Doomhowl-Interactive/llvm-project
0
0
Fork 0
Code Issues Pull Requests Actions 5 Packages Projects Releases Wiki Activity
Files
main
llvm-project/llvm/test/Transforms/LoopVectorize/minmax-extra-outside-users.ll
Madhur Amilkanthwar 2a7313a2ff [LV] Relax OutOfLoopUses check in getMinMaxRecurrence() (#189906) 
For more context of this patch, please see #182473. By relaxing,
`OutOfLoopUses` check, we can recognize more case for minmax reduction
and thus vectorize.
2026-04-23 10:43:41 +05:30

141 lines
7.5 KiB
LLVM
Raw Permalink Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt -p loop-vectorize -force-vector-width=4 -S %s | FileCheck %s
define i32 @smin_three_exit_users(ptr %src, i64 %n) {
; CHECK-LABEL: define i32 @smin_three_exit_users(
; 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: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 2147483647), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP2]] = select <4 x i1> [[TMP1]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP4:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP2]])
; 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 i32 [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ 2147483647, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MIN:%.*]] = phi i32 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MIN_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[V:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[MIN]], [[V]]
; CHECK-NEXT: [[MIN_NEXT]] = select i1 [[CMP]], i32 [[MIN]], i32 [[V]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[R0:%.*]] = phi i32 [ [[MIN_NEXT]], %[[LOOP]] ], [ [[TMP4]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[R1:%.*]] = phi i32 [ [[MIN_NEXT]], %[[LOOP]] ], [ [[TMP4]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[R2:%.*]] = phi i32 [ [[MIN_NEXT]], %[[LOOP]] ], [ [[TMP4]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[SUM0:%.*]] = add i32 [[R0]], [[R1]]
; CHECK-NEXT: [[SUM1:%.*]] = add i32 [[SUM0]], [[R2]]
; CHECK-NEXT: ret i32 [[SUM1]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%min = phi i32 [ 2147483647, %entry ], [ %min.next, %loop ]
%gep = getelementptr inbounds i32, ptr %src, i64 %iv
%v = load i32, ptr %gep, align 4
%cmp = icmp slt i32 %min, %v
%min.next = select i1 %cmp, i32 %min, i32 %v
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
%r0 = phi i32 [ %min.next, %loop ]
%r1 = phi i32 [ %min.next, %loop ]
%r2 = phi i32 [ %min.next, %loop ]
%sum0 = add i32 %r0, %r1
%sum1 = add i32 %sum0, %r2
ret i32 %sum1
}
define i32 @smax_three_exit_users(ptr %src, i64 %n) {
; CHECK-LABEL: define i32 @smax_three_exit_users(
; 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: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -2147483648), %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = icmp sgt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP2]] = select <4 x i1> [[TMP1]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP4:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP2]])
; 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 i32 [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ -2147483648, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi i32 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[V:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 [[MAX]], [[V]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], i32 [[MAX]], i32 [[V]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[R0:%.*]] = phi i32 [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP4]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[R1:%.*]] = phi i32 [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP4]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[R2:%.*]] = phi i32 [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP4]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[SUM0:%.*]] = add i32 [[R0]], [[R1]]
; CHECK-NEXT: [[SUM1:%.*]] = add i32 [[SUM0]], [[R2]]
; CHECK-NEXT: ret i32 [[SUM1]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi i32 [ -2147483648, %entry ], [ %max.next, %loop ]
%gep = getelementptr inbounds i32, ptr %src, i64 %iv
%v = load i32, ptr %gep, align 4
%cmp = icmp sgt i32 %max, %v
%max.next = select i1 %cmp, i32 %max, i32 %v
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
%r0 = phi i32 [ %max.next, %loop ]
%r1 = phi i32 [ %max.next, %loop ]
%r2 = phi i32 [ %max.next, %loop ]
%sum0 = add i32 %r0, %r1
%sum1 = add i32 %sum0, %r2
ret i32 %sum1
}
Reference in New Issue View Git Blame Copy Permalink