6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-15 04:30:12 +00:00
Code Issues Projects Releases Wiki Activity
9ecdca9115
llvm-6502/test/Transforms/ObjCARC/allocas.ll

501 lines
14 KiB
LLVM
Raw Normal View History

[objc-arc-opts] In the presense of an alloca unconditionally remove RR pairs if and only if we are both KnownSafeBU/KnownSafeTD rather than just either or. In the presense of a block being initialized, the frontend will emit the objc_retain on the original pointer and the release on the pointer loaded from the alloca. The optimizer will through the provenance analysis realize that the two are related (albiet different), but since we only require KnownSafe in one direction, will match the inner retain on the original pointer with the guard release on the original pointer. This is fixed by ensuring that in the presense of allocas we only unconditionally remove pointers if both our retain and our release are KnownSafe (i.e. we are KnownSafe in both directions) since we must deal with the possibility that the frontend will emit what (to the optimizer) appears to be unbalanced retain/releases. An example of the miscompile is: %A = alloca retain(%x) retain(%x) <--- Inner Retain store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) release(%x) <--- Guarding Release getting optimized to: %A = alloca retain(%x) store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181743 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-13 23:49:42 +00:00
; RUN: opt -objc-arc -S < %s | FileCheck %s
declare i8* @objc_retain(i8*)
declare i8* @objc_retainAutoreleasedReturnValue(i8*)
declare void @objc_release(i8*)
declare i8* @objc_autorelease(i8*)
declare i8* @objc_autoreleaseReturnValue(i8*)
declare void @objc_autoreleasePoolPop(i8*)
declare i8* @objc_autoreleasePoolPush()
declare i8* @objc_retainBlock(i8*)
declare i8* @objc_retainedObject(i8*)
declare i8* @objc_unretainedObject(i8*)
declare i8* @objc_unretainedPointer(i8*)
declare void @use_pointer(i8*)
declare void @callee()
declare void @callee_fnptr(void ()*)
declare void @invokee()
declare i8* @returner()
[objc-arc] KnownSafe does not imply that it is safe to perform code motion across CFG edges since even if it is safe to remove RR pairs, we may still be able to move a retain/release into a loop. rdar://13949644 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182670 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-24 20:44:05 +00:00
declare i8* @returner1()
declare i8* @returner2()
[objc-arc-opts] In the presense of an alloca unconditionally remove RR pairs if and only if we are both KnownSafeBU/KnownSafeTD rather than just either or. In the presense of a block being initialized, the frontend will emit the objc_retain on the original pointer and the release on the pointer loaded from the alloca. The optimizer will through the provenance analysis realize that the two are related (albiet different), but since we only require KnownSafe in one direction, will match the inner retain on the original pointer with the guard release on the original pointer. This is fixed by ensuring that in the presense of allocas we only unconditionally remove pointers if both our retain and our release are KnownSafe (i.e. we are KnownSafe in both directions) since we must deal with the possibility that the frontend will emit what (to the optimizer) appears to be unbalanced retain/releases. An example of the miscompile is: %A = alloca retain(%x) retain(%x) <--- Inner Retain store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) release(%x) <--- Guarding Release getting optimized to: %A = alloca retain(%x) store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181743 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-13 23:49:42 +00:00
declare void @bar(i32 ()*)
declare void @use_alloca(i8**)
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
declare void @llvm.dbg.value(metadata, i64, metadata, metadata)
[objc-arc-opts] In the presense of an alloca unconditionally remove RR pairs if and only if we are both KnownSafeBU/KnownSafeTD rather than just either or. In the presense of a block being initialized, the frontend will emit the objc_retain on the original pointer and the release on the pointer loaded from the alloca. The optimizer will through the provenance analysis realize that the two are related (albiet different), but since we only require KnownSafe in one direction, will match the inner retain on the original pointer with the guard release on the original pointer. This is fixed by ensuring that in the presense of allocas we only unconditionally remove pointers if both our retain and our release are KnownSafe (i.e. we are KnownSafe in both directions) since we must deal with the possibility that the frontend will emit what (to the optimizer) appears to be unbalanced retain/releases. An example of the miscompile is: %A = alloca retain(%x) retain(%x) <--- Inner Retain store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) release(%x) <--- Guarding Release getting optimized to: %A = alloca retain(%x) store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181743 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-13 23:49:42 +00:00
declare i8* @objc_msgSend(i8*, i8*, ...)
Fix known typos Sweep the codebase for common typos. Includes some changes to visible function names that were misspelt. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200018 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-24 17:20:08 +00:00
; In the presence of allocas, unconditionally remove retain/release pairs only
[objc-arc-opts] In the presense of an alloca unconditionally remove RR pairs if and only if we are both KnownSafeBU/KnownSafeTD rather than just either or. In the presense of a block being initialized, the frontend will emit the objc_retain on the original pointer and the release on the pointer loaded from the alloca. The optimizer will through the provenance analysis realize that the two are related (albiet different), but since we only require KnownSafe in one direction, will match the inner retain on the original pointer with the guard release on the original pointer. This is fixed by ensuring that in the presense of allocas we only unconditionally remove pointers if both our retain and our release are KnownSafe (i.e. we are KnownSafe in both directions) since we must deal with the possibility that the frontend will emit what (to the optimizer) appears to be unbalanced retain/releases. An example of the miscompile is: %A = alloca retain(%x) retain(%x) <--- Inner Retain store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) release(%x) <--- Guarding Release getting optimized to: %A = alloca retain(%x) store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181743 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-13 23:49:42 +00:00
; if they are known safe in both directions. This prevents matching up an inner
; retain with the boundary guarding release in the following situation:
;
; %A = alloca
; retain(%x)
; retain(%x) <--- Inner Retain
; store %x, %A
; %y = load %A
; ... DO STUFF ...
; release(%y)
; release(%x) <--- Guarding Release
;
; rdar://13750319
; CHECK: define void @test1a(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test1a(i8* %x) {
entry:
%A = alloca i8*
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
store i8* %x, i8** %A, align 8
%y = load i8** %A
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test1b(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test1b(i8* %x) {
entry:
%A = alloca i8*
%gep = getelementptr i8** %A, i32 0
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
store i8* %x, i8** %gep, align 8
%y = load i8** %A
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test1c(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test1c(i8* %x) {
entry:
%A = alloca i8*, i32 3
%gep = getelementptr i8** %A, i32 2
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
store i8* %x, i8** %gep, align 8
%y = load i8** %gep
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test1d(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test1d(i8* %x) {
entry:
br i1 undef, label %use_allocaA, label %use_allocaB
use_allocaA:
%allocaA = alloca i8*
br label %exit
use_allocaB:
%allocaB = alloca i8*
br label %exit
exit:
%A = phi i8** [ %allocaA, %use_allocaA ], [ %allocaB, %use_allocaB ]
%gep = getelementptr i8** %A, i32 0
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
store i8* %x, i8** %gep, align 8
%y = load i8** %gep
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test1e(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test1e(i8* %x) {
entry:
br i1 undef, label %use_allocaA, label %use_allocaB
use_allocaA:
%allocaA = alloca i8*, i32 4
br label %exit
use_allocaB:
%allocaB = alloca i8*, i32 4
br label %exit
exit:
%A = phi i8** [ %allocaA, %use_allocaA ], [ %allocaB, %use_allocaB ]
%gep = getelementptr i8** %A, i32 2
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
store i8* %x, i8** %gep, align 8
%y = load i8** %gep
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test1f(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test1f(i8* %x) {
entry:
%allocaOne = alloca i8*
%allocaTwo = alloca i8*
%A = select i1 undef, i8** %allocaOne, i8** %allocaTwo
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
store i8* %x, i8** %A, align 8
%y = load i8** %A
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
[objc-arc] Make sure that multiple owners is propogated correctly through the pass via the usage of a global data structure. rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182669 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-24 20:44:02 +00:00
; Make sure that if a store is in a different basic block we handle known safe
; conservatively.
; CHECK: define void @test2a(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test2a(i8* %x) {
entry:
%A = alloca i8*
store i8* %x, i8** %A, align 8
%y = load i8** %A
br label %bb1
bb1:
br label %bb2
bb2:
br label %bb3
bb3:
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test2b(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test2b(i8* %x) {
entry:
%A = alloca i8*
%gep1 = getelementptr i8** %A, i32 0
store i8* %x, i8** %gep1, align 8
%gep2 = getelementptr i8** %A, i32 0
%y = load i8** %gep2
br label %bb1
bb1:
br label %bb2
bb2:
br label %bb3
bb3:
tail call i8* @objc_retain(i8* %x)
tail call i8* @objc_retain(i8* %x)
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
; CHECK: define void @test2c(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test2c(i8* %x) {
entry:
%A = alloca i8*, i32 3
%gep1 = getelementptr i8** %A, i32 2
store i8* %x, i8** %gep1, align 8
%gep2 = getelementptr i8** %A, i32 2
%y = load i8** %gep2
tail call i8* @objc_retain(i8* %x)
br label %bb1
bb1:
br label %bb2
bb2:
br label %bb3
bb3:
tail call i8* @objc_retain(i8* %x)
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
[objc-arc-opts] In the presense of an alloca unconditionally remove RR pairs if and only if we are both KnownSafeBU/KnownSafeTD rather than just either or. In the presense of a block being initialized, the frontend will emit the objc_retain on the original pointer and the release on the pointer loaded from the alloca. The optimizer will through the provenance analysis realize that the two are related (albiet different), but since we only require KnownSafe in one direction, will match the inner retain on the original pointer with the guard release on the original pointer. This is fixed by ensuring that in the presense of allocas we only unconditionally remove pointers if both our retain and our release are KnownSafe (i.e. we are KnownSafe in both directions) since we must deal with the possibility that the frontend will emit what (to the optimizer) appears to be unbalanced retain/releases. An example of the miscompile is: %A = alloca retain(%x) retain(%x) <--- Inner Retain store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) release(%x) <--- Guarding Release getting optimized to: %A = alloca retain(%x) store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181743 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-13 23:49:42 +00:00
[objc-arc] KnownSafe does not imply that it is safe to perform code motion across CFG edges since even if it is safe to remove RR pairs, we may still be able to move a retain/release into a loop. rdar://13949644 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182670 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-24 20:44:05 +00:00
; CHECK: define void @test2d(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_retain(i8* %x)
; CHECK: @objc_release(i8* %y)
; CHECK: @objc_release(i8* %x)
; CHECK: ret void
; CHECK: }
define void @test2d(i8* %x) {
entry:
tail call i8* @objc_retain(i8* %x)
br label %bb1
bb1:
%Abb1 = alloca i8*, i32 3
%gepbb11 = getelementptr i8** %Abb1, i32 2
store i8* %x, i8** %gepbb11, align 8
%gepbb12 = getelementptr i8** %Abb1, i32 2
%ybb1 = load i8** %gepbb12
br label %bb3
bb2:
%Abb2 = alloca i8*, i32 4
%gepbb21 = getelementptr i8** %Abb2, i32 2
store i8* %x, i8** %gepbb21, align 8
%gepbb22 = getelementptr i8** %Abb2, i32 2
%ybb2 = load i8** %gepbb22
br label %bb3
bb3:
%A = phi i8** [ %Abb1, %bb1 ], [ %Abb2, %bb2 ]
%y = phi i8* [ %ybb1, %bb1 ], [ %ybb2, %bb2 ]
tail call i8* @objc_retain(i8* %x)
call void @use_alloca(i8** %A)
call void @objc_release(i8* %y), !clang.imprecise_release !0
call void @use_pointer(i8* %x)
call void @objc_release(i8* %x), !clang.imprecise_release !0
ret void
}
Fix known typos Sweep the codebase for common typos. Includes some changes to visible function names that were misspelt. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200018 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-24 17:20:08 +00:00
; Make sure in the presence of allocas, if we find a cfghazard we do not perform
[objc-arc] KnownSafe does not imply that it is safe to perform code motion across CFG edges since even if it is safe to remove RR pairs, we may still be able to move a retain/release into a loop. rdar://13949644 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@182670 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-24 20:44:05 +00:00
; code motion even if we are known safe. These two concepts are separate and
; should be treated as such.
;
; rdar://13949644
; CHECK: define void @test3a() {
; CHECK: entry:
; CHECK: @objc_retainAutoreleasedReturnValue
; CHECK: @objc_retain
; CHECK: @objc_retain
; CHECK: @objc_retain
; CHECK: @objc_retain
; CHECK: arraydestroy.body:
; CHECK: @objc_release
; CHECK-NOT: @objc_release
; CHECK: arraydestroy.done:
; CHECK-NOT: @objc_release
; CHECK: arraydestroy.body1:
; CHECK: @objc_release
; CHECK-NOT: @objc_release
; CHECK: arraydestroy.done1:
; CHECK: @objc_release
; CHECK: ret void
; CHECK: }
define void @test3a() {
entry:
%keys = alloca [2 x i8*], align 16
%objs = alloca [2 x i8*], align 16
%call1 = call i8* @returner()
%tmp0 = tail call i8* @objc_retainAutoreleasedReturnValue(i8* %call1)
%objs.begin = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 0
tail call i8* @objc_retain(i8* %call1)
store i8* %call1, i8** %objs.begin, align 8
%objs.elt = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 1
tail call i8* @objc_retain(i8* %call1)
store i8* %call1, i8** %objs.elt
%call2 = call i8* @returner1()
%call3 = call i8* @returner2()
%keys.begin = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 0
tail call i8* @objc_retain(i8* %call2)
store i8* %call2, i8** %keys.begin, align 8
%keys.elt = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 1
tail call i8* @objc_retain(i8* %call3)
store i8* %call3, i8** %keys.elt
%gep = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 2
br label %arraydestroy.body
arraydestroy.body:
%arraydestroy.elementPast = phi i8** [ %gep, %entry ], [ %arraydestroy.element, %arraydestroy.body ]
%arraydestroy.element = getelementptr inbounds i8** %arraydestroy.elementPast, i64 -1
%destroy_tmp = load i8** %arraydestroy.element, align 8
call void @objc_release(i8* %destroy_tmp), !clang.imprecise_release !0
%objs_ptr = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 0
%arraydestroy.cmp = icmp eq i8** %arraydestroy.element, %objs_ptr
br i1 %arraydestroy.cmp, label %arraydestroy.done, label %arraydestroy.body
arraydestroy.done:
%gep1 = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 2
br label %arraydestroy.body1
arraydestroy.body1:
%arraydestroy.elementPast1 = phi i8** [ %gep1, %arraydestroy.done ], [ %arraydestroy.element1, %arraydestroy.body1 ]
%arraydestroy.element1 = getelementptr inbounds i8** %arraydestroy.elementPast1, i64 -1
%destroy_tmp1 = load i8** %arraydestroy.element1, align 8
call void @objc_release(i8* %destroy_tmp1), !clang.imprecise_release !0
%keys_ptr = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 0
%arraydestroy.cmp1 = icmp eq i8** %arraydestroy.element1, %keys_ptr
br i1 %arraydestroy.cmp1, label %arraydestroy.done1, label %arraydestroy.body1
arraydestroy.done1:
call void @objc_release(i8* %call1), !clang.imprecise_release !0
ret void
}
; Make sure that even though we stop said code motion we still allow for
; pointers to be removed if we are known safe in both directions.
;
; rdar://13949644
; CHECK: define void @test3b() {
; CHECK: entry:
; CHECK: @objc_retainAutoreleasedReturnValue
; CHECK: @objc_retain
; CHECK: @objc_retain
; CHECK: @objc_retain
; CHECK: @objc_retain
; CHECK: arraydestroy.body:
; CHECK: @objc_release
; CHECK-NOT: @objc_release
; CHECK: arraydestroy.done:
; CHECK-NOT: @objc_release
; CHECK: arraydestroy.body1:
; CHECK: @objc_release
; CHECK-NOT: @objc_release
; CHECK: arraydestroy.done1:
; CHECK: @objc_release
; CHECK: ret void
; CHECK: }
define void @test3b() {
entry:
%keys = alloca [2 x i8*], align 16
%objs = alloca [2 x i8*], align 16
%call1 = call i8* @returner()
%tmp0 = tail call i8* @objc_retainAutoreleasedReturnValue(i8* %call1)
%tmp1 = tail call i8* @objc_retain(i8* %call1)
%objs.begin = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 0
tail call i8* @objc_retain(i8* %call1)
store i8* %call1, i8** %objs.begin, align 8
%objs.elt = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 1
tail call i8* @objc_retain(i8* %call1)
store i8* %call1, i8** %objs.elt
%call2 = call i8* @returner1()
%call3 = call i8* @returner2()
%keys.begin = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 0
tail call i8* @objc_retain(i8* %call2)
store i8* %call2, i8** %keys.begin, align 8
%keys.elt = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 1
tail call i8* @objc_retain(i8* %call3)
store i8* %call3, i8** %keys.elt
%gep = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 2
br label %arraydestroy.body
arraydestroy.body:
%arraydestroy.elementPast = phi i8** [ %gep, %entry ], [ %arraydestroy.element, %arraydestroy.body ]
%arraydestroy.element = getelementptr inbounds i8** %arraydestroy.elementPast, i64 -1
%destroy_tmp = load i8** %arraydestroy.element, align 8
call void @objc_release(i8* %destroy_tmp), !clang.imprecise_release !0
%objs_ptr = getelementptr inbounds [2 x i8*]* %objs, i64 0, i64 0
%arraydestroy.cmp = icmp eq i8** %arraydestroy.element, %objs_ptr
br i1 %arraydestroy.cmp, label %arraydestroy.done, label %arraydestroy.body
arraydestroy.done:
%gep1 = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 2
br label %arraydestroy.body1
arraydestroy.body1:
%arraydestroy.elementPast1 = phi i8** [ %gep1, %arraydestroy.done ], [ %arraydestroy.element1, %arraydestroy.body1 ]
%arraydestroy.element1 = getelementptr inbounds i8** %arraydestroy.elementPast1, i64 -1
%destroy_tmp1 = load i8** %arraydestroy.element1, align 8
call void @objc_release(i8* %destroy_tmp1), !clang.imprecise_release !0
%keys_ptr = getelementptr inbounds [2 x i8*]* %keys, i64 0, i64 0
%arraydestroy.cmp1 = icmp eq i8** %arraydestroy.element1, %keys_ptr
br i1 %arraydestroy.cmp1, label %arraydestroy.done1, label %arraydestroy.body1
arraydestroy.done1:
call void @objc_release(i8* %call1), !clang.imprecise_release !0
call void @objc_release(i8* %call1), !clang.imprecise_release !0
ret void
}
[objc-arc-opts] In the presense of an alloca unconditionally remove RR pairs if and only if we are both KnownSafeBU/KnownSafeTD rather than just either or. In the presense of a block being initialized, the frontend will emit the objc_retain on the original pointer and the release on the pointer loaded from the alloca. The optimizer will through the provenance analysis realize that the two are related (albiet different), but since we only require KnownSafe in one direction, will match the inner retain on the original pointer with the guard release on the original pointer. This is fixed by ensuring that in the presense of allocas we only unconditionally remove pointers if both our retain and our release are KnownSafe (i.e. we are KnownSafe in both directions) since we must deal with the possibility that the frontend will emit what (to the optimizer) appears to be unbalanced retain/releases. An example of the miscompile is: %A = alloca retain(%x) retain(%x) <--- Inner Retain store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) release(%x) <--- Guarding Release getting optimized to: %A = alloca retain(%x) store %x, %A %y = load %A ... DO STUFF ... release(%y) call void @use(%x) rdar://13750319 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181743 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-13 23:49:42 +00:00
!0 = metadata !{}
declare i32 @__gxx_personality_v0(...)
Reference in New Issue Copy Permalink