP_LOAD:%.*]] = call <vscale x 64 x i8> @llvm.vp.load.nxv64i8.p0(ptr align 1 [[TMP5]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], i32 [[INTERLEAVE_EVL]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[WIDE_VP_LOAD:%.*]] = call <vscale x 64 x i8> @llvm.vp.load.nxv64i8.p0(ptr align 1 [[TMP4]], <vscale x 64 x i1> [[INTERLEAVED_MASK]], i32 [[INTERLEAVE_EVL]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_VP_LOAD]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP10:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP6:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP8:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP9:%.*]] = icmp slt <vscale x 16 x i8> [[TMP5]], [[TMP6]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP10:%.*]] = select <vscale x 16 x i1> [[TMP9]], <vscale x 16 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP5]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP11:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP10]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP12:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP13:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP12]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP14:%.*]] = sext i32 [[TMP3]] to i64
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP14]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL3:%.*]] = shl nuw nsw i32 [[TMP1]], 2
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP12:%.*]] = icmp slt <vscale x 16 x i8> [[TMP7]], [[TMP8]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP13:%.*]] = select <vscale x 16 x i1> [[TMP12]], <vscale x 16 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP7]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP14:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[Q]], i32 [[TMP3]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVE_EVL3:%.*]] = mul nuw nsw i32 [[TMP1]], 4
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_MASK4:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP2]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: call void @llvm.vp.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr align 1 [[TMP15]], <vscale x 64 x i1> [[INTERLEAVED_MASK4]], i32 [[INTERLEAVE_EVL3]])
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[CURRENT_ITERATION_NEXT]] = add nuw i32 [[TMP1]], [[INDEX]]
|
||||
; PREDICATED_DATA-WITH-EVL-NEXT: [[AVL_NEXT]] = sub nuw i32 [[AVL]], [[TMP1]]
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
||||
; RUN: opt -S -mtriple=s390x-unknown-linux -mcpu=z13 -passes=loop-vectorize,instcombine -force-vector-width=2 < %s | FileCheck %s
|
||||
; RUN: opt -S -mtriple=s390x-unknown-linux -mcpu=z13 -passes=loop-vectorize -force-vector-width=2 < %s | FileCheck %s
|
||||
;
|
||||
; Test that loop vectorizer does not generate vector addresses that must then
|
||||
; always be extracted.
|
||||
|
|
@ -14,53 +14,53 @@ define void @foo(ptr nocapture %A) {
|
|||
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; CHECK: vector.body:
|
||||
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; CHECK-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
|
||||
; CHECK-NEXT: [[DOTIDX1:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX1]]
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i64 16
|
||||
; CHECK-NEXT: [[DOTIDX2:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP31:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX2]]
|
||||
; CHECK-NEXT: [[TMP32:%.*]] = getelementptr i8, ptr [[TMP31]], i64 32
|
||||
; CHECK-NEXT: [[DOTIDX3:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX3]]
|
||||
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i64 48
|
||||
; CHECK-NEXT: [[DOTIDX4:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX4]]
|
||||
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i64 64
|
||||
; CHECK-NEXT: [[DOTIDX5:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX5]]
|
||||
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i64 80
|
||||
; CHECK-NEXT: [[DOTIDX6:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX6]]
|
||||
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP11]], i64 96
|
||||
; CHECK-NEXT: [[DOTIDX7:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX7]]
|
||||
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i64 112
|
||||
; CHECK-NEXT: [[DOTIDX8:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX8]]
|
||||
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP15]], i64 128
|
||||
; CHECK-NEXT: [[DOTIDX9:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX9]]
|
||||
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i64 144
|
||||
; CHECK-NEXT: [[DOTIDX10:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX10]]
|
||||
; CHECK-NEXT: [[TMP20:%.*]] = getelementptr i8, ptr [[TMP19]], i64 160
|
||||
; CHECK-NEXT: [[DOTIDX11:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX11]]
|
||||
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr i8, ptr [[TMP21]], i64 176
|
||||
; CHECK-NEXT: [[DOTIDX12:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX12]]
|
||||
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr i8, ptr [[TMP23]], i64 192
|
||||
; CHECK-NEXT: [[DOTIDX13:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX13]]
|
||||
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr i8, ptr [[TMP25]], i64 208
|
||||
; CHECK-NEXT: [[DOTIDX14:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP27:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX14]]
|
||||
; CHECK-NEXT: [[TMP28:%.*]] = getelementptr i8, ptr [[TMP27]], i64 224
|
||||
; CHECK-NEXT: [[DOTIDX15:%.*]] = shl i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP29:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX15]]
|
||||
; CHECK-NEXT: [[TMP30:%.*]] = getelementptr i8, ptr [[TMP29]], i64 240
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 1
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 2
|
||||
; CHECK-NEXT: [[TMP31:%.*]] = add i64 [[INDEX]], 3
|
||||
; CHECK-NEXT: [[TMP33:%.*]] = add i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP34:%.*]] = add i64 [[INDEX]], 5
|
||||
; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 6
|
||||
; CHECK-NEXT: [[TMP35:%.*]] = add i64 [[INDEX]], 7
|
||||
; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 8
|
||||
; CHECK-NEXT: [[TMP36:%.*]] = add i64 [[INDEX]], 9
|
||||
; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 10
|
||||
; CHECK-NEXT: [[TMP37:%.*]] = add i64 [[INDEX]], 11
|
||||
; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 12
|
||||
; CHECK-NEXT: [[TMP38:%.*]] = add i64 [[INDEX]], 13
|
||||
; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[INDEX]], 14
|
||||
; CHECK-NEXT: [[TMP39:%.*]] = add i64 [[INDEX]], 15
|
||||
; CHECK-NEXT: [[TMP15:%.*]] = shl nsw i64 [[INDEX]], 2
|
||||
; CHECK-NEXT: [[TMP40:%.*]] = shl nsw i64 [[TMP0]], 2
|
||||
; CHECK-NEXT: [[TMP17:%.*]] = shl nsw i64 [[TMP1]], 2
|
||||
; CHECK-NEXT: [[TMP41:%.*]] = shl nsw i64 [[TMP31]], 2
|
||||
; CHECK-NEXT: [[TMP19:%.*]] = shl nsw i64 [[TMP33]], 2
|
||||
; CHECK-NEXT: [[TMP42:%.*]] = shl nsw i64 [[TMP34]], 2
|
||||
; CHECK-NEXT: [[TMP21:%.*]] = shl nsw i64 [[TMP5]], 2
|
||||
; CHECK-NEXT: [[TMP43:%.*]] = shl nsw i64 [[TMP35]], 2
|
||||
; CHECK-NEXT: [[TMP23:%.*]] = shl nsw i64 [[TMP7]], 2
|
||||
; CHECK-NEXT: [[TMP44:%.*]] = shl nsw i64 [[TMP36]], 2
|
||||
; CHECK-NEXT: [[TMP25:%.*]] = shl nsw i64 [[TMP9]], 2
|
||||
; CHECK-NEXT: [[TMP45:%.*]] = shl nsw i64 [[TMP37]], 2
|
||||
; CHECK-NEXT: [[TMP27:%.*]] = shl nsw i64 [[TMP11]], 2
|
||||
; CHECK-NEXT: [[TMP46:%.*]] = shl nsw i64 [[TMP38]], 2
|
||||
; CHECK-NEXT: [[TMP29:%.*]] = shl nsw i64 [[TMP13]], 2
|
||||
; CHECK-NEXT: [[TMP47:%.*]] = shl nsw i64 [[TMP39]], 2
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP15]]
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP40]]
|
||||
; CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
|
||||
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP41]]
|
||||
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP19]]
|
||||
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP42]]
|
||||
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP21]]
|
||||
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP43]]
|
||||
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP23]]
|
||||
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP44]]
|
||||
; CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP25]]
|
||||
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP45]]
|
||||
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP27]]
|
||||
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP46]]
|
||||
; CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP29]]
|
||||
; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP47]]
|
||||
; CHECK-NEXT: store i32 4, ptr [[TMP2]], align 4
|
||||
; CHECK-NEXT: store i32 4, ptr [[TMP3]], align 4
|
||||
; CHECK-NEXT: store i32 4, ptr [[TMP32]], align 4
|
||||
|
|
@ -113,21 +113,21 @@ define void @foo1(ptr nocapture noalias %A, ptr nocapture %PtrPtr) {
|
|||
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; CHECK: vector.body:
|
||||
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds [8 x i8], ptr [[PTRPTR:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[TMP11]], i64 8
|
||||
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP15]], i64 16
|
||||
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr i8, ptr [[TMP23]], i64 24
|
||||
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr i8, ptr [[TMP25]], i64 32
|
||||
; CHECK-NEXT: [[TMP39:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP43:%.*]] = getelementptr i8, ptr [[TMP39]], i64 40
|
||||
; CHECK-NEXT: [[TMP44:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP44]], i64 48
|
||||
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr [8 x i8], ptr [[PTRPTR]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i64 56
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 1
|
||||
; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 2
|
||||
; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 3
|
||||
; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[INDEX]], 5
|
||||
; CHECK-NEXT: [[TMP23:%.*]] = add i64 [[INDEX]], 6
|
||||
; CHECK-NEXT: [[TMP39:%.*]] = add i64 [[INDEX]], 7
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP0]]
|
||||
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP7]]
|
||||
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP11]]
|
||||
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP13]]
|
||||
; CHECK-NEXT: [[TMP43:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP15]]
|
||||
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP23]]
|
||||
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds ptr, ptr [[PTRPTR]], i64 [[TMP39]]
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP1]], align 8
|
||||
; CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TMP2]], align 8
|
||||
; CHECK-NEXT: [[TMP17:%.*]] = load ptr, ptr [[TMP16]], align 8
|
||||
|
|
@ -138,24 +138,24 @@ define void @foo1(ptr nocapture noalias %A, ptr nocapture %PtrPtr) {
|
|||
; CHECK-NEXT: [[TMP22:%.*]] = load ptr, ptr [[TMP14]], align 8
|
||||
; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 4
|
||||
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[TMP4]], align 4
|
||||
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP5]], i64 0
|
||||
; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x i32> [[TMP7]], i32 [[TMP6]], i64 1
|
||||
; CHECK-NEXT: [[TMP25:%.*]] = insertelement <2 x i32> poison, i32 [[TMP5]], i32 0
|
||||
; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x i32> [[TMP25]], i32 [[TMP6]], i32 1
|
||||
; CHECK-NEXT: [[TMP27:%.*]] = load i32, ptr [[TMP17]], align 4
|
||||
; CHECK-NEXT: [[TMP28:%.*]] = load i32, ptr [[TMP18]], align 4
|
||||
; CHECK-NEXT: [[TMP29:%.*]] = insertelement <2 x i32> poison, i32 [[TMP27]], i64 0
|
||||
; CHECK-NEXT: [[TMP30:%.*]] = insertelement <2 x i32> [[TMP29]], i32 [[TMP28]], i64 1
|
||||
; CHECK-NEXT: [[TMP29:%.*]] = insertelement <2 x i32> poison, i32 [[TMP27]], i32 0
|
||||
; CHECK-NEXT: [[TMP30:%.*]] = insertelement <2 x i32> [[TMP29]], i32 [[TMP28]], i32 1
|
||||
; CHECK-NEXT: [[TMP31:%.*]] = load i32, ptr [[TMP19]], align 4
|
||||
; CHECK-NEXT: [[TMP32:%.*]] = load i32, ptr [[TMP20]], align 4
|
||||
; CHECK-NEXT: [[TMP33:%.*]] = insertelement <2 x i32> poison, i32 [[TMP31]], i64 0
|
||||
; CHECK-NEXT: [[TMP34:%.*]] = insertelement <2 x i32> [[TMP33]], i32 [[TMP32]], i64 1
|
||||
; CHECK-NEXT: [[TMP33:%.*]] = insertelement <2 x i32> poison, i32 [[TMP31]], i32 0
|
||||
; CHECK-NEXT: [[TMP34:%.*]] = insertelement <2 x i32> [[TMP33]], i32 [[TMP32]], i32 1
|
||||
; CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[TMP21]], align 4
|
||||
; CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[TMP22]], align 4
|
||||
; CHECK-NEXT: [[TMP37:%.*]] = insertelement <2 x i32> poison, i32 [[TMP35]], i64 0
|
||||
; CHECK-NEXT: [[TMP38:%.*]] = insertelement <2 x i32> [[TMP37]], i32 [[TMP36]], i64 1
|
||||
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i64 8
|
||||
; CHECK-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i64 16
|
||||
; CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i64 24
|
||||
; CHECK-NEXT: [[TMP37:%.*]] = insertelement <2 x i32> poison, i32 [[TMP35]], i32 0
|
||||
; CHECK-NEXT: [[TMP38:%.*]] = insertelement <2 x i32> [[TMP37]], i32 [[TMP36]], i32 1
|
||||
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 2
|
||||
; CHECK-NEXT: [[TMP41:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 4
|
||||
; CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 6
|
||||
; CHECK-NEXT: store <2 x i32> [[TMP8]], ptr [[TMP9]], align 4
|
||||
; CHECK-NEXT: store <2 x i32> [[TMP30]], ptr [[TMP40]], align 4
|
||||
; CHECK-NEXT: store <2 x i32> [[TMP34]], ptr [[TMP41]], align 4
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "scalar.ph\:" --version 5
|
||||
; REQUIRES: asserts
|
||||
; RUN: opt < %s -aa-pipeline=basic-aa -passes=loop-vectorize,instcombine -S -debug-only=loop-vectorize 2>%t -S | FileCheck %s
|
||||
; RUN: opt < %s -aa-pipeline=basic-aa -passes=loop-vectorize -S -debug-only=loop-vectorize 2>%t -S | FileCheck %s
|
||||
; RUN: cat %t | FileCheck %s --check-prefix=DEBUG
|
||||
; RUN: opt < %s -passes=loop-vectorize -force-vector-width=2 -S | FileCheck %s -check-prefix=FORCE
|
||||
|
||||
|
|
@ -36,13 +36,12 @@ define void @PR31671(float %x, ptr %d) #0 {
|
|||
; CHECK: [[VECTOR_BODY]]:
|
||||
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
|
||||
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <16 x i64> [ <i64 0, i64 5, i64 10, i64 15, i64 20, i64 25, i64 30, i64 35, i64 40, i64 45, i64 50, i64 55, i64 60, i64 65, i64 70, i64 75>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
|
||||
; CHECK-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds nuw i8, ptr [[D]], i64 128016
|
||||
; CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[INDEX]], 20
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[DOTSPLIT]], i64 [[DOTIDX]]
|
||||
; CHECK-NEXT: [[TMP6:%.*]] = mul i64 [[INDEX]], 5
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[DATA:%.*]], ptr [[D]], i64 0, i32 3, i64 [[TMP6]]
|
||||
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <80 x float>, ptr [[TMP0]], align 4
|
||||
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <80 x float> [[WIDE_VEC]], <80 x float> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = fmul <16 x float> [[BROADCAST_SPLAT]], [[STRIDED_VEC]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [4 x i8], ptr [[D]], <16 x i64> [[VEC_IND]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[DATA]], ptr [[D]], i64 0, i32 0, <16 x i64> [[VEC_IND]]
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = extractelement <16 x ptr> [[TMP2]], i64 0
|
||||
; CHECK-NEXT: [[WIDE_VEC1:%.*]] = load <80 x float>, ptr [[TMP3]], align 4
|
||||
; CHECK-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <80 x float> [[WIDE_VEC1]], <80 x float> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
|
||||
|
|
@ -179,7 +178,9 @@ define void @PR40816() #1 {
|
|||
; CHECK: [[MIDDLE_BLOCK]]:
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = phi i32 [ 0, %[[PRED_STORE_CONTINUE6]] ], [ [[INC:%.*]], %[[MIDDLE_BLOCK]] ]
|
||||
; CHECK-NEXT: store i32 [[TMP0]], ptr @b, align 1
|
||||
; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[TMP0]], 2
|
||||
; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds [3 x i32], ptr @a, i32 0, i32 [[TMP0]]
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX1]], align 1
|
||||
; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[TMP1]], 0
|
||||
; CHECK-NEXT: [[INC]] = add nuw nsw i32 [[TMP0]], 1
|
||||
; CHECK-NEXT: br i1 [[CMP2]], label %[[SCALAR_PH:.*]], label %[[MIDDLE_BLOCK]]
|
||||
; CHECK: [[SCALAR_PH]]:
|
||||
|
|
|
|||
|
|
@ -1,9 +1,9 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize,instcombine < %s | FileCheck %s --check-prefix=SSE
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize,instcombine -mcpu=sandybridge < %s | FileCheck %s --check-prefix=AVX1
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize,instcombine -mcpu=haswell < %s | FileCheck %s --check-prefix=AVX2
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize,instcombine -mcpu=slm < %s | FileCheck %s --check-prefix=SSE
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize,instcombine -mcpu=atom < %s | FileCheck %s --check-prefix=ATOM
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize < %s | FileCheck %s --check-prefix=SSE
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize -mcpu=sandybridge < %s | FileCheck %s --check-prefix=AVX1
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize -mcpu=haswell < %s | FileCheck %s --check-prefix=AVX2
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize -mcpu=slm < %s | FileCheck %s --check-prefix=SSE
|
||||
; RUN: opt -S -mtriple=x86_64-pc_linux -passes=loop-vectorize -mcpu=atom < %s | FileCheck %s --check-prefix=ATOM
|
||||
|
||||
define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
||||
; SSE-LABEL: @foo(
|
||||
|
|
@ -13,11 +13,11 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; SSE-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; SSE: vector.body:
|
||||
; SSE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; SSE-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[INDEX]], 3
|
||||
; SSE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX]]
|
||||
; SSE-NEXT: [[DOTIDX5:%.*]] = shl i64 [[INDEX]], 3
|
||||
; SSE-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX5]]
|
||||
; SSE-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i64 32
|
||||
; SSE-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 4
|
||||
; SSE-NEXT: [[TMP1:%.*]] = shl nsw i64 [[INDEX]], 1
|
||||
; SSE-NEXT: [[TMP9:%.*]] = shl nsw i64 [[TMP0]], 1
|
||||
; SSE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP1]]
|
||||
; SSE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP9]]
|
||||
; SSE-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
|
||||
; SSE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
|
||||
; SSE-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
|
||||
|
|
@ -26,8 +26,8 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; SSE-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <8 x i32> [[WIDE_VEC2]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
|
||||
; SSE-NEXT: [[TMP4:%.*]] = add nsw <4 x i32> [[STRIDED_VEC1]], [[STRIDED_VEC]]
|
||||
; SSE-NEXT: [[TMP5:%.*]] = add nsw <4 x i32> [[STRIDED_VEC4]], [[STRIDED_VEC3]]
|
||||
; SSE-NEXT: [[TMP6:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; SSE-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP6]], i64 16
|
||||
; SSE-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; SSE-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i64 4
|
||||
; SSE-NEXT: store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4
|
||||
; SSE-NEXT: store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
|
||||
; SSE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
|
||||
|
|
@ -45,17 +45,17 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; AVX1-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; AVX1: vector.body:
|
||||
; AVX1-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; AVX1-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[INDEX]], 3
|
||||
; AVX1-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX]]
|
||||
; AVX1-NEXT: [[DOTIDX11:%.*]] = shl i64 [[INDEX]], 3
|
||||
; AVX1-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX11]]
|
||||
; AVX1-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[TMP1]], i64 32
|
||||
; AVX1-NEXT: [[DOTIDX12:%.*]] = shl i64 [[INDEX]], 3
|
||||
; AVX1-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX12]]
|
||||
; AVX1-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[TMP3]], i64 64
|
||||
; AVX1-NEXT: [[DOTIDX13:%.*]] = shl i64 [[INDEX]], 3
|
||||
; AVX1-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX13]]
|
||||
; AVX1-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[TMP5]], i64 96
|
||||
; AVX1-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 4
|
||||
; AVX1-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 8
|
||||
; AVX1-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 12
|
||||
; AVX1-NEXT: [[TMP3:%.*]] = shl nsw i64 [[INDEX]], 1
|
||||
; AVX1-NEXT: [[TMP4:%.*]] = shl nsw i64 [[TMP0]], 1
|
||||
; AVX1-NEXT: [[TMP5:%.*]] = shl nsw i64 [[TMP1]], 1
|
||||
; AVX1-NEXT: [[TMP19:%.*]] = shl nsw i64 [[TMP2]], 1
|
||||
; AVX1-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP3]]
|
||||
; AVX1-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP4]]
|
||||
; AVX1-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP5]]
|
||||
; AVX1-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP19]]
|
||||
; AVX1-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP6]], align 4
|
||||
; AVX1-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
|
||||
; AVX1-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
|
||||
|
|
@ -72,10 +72,10 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; AVX1-NEXT: [[TMP11:%.*]] = add nsw <4 x i32> [[STRIDED_VEC4]], [[STRIDED_VEC3]]
|
||||
; AVX1-NEXT: [[TMP12:%.*]] = add nsw <4 x i32> [[STRIDED_VEC7]], [[STRIDED_VEC6]]
|
||||
; AVX1-NEXT: [[TMP13:%.*]] = add nsw <4 x i32> [[STRIDED_VEC10]], [[STRIDED_VEC9]]
|
||||
; AVX1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; AVX1-NEXT: [[TMP15:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP14]], i64 16
|
||||
; AVX1-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP14]], i64 32
|
||||
; AVX1-NEXT: [[TMP17:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP14]], i64 48
|
||||
; AVX1-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; AVX1-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 4
|
||||
; AVX1-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 8
|
||||
; AVX1-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 12
|
||||
; AVX1-NEXT: store <4 x i32> [[TMP10]], ptr [[TMP14]], align 4
|
||||
; AVX1-NEXT: store <4 x i32> [[TMP11]], ptr [[TMP15]], align 4
|
||||
; AVX1-NEXT: store <4 x i32> [[TMP12]], ptr [[TMP16]], align 4
|
||||
|
|
@ -95,17 +95,17 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; AVX2-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; AVX2: vector.body:
|
||||
; AVX2-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; AVX2-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[INDEX]], 3
|
||||
; AVX2-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX]]
|
||||
; AVX2-NEXT: [[DOTIDX11:%.*]] = shl i64 [[INDEX]], 3
|
||||
; AVX2-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX11]]
|
||||
; AVX2-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[TMP1]], i64 64
|
||||
; AVX2-NEXT: [[DOTIDX12:%.*]] = shl i64 [[INDEX]], 3
|
||||
; AVX2-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX12]]
|
||||
; AVX2-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[TMP3]], i64 128
|
||||
; AVX2-NEXT: [[DOTIDX13:%.*]] = shl i64 [[INDEX]], 3
|
||||
; AVX2-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[DOTIDX13]]
|
||||
; AVX2-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[TMP5]], i64 192
|
||||
; AVX2-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 8
|
||||
; AVX2-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 16
|
||||
; AVX2-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 24
|
||||
; AVX2-NEXT: [[TMP3:%.*]] = shl nsw i64 [[INDEX]], 1
|
||||
; AVX2-NEXT: [[TMP4:%.*]] = shl nsw i64 [[TMP0]], 1
|
||||
; AVX2-NEXT: [[TMP5:%.*]] = shl nsw i64 [[TMP1]], 1
|
||||
; AVX2-NEXT: [[TMP19:%.*]] = shl nsw i64 [[TMP2]], 1
|
||||
; AVX2-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP3]]
|
||||
; AVX2-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP4]]
|
||||
; AVX2-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP5]]
|
||||
; AVX2-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP19]]
|
||||
; AVX2-NEXT: [[WIDE_VEC:%.*]] = load <16 x i32>, ptr [[TMP6]], align 4
|
||||
; AVX2-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
|
||||
; AVX2-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
|
||||
|
|
@ -122,10 +122,10 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; AVX2-NEXT: [[TMP11:%.*]] = add nsw <8 x i32> [[STRIDED_VEC4]], [[STRIDED_VEC3]]
|
||||
; AVX2-NEXT: [[TMP12:%.*]] = add nsw <8 x i32> [[STRIDED_VEC7]], [[STRIDED_VEC6]]
|
||||
; AVX2-NEXT: [[TMP13:%.*]] = add nsw <8 x i32> [[STRIDED_VEC10]], [[STRIDED_VEC9]]
|
||||
; AVX2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; AVX2-NEXT: [[TMP15:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP14]], i64 32
|
||||
; AVX2-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP14]], i64 64
|
||||
; AVX2-NEXT: [[TMP17:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP14]], i64 96
|
||||
; AVX2-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; AVX2-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 8
|
||||
; AVX2-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 16
|
||||
; AVX2-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 24
|
||||
; AVX2-NEXT: store <8 x i32> [[TMP10]], ptr [[TMP14]], align 4
|
||||
; AVX2-NEXT: store <8 x i32> [[TMP11]], ptr [[TMP15]], align 4
|
||||
; AVX2-NEXT: store <8 x i32> [[TMP12]], ptr [[TMP16]], align 4
|
||||
|
|
@ -145,14 +145,14 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
|
|||
; ATOM-NEXT: ret void
|
||||
; ATOM: for.body:
|
||||
; ATOM-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
|
||||
; ATOM-NEXT: [[TMP0:%.*]] = shl nuw nsw i64 [[INDVARS_IV]], 1
|
||||
; ATOM-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw [4 x i8], ptr [[B:%.*]], i64 [[TMP0]]
|
||||
; ATOM-NEXT: [[TMP0:%.*]] = shl nsw i64 [[INDVARS_IV]], 1
|
||||
; ATOM-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP0]]
|
||||
; ATOM-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
|
||||
; ATOM-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw [4 x i8], ptr [[B]], i64 [[TMP0]]
|
||||
; ATOM-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i64 4
|
||||
; ATOM-NEXT: [[TMP2:%.*]] = or disjoint i64 [[TMP0]], 1
|
||||
; ATOM-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP2]]
|
||||
; ATOM-NEXT: [[TMP3:%.*]] = load i32, ptr [[ARRAYIDX3]], align 4
|
||||
; ATOM-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP3]], [[TMP1]]
|
||||
; ATOM-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds nuw [4 x i8], ptr [[A:%.*]], i64 [[INDVARS_IV]]
|
||||
; ATOM-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDVARS_IV]]
|
||||
; ATOM-NEXT: store i32 [[ADD4]], ptr [[ARRAYIDX6]], align 4
|
||||
; ATOM-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
|
||||
; ATOM-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1024
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
||||
; RUN: opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s
|
||||
; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s
|
||||
|
||||
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64-unknown-linux-gnu"
|
||||
|
|
@ -19,19 +19,19 @@ define void @loop(ptr nocapture %a, ptr nocapture %b) {
|
|||
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
|
||||
; CHECK: for.body:
|
||||
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
|
||||
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x i8], ptr [[B:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
|
||||
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
|
||||
; CHECK-NEXT: [[IDXPROM3:%.*]] = sext i32 [[TMP1]] to i64
|
||||
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x i8], ptr [[A]], i64 [[IDXPROM3]]
|
||||
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IDXPROM3]]
|
||||
; CHECK-NEXT: store i32 [[TMP0]], ptr [[ARRAYIDX4]], align 4
|
||||
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
|
||||
; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x i8], ptr [[B]], i64 [[INDVARS_IV_NEXT]]
|
||||
; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[ARRAYIDX6]], align 4
|
||||
; CHECK-NEXT: store i32 [[TMP2]], ptr [[ARRAYIDX2]], align 4
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = and i64 [[INDVARS_IV_NEXT]], 4294967295
|
||||
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[TMP3]], 512
|
||||
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
|
||||
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], 512
|
||||
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
|
||||
; CHECK: for.end:
|
||||
; CHECK-NEXT: ret void
|
||||
|
|
@ -71,39 +71,39 @@ define void @parallel_loop(ptr nocapture %a, ptr nocapture %b) {
|
|||
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; CHECK: vector.body:
|
||||
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [4 x i8], ptr [[B:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4, !llvm.access.group [[ACC_GRP0:![0-9]+]]
|
||||
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr [4 x i8], ptr [[A]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[TMP2]], i64 4
|
||||
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr [4 x i8], ptr [[A]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr i8, ptr [[TMP4]], i64 8
|
||||
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr [4 x i8], ptr [[A]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i64 12
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 1
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 2
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 3
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4, !llvm.access.group [[ACC_GRP0:![0-9]+]]
|
||||
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP0]]
|
||||
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP1]]
|
||||
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP2]]
|
||||
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP4]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP26]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP7]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP6]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP7]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[TMP12:%.*]] = sext i32 [[TMP8]] to i64
|
||||
; CHECK-NEXT: [[TMP13:%.*]] = sext i32 [[TMP9]] to i64
|
||||
; CHECK-NEXT: [[TMP14:%.*]] = sext i32 [[TMP10]] to i64
|
||||
; CHECK-NEXT: [[TMP15:%.*]] = sext i32 [[TMP11]] to i64
|
||||
; CHECK-NEXT: [[TMP16:%.*]] = sext i32 [[TMP12]] to i64
|
||||
; CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds [4 x i8], ptr [[A]], i64 [[TMP13]]
|
||||
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8], ptr [[A]], i64 [[TMP14]]
|
||||
; CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8], ptr [[A]], i64 [[TMP15]]
|
||||
; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds [4 x i8], ptr [[A]], i64 [[TMP16]]
|
||||
; CHECK-NEXT: [[TMP31:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 0
|
||||
; CHECK-NEXT: store i32 [[TMP31]], ptr [[TMP20]], align 4, !llvm.access.group [[ACC_GRP1:![0-9]+]]
|
||||
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP12]]
|
||||
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP13]]
|
||||
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]
|
||||
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP15]]
|
||||
; CHECK-NEXT: [[TMP20:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 0
|
||||
; CHECK-NEXT: store i32 [[TMP20]], ptr [[TMP16]], align 4, !llvm.access.group [[ACC_GRP1:![0-9]+]]
|
||||
; CHECK-NEXT: [[TMP21:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 1
|
||||
; CHECK-NEXT: store i32 [[TMP21]], ptr [[TMP24]], align 4, !llvm.access.group [[ACC_GRP1]]
|
||||
; CHECK-NEXT: store i32 [[TMP21]], ptr [[TMP17]], align 4, !llvm.access.group [[ACC_GRP1]]
|
||||
; CHECK-NEXT: [[TMP22:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 2
|
||||
; CHECK-NEXT: store i32 [[TMP22]], ptr [[TMP28]], align 4, !llvm.access.group [[ACC_GRP1]]
|
||||
; CHECK-NEXT: store i32 [[TMP22]], ptr [[TMP18]], align 4, !llvm.access.group [[ACC_GRP1]]
|
||||
; CHECK-NEXT: [[TMP23:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 3
|
||||
; CHECK-NEXT: store i32 [[TMP23]], ptr [[TMP30]], align 4, !llvm.access.group [[ACC_GRP1]]
|
||||
; CHECK-NEXT: [[TMP27:%.*]] = getelementptr [4 x i8], ptr [[B]], i64 [[INDEX]]
|
||||
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr i8, ptr [[TMP27]], i64 4
|
||||
; CHECK-NEXT: store i32 [[TMP23]], ptr [[TMP19]], align 4, !llvm.access.group [[ACC_GRP1]]
|
||||
; CHECK-NEXT: [[TMP24:%.*]] = add i64 [[INDEX]], 1
|
||||
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP24]]
|
||||
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP25]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: store <4 x i32> [[WIDE_LOAD1]], ptr [[TMP5]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: store <4 x i32> [[WIDE_LOAD1]], ptr [[TMP4]], align 4, !llvm.access.group [[ACC_GRP0]]
|
||||
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
|
||||
; CHECK-NEXT: [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
|
||||
; CHECK-NEXT: br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
|
||||
|
|
@ -148,19 +148,19 @@ define void @mixed_metadata(ptr nocapture %a, ptr nocapture %b) {
|
|||
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
|
||||
; CHECK: for.body:
|
||||
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
|
||||
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x i8], ptr [[B:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
|
||||
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x i8], ptr [[A:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDVARS_IV]]
|
||||
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4, !llvm.access.group [[ACC_GRP6]]
|
||||
; CHECK-NEXT: [[IDXPROM3:%.*]] = sext i32 [[TMP1]] to i64
|
||||
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x i8], ptr [[A]], i64 [[IDXPROM3]]
|
||||
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IDXPROM3]]
|
||||
; CHECK-NEXT: store i32 [[TMP0]], ptr [[ARRAYIDX4]], align 4, !llvm.access.group [[ACC_GRP7:![0-9]+]]
|
||||
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
|
||||
; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x i8], ptr [[B]], i64 [[INDVARS_IV_NEXT]]
|
||||
; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT]]
|
||||
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[ARRAYIDX6]], align 4, !llvm.access.group [[ACC_GRP6]]
|
||||
; CHECK-NEXT: store i32 [[TMP2]], ptr [[ARRAYIDX2]], align 4, !llvm.access.group [[ACC_GRP6]]
|
||||
; CHECK-NEXT: [[TMP3:%.*]] = and i64 [[INDVARS_IV_NEXT]], 4294967295
|
||||
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[TMP3]], 512
|
||||
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
|
||||
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], 512
|
||||
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
|
||||
; CHECK: for.end:
|
||||
; CHECK-NEXT: ret void
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
||||
; RUN: opt %s -passes=loop-vectorize,instcombine -mtriple=x86_64-unknown-linux-gnu -mattr=avx512vl,avx512dq,avx512bw -S | FileCheck %s
|
||||
; RUN: opt %s -passes=loop-vectorize -mtriple=x86_64-unknown-linux-gnu -mattr=avx512vl,avx512dq,avx512bw -S | FileCheck %s
|
||||
|
||||
@bytes = global [128 x i8] zeroinitializer, align 16
|
||||
|
||||
|
|
@ -13,14 +13,15 @@ define zeroext i8 @sum() {
|
|||
; CHECK: vector.ph:
|
||||
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
|
||||
; CHECK: vector.body:
|
||||
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <64 x i8>, ptr @bytes, align 1
|
||||
; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <64 x i8>, ptr getelementptr inbounds (i8, ptr @bytes, i64 64), align 1
|
||||
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = add <64 x i8> [[WIDE_LOAD1]], zeroinitializer
|
||||
; CHECK-NEXT: br label [[MIDDLE_BLOCK:%.*]]
|
||||
; CHECK: middle.block:
|
||||
; CHECK-NEXT: br label [[FOR_END:%.*]]
|
||||
; CHECK: for.end:
|
||||
; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <64 x i8>, ptr getelementptr inbounds nuw (i8, ptr @bytes, i64 64), align 1
|
||||
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <64 x i8>, ptr @bytes, align 1
|
||||
; CHECK-NEXT: [[BIN_RDX:%.*]] = add <64 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD]]
|
||||
; CHECK-NEXT: [[TMP0:%.*]] = call i8 @llvm.vector.reduce.add.v64i8(<64 x i8> [[BIN_RDX]])
|
||||
; CHECK-NEXT: br label [[FOR_END:%.*]]
|
||||
; CHECK: for.end:
|
||||
; CHECK-NEXT: ret i8 [[TMP0]]
|
||||
;
|
||||
entry:
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load diff
Loading…
Reference in a new issue