shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
Florian Hahn 77fb848894
Reapply "[LV] Simplify and unify resume value handling for epilogue vec." (#187504) 
This reverts commit cdaf29f84dd0abbd1f961982799059c92d76625b.

This version skips removeBranchOnConst when vectorizing the epilogue, as
it may trigger folds that remove the resume phi used as resume value
from the epilogue.

This fixes https://github.com/llvm/llvm-project/issues/187323.

Original message:
This patch tries to drastically simplify resume value handling for the
scalar loop when vectorizing the epilogue.

It uses a simpler, uniform approach for updating all resume values in
the scalar loop:

1. Create ResumeForEpilogue recipes for all scalar resume phis in the
main loop (the epilogue plan will have exactly the same scalar resume
phis, in exactly the same order)
2. Update ::execute for ResumeForEpilogue to set the underlying value
when executing. This is not super clean, but allows easy lookup of the
generated IR value when we update the resume phis in the epilogue. Once
we connect the 2 plans together explicitly, this can be removed.
3. Use the list of ResumeForEpilogue VPInstructions from the main loop
to update the resume/bypass values from the epilogue.

This simplifies the code quite a bit, makes it more robust (should fix
https://github.com/llvm/llvm-project/issues/179407) and also fixes a
mis-compile in the existing tests (see change in

llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-sub-epilogue-vec.ll,
where previously we would incorrectly resume using the start value when
the epilogue iteration check failed)

In some cases, we get simpler code, due to additional CSE, in some cases
the induction end value computations get moved from the epilogue
iteration check to the vector preheader. We could try to sink the
instructions as cleanup, but it is probably not worth the trouble.

Fixes https://github.com/llvm/llvm-project/issues/179407.

PR for recommit https://github.com/llvm/llvm-project/pull/188134
2026-03-23 22:09:40 +00:00

314 lines
20 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=loop-vectorize -mcpu=znver2 -S %s | FileCheck %s
target triple = "x86_64-unknown-linux-gnu"
define void @test_pr59459(i64 %iv.start, ptr %arr) {
; CHECK-LABEL: @test_pr59459(
; CHECK-NEXT: iter.check:
; CHECK-NEXT: [[TMP0:%.*]] = trunc i64 [[IV_START:%.*]] to i32
; CHECK-NEXT: [[SMAX1:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP0]], i32 92)
; CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[SMAX1]], [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP4:%.*]] = trunc i64 [[IV_START]] to i32
; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP4]], i32 92)
; CHECK-NEXT: [[TMP5:%.*]] = sub i32 [[SMAX]], [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = add i32 [[TMP4]], -1
; CHECK-NEXT: [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
; CHECK-NEXT: [[TMP8:%.*]] = icmp ult i32 [[TMP7]], [[TMP6]]
; CHECK-NEXT: br i1 [[TMP8]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK: vector.main.loop.iter.check:
; CHECK-NEXT: [[MIN_ITERS_CHECK2:%.*]] = icmp ult i64 [[TMP3]], 16
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK2]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 16
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
; CHECK-NEXT: [[IND_END:%.*]] = add i64 [[IV_START]], [[N_VEC]]
; CHECK-NEXT: [[TMP9:%.*]] = trunc i64 [[IV_START]] to i32
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <16 x i32> poison, i32 [[TMP9]], i64 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <16 x i32> [[DOTSPLATINSERT]], <16 x i32> poison, <16 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <16 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <16 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 [[IV_START]], [[INDEX]]
; CHECK-NEXT: [[TMP10:%.*]] = trunc i64 [[OFFSET_IDX]] to i32
; CHECK-NEXT: [[TMP12:%.*]] = add i32 [[TMP10]], -1
; CHECK-NEXT: [[TMP13:%.*]] = mul <16 x i32> [[VEC_IND]], splat (i32 196608)
; CHECK-NEXT: [[TMP14:%.*]] = lshr exact <16 x i32> [[TMP13]], splat (i32 16)
; CHECK-NEXT: [[TMP15:%.*]] = trunc <16 x i32> [[TMP14]] to <16 x i16>
; CHECK-NEXT: [[TMP16:%.*]] = zext i32 [[TMP12]] to i64
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr i16, ptr [[ARR:%.*]], i64 [[TMP16]]
; CHECK-NEXT: store <16 x i16> [[TMP15]], ptr [[TMP17]], align 2
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <16 x i32> [[VEC_IND]], splat (i32 16)
; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK: vec.epilog.iter.check:
; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_MOD_VF]], 4
; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]], !prof [[PROF3:![0-9]+]]
; CHECK: vec.epilog.ph:
; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ [[IV_START]], [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT: [[N_MOD_VF3:%.*]] = urem i64 [[TMP3]], 4
; CHECK-NEXT: [[N_VEC4:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF3]]
; CHECK-NEXT: [[IND_END5:%.*]] = add i64 [[IV_START]], [[N_VEC4]]
; CHECK-NEXT: [[TMP20:%.*]] = trunc i64 [[BC_RESUME_VAL]] to i32
; CHECK-NEXT: [[DOTSPLATINSERT9:%.*]] = insertelement <4 x i32> poison, i32 [[TMP20]], i64 0
; CHECK-NEXT: [[DOTSPLAT10:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT9]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION11:%.*]] = add <4 x i32> [[DOTSPLAT10]], <i32 0, i32 1, i32 2, i32 3>
; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK: vec.epilog.vector.body:
; CHECK-NEXT: [[INDEX8:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT15:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND12:%.*]] = phi <4 x i32> [ [[INDUCTION11]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT13:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT: [[OFFSET_IDX14:%.*]] = add i64 [[IV_START]], [[INDEX8]]
; CHECK-NEXT: [[TMP21:%.*]] = trunc i64 [[OFFSET_IDX14]] to i32
; CHECK-NEXT: [[TMP23:%.*]] = add i32 [[TMP21]], -1
; CHECK-NEXT: [[TMP24:%.*]] = mul <4 x i32> [[VEC_IND12]], splat (i32 196608)
; CHECK-NEXT: [[TMP25:%.*]] = lshr exact <4 x i32> [[TMP24]], splat (i32 16)
; CHECK-NEXT: [[TMP26:%.*]] = trunc <4 x i32> [[TMP25]] to <4 x i16>
; CHECK-NEXT: [[TMP27:%.*]] = zext i32 [[TMP23]] to i64
; CHECK-NEXT: [[TMP28:%.*]] = getelementptr i16, ptr [[ARR]], i64 [[TMP27]]
; CHECK-NEXT: store <4 x i16> [[TMP26]], ptr [[TMP28]], align 2
; CHECK-NEXT: [[INDEX_NEXT15]] = add nuw i64 [[INDEX8]], 4
; CHECK-NEXT: [[VEC_IND_NEXT13]] = add <4 x i32> [[VEC_IND12]], splat (i32 4)
; CHECK-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT15]], [[N_VEC4]]
; CHECK-NEXT: br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: vec.epilog.middle.block:
; CHECK-NEXT: [[CMP_N16:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC4]]
; CHECK-NEXT: br i1 [[CMP_N16]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK: vec.epilog.scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL7:%.*]] = phi i64 [ [[IND_END5]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ [[IV_START]], [[VECTOR_SCEVCHECK]] ], [ [[IV_START]], [[ITER_CHECK:%.*]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL7]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_TRUNC:%.*]] = trunc i64 [[IV]] to i32
; CHECK-NEXT: [[STORE_IDX:%.*]] = add i32 [[IV_TRUNC]], -1
; CHECK-NEXT: [[X:%.*]] = mul i32 [[IV_TRUNC]], 196608
; CHECK-NEXT: [[Y:%.*]] = lshr exact i32 [[X]], 16
; CHECK-NEXT: [[STORE_VAL:%.*]] = trunc i32 [[Y]] to i16
; CHECK-NEXT: [[STORE_IDX_WIDE:%.*]] = zext i32 [[STORE_IDX]] to i64
; CHECK-NEXT: [[ADDR:%.*]] = getelementptr i16, ptr [[ARR]], i64 [[STORE_IDX_WIDE]]
; CHECK-NEXT: store i16 [[STORE_VAL]], ptr [[ADDR]], align 2
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp sgt i32 [[IV_TRUNC]], 91
; CHECK-NEXT: br i1 [[LOOP_COND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ %iv.start, %entry ], [ %iv.next, %loop ]
%iv.trunc = trunc i64 %iv to i32
%store.idx = add i32 %iv.trunc, -1
%x = mul i32 %iv.trunc, 196608
%y = lshr exact i32 %x, 16
%store.val = trunc i32 %y to i16
%store.idx.wide = zext i32 %store.idx to i64
%addr = getelementptr i16, ptr %arr, i64 %store.idx.wide
store i16 %store.val, ptr %addr
%iv.next = add nuw nsw i64 %iv, 1
%loop.cond = icmp sgt i32 %iv.trunc, 91
br i1 %loop.cond, label %exit, label %loop
exit:
ret void
}
define void @test_induction_step_needs_expansion(ptr noalias %j, ptr %k, i64 %l, i16 %off) {
; CHECK-LABEL: @test_induction_step_needs_expansion(
; CHECK-NEXT: iter.check:
; CHECK-NEXT: [[TMP0:%.*]] = sub i16 0, [[OFF:%.*]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[L:%.*]], 8
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK: vector.main.loop.iter.check:
; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[L]], 64
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[L]], 64
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[L]], [[N_MOD_VF]]
; CHECK-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <16 x i16> poison, i16 [[TMP0]], i64 0
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <16 x i16> [[DOTSPLATINSERT2]], <16 x i16> poison, <16 x i32> zeroinitializer
; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i16
; CHECK-NEXT: [[IND_END:%.*]] = mul i16 [[DOTCAST]], [[TMP0]]
; CHECK-NEXT: [[TMP1:%.*]] = mul <16 x i16> splat (i16 16), [[TMP2]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <16 x i16> poison, i16 [[OFF]], i64 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <16 x i16> [[DOTSPLATINSERT]], <16 x i16> poison, <16 x i32> zeroinitializer
; CHECK-NEXT: [[TMP11:%.*]] = mul <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8, i16 9, i16 10, i16 11, i16 12, i16 13, i16 14, i16 15>, [[TMP2]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <16 x i16> [ [[TMP11]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <16 x i16> [[VEC_IND]], [[TMP1]]
; CHECK-NEXT: [[STEP_ADD_2:%.*]] = add <16 x i16> [[STEP_ADD]], [[TMP1]]
; CHECK-NEXT: [[STEP_ADD_3:%.*]] = add <16 x i16> [[STEP_ADD_2]], [[TMP1]]
; CHECK-NEXT: [[TMP4:%.*]] = sub <16 x i16> [[VEC_IND]], [[DOTSPLAT]]
; CHECK-NEXT: [[TMP5:%.*]] = sub <16 x i16> [[STEP_ADD]], [[DOTSPLAT]]
; CHECK-NEXT: [[TMP6:%.*]] = sub <16 x i16> [[STEP_ADD_2]], [[DOTSPLAT]]
; CHECK-NEXT: [[TMP7:%.*]] = sub <16 x i16> [[STEP_ADD_3]], [[DOTSPLAT]]
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[K:%.*]], i64 [[INDEX]]
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i64 16
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i64 32
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i64 48
; CHECK-NEXT: store <16 x i16> [[TMP4]], ptr [[TMP8]], align 2
; CHECK-NEXT: store <16 x i16> [[TMP5]], ptr [[TMP10]], align 2
; CHECK-NEXT: store <16 x i16> [[TMP6]], ptr [[TMP21]], align 2
; CHECK-NEXT: store <16 x i16> [[TMP7]], ptr [[TMP12]], align 2
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 64
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <16 x i16> [[STEP_ADD_3]], [[TMP1]]
; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[L]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK: vec.epilog.iter.check:
; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_MOD_VF]], 8
; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]], !prof [[PROF7:![0-9]+]]
; CHECK: vec.epilog.ph:
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i16 [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT: [[N_MOD_VF4:%.*]] = urem i64 [[L]], 8
; CHECK-NEXT: [[N_VEC5:%.*]] = sub i64 [[L]], [[N_MOD_VF4]]
; CHECK-NEXT: [[DOTCAST7:%.*]] = trunc i64 [[N_VEC5]] to i16
; CHECK-NEXT: [[IND_END8:%.*]] = mul i16 [[DOTCAST7]], [[TMP0]]
; CHECK-NEXT: [[DOTSPLATINSERT13:%.*]] = insertelement <8 x i16> poison, i16 [[OFF]], i64 0
; CHECK-NEXT: [[DOTSPLAT14:%.*]] = shufflevector <8 x i16> [[DOTSPLATINSERT13]], <8 x i16> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: [[DOTSPLATINSERT15:%.*]] = insertelement <8 x i16> poison, i16 [[BC_RESUME_VAL]], i64 0
; CHECK-NEXT: [[DOTSPLAT16:%.*]] = shufflevector <8 x i16> [[DOTSPLATINSERT15]], <8 x i16> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: [[DOTSPLATINSERT14:%.*]] = insertelement <8 x i16> poison, i16 [[TMP0]], i64 0
; CHECK-NEXT: [[DOTSPLAT15:%.*]] = shufflevector <8 x i16> [[DOTSPLATINSERT14]], <8 x i16> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: [[TMP22:%.*]] = mul <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, [[DOTSPLAT15]]
; CHECK-NEXT: [[INDUCTION17:%.*]] = add <8 x i16> [[DOTSPLAT16]], [[TMP22]]
; CHECK-NEXT: [[TMP15:%.*]] = shl i16 [[TMP0]], 3
; CHECK-NEXT: [[BROADCAST_SPLATINSERT22:%.*]] = insertelement <8 x i16> poison, i16 [[TMP15]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT23:%.*]] = shufflevector <8 x i16> [[BROADCAST_SPLATINSERT22]], <8 x i16> poison, <8 x i32> zeroinitializer
; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK: vec.epilog.vector.body:
; CHECK-NEXT: [[INDEX12:%.*]] = phi i64 [ [[BC_RESUME_VAL1]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT24:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND20:%.*]] = phi <8 x i16> [ [[INDUCTION17]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT21:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP17:%.*]] = sub <8 x i16> [[VEC_IND20]], [[DOTSPLAT14]]
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i16, ptr [[K]], i64 [[INDEX12]]
; CHECK-NEXT: store <8 x i16> [[TMP17]], ptr [[TMP18]], align 2
; CHECK-NEXT: [[INDEX_NEXT24]] = add nuw i64 [[INDEX12]], 8
; CHECK-NEXT: [[VEC_IND_NEXT21]] = add <8 x i16> [[VEC_IND20]], [[BROADCAST_SPLAT23]]
; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT24]], [[N_VEC5]]
; CHECK-NEXT: br i1 [[TMP20]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: vec.epilog.middle.block:
; CHECK-NEXT: [[CMP_N25:%.*]] = icmp eq i64 [[L]], [[N_VEC5]]
; CHECK-NEXT: br i1 [[CMP_N25]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK: vec.epilog.scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL6:%.*]] = phi i64 [ [[N_VEC5]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
; CHECK-NEXT: [[BC_RESUME_VAL11:%.*]] = phi i16 [ [[IND_END8]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL6]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[P_09:%.*]] = phi i16 [ [[BC_RESUME_VAL11]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[ADD]] = sub i16 [[P_09]], [[OFF]]
; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, ptr [[K]], i64 [[IV]]
; CHECK-NEXT: store i16 [[ADD]], ptr [[ARRAYIDX3]], align 2
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[L]]
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%p.09 = phi i16 [ 0, %entry ], [ %add, %loop ]
%add = sub i16 %p.09, %off
%arrayidx3 = getelementptr inbounds i16, ptr %k, i64 %iv
store i16 %add, ptr %arrayidx3, align 2
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %l
br i1 %ec , label %exit, label %loop
exit:
ret void
}
; Test case for https://github.com/llvm/llvm-project/issues/151686.
define i8 @multiple_inductions_start_at_0() {
; CHECK-LABEL: @multiple_inductions_start_at_0(
; CHECK-NEXT: iter.check:
; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK: vector.main.loop.iter.check:
; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[STEP_ADD_3:%.*]] = phi <32 x i8> [ zeroinitializer, [[VECTOR_PH]] ], [ [[STEP_ADD_3]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 128
; CHECK-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
; CHECK-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <32 x i8> [[STEP_ADD_3]], i32 31
; CHECK-NEXT: br i1 false, label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK: vec.epilog.iter.check:
; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]], !prof [[PROF11:![0-9]+]]
; CHECK: vec.epilog.ph:
; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i32 [ 1024, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT: [[TMP2:%.*]] = trunc i32 [[BC_RESUME_VAL]] to i8
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i8> poison, i8 [[TMP2]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK: vec.epilog.vector.body:
; CHECK-NEXT: [[INDEX1:%.*]] = phi i32 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND2:%.*]] = phi <4 x i8> [ [[BROADCAST_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND2]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT: [[INDEX_NEXT3]] = add nuw i32 [[INDEX1]], 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT3]], 1052
; CHECK-NEXT: br i1 [[TMP3]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: vec.epilog.middle.block:
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x i8> [[VEC_IND2]], i32 3
; CHECK-NEXT: br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK: vec.epilog.scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL5:%.*]] = phi i32 [ 1052, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 1024, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
; CHECK-NEXT: [[BC_RESUME_VAL6:%.*]] = phi i32 [ -469762048, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 0, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL5]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL6]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[ADD]] = add i32 [[IV_2]], -16777216
; CHECK-NEXT: [[TRUNC:%.*]] = trunc i32 [[IV_2]] to i8
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp ugt i32 [[IV]], 1050
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[RES:%.*]] = phi i8 [ [[TRUNC]], [[LOOP]] ], [ [[TMP1]], [[MIDDLE_BLOCK]] ], [ [[TMP4]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i8 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
%iv.2 = phi i32 [ 0, %entry ], [ %add, %loop ]
%add = add i32 %iv.2, -16777216
%trunc = trunc i32 %iv.2 to i8
%iv.next = add i32 %iv, 1
%ec = icmp ugt i32 %iv, 1050
br i1 %ec, label %exit, label %loop
exit:
%res = phi i8 [ %trunc, %loop ]
ret i8 %res
}
Reference in New Issue View Git Blame Copy Permalink