[llvm] [Loads] Respect UseDerefAtPointSemantics in isDerefAndAlignedPointer. (PR #123196)
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Thu Jan 16 07:55:29 PST 2025
https://github.com/fhahn updated https://github.com/llvm/llvm-project/pull/123196
>From 2c29b4af5d427e35b2649b72d274693ba54c0b66 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Thu, 16 Jan 2025 12:53:11 +0000
Subject: [PATCH 1/2] [Loads] Respect UseDerefAtPointSemantics in
isDerefAndAlignedPointer.
If a pointer gets freed, it may not be dereferenceable any longer, even
though there is a dominating dereferenceable assumption. As first step,
only consider assumptions if the function doesn't free memory if
UseDerefAtPointSemantics is used.
---
llvm/lib/Analysis/Loads.cpp | 5 +++-
llvm/lib/IR/Value.cpp | 2 +-
...able-info-from-assumption-constant-size.ll | 23 +++++++++++++++----
3 files changed, 23 insertions(+), 7 deletions(-)
diff --git a/llvm/lib/Analysis/Loads.cpp b/llvm/lib/Analysis/Loads.cpp
index 7bbd469bd035d3..4cc93a8daeb0f1 100644
--- a/llvm/lib/Analysis/Loads.cpp
+++ b/llvm/lib/Analysis/Loads.cpp
@@ -25,6 +25,8 @@
using namespace llvm;
+extern cl::opt<unsigned> UseDerefAtPointSemantics;
+
static bool isAligned(const Value *Base, Align Alignment,
const DataLayout &DL) {
return Base->getPointerAlignment(DL) >= Alignment;
@@ -168,7 +170,8 @@ static bool isDereferenceableAndAlignedPointer(
Size, DL, CtxI, AC, DT, TLI,
Visited, MaxDepth);
- if (CtxI) {
+ if (CtxI &&
+ (!UseDerefAtPointSemantics || CtxI->getFunction()->doesNotFreeMemory())) {
/// Look through assumes to see if both dereferencability and alignment can
/// be proven by an assume if needed.
RetainedKnowledge AlignRK;
diff --git a/llvm/lib/IR/Value.cpp b/llvm/lib/IR/Value.cpp
index 65b63955b6f6d5..eddb67282fca46 100644
--- a/llvm/lib/IR/Value.cpp
+++ b/llvm/lib/IR/Value.cpp
@@ -36,7 +36,7 @@
using namespace llvm;
-static cl::opt<bool> UseDerefAtPointSemantics(
+cl::opt<bool> UseDerefAtPointSemantics(
"use-dereferenceable-at-point-semantics", cl::Hidden, cl::init(false),
cl::desc("Deref attributes and metadata infer facts at definition only"));
diff --git a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
index 572511a5ffb926..7b41c9b3197aa6 100644
--- a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
+++ b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
@@ -1411,7 +1411,7 @@ exit:
}
; %a may be freeed between the dereferenceable assumption and accesses.
-; FIXME: It is not safe to use with -use-dereferenceable-at-point-semantics.
+; It is not safe to use with -use-dereferenceable-at-point-semantics.
define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
; CHECK-LABEL: define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) {
@@ -1422,16 +1422,29 @@ define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_
; 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: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT: [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[A]], align 4
+; CHECK-NEXT: [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
+; CHECK-NEXT: br i1 [[TMP5]], label %[[PRED_LOAD_IF:.*]], label %[[PRED_LOAD_CONTINUE:.*]]
+; CHECK: [[PRED_LOAD_IF]]:
; CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[A]], align 4
-; CHECK-NEXT: [[TMP13:%.*]] = insertelement <2 x i32> poison, i32 [[TMP5]], i32 0
-; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> [[TMP13]], i32 [[TMP15]], i32 1
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP15]], i32 0
+; CHECK-NEXT: br label %[[PRED_LOAD_CONTINUE]]
+; CHECK: [[PRED_LOAD_CONTINUE]]:
+; CHECK-NEXT: [[TMP12:%.*]] = phi <2 x i32> [ poison, %[[VECTOR_BODY]] ], [ [[TMP7]], %[[PRED_LOAD_IF]] ]
+; CHECK-NEXT: [[TMP13:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
+; CHECK-NEXT: br i1 [[TMP13]], label %[[PRED_LOAD_IF1:.*]], label %[[PRED_LOAD_CONTINUE2]]
+; CHECK: [[PRED_LOAD_IF1]]:
+; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[A]], align 4
+; CHECK-NEXT: [[TMP16:%.*]] = insertelement <2 x i32> [[TMP12]], i32 [[TMP14]], i32 1
+; CHECK-NEXT: br label %[[PRED_LOAD_CONTINUE2]]
+; CHECK: [[PRED_LOAD_CONTINUE2]]:
+; CHECK-NEXT: [[TMP11:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP11]]
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
>From 4a46cd0b2c264f4456c2912e21ac17acade2ff92 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Thu, 16 Jan 2025 15:51:52 +0000
Subject: [PATCH 2/2] !fixup use canBeFreed.
---
llvm/lib/Analysis/Loads.cpp | 3 +-
...able-info-from-assumption-constant-size.ll | 131 +++++++++++++++---
2 files changed, 116 insertions(+), 18 deletions(-)
diff --git a/llvm/lib/Analysis/Loads.cpp b/llvm/lib/Analysis/Loads.cpp
index 4cc93a8daeb0f1..c2b4ee1fd0aa2e 100644
--- a/llvm/lib/Analysis/Loads.cpp
+++ b/llvm/lib/Analysis/Loads.cpp
@@ -170,8 +170,7 @@ static bool isDereferenceableAndAlignedPointer(
Size, DL, CtxI, AC, DT, TLI,
Visited, MaxDepth);
- if (CtxI &&
- (!UseDerefAtPointSemantics || CtxI->getFunction()->doesNotFreeMemory())) {
+ if (CtxI && (!UseDerefAtPointSemantics || !V->canBeFreed())) {
/// Look through assumes to see if both dereferencability and alignment can
/// be proven by an assume if needed.
RetainedKnowledge AlignRK;
diff --git a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
index 7b41c9b3197aa6..ba11bbf8e82e70 100644
--- a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
+++ b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
@@ -3,7 +3,7 @@
declare void @llvm.assume(i1)
-define void @deref_assumption_in_header_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_header_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_header_constant_trip_count(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1:[0-9]+]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -104,7 +104,7 @@ exit:
ret void
}
-define void @align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -181,7 +181,7 @@ exit:
ret void
}
-define void @deref_assumption_too_small_in_header_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_too_small_in_header_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_too_small_in_header_constant_trip_count(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -282,7 +282,7 @@ exit:
ret void
}
-define void @deref_assumption_in_header_constant_trip_count_align_1(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_header_constant_trip_count_align_1(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_header_constant_trip_count_align_1(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -383,7 +383,7 @@ exit:
ret void
}
-define void @deref_assumption_in_header_constant_trip_count_align_via_arg_attribute(ptr noalias align 4 %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_header_constant_trip_count_align_via_arg_attribute(ptr noalias align 4 %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_header_constant_trip_count_align_via_arg_attribute(
; CHECK-SAME: ptr noalias align 4 [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -484,7 +484,7 @@ exit:
ret void
}
-define void @deref_assumption_in_header_constant_trip_count_align_not_known(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_header_constant_trip_count_align_not_known(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_header_constant_trip_count_align_not_known(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -585,7 +585,7 @@ exit:
ret void
}
-define void @deref_assumption_in_then_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_then_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_then_constant_trip_count(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -682,7 +682,7 @@ exit:
ret void
}
-define void @deref_assumption_in_latch_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_latch_constant_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_latch_constant_trip_count(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -785,7 +785,7 @@ exit:
ret void
}
-define void @deref_assumption_in_header_variable_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) nofree {
+define void @deref_assumption_in_header_variable_trip_count(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_header_variable_trip_count(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]], i64 [[N:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -890,7 +890,7 @@ exit:
ret void
}
-define void @deref_assumption_in_preheader_constant_trip_count_align_1(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_preheader_constant_trip_count_align_1(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_preheader_constant_trip_count_align_1(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -968,7 +968,7 @@ exit:
ret void
}
-define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_1(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_1(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_1(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -1063,7 +1063,7 @@ exit:
ret void
}
-define void @align_and_deref_assumption_in_preheader_constant_trip_count_align_4(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @align_and_deref_assumption_in_preheader_constant_trip_count_align_4(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @align_and_deref_assumption_in_preheader_constant_trip_count_align_4(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -1142,7 +1142,7 @@ exit:
}
-define void @deref_assumption_in_preheader_constant_trip_count_align_4_known_via_argument_attr(ptr noalias align 4 %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_preheader_constant_trip_count_align_4_known_via_argument_attr(ptr noalias align 4 %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_preheader_constant_trip_count_align_4_known_via_argument_attr(
; CHECK-SAME: ptr noalias align 4 [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -1220,7 +1220,7 @@ exit:
ret void
}
-define void @deref_assumption_in_preheader_constant_trip_count_align_4_not_known(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_in_preheader_constant_trip_count_align_4_not_known(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_in_preheader_constant_trip_count_align_4_not_known(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -1315,7 +1315,7 @@ exit:
ret void
}
-define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_4(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree {
+define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_4(ptr noalias %a, ptr noalias %b, ptr noalias %c) nofree nosync{
; CHECK-LABEL: define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_4(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
; CHECK-NEXT: [[ENTRY:.*]]:
@@ -1410,7 +1410,7 @@ exit:
ret void
}
-; %a may be freeed between the dereferenceable assumption and accesses.
+; %a may be freed between the dereferenceable assumption and accesses.
; It is not safe to use with -use-dereferenceable-at-point-semantics.
define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
; CHECK-LABEL: define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(
@@ -1504,6 +1504,103 @@ exit:
ret void
}
+; %a may be freed between the dereferenceable assumption and accesses.
+; It is not safe to use with -use-dereferenceable-at-point-semantics.
+define void @may_free_local_ptr_align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(ptr noalias %b, ptr noalias %c) nofree nosync {
+; CHECK-LABEL: define void @may_free_local_ptr_align_deref_assumption_in_header_constant_trip_count_loop_invariant_ptr(
+; CHECK-SAME: ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR1]] {
+; CHECK-NEXT: [[ENTRY:.*]]:
+; CHECK-NEXT: [[A:%.*]] = call ptr @get_ptr()
+; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[A]], i64 4), "dereferenceable"(ptr [[A]], i64 4) ]
+; CHECK-NEXT: call void @may_free()
+; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], 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:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
+; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
+; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
+; CHECK-NEXT: [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
+; CHECK-NEXT: br i1 [[TMP5]], label %[[PRED_LOAD_IF:.*]], label %[[PRED_LOAD_CONTINUE:.*]]
+; CHECK: [[PRED_LOAD_IF]]:
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[A]], align 4
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP6]], i32 0
+; CHECK-NEXT: br label %[[PRED_LOAD_CONTINUE]]
+; CHECK: [[PRED_LOAD_CONTINUE]]:
+; CHECK-NEXT: [[TMP8:%.*]] = phi <2 x i32> [ poison, %[[VECTOR_BODY]] ], [ [[TMP7]], %[[PRED_LOAD_IF]] ]
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
+; CHECK-NEXT: br i1 [[TMP9]], label %[[PRED_LOAD_IF1:.*]], label %[[PRED_LOAD_CONTINUE2]]
+; CHECK: [[PRED_LOAD_IF1]]:
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[A]], align 4
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> [[TMP8]], i32 [[TMP10]], i32 1
+; CHECK-NEXT: br label %[[PRED_LOAD_CONTINUE2]]
+; CHECK: [[PRED_LOAD_CONTINUE2]]:
+; CHECK-NEXT: [[TMP12:%.*]] = phi <2 x i32> [ [[TMP8]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP11]], %[[PRED_LOAD_IF1]] ]
+; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP12]]
+; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
+; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
+; CHECK-NEXT: store <2 x i32> [[PREDPHI]], ptr [[TMP14]], align 4
+; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT: br i1 [[TMP15]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; CHECK: [[MIDDLE_BLOCK]]:
+; CHECK-NEXT: br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK: [[SCALAR_PH]]:
+; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT: br label %[[LOOP_HEADER:.*]]
+; CHECK: [[LOOP_HEADER]]:
+; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
+; CHECK-NEXT: [[GEP_B:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; CHECK-NEXT: [[L_B:%.*]] = load i32, ptr [[GEP_B]], align 4
+; CHECK-NEXT: [[C_1:%.*]] = icmp sge i32 [[L_B]], 0
+; CHECK-NEXT: br i1 [[C_1]], label %[[LOOP_LATCH]], label %[[LOOP_THEN:.*]]
+; CHECK: [[LOOP_THEN]]:
+; CHECK-NEXT: [[L_A:%.*]] = load i32, ptr [[A]], align 4
+; CHECK-NEXT: br label %[[LOOP_LATCH]]
+; CHECK: [[LOOP_LATCH]]:
+; CHECK-NEXT: [[MERGE:%.*]] = phi i32 [ [[L_A]], %[[LOOP_THEN]] ], [ [[L_B]], %[[LOOP_HEADER]] ]
+; CHECK-NEXT: [[GEP_C:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]]
+; CHECK-NEXT: store i32 [[MERGE]], ptr [[GEP_C]], align 4
+; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP35:![0-9]+]]
+; CHECK: [[EXIT]]:
+; CHECK-NEXT: ret void
+;
+entry:
+ %a = call ptr @get_ptr()
+ call void @llvm.assume(i1 true) [ "align"(ptr %a, i64 4), "dereferenceable"(ptr %a, i64 4) ]
+ call void @may_free()
+ br label %loop.header
+
+loop.header:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
+ %gep.b = getelementptr inbounds i32, ptr %b, i64 %iv
+ %l.b = load i32, ptr %gep.b, align 4
+ %c.1 = icmp sge i32 %l.b, 0
+ br i1 %c.1, label %loop.latch, label %loop.then
+
+loop.then:
+ %l.a = load i32, ptr %a, align 4
+ br label %loop.latch
+
+loop.latch:
+ %merge = phi i32 [ %l.a, %loop.then ], [ %l.b, %loop.header ]
+ %gep.c = getelementptr inbounds i32, ptr %c, i64 %iv
+ store i32 %merge, ptr %gep.c, align 4
+ %iv.next = add nuw nsw i64 %iv, 1
+ %ec = icmp eq i64 %iv.next, 1000
+ br i1 %ec, label %exit, label %loop.header
+
+exit:
+ ret void
+}
+
+declare ptr @get_ptr()
declare void @may_free()
;.
@@ -1541,4 +1638,6 @@ declare void @may_free()
; CHECK: [[LOOP31]] = distinct !{[[LOOP31]], [[META2]], [[META1]]}
; CHECK: [[LOOP32]] = distinct !{[[LOOP32]], [[META1]], [[META2]]}
; CHECK: [[LOOP33]] = distinct !{[[LOOP33]], [[META2]], [[META1]]}
+; CHECK: [[LOOP34]] = distinct !{[[LOOP34]], [[META1]], [[META2]]}
+; CHECK: [[LOOP35]] = distinct !{[[LOOP35]], [[META2]], [[META1]]}
;.
More information about the llvm-commits
mailing list