[llvm] 686600b - [LV] Add test showing missed removal of implied predicate.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Sun May 26 17:24:29 PDT 2024
Author: Florian Hahn
Date: 2024-05-26T17:23:14-07:00
New Revision: 686600b521aa3ca5fb9f015d25636067e2c5c9fa
URL: https://github.com/llvm/llvm-project/commit/686600b521aa3ca5fb9f015d25636067e2c5c9fa
DIFF: https://github.com/llvm/llvm-project/commit/686600b521aa3ca5fb9f015d25636067e2c5c9fa.diff
LOG: [LV] Add test showing missed removal of implied predicate.
Tests for https://github.com/llvm/llvm-project/pull/93397
Added:
Modified:
llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
Removed:
################################################################################
diff --git a/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll b/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
index 6ae6645378b32..63b3bb14e550d 100644
--- a/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
+++ b/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
@@ -149,3 +149,111 @@ for.cond: ; preds = %for.body, %entry
for.end: ; preds = %for.cond
ret void
}
+
+ at h = global i64 0
+
+; TODO: Currently we generate SCEV check code for the same predicate twice.
+define void @implied_wrap_predicate(ptr %A, ptr %B, ptr %C) {
+; CHECK-LABEL: define void @implied_wrap_predicate
+; CHECK-SAME: (ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]]) {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[A3:%.*]] = ptrtoint ptr [[A]] to i64
+; CHECK-NEXT: [[C2:%.*]] = ptrtoint ptr [[C]] to i64
+; CHECK-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[A3]], 16
+; CHECK-NEXT: [[UMAX4:%.*]] = call i64 @llvm.umax.i64(i64 [[TMP0]], i64 add (i64 ptrtoint (ptr @h to i64), i64 1))
+; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[UMAX4]], -9
+; CHECK-NEXT: [[TMP2:%.*]] = sub i64 [[TMP1]], [[A3]]
+; CHECK-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP2]], 3
+; CHECK-NEXT: [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 1
+; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP4]], 4
+; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK: vector.scevcheck:
+; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[A1]], 16
+; CHECK-NEXT: [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[TMP5]], i64 add (i64 ptrtoint (ptr @h to i64), i64 1))
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[UMAX]], -9
+; CHECK-NEXT: [[TMP7:%.*]] = sub i64 [[TMP6]], [[A1]]
+; CHECK-NEXT: [[TMP8:%.*]] = lshr i64 [[TMP7]], 3
+; CHECK-NEXT: [[TMP9:%.*]] = trunc i64 [[TMP8]] to i16
+; CHECK-NEXT: [[TMP10:%.*]] = add i16 1, [[TMP9]]
+; CHECK-NEXT: [[TMP11:%.*]] = icmp ult i16 [[TMP10]], 1
+; CHECK-NEXT: [[TMP12:%.*]] = icmp ugt i64 [[TMP8]], 65535
+; CHECK-NEXT: [[TMP13:%.*]] = or i1 [[TMP11]], [[TMP12]]
+; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[TMP8]] to i16
+; CHECK-NEXT: [[TMP15:%.*]] = add i16 2, [[TMP14]]
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ult i16 [[TMP15]], 2
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ugt i64 [[TMP8]], 65535
+; CHECK-NEXT: [[TMP18:%.*]] = or i1 [[TMP16]], [[TMP17]]
+; CHECK-NEXT: [[TMP19:%.*]] = trunc i64 [[TMP8]] to i16
+; CHECK-NEXT: [[TMP20:%.*]] = add i16 1, [[TMP19]]
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ult i16 [[TMP20]], 1
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ugt i64 [[TMP8]], 65535
+; CHECK-NEXT: [[TMP23:%.*]] = or i1 [[TMP21]], [[TMP22]]
+; CHECK-NEXT: [[TMP24:%.*]] = or i1 [[TMP13]], [[TMP18]]
+; CHECK-NEXT: [[TMP25:%.*]] = or i1 [[TMP24]], [[TMP23]]
+; CHECK-NEXT: br i1 [[TMP25]], label [[SCALAR_PH]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK: vector.memcheck:
+; CHECK-NEXT: [[TMP26:%.*]] = sub i64 [[C2]], [[A3]]
+; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP26]], 32
+; CHECK-NEXT: br i1 [[DIFF_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK: vector.ph:
+; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP4]], 4
+; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP4]], [[N_MOD_VF]]
+; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i16
+; CHECK-NEXT: [[IND_END:%.*]] = add i16 1, [[DOTCAST]]
+; CHECK-NEXT: [[IND_END5:%.*]] = add i64 1, [[N_VEC]]
+; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
+; CHECK: vector.body:
+; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
+; CHECK-NEXT: [[TMP27:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP27]]
+; CHECK-NEXT: [[TMP29:%.*]] = getelementptr i64, ptr [[TMP28]], i32 0
+; CHECK-NEXT: store <4 x i64> zeroinitializer, ptr [[TMP29]], align 4
+; CHECK-NEXT: [[TMP30:%.*]] = getelementptr i64, ptr [[C]], i64 [[TMP27]]
+; CHECK-NEXT: [[TMP31:%.*]] = getelementptr i64, ptr [[TMP30]], i32 0
+; CHECK-NEXT: store <4 x i64> zeroinitializer, ptr [[TMP31]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT: [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK: middle.block:
+; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP4]], [[N_VEC]]
+; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK: scalar.ph:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i16 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ], [ 1, [[VECTOR_SCEVCHECK]] ], [ 1, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT: [[BC_RESUME_VAL6:%.*]] = phi i64 [ [[IND_END5]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ], [ 1, [[VECTOR_SCEVCHECK]] ], [ 1, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT: [[IV_EXT:%.*]] = phi i64 [ [[BC_RESUME_VAL6]], [[SCALAR_PH]] ], [ [[IV_EXT_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT: [[GEP_A:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV_EXT]]
+; CHECK-NEXT: store i64 0, ptr [[GEP_A]], align 4
+; CHECK-NEXT: [[GEP_C:%.*]] = getelementptr i64, ptr [[C]], i64 [[IV_EXT]]
+; CHECK-NEXT: store i64 0, ptr [[GEP_C]], align 4
+; CHECK-NEXT: [[IV_NEXT]] = add i16 [[IV]], 1
+; CHECK-NEXT: [[IV_EXT_NEXT]] = zext i16 [[IV_NEXT]] to i64
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV_EXT_NEXT]]
+; CHECK-NEXT: [[CMP:%.*]] = icmp ugt ptr [[GEP]], @h
+; CHECK-NEXT: br i1 [[CMP]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK: exit:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %loop
+
+loop:
+ %iv = phi i16 [ 1, %entry ], [ %iv.next, %loop ]
+ %iv.ext = phi i64 [ 1, %entry ], [ %iv.ext.next, %loop ]
+ %gep.A = getelementptr i64, ptr %A, i64 %iv.ext
+ store i64 0, ptr %gep.A
+ %gep.C = getelementptr i64, ptr %C, i64 %iv.ext
+ store i64 0, ptr %gep.C
+ %iv.next = add i16 %iv, 1
+ %iv.ext.next = zext i16 %iv.next to i64
+ %gep = getelementptr i64, ptr %A, i64 %iv.ext.next
+ %cmp = icmp ugt ptr %gep, @h
+ br i1 %cmp, label %exit, label %loop
+
+exit:
+ ret void
+}
More information about the llvm-commits
mailing list