[LV][NFC] Remove undef values in some test cases #164401
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Split off from PR llvm#163525, this standalone patch replaces simple cases where undef is used as a value for arithmetic or getelementptr instructions. This will reduce the likelihood of contributors hitting the `undef deprecator` warning in github.
david-arm
changed the title
|
@llvm/pr-subscribers-llvm-transforms Author: David Sherwood (david-arm) ChangesSplit off from PR #163525, this standalone patch replaces simple cases where undef is used as a value for arithmetic or getelementptr instructions. This will reduce the likelihood of contributors hitting the Patch is 33.93 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/164401.diff 13 Files Affected:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-cg-bug.ll b/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-cg-bug.ll
index e6b74062ad765..a33f8eb920039 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-cg-bug.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-cg-bug.ll
@@ -35,7 +35,7 @@ target triple = "x86_64-unknown-linux-gnu"
; This test was originally vectorized, but now SCEV is smart enough to prove
; that its trip count is 1, so it gets ignored by vectorizer.
; Function Attrs: uwtable
-define void @test_01(i1 %arg) {
+define void @test_01(ptr addrspace(1) %p, i1 %arg) {
br label %.outer
; <label>:1: ; preds = %2
@@ -57,8 +57,8 @@ define void @test_01(i1 %arg) {
%8 = phi i32 [ %.ph2, %.outer ], [ %7, %6 ]
%9 = add i32 %8, 2
%10 = zext i32 %9 to i64
- %11 = getelementptr inbounds i32, ptr addrspace(1) undef, i64 %10
- %12 = ashr i32 undef, %4
+ %11 = getelementptr inbounds i32, ptr addrspace(1) %p, i64 %10
+ %12 = ashr i32 12, %4
store i32 %12, ptr addrspace(1) %11, align 4
%13 = add i32 %7, 1
%14 = icmp sgt i32 %13, 61
@@ -74,7 +74,7 @@ define void @test_01(i1 %arg) {
; CHECK: store <4 x i32>
; Function Attrs: uwtable
-define void @test_02(i1 %arg) {
+define void @test_02(ptr addrspace(1) %p, i1 %arg) {
br label %.outer
; <label>:1: ; preds = %2
@@ -96,8 +96,8 @@ define void @test_02(i1 %arg) {
%8 = phi i32 [ %.ph2, %.outer ], [ %7, %6 ]
%9 = add i32 %8, 2
%10 = zext i32 %9 to i64
- %11 = getelementptr inbounds i32, ptr addrspace(1) undef, i64 %10
- %12 = ashr i32 undef, %4
+ %11 = getelementptr inbounds i32, ptr addrspace(1) %p, i64 %10
+ %12 = ashr i32 12, %4
store i32 %12, ptr addrspace(1) %11, align 4
%13 = add i32 %7, 1
%14 = icmp sgt i32 %13, 610
diff --git a/llvm/test/Transforms/LoopVectorize/X86/cost-model-assert.ll b/llvm/test/Transforms/LoopVectorize/X86/cost-model-assert.ll
index 4cff8753ba9b1..239366c59470e 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/cost-model-assert.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/cost-model-assert.ll
@@ -11,9 +11,9 @@
target datalayout = "e-m:w-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-w64-windows-gnu"
-define void @cff_index_load_offsets(i1 %cond, i8 %x, ptr %p) #0 {
+define void @cff_index_load_offsets(i1 %cond, i8 %x, ptr %p, ptr %pend) #0 {
; CHECK-LABEL: define void @cff_index_load_offsets(
-; CHECK-SAME: i1 [[COND:%.*]], i8 [[X:%.*]], ptr [[P:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-SAME: i1 [[COND:%.*]], i8 [[X:%.*]], ptr [[P:%.*]], ptr [[PEND:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: br i1 [[COND]], label %[[IF_THEN:.*]], label %[[EXIT:.*]]
; CHECK: [[IF_THEN]]:
@@ -26,14 +26,14 @@ define void @cff_index_load_offsets(i1 %cond, i8 %x, ptr %p) #0 {
; CHECK-NEXT: [[CONV73:%.*]] = zext i8 [[TMP0]] to i32
; CHECK-NEXT: [[SHL74:%.*]] = shl nuw nsw i32 [[CONV73]], 16
; CHECK-NEXT: [[OR75:%.*]] = or i32 [[SHL74]], [[SHL71]]
-; CHECK-NEXT: [[TMP1:%.*]] = load i8, ptr undef, align 1, !tbaa [[CHAR_TBAA1]]
-; CHECK-NEXT: [[SHL78:%.*]] = shl nuw nsw i32 undef, 8
+; CHECK-NEXT: [[TMP1:%.*]] = load i8, ptr [[P]], align 1, !tbaa [[CHAR_TBAA1]]
+; CHECK-NEXT: [[SHL78:%.*]] = shl nuw nsw i32 12, 8
; CHECK-NEXT: [[OR79:%.*]] = or i32 [[OR75]], [[SHL78]]
; CHECK-NEXT: [[CONV81:%.*]] = zext i8 [[TMP1]] to i32
; CHECK-NEXT: [[OR83:%.*]] = or i32 [[OR79]], [[CONV81]]
-; CHECK-NEXT: store i32 [[OR83]], ptr undef, align 4, !tbaa [[LONG_TBAA4:![0-9]+]]
+; CHECK-NEXT: store i32 [[OR83]], ptr [[P]], align 4, !tbaa [[LONG_TBAA4:![0-9]+]]
; CHECK-NEXT: [[ADD_PTR86]] = getelementptr inbounds i8, ptr [[P_359]], i64 4
-; CHECK-NEXT: [[CMP66:%.*]] = icmp ult ptr [[ADD_PTR86]], undef
+; CHECK-NEXT: [[CMP66:%.*]] = icmp ult ptr [[ADD_PTR86]], [[PEND]]
; CHECK-NEXT: br i1 [[CMP66]], label %[[FOR_BODY68]], label %[[SW_EPILOG:.*]]
; CHECK: [[SW_EPILOG]]:
; CHECK-NEXT: unreachable
@@ -54,14 +54,14 @@ for.body68: ; preds = %for.body68, %if.the
%conv73 = zext i8 %0 to i32
%shl74 = shl nuw nsw i32 %conv73, 16
%or75 = or i32 %shl74, %shl71
- %1 = load i8, ptr undef, align 1, !tbaa !1
- %shl78 = shl nuw nsw i32 undef, 8
+ %1 = load i8, ptr %p, align 1, !tbaa !1
+ %shl78 = shl nuw nsw i32 12, 8
%or79 = or i32 %or75, %shl78
%conv81 = zext i8 %1 to i32
%or83 = or i32 %or79, %conv81
- store i32 %or83, ptr undef, align 4, !tbaa !4
+ store i32 %or83, ptr %p, align 4, !tbaa !4
%add.ptr86 = getelementptr inbounds i8, ptr %p.359, i64 4
- %cmp66 = icmp ult ptr %add.ptr86, undef
+ %cmp66 = icmp ult ptr %add.ptr86, %pend
br i1 %cmp66, label %for.body68, label %sw.epilog
sw.epilog: ; preds = %for.body68
diff --git a/llvm/test/Transforms/LoopVectorize/X86/interleaved-accesses-large-gap.ll b/llvm/test/Transforms/LoopVectorize/X86/interleaved-accesses-large-gap.ll
index e75d469506376..acec9e47a94ee 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/interleaved-accesses-large-gap.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/interleaved-accesses-large-gap.ll
@@ -41,8 +41,8 @@ for.cond.cleanup: ; preds = %for.body
; Make sure interleave groups with a key being the special 'empty' value for
; the map do not cause a crash.
-define void @test_gap_empty_key() {
-; CHECK-LABEL: @test_gap_empty_key()
+define void @test_gap_empty_key(ptr %p) {
+; CHECK-LABEL: @test_gap_empty_key(ptr %p)
; CHECK-NEXT: entry:
; CHECK-NEXT: br label %for.body
@@ -57,7 +57,7 @@ entry:
for.body:
%iv = phi i64 [ 1, %entry ], [ %iv.next, %for.body ]
%iv.next = add nsw i64 %iv, 1
- %arrayidx = getelementptr inbounds [3 x i32], ptr undef, i64 0, i64 %iv.next
+ %arrayidx = getelementptr inbounds [3 x i32], ptr %p, i64 0, i64 %iv.next
%G2 = getelementptr i32, ptr %arrayidx, i64 %iv.next
%G9 = getelementptr i32, ptr %G2, i32 -2147483647
store i32 0, ptr %G2
@@ -71,8 +71,8 @@ exit:
; Make sure interleave groups with a key being the special 'tombstone' value for
; the map do not cause a crash.
-define void @test_tombstone_key() {
-; CHECK-LABEL: @test_tombstone_key()
+define void @test_tombstone_key(ptr %p) {
+; CHECK-LABEL: @test_tombstone_key(ptr %p)
; CHECK-NEXT: entry:
; CHECK-NEXT: br label %for.body
@@ -87,7 +87,7 @@ entry:
for.body:
%iv = phi i64 [ 1, %entry ], [ %iv.next, %for.body ]
%iv.next = add nsw i64 %iv, 1
- %arrayidx = getelementptr inbounds [3 x i32], ptr undef, i64 0, i64 %iv.next
+ %arrayidx = getelementptr inbounds [3 x i32], ptr %p, i64 0, i64 %iv.next
%G2 = getelementptr i32, ptr %arrayidx, i64 %iv.next
%G9 = getelementptr i32, ptr %G2, i32 -2147483648
store i32 0, ptr %G2
diff --git a/llvm/test/Transforms/LoopVectorize/demanded-bits-of-pointer-instruction.ll b/llvm/test/Transforms/LoopVectorize/demanded-bits-of-pointer-instruction.ll
index 41756ffb64e6c..8744e45344242 100644
--- a/llvm/test/Transforms/LoopVectorize/demanded-bits-of-pointer-instruction.ll
+++ b/llvm/test/Transforms/LoopVectorize/demanded-bits-of-pointer-instruction.ll
@@ -4,13 +4,13 @@
; Only make sure we do not crash.
; CHECK: @test
-define void @test(ptr %ptr, ptr %ptr_end) {
+define void @test(i8 %v, ptr %ptr, ptr %ptr_end) {
start:
br label %loop
loop:
%ptr2 = phi ptr [ %ptr3, %loop ], [ %ptr, %start ]
- %x = sext i8 undef to i64
+ %x = sext i8 %v to i64
%ptr3 = getelementptr inbounds i8, ptr %ptr2, i64 1
%cmp = icmp ult ptr %ptr3, %ptr_end
br i1 %cmp, label %loop, label %end
diff --git a/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll b/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll
index c164c4a46bd94..e7913c583b938 100644
--- a/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll
+++ b/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll
@@ -384,15 +384,15 @@ for.inc26:
; conditional store to remain scalar. Since we can only type-shrink vector
; types, we shouldn't try to represent the expression in a smaller type.
;
-define void @minimal_bit_widths(i1 %c) {
+define void @minimal_bit_widths(ptr %p, i1 %c) {
; UNROLL-LABEL: @minimal_bit_widths(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
; UNROLL-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
-; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr undef, i64 [[INDEX]]
-; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr undef, i64 [[TMP1]]
+; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 [[INDEX]]
+; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP1]]
; UNROLL-NEXT: [[TMP4:%.*]] = load i8, ptr [[TMP2]], align 1
; UNROLL-NEXT: [[TMP5:%.*]] = load i8, ptr [[TMP3]], align 1
; UNROLL-NEXT: br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE2]]
@@ -415,8 +415,8 @@ define void @minimal_bit_widths(i1 %c) {
; UNROLL-NOSIMPLIFY: vector.body:
; UNROLL-NOSIMPLIFY-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
; UNROLL-NOSIMPLIFY-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
-; UNROLL-NOSIMPLIFY-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr undef, i64 [[INDEX]]
-; UNROLL-NOSIMPLIFY-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr undef, i64 [[TMP1]]
+; UNROLL-NOSIMPLIFY-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 [[INDEX]]
+; UNROLL-NOSIMPLIFY-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP1]]
; UNROLL-NOSIMPLIFY-NEXT: [[TMP4:%.*]] = load i8, ptr [[TMP2]], align 1
; UNROLL-NOSIMPLIFY-NEXT: [[TMP5:%.*]] = load i8, ptr [[TMP3]], align 1
; UNROLL-NOSIMPLIFY-NEXT: br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
@@ -442,16 +442,16 @@ define void @minimal_bit_widths(i1 %c) {
; VEC-NEXT: br label [[VECTOR_BODY:%.*]]
; VEC: vector.body:
; VEC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
-; VEC-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr undef, i64 [[INDEX]]
+; VEC-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 [[INDEX]]
; VEC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP1]], align 1
; VEC-NEXT: br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE2]]
; VEC: pred.store.if:
; VEC-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 0
-; VEC-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr undef, i64 [[TMP8]]
+; VEC-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP8]]
; VEC-NEXT: [[TMP4:%.*]] = extractelement <2 x i8> [[WIDE_LOAD]], i32 0
; VEC-NEXT: store i8 [[TMP4]], ptr [[TMP3]], align 1
; VEC-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 1
-; VEC-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr undef, i64 [[TMP5]]
+; VEC-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]]
; VEC-NEXT: [[TMP7:%.*]] = extractelement <2 x i8> [[WIDE_LOAD]], i32 1
; VEC-NEXT: store i8 [[TMP7]], ptr [[TMP6]], align 1
; VEC-NEXT: br label [[PRED_STORE_CONTINUE2]]
@@ -468,7 +468,7 @@ entry:
for.body:
%tmp0 = phi i64 [ %tmp6, %for.inc ], [ 0, %entry ]
%tmp1 = phi i64 [ %tmp7, %for.inc ], [ 1000, %entry ]
- %tmp2 = getelementptr i8, ptr undef, i64 %tmp0
+ %tmp2 = getelementptr i8, ptr %p, i64 %tmp0
%tmp3 = load i8, ptr %tmp2, align 1
br i1 %c, label %if.then, label %for.inc
diff --git a/llvm/test/Transforms/LoopVectorize/nsw-crash.ll b/llvm/test/Transforms/LoopVectorize/nsw-crash.ll
index 106054d989776..d87d9b155db1c 100644
--- a/llvm/test/Transforms/LoopVectorize/nsw-crash.ll
+++ b/llvm/test/Transforms/LoopVectorize/nsw-crash.ll
@@ -3,7 +3,7 @@
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"
-define void @test(i1 %arg) {
+define void @test(ptr %p, i1 %arg) {
entry:
br i1 %arg, label %while.end, label %while.body.lr.ph
@@ -11,7 +11,7 @@ while.body.lr.ph:
br label %while.body
while.body:
- %it.sroa.0.091 = phi ptr [ undef, %while.body.lr.ph ], [ %incdec.ptr.i, %while.body ]
+ %it.sroa.0.091 = phi ptr [ %p, %while.body.lr.ph ], [ %incdec.ptr.i, %while.body ]
%incdec.ptr.i = getelementptr inbounds i32, ptr %it.sroa.0.091, i64 1
%inc32 = add i32 undef, 1 ; <------------- Make sure we don't set NSW flags to the undef.
%cmp.i11 = icmp eq ptr %incdec.ptr.i, undef
diff --git a/llvm/test/Transforms/LoopVectorize/optsize.ll b/llvm/test/Transforms/LoopVectorize/optsize.ll
index 763072ab16f73..fedd7e5bb797c 100644
--- a/llvm/test/Transforms/LoopVectorize/optsize.ll
+++ b/llvm/test/Transforms/LoopVectorize/optsize.ll
@@ -258,17 +258,12 @@ define void @pr43371() optsize {
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 0, i16 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP0:%.*]] = add <2 x i16> undef, [[VEC_IND]]
-; CHECK-NEXT: [[TMP1:%.*]] = zext <2 x i16> [[TMP0]] to <2 x i32>
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 0
-; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[TMP1]], i32 1
-; CHECK-NEXT: [[TMP3:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP2]]
+; CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; CHECK-NEXT: [[TMP0:%.*]] = add i16 3, [[OFFSET_IDX]]
+; CHECK-NEXT: [[TMP4:%.*]] = zext i16 [[TMP0]] to i32
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP4]]
-; CHECK-NEXT: store i16 0, ptr [[TMP3]], align 1
-; CHECK-NEXT: store i16 0, ptr [[TMP5]], align 1
+; CHECK-NEXT: store <2 x i16> zeroinitializer, ptr [[TMP5]], align 1
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
@@ -284,17 +279,12 @@ define void @pr43371() optsize {
; PGSO-NEXT: br label %[[VECTOR_BODY:.*]]
; PGSO: [[VECTOR_BODY]]:
; PGSO-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PGSO-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 0, i16 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PGSO-NEXT: [[TMP0:%.*]] = add <2 x i16> undef, [[VEC_IND]]
-; PGSO-NEXT: [[TMP1:%.*]] = zext <2 x i16> [[TMP0]] to <2 x i32>
-; PGSO-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 0
-; PGSO-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[TMP1]], i32 1
-; PGSO-NEXT: [[TMP3:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP2]]
+; PGSO-NEXT: [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; PGSO-NEXT: [[TMP0:%.*]] = add i16 3, [[OFFSET_IDX]]
+; PGSO-NEXT: [[TMP4:%.*]] = zext i16 [[TMP0]] to i32
; PGSO-NEXT: [[TMP5:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP4]]
-; PGSO-NEXT: store i16 0, ptr [[TMP3]], align 1
-; PGSO-NEXT: store i16 0, ptr [[TMP5]], align 1
+; PGSO-NEXT: store <2 x i16> zeroinitializer, ptr [[TMP5]], align 1
; PGSO-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; PGSO-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
; PGSO-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
; PGSO-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; PGSO: [[MIDDLE_BLOCK]]:
@@ -310,17 +300,12 @@ define void @pr43371() optsize {
; NPGSO-NEXT: br label %[[VECTOR_BODY:.*]]
; NPGSO: [[VECTOR_BODY]]:
; NPGSO-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; NPGSO-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 0, i16 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; NPGSO-NEXT: [[TMP0:%.*]] = add <2 x i16> undef, [[VEC_IND]]
-; NPGSO-NEXT: [[TMP1:%.*]] = zext <2 x i16> [[TMP0]] to <2 x i32>
-; NPGSO-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 0
-; NPGSO-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[TMP1]], i32 1
-; NPGSO-NEXT: [[TMP3:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP2]]
+; NPGSO-NEXT: [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; NPGSO-NEXT: [[TMP0:%.*]] = add i16 3, [[OFFSET_IDX]]
+; NPGSO-NEXT: [[TMP4:%.*]] = zext i16 [[TMP0]] to i32
; NPGSO-NEXT: [[TMP5:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP4]]
-; NPGSO-NEXT: store i16 0, ptr [[TMP3]], align 1
-; NPGSO-NEXT: store i16 0, ptr [[TMP5]], align 1
+; NPGSO-NEXT: store <2 x i16> zeroinitializer, ptr [[TMP5]], align 1
; NPGSO-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; NPGSO-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
; NPGSO-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
; NPGSO-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; NPGSO: [[MIDDLE_BLOCK]]:
@@ -340,7 +325,7 @@ for.cond.cleanup28:
for.body29:
%i24.0170 = phi i16 [ 0, %entry], [ %inc37, %for.body29]
- %add33 = add i16 undef, %i24.0170
+ %add33 = add i16 3, %i24.0170
%idxprom34 = zext i16 %add33 to i32
%arrayidx35 = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 %idxprom34
store i16 0, ptr %arrayidx35, align 1
@@ -359,17 +344,12 @@ define void @pr43371_pgso() !prof !14 {
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 0, i16 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT: [[TMP0:%.*]] = add <2 x i16> undef, [[VEC_IND]]
-; CHECK-NEXT: [[TMP1:%.*]] = zext <2 x i16> [[TMP0]] to <2 x i32>
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 0
-; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[TMP1]], i32 1
-; CHECK-NEXT: [[TMP3:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP2]]
+; CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; CHECK-NEXT: [[TMP0:%.*]] = add i16 3, [[OFFSET_IDX]]
+; CHECK-NEXT: [[TMP4:%.*]] = zext i16 [[TMP0]] to i32
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP4]]
-; CHECK-NEXT: store i16 0, ptr [[TMP3]], align 1
-; CHECK-NEXT: store i16 0, ptr [[TMP5]], align 1
+; CHECK-NEXT: store <2 x i16> zeroinitializer, ptr [[TMP5]], align 1
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
@@ -385,17 +365,12 @@ define void @pr43371_pgso() !prof !14 {
; PGSO-NEXT: br label %[[VECTOR_BODY:.*]]
; PGSO: [[VECTOR_BODY]]:
; PGSO-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PGSO-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 0, i16 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; PGSO-NEXT: [[TMP0:%.*]] = add <2 x i16> undef, [[VEC_IND]]
-; PGSO-NEXT: [[TMP1:%.*]] = zext <2 x i16> [[TMP0]] to <2 x i32>
-; PGSO-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 0
-; PGSO-NEXT: [[TMP4:%.*]] = extractelement <2 x i32> [[TMP1]], i32 1
-; PGSO-NEXT: [[TMP3:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP2]]
+; PGSO-NEXT: [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; PGSO-NEXT: [[TMP0:%.*]] = ad...
[truncated]
|
You can test this locally with the following command:git diff -U0 --pickaxe-regex -S '([^a-zA-Z0-9#_-]undef([^a-zA-Z0-9_-]|$)|UndefValue::get)' 'HEAD~1' HEAD llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-cg-bug.ll llvm/test/Transforms/LoopVectorize/X86/cost-model-assert.ll llvm/test/Transforms/LoopVectorize/X86/interleaved-accesses-large-gap.ll llvm/test/Transforms/LoopVectorize/demanded-bits-of-pointer-instruction.ll llvm/test/Transforms/LoopVectorize/if-pred-stores.ll llvm/test/Transforms/LoopVectorize/nsw-crash.ll llvm/test/Transforms/LoopVectorize/optsize.ll llvm/test/Transforms/LoopVectorize/pr32859.ll llvm/test/Transforms/LoopVectorize/pr36311.ll llvm/test/Transforms/LoopVectorize/reduction-ptr.ll llvm/test/Transforms/LoopVectorize/reduction-small-size.ll llvm/test/Transforms/LoopVectorize/runtime-drop-crash.ll llvm/test/Transforms/LoopVectorize/undef-inst-bug.llThe following files introduce new uses of undef:
Undef is now deprecated and should only be used in the rare cases where no replacement is possible. For example, a load of uninitialized memory yields In tests, avoid using For example, this is considered a bad practice: define void @fn() {
...
br i1 undef, ...
}Please use the following instead: define void @fn(i1 %cond) {
...
br i1 %cond, ...
}Please refer to the Undefined Behavior Manual for more information. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Split off from PR #163525, this standalone patch replaces simple cases where undef is used as a value for arithmetic or getelementptr instructions. This will reduce the likelihood of contributors hitting the
undef deprecatorwarning in github.