shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/select-umin-first-index.ll
Florian Hahn 4042975b63
[LV] Support argmin/argmax with strict predicates. (#170223) 
Extend handleMultiUseReductions to support strict predicates (>, <),
matching the first index instead of the last for non-strict predicates.

Builds on top of https://github.com/llvm/llvm-project/pull/141431.

FindLast reductions with strict predicates are adjusted to compute the
correct result as follows:

1. Find the first canonical indices corresponding to partial min/max
   values, using loop reductions.
2. Find which of the partial min/max values are equal to the overall
    min/max value.
3. Select among the canonical indices those corresponding to the overall
    min/max value.
4. Find the first canonical index of overall min/max and scale it back to
    the original IV using VPDerivedIVRecipe.
5. If the overall min/max equals the starting min/max, the condition in
    the loop was always false, due to being strict; return the original start 
    value in that case.
2026-02-19 10:52:27 +00:00

880 lines
45 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
; Test cases for selecting the index with the minimum value.
define i64 @test_vectorize_select_umin_idx(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ poison, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i64> [ splat (i64 -1), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[GEP1]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <4 x i64> [[VEC_PHI2]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP2]] = call <4 x i64> @llvm.umin.v4i64(<4 x i64> [[VEC_PHI2]], <4 x i64> [[WIDE_LOAD]])
; CHECK-NEXT: [[TMP3]] = select <4 x i1> [[TMP1]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV1]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP2]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP5]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP2]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[TMP3]], <4 x i64> splat (i64 -1)
; CHECK-NEXT: [[TMP8:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP7]])
; CHECK-NEXT: [[RDX_SELECT_CMP:%.*]] = icmp eq i64 [[TMP5]], -1
; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 0, i64 [[TMP8]]
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX2:%.*]] = phi i64 [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ [[BC_MERGE_RDX2]], %[[SCALAR_PH]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ], [ [[TMP10]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_signed_sentinel_possible(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_signed_sentinel_possible(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: br label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; 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 <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ poison, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i64> [ splat (i64 -2), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP0]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <4 x i64> [[VEC_PHI1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP2]] = call <4 x i64> @llvm.umin.v4i64(<4 x i64> [[VEC_PHI1]], <4 x i64> [[WIDE_LOAD]])
; CHECK-NEXT: [[TMP3]] = select <4 x i1> [[TMP1]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
; CHECK-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP2]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP5]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP2]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[TMP3]], <4 x i64> splat (i64 -1)
; CHECK-NEXT: [[TMP8:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP7]])
; CHECK-NEXT: [[RDX_SELECT_CMP:%.*]] = icmp eq i64 [[TMP5]], -2
; CHECK-NEXT: [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 0, i64 [[TMP8]]
; CHECK-NEXT: br label %[[EXIT:.*]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret i64 [[RDX_SELECT]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -2, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, 100
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_cond_flipped(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_cond_flipped(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ poison, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i64> [ splat (i64 -1), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[GEP1]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp ult <4 x i64> [[WIDE_LOAD]], [[VEC_PHI2]]
; CHECK-NEXT: [[TMP2]] = call <4 x i64> @llvm.umin.v4i64(<4 x i64> [[VEC_PHI2]], <4 x i64> [[WIDE_LOAD]])
; CHECK-NEXT: [[TMP3]] = select <4 x i1> [[TMP1]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV1]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP2]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP5]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP2]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[TMP3]], <4 x i64> splat (i64 -1)
; CHECK-NEXT: [[TMP8:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP7]])
; CHECK-NEXT: [[RDX_SELECT_CMP:%.*]] = icmp eq i64 [[TMP5]], -1
; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 0, i64 [[TMP8]]
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX2:%.*]] = phi i64 [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ [[BC_MERGE_RDX2]], %[[SCALAR_PH]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[L]], [[MIN_VAL]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ], [ [[TMP10]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ult i64 %l, %min.val
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_select_ops_flipped(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_select_ops_flipped(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ 100, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP1]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[L]], [[MIN_VAL]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[MIN_IDX]], i64 [[IV1]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ 100, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ult i64 %l, %min.val
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %min.idx, i64 %iv
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_via_select_idx(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_via_select_idx(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ 100, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = select i1 [[CMP]], i64 [[L]], i64 [[MIN_VAL]]
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ 100, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = select i1 %cmp, i64 %l, i64 %min.val
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_all_exit_inst(ptr %src, ptr %umin, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_all_exit_inst(
; CHECK-SAME: ptr [[SRC:%.*]], ptr [[UMIN:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -20, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: [[RES_UMIN:%.*]] = phi i64 [ [[MIN_VAL_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: store i64 [[RES_UMIN]], ptr [[UMIN]], align 4
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -20, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
%res.umin = phi i64 [ %min.val.next, %loop ]
store i64 %res.umin, ptr %umin
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_min_ops_switched(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_min_ops_switched(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ poison, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i64> [ splat (i64 -1), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[GEP1]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <4 x i64> [[VEC_PHI2]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP2]] = call <4 x i64> @llvm.umin.v4i64(<4 x i64> [[WIDE_LOAD]], <4 x i64> [[VEC_PHI2]])
; CHECK-NEXT: [[TMP3]] = select <4 x i1> [[TMP1]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV1]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP2]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP5]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP2]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[TMP3]], <4 x i64> splat (i64 -1)
; CHECK-NEXT: [[TMP8:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP7]])
; CHECK-NEXT: [[RDX_SELECT_CMP:%.*]] = icmp eq i64 [[TMP5]], -1
; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 0, i64 [[TMP8]]
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX2:%.*]] = phi i64 [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ [[BC_MERGE_RDX2]], %[[SCALAR_PH]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[L]], i64 [[MIN_VAL]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ], [ [[TMP10]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %l, i64 %min.val)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_not_vectorize_select_no_min_reduction(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_not_vectorize_select_no_min_reduction(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[RED_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[RED_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[RED_VAL]], [[L]]
; CHECK-NEXT: [[RED_VAL_NEXT]] = add i64 [[RED_VAL]], [[L]]
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%red.val = phi i64 [ -1, %entry ], [ %red.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %red.val, %l
%red.val.next = add i64 %red.val, %l
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_cmp_and_umin_use_different_values(ptr %src, i64 %x, i64 %n) {
; CHECK-LABEL: define i64 @test_cmp_and_umin_use_different_values(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[X:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[X]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %x
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i32 @test_vectorize_select_umin_idx_with_trunc(ptr %src, i64 %n) {
; CHECK-LABEL: define i32 @test_vectorize_select_umin_idx_with_trunc(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[TRUNC:%.*]] = trunc i64 [[IV]] to i32
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i32 [[TRUNC]], i32 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i32 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i32 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i32 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%trunc = trunc i64 %iv to i32
%min.idx.next = select i1 %cmp, i32 %trunc, i32 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i32 [ %min.idx.next, %loop ]
ret i32 %res
}
define i32 @test_vectorize_select_umin_idx_with_trunc_valid(ptr %src, i64 %n) {
; CHECK-LABEL: define i32 @test_vectorize_select_umin_idx_with_trunc_valid(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[TRUNC:%.*]] = trunc i64 [[IV]] to i32
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i32 [[TRUNC]], i32 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 100
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i32 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i32 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min.idx = phi i32 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%trunc = trunc i64 %iv to i32
%min.idx.next = select i1 %cmp, i32 %trunc, i32 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, 100
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i32 [ %min.idx.next, %loop ]
ret i32 %res
}
define ptr @test_with_ptr_index(ptr %start, ptr %end) {
; CHECK-LABEL: define ptr @test_with_ptr_index(
; CHECK-SAME: ptr [[START:%.*]], ptr [[END:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi ptr [ [[START]], %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi ptr [ null, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[IV]], align 4
; CHECK-NEXT: [[CMP7_US:%.*]] = icmp ult i64 [[L]], [[MIN_VAL]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP7_US]], ptr [[IV]], ptr [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = getelementptr i32, ptr [[IV]], i64 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq ptr [[IV_NEXT]], [[END]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi ptr [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret ptr [[RES]]
;
entry:
br label %loop
loop:
%iv = phi ptr [ %start, %entry ], [ %iv.next, %loop ]
%min.idx = phi ptr [ null, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%l = load i64, ptr %iv
%cmp7.us = icmp ult i64 %l, %min.val
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp7.us, ptr %iv, ptr %min.idx
%iv.next = getelementptr i32, ptr %iv, i64 1
%exitcond.not = icmp eq ptr %iv.next, %end
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi ptr [ %min.idx.next, %loop ]
ret ptr %res
}
define i64 @test_no_vectorize_select_iv_decrement(ptr %src) {
; CHECK-LABEL: define i64 @test_no_vectorize_select_iv_decrement(
; CHECK-SAME: ptr [[SRC:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 1000, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], -1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 0
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 1000, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, -1
%exitcond.not = icmp eq i64 %iv.next, 0
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_no_vectorize_select_iv_sub(ptr %src) {
; CHECK-LABEL: define i64 @test_no_vectorize_select_iv_sub(
; CHECK-SAME: ptr [[SRC:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 1000, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = sub i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 0
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 1000, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = sub i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, 0
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_no_vectorize_select_iv_mul(ptr %src) {
; CHECK-LABEL: define i64 @test_no_vectorize_select_iv_mul(
; CHECK-SAME: ptr [[SRC:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 1, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = mul i64 [[IV]], 2
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 128
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = mul i64 %iv, 2
%exitcond.not = icmp eq i64 %iv.next, 128
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_wraps(ptr %src, i64 %n, i64 %start) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_wraps(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]], i64 [[START:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[IDX:%.*]] = phi i64 [ [[START]], %[[ENTRY]] ], [ [[IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ -1, %[[ENTRY]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IDX]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[IDX_NEXT]] = add i64 [[IDX]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%idx = phi i64 [ %start, %entry ], [ %idx.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %idx, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%idx.next = add i64 %idx, 1
%exitcond.not = icmp eq i64 %iv.next, %n
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
define i64 @test_vectorize_select_umin_idx_iv_start_different(ptr %src, i64 %n) {
; CHECK-LABEL: define i64 @test_vectorize_select_umin_idx_iv_start_different(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: br label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; 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 <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ poison, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i64> [ splat (i64 -1), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[IV:%.*]] = add i64 10, [[INDEX]]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[GEP]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt <4 x i64> [[VEC_PHI1]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP2]] = call <4 x i64> @llvm.umin.v4i64(<4 x i64> [[VEC_PHI1]], <4 x i64> [[WIDE_LOAD]])
; CHECK-NEXT: [[TMP3]] = select <4 x i1> [[TMP1]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 9988
; CHECK-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP2]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP5]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP2]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[TMP3]], <4 x i64> splat (i64 -1)
; CHECK-NEXT: [[TMP8:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP7]])
; CHECK-NEXT: [[DERIVED_IV_RESULT:%.*]] = add i64 10, [[TMP8]]
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[TMP5]], -1
; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i64 0, i64 [[DERIVED_IV_RESULT]]
; CHECK-NEXT: br label %[[SCALAR_PH:.*]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ 9998, %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_IDX:%.*]] = phi i64 [ [[TMP10]], %[[SCALAR_PH]] ], [ [[MIN_IDX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN_VAL:%.*]] = phi i64 [ [[TMP5]], %[[SCALAR_PH]] ], [ [[MIN_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i64, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load i64, ptr [[GEP1]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[MIN_VAL]], [[L]]
; CHECK-NEXT: [[MIN_VAL_NEXT]] = tail call i64 @llvm.umin.i64(i64 [[MIN_VAL]], i64 [[L]])
; CHECK-NEXT: [[MIN_IDX_NEXT]] = select i1 [[CMP]], i64 [[IV1]], i64 [[MIN_IDX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 10000
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[RES:%.*]] = phi i64 [ [[MIN_IDX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret i64 [[RES]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 10, %entry ], [ %iv.next, %loop ]
%min.idx = phi i64 [ 0, %entry ], [ %min.idx.next, %loop ]
%min.val = phi i64 [ -1, %entry ], [ %min.val.next, %loop ]
%gep = getelementptr i64, ptr %src, i64 %iv
%l = load i64, ptr %gep
%cmp = icmp ugt i64 %min.val, %l
%min.val.next = tail call i64 @llvm.umin.i64(i64 %min.val, i64 %l)
%min.idx.next = select i1 %cmp, i64 %iv, i64 %min.idx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, 10000
br i1 %exitcond.not, label %exit, label %loop
exit:
%res = phi i64 [ %min.idx.next, %loop ]
ret i64 %res
}
declare i64 @llvm.umin.i64(i64, i64)
declare i16 @llvm.umin.i16(i16, i16)
Reference in New Issue View Git Blame Copy Permalink