maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/Attributor/liveness.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1194 lines
35 KiB
LLVM
Raw Normal View History

[Attributor][FIX] Transform invoke of nounwind to call + br %normal_dest Even if the invoked function may-return, we can replace it with a call and branch if it is nounwind. We had almost everything in place to do this but did not which actually caused a crash when we removed the landingpad from the actually dead unwind block. Exposed by the IPConstantProp tests.
2019-11-02 10:59:32 +08:00
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --scrub-attributes
[Attributor] Allow PHI nodes in AAValueConstantRangeFloating Traversing PHI nodes is natural with the genericValueTraversal but also a bit tricky. The problem is similar to the ones we have seen in AAAlign and AADereferenceable, namely that we continue to increase the range in each iteration. We use a pessimistic approach here to stop the iterations. Nevertheless, optimistic information can now be propagated through a PHI node.
2020-02-10 10:14:35 +08:00
; RUN: opt -attributor --attributor-disable=false -attributor-max-iterations-verify -attributor-annotate-decl-cs -attributor-max-iterations=6 -S < %s | FileCheck %s --check-prefixes=CHECK,MODULE,ALL_BUT_OLD_CGSCCC
; RUN: opt -attributor-cgscc --attributor-disable=false -attributor-annotate-decl-cs -attributor-max-iterations=6 -S < %s | FileCheck %s --check-prefixes=CHECK,CGSCC
; RUN: opt -passes=attributor --attributor-disable=false -attributor-max-iterations-verify -attributor-annotate-decl-cs -attributor-max-iterations=6 -S < %s | FileCheck %s --check-prefixes=CHECK,MODULE,ALL_BUT_OLD_CGSCCC
; RUN: opt -passes='attributor-cgscc' --attributor-disable=false -attributor-annotate-decl-cs -attributor-max-iterations=6 -S < %s | FileCheck %s --check-prefixes=CHECK,CGSCC,ALL_BUT_OLD_CGSCCC
[Utils] Allow "on-the-fly" argument changes for update_test_check scripts Update test scripts were limited because they performed a single action on the entire file and if that action was controlled by arguments, like the one introduced in D68819, there was no record of it. This patch introduces the capability of changing the arguments passed to the script "on-the-fly" while processing a test file. In addition, an "on/off" switch was added so that processing can be disabled for parts of the file where the content is simply copied. The last extension is a record of the invocation arguments in the auto generated NOTE. These arguments are also picked up in a subsequent invocation, allowing updates with special options enabled without user interaction. To change the arguments the string `UTC_ARGS:` has to be present in a line, followed by "additional command line arguments". That is everything that follows `UTC_ARGS:` will be added to a growing list of "command line arguments" which is reparsed after every update. Reviewed By: arichardson Differential Revision: https://reviews.llvm.org/D69701
2019-11-01 11:45:17 +08:00
; UTC_ARGS: --disable
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; ALL_BUT_OLD_CGSCCC: @dead_with_blockaddress_users.l = constant [2 x i8*] [i8* inttoptr (i32 1 to i8*), i8* inttoptr (i32 1 to i8*)]
[Attributor] Ignore BlockAddress users in call site traversal BlockAddress users will not "call" the function so they do not qualify as call sites in the first place. When we delete a function with BlockAddress users we need to first remove the body so they are properly discarded.
2019-11-02 11:35:18 +08:00
@dead_with_blockaddress_users.l = constant [2 x i8*] [i8* blockaddress(@dead_with_blockaddress_users, %lab0), i8* blockaddress(@dead_with_blockaddress_users, %end)]
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
declare void @no_return_call() nofree noreturn nounwind nosync
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
declare void @normal_call() readnone
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
declare i32 @foo()
[Attributor][FIX] Transform invoke of nounwind to call + br %normal_dest Even if the invoked function may-return, we can replace it with a call and branch if it is nounwind. We had almost everything in place to do this but did not which actually caused a crash when we removed the landingpad from the actually dead unwind block. Exposed by the IPConstantProp tests.
2019-11-02 10:59:32 +08:00
declare i32 @foo_nounwind() nounwind
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
declare i32 @foo_noreturn_nounwind() noreturn nounwind
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
declare i32 @foo_noreturn() noreturn
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
declare i32 @bar() nosync readnone
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
[Attributor] Liveness for internal functions. For an internal function, if all its call sites are dead, the body of the function is considered dead. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D66155 llvm-svn: 369470
2019-08-21 07:16:57 +08:00
; This internal function has no live call sites, so all its BBs are considered dead,
; and nothing should be deduced for it.
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE-NOT: define internal i32 @dead_internal_func(i32 %0)
[Attributor] Liveness for internal functions. For an internal function, if all its call sites are dead, the body of the function is considered dead. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D66155 llvm-svn: 369470
2019-08-21 07:16:57 +08:00
define internal i32 @dead_internal_func(i32 %0) {
%2 = icmp slt i32 %0, 1
br i1 %2, label %3, label %5
; <label>:3: ; preds = %5, %1
%4 = phi i32 [ 1, %1 ], [ %8, %5 ]
ret i32 %4
; <label>:5: ; preds = %1, %5
%6 = phi i32 [ %9, %5 ], [ 1, %1 ]
%7 = phi i32 [ %8, %5 ], [ 1, %1 ]
%8 = mul nsw i32 %6, %7
%9 = add nuw nsw i32 %6, 1
%10 = icmp eq i32 %6, %0
br i1 %10, label %3, label %5
}
[Attributor] Derive memory location attributes (argmemonly, ...) In addition to memory behavior attributes (readonly/writeonly) we now derive memory location attributes (argmemonly/inaccessiblememonly/...). The former is part of AAMemoryBehavior and the latter part of AAMemoryLocation. While they are similar in nature it got messy when they were put in a single AA. Location attributes for arguments and floating values will follow later. Note that both memory attributes kinds can derive readnone. If there are no accesses AAMemoryBehavior will derive readnone. If there are accesses but only to stack (=local) locations AAMemoryLocation will derive readnone. Reviewed By: uenoku Differential Revision: https://reviews.llvm.org/D73426
2020-01-26 16:51:57 +08:00
; CHECK: Function Attrs: argmemonly nofree norecurse nounwind uwtable willreturn
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
define i32 @volatile_load(i32*) norecurse nounwind uwtable {
%2 = load volatile i32, i32* %0, align 4
ret i32 %2
}
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE-NOT: internal_load
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
define internal i32 @internal_load(i32*) norecurse nounwind uwtable {
%2 = load i32, i32* %0, align 4
ret i32 %2
}
; TEST 1: Only first block is live.
[Attributor] Deduce the "no-return" attribute for functions A function is "no-return" if we never reach a return instruction, either because there are none or the ones that exist are dead. Test have been adjusted: - either noreturn was added, or - noreturn was avoided by modifying the code. The new noreturn_{sync,async} test make sure we do handle invoke instructions with a noreturn (and potentially nowunwind) callee correctly, even in the presence of potential asynchronous exceptions. llvm-svn: 367948
2019-08-06 07:22:05 +08:00
; CHECK: Function Attrs: nofree noreturn nosync nounwind
[Attributor] Use fine-grained liveness in all helpers We used coarse-grained liveness before, thus we looked if the instruction was executed, but we did not use fine-grained liveness, hence if the instruction was needed or could be deleted even if the surrounding ones are live. This patches introduces this level of liveness checks together with other liveness queries, e.g., for uses. For more control we enforce that all liveness queries go through the Attributor. Test have been adjusted to reflect the changes or augmented to prevent deletion of the parts we want to check. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73313
2020-01-12 15:09:22 +08:00
; MODULE-NEXT: define i32 @first_block_no_return(i32 %a, i32* nocapture nofree nonnull readnone %ptr1, i32* nocapture nofree readnone %ptr2)
; CGSCC-NEXT: define i32 @first_block_no_return(i32 %a, i32* nocapture nofree nonnull readnone align 4 dereferenceable(4) %ptr1, i32* nocapture nofree readnone %ptr2)
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
define i32 @first_block_no_return(i32 %a, i32* nonnull %ptr1, i32* %ptr2) #0 {
entry:
call i32 @internal_load(i32* %ptr1)
call void @no_return_call()
; CHECK: call void @no_return_call()
; CHECK-NEXT: unreachable
[Attributor] Liveness for internal functions. For an internal function, if all its call sites are dead, the body of the function is considered dead. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D66155 llvm-svn: 369470
2019-08-21 07:16:57 +08:00
call i32 @dead_internal_func(i32 10)
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call i32 @internal_load(i32* %ptr2)
%load = call i32 @volatile_load(i32* %ptr1)
call void @normal_call()
%call = call i32 @foo()
br label %cond.end
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %cond.true
%cond = phi i32 [ %call, %cond.true ], [ %call1, %cond.false ]
ret i32 %cond
}
; TEST 2: cond.true is dead, but cond.end is not, since cond.false is live
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; This is just an example. For example we can put a sync call in a
; dead block and check if it is deduced.
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; CHECK: Function Attrs: nosync
NoFree argument attribute. Summary: Deducing nofree atrribute for function arguments. Reviewers: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67886
2019-11-02 23:35:38 +08:00
; CHECK-NEXT: define i32 @dead_block_present(i32 %a, i32* nocapture nofree readnone %ptr1)
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
define i32 @dead_block_present(i32 %a, i32* %ptr1) #0 {
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @no_return_call()
; CHECK: call void @no_return_call()
; CHECK-NEXT: unreachable
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
%call = call i32 @volatile_load(i32* %ptr1)
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
br label %cond.end
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %cond.true
[Attributor] Use the delete API for liveness Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66833 llvm-svn: 370818
2019-09-04 04:42:16 +08:00
; CHECK: cond.end:
; CHECK-NEXT: ret i32 %call1
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
%cond = phi i32 [ %call, %cond.true ], [ %call1, %cond.false ]
ret i32 %cond
}
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; TEST 3: both cond.true and cond.false are dead, therfore cond.end is dead as well.
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
define i32 @all_dead(i32 %a) #0 {
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @no_return_call()
; CHECK: call void @no_return_call()
; CHECK-NEXT: unreachable
[Attributor] Liveness for internal functions. For an internal function, if all its call sites are dead, the body of the function is considered dead. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D66155 llvm-svn: 369470
2019-08-21 07:16:57 +08:00
call i32 @dead_internal_func(i32 10)
[Attributor] Use the delete API for liveness Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66833 llvm-svn: 370818
2019-09-04 04:42:16 +08:00
; CHECK-NOT: call
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
%call = call i32 @foo()
br label %cond.end
cond.false: ; preds = %entry
call void @no_return_call()
; CHECK: call void @no_return_call()
; CHECK-NEXT: unreachable
[Attributor] Liveness for internal functions. For an internal function, if all its call sites are dead, the body of the function is considered dead. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D66155 llvm-svn: 369470
2019-08-21 07:16:57 +08:00
call i32 @dead_internal_func(i32 10)
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %cond.true
%cond = phi i32 [ %call, %cond.true ], [ %call1, %cond.false ]
ret i32 %cond
}
declare i32 @__gxx_personality_v0(...)
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; TEST 4: All blocks are live.
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; CHECK: define i32 @all_live(i32 %a)
define i32 @all_live(i32 %a) #0 {
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @normal_call()
%call = call i32 @foo_noreturn()
br label %cond.end
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %cond.true
%cond = phi i32 [ %call, %cond.true ], [ %call1, %cond.false ]
ret i32 %cond
}
[Attributor][FIX] Transform invoke of nounwind to call + br %normal_dest Even if the invoked function may-return, we can replace it with a call and branch if it is nounwind. We had almost everything in place to do this but did not which actually caused a crash when we removed the landingpad from the actually dead unwind block. Exposed by the IPConstantProp tests.
2019-11-02 10:59:32 +08:00
; TEST 5.1 noreturn invoke instruction with a unreachable normal successor block.
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; CHECK: define i32 @invoke_noreturn(i32 %a)
define i32 @invoke_noreturn(i32 %a) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @normal_call()
%call = invoke i32 @foo_noreturn() to label %continue
unwind label %cleanup
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
; CHECK: %call = invoke i32 @foo_noreturn()
[Attributor][FIX] Avoid splitting blocks if possible Before, we eagerly split blocks even if it was not necessary, e.g., they had a single unreachable instruction and only a single predecessor. llvm-svn: 374703
2019-10-13 13:27:09 +08:00
; CHECK-NEXT: to label %continue unwind label %cleanup
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %continue
%cond = phi i32 [ %call, %continue ], [ %call1, %cond.false ]
ret i32 %cond
continue:
[Attributor][FIX] Avoid splitting blocks if possible Before, we eagerly split blocks even if it was not necessary, e.g., they had a single unreachable instruction and only a single predecessor. llvm-svn: 374703
2019-10-13 13:27:09 +08:00
; CHECK: continue:
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
; CHECK-NEXT: unreachable
br label %cond.end
cleanup:
%res = landingpad { i8*, i32 }
catch i8* null
ret i32 0
}
[Attributor][FIX] Transform invoke of nounwind to call + br %normal_dest Even if the invoked function may-return, we can replace it with a call and branch if it is nounwind. We had almost everything in place to do this but did not which actually caused a crash when we removed the landingpad from the actually dead unwind block. Exposed by the IPConstantProp tests.
2019-11-02 10:59:32 +08:00
; TEST 5.2 noreturn invoke instruction replaced by a call and an unreachable instruction
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
; put after it.
; CHECK: define i32 @invoke_noreturn_nounwind(i32 %a)
define i32 @invoke_noreturn_nounwind(i32 %a) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @normal_call()
%call = invoke i32 @foo_noreturn_nounwind() to label %continue
unwind label %cleanup
[Attributor][Fix] Do not remove instructions during manifestation When we remove instructions cached references could still be live. This patch avoids removing invoke instructions that are replaced by calls and instead keeps them around but in a dead block. llvm-svn: 367933
2019-08-06 05:35:02 +08:00
; CHECK: call void @normal_call()
; CHECK-NEXT: call i32 @foo_noreturn_nounwind()
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
; CHECK-NEXT: unreachable
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
[Attributor] Use the delete API for liveness Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66833 llvm-svn: 370818
2019-09-04 04:42:16 +08:00
; CHECK-NOT: @foo_noreturn_nounwind()
[Attributor][Fix] Do not remove instructions during manifestation When we remove instructions cached references could still be live. This patch avoids removing invoke instructions that are replaced by calls and instead keeps them around but in a dead block. llvm-svn: 367933
2019-08-06 05:35:02 +08:00
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %continue
%cond = phi i32 [ %call, %continue ], [ %call1, %cond.false ]
ret i32 %cond
continue:
br label %cond.end
cleanup:
%res = landingpad { i8*, i32 }
catch i8* null
ret i32 0
}
[Attributor][FIX] Transform invoke of nounwind to call + br %normal_dest Even if the invoked function may-return, we can replace it with a call and branch if it is nounwind. We had almost everything in place to do this but did not which actually caused a crash when we removed the landingpad from the actually dead unwind block. Exposed by the IPConstantProp tests.
2019-11-02 10:59:32 +08:00
; TEST 5.3 unounwind invoke instruction replaced by a call and a branch instruction put after it.
define i32 @invoke_nounwind(i32 %a) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
; CHECK-LABEL: define {{[^@]+}}@invoke_nounwind
; CHECK: cond.true:
; CHECK-NEXT: call void @normal_call()
; CHECK-NEXT: [[CALL:%.*]] = call i32 @foo_nounwind()
; CHECK-NEXT: br label [[CONTINUE:%.*]]
; CHECK: continue:
; CHECK-NEXT: br label [[COND_END:%.*]]
;
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @normal_call()
%call = invoke i32 @foo_nounwind() to label %continue
unwind label %cleanup
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %cond.end
cond.end: ; preds = %cond.false, %continue
%cond = phi i32 [ %call, %continue ], [ %call1, %cond.false ]
ret i32 %cond
continue:
br label %cond.end
cleanup:
%res = landingpad { i8*, i32 }
catch i8* null
ret i32 0
}
[Utils] Allow "on-the-fly" argument changes for update_test_check scripts Update test scripts were limited because they performed a single action on the entire file and if that action was controlled by arguments, like the one introduced in D68819, there was no record of it. This patch introduces the capability of changing the arguments passed to the script "on-the-fly" while processing a test file. In addition, an "on/off" switch was added so that processing can be disabled for parts of the file where the content is simply copied. The last extension is a record of the invocation arguments in the auto generated NOTE. These arguments are also picked up in a subsequent invocation, allowing updates with special options enabled without user interaction. To change the arguments the string `UTC_ARGS:` has to be present in a line, followed by "additional command line arguments". That is everything that follows `UTC_ARGS:` will be added to a growing list of "command line arguments" which is reparsed after every update. Reviewed By: arichardson Differential Revision: https://reviews.llvm.org/D69701
2019-11-01 11:45:17 +08:00
; UTC_ARGS: --enable
[Attributor][FIX] Carefully change invokes to calls (after manifest) Before we manually inserted unreachable early but that could lead to broken PHI nodes. Now we use the existing late modification functionality.
2020-01-08 06:01:57 +08:00
; TEST 5.4 unounwind invoke instruction replaced by a call and a branch instruction put after it.
define i32 @invoke_nounwind_phi(i32 %a) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
; CHECK-LABEL: define {{[^@]+}}@invoke_nounwind_phi
; CHECK-SAME: (i32 [[A:%.*]]) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*)
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[A]], 0
; CHECK-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
; CHECK: cond.true:
; CHECK-NEXT: call void @normal_call()
; CHECK-NEXT: [[CALL:%.*]] = call i32 @foo_nounwind()
; CHECK-NEXT: br label [[CONTINUE:%.*]]
; CHECK: cond.false:
; CHECK-NEXT: call void @normal_call()
; CHECK-NEXT: [[CALL1:%.*]] = call i32 @bar()
; CHECK-NEXT: br label [[CONTINUE]]
; CHECK: continue:
; CHECK-NEXT: [[P:%.*]] = phi i32 [ 0, [[COND_TRUE]] ], [ 1, [[COND_FALSE]] ]
; CHECK-NEXT: ret i32 [[P]]
; CHECK: cleanup:
; CHECK-NEXT: unreachable
;
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @normal_call()
%call = invoke i32 @foo_nounwind() to label %continue
unwind label %cleanup
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %continue
continue:
%p = phi i32 [ 0, %cond.true ], [ 1, %cond.false ]
ret i32 %p
cleanup:
%res = landingpad { i8*, i32 } catch i8* null
ret i32 0
}
; TEST 5.5 unounwind invoke instruction replaced by a call and a branch instruction put after it.
define i32 @invoke_nounwind_phi_dom(i32 %a) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
; CHECK-LABEL: define {{[^@]+}}@invoke_nounwind_phi_dom
; CHECK-SAME: (i32 [[A:%.*]]) personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*)
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[A]], 0
; CHECK-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
; CHECK: cond.true:
; CHECK-NEXT: call void @normal_call()
; CHECK-NEXT: [[CALL:%.*]] = call i32 @foo_nounwind()
; CHECK-NEXT: br label [[CONTINUE:%.*]]
; CHECK: cond.false:
; CHECK-NEXT: call void @normal_call()
; CHECK-NEXT: [[CALL1:%.*]] = call i32 @bar()
; CHECK-NEXT: br label [[CONTINUE]]
; CHECK: continue:
; CHECK-NEXT: [[P:%.*]] = phi i32 [ [[CALL]], [[COND_TRUE]] ], [ [[CALL1]], [[COND_FALSE]] ]
; CHECK-NEXT: ret i32 [[P]]
; CHECK: cleanup:
; CHECK-NEXT: unreachable
;
entry:
%cmp = icmp eq i32 %a, 0
br i1 %cmp, label %cond.true, label %cond.false
cond.true: ; preds = %entry
call void @normal_call()
%call = invoke i32 @foo_nounwind() to label %continue
unwind label %cleanup
cond.false: ; preds = %entry
call void @normal_call()
%call1 = call i32 @bar()
br label %continue
continue:
%p = phi i32 [ %call, %cond.true ], [ %call1, %cond.false ]
ret i32 %p
cleanup:
%res = landingpad { i8*, i32 } catch i8* null
ret i32 0
}
[Utils] Allow "on-the-fly" argument changes for update_test_check scripts Update test scripts were limited because they performed a single action on the entire file and if that action was controlled by arguments, like the one introduced in D68819, there was no record of it. This patch introduces the capability of changing the arguments passed to the script "on-the-fly" while processing a test file. In addition, an "on/off" switch was added so that processing can be disabled for parts of the file where the content is simply copied. The last extension is a record of the invocation arguments in the auto generated NOTE. These arguments are also picked up in a subsequent invocation, allowing updates with special options enabled without user interaction. To change the arguments the string `UTC_ARGS:` has to be present in a line, followed by "additional command line arguments". That is everything that follows `UTC_ARGS:` will be added to a growing list of "command line arguments" which is reparsed after every update. Reviewed By: arichardson Differential Revision: https://reviews.llvm.org/D69701
2019-11-01 11:45:17 +08:00
; UTC_ARGS: --disable
[Attributor][FIX] Transform invoke of nounwind to call + br %normal_dest Even if the invoked function may-return, we can replace it with a call and branch if it is nounwind. We had almost everything in place to do this but did not which actually caused a crash when we removed the landingpad from the actually dead unwind block. Exposed by the IPConstantProp tests.
2019-11-02 10:59:32 +08:00
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; TEST 6: Undefined behvior, taken from LangRef.
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; FIXME: Should be able to detect undefined behavior.
NoFree argument attribute. Summary: Deducing nofree atrribute for function arguments. Reviewers: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67886
2019-11-02 23:35:38 +08:00
; CHECK: define void @ub(i32* nocapture nofree writeonly %0)
[Attributor][Fix] Resolve various liveness issues Summary: This contains various fixes: - Explicitly determine and return the next noreturn instruction. - If an invoke calls a noreturn function which is not nounwind we keep the unwind destination live. This also means we require an invoke. Though we can still add the unreachable to the normal destination block. - Check if the return instructions are dead after we look for calls to avoid triggering an optimistic fixpoint in the presence of assumed liveness information. - Make the interface work with "const" pointers. - Some simplifications While additional tests are included, full coverage is achieved only with D59978. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D65701 llvm-svn: 367791
2019-08-05 02:38:53 +08:00
define void @ub(i32* %0) {
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
%poison = sub nuw i32 0, 1 ; Results in a poison value.
%still_poison = and i32 %poison, 0 ; 0, but also poison.
%poison_yet_again = getelementptr i32, i32* %0, i32 %still_poison
store i32 0, i32* %poison_yet_again ; Undefined behavior due to store to poison.
ret void
}
define void @inf_loop() #0 {
entry:
br label %while.body
while.body: ; preds = %entry, %while.body
br label %while.body
}
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; TEST 7: Infinite loop.
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; FIXME: Detect infloops, and mark affected blocks dead.
define i32 @test5(i32, i32) #0 {
%3 = icmp sgt i32 %0, %1
br i1 %3, label %cond.if, label %cond.elseif
cond.if: ; preds = %2
%4 = tail call i32 @bar()
br label %cond.end
cond.elseif: ; preds = %2
call void @inf_loop()
%5 = icmp slt i32 %0, %1
br i1 %5, label %cond.end, label %cond.else
cond.else: ; preds = %cond.elseif
%6 = tail call i32 @foo()
br label %cond.end
cond.end: ; preds = %cond.if, %cond.else, %cond.elseif
%7 = phi i32 [ %1, %cond.elseif ], [ 0, %cond.else ], [ 0, %cond.if ]
ret i32 %7
}
define void @rec() #0 {
entry:
call void @rec()
ret void
}
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; TEST 8: Recursion
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; FIXME: everything after first block should be marked dead
; and unreachable should be put after call to @rec().
define i32 @test6(i32, i32) #0 {
call void @rec()
%3 = icmp sgt i32 %0, %1
br i1 %3, label %cond.if, label %cond.elseif
cond.if: ; preds = %2
%4 = tail call i32 @bar()
br label %cond.end
cond.elseif: ; preds = %2
call void @rec()
%5 = icmp slt i32 %0, %1
br i1 %5, label %cond.end, label %cond.else
cond.else: ; preds = %cond.elseif
%6 = tail call i32 @foo()
br label %cond.end
cond.end: ; preds = %cond.if, %cond.else, %cond.elseif
%7 = phi i32 [ %1, %cond.elseif ], [ 0, %cond.else ], [ 0, %cond.if ]
ret i32 %7
}
[Attributor] Using liveness in other attributes. Modifying other AbstractAttributes to use Liveness AA and skip dead instructions. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential revision: https://reviews.llvm.org/D65243 llvm-svn: 367725
2019-08-03 05:31:22 +08:00
; TEST 9: Recursion
Fixing build error from commit 95cbc3d [Attributor] Liveness analysis. Liveness analysis abstract attribute used to indicate which BasicBlocks are dead and can therefore be ignored. Right now we are only looking at noreturn calls. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D64162 llvm-svn: 366769
2019-07-23 07:58:23 +08:00
; FIXME: contains recursive call to itself in cond.elseif block
define i32 @test7(i32, i32) #0 {
%3 = icmp sgt i32 %0, %1
br i1 %3, label %cond.if, label %cond.elseif
cond.if: ; preds = %2
%4 = tail call i32 @bar()
br label %cond.end
cond.elseif: ; preds = %2
%5 = tail call i32 @test7(i32 %0, i32 %1)
%6 = icmp slt i32 %0, %1
br i1 %6, label %cond.end, label %cond.else
cond.else: ; preds = %cond.elseif
%7 = tail call i32 @foo()
br label %cond.end
cond.end: ; preds = %cond.if, %cond.else, %cond.elseif
%8 = phi i32 [ %1, %cond.elseif ], [ 0, %cond.else ], [ 0, %cond.if ]
ret i32 %8
}
[Attributor] Liveness for internal functions. For an internal function, if all its call sites are dead, the body of the function is considered dead. Reviewers: jdoerfert, uenoku Subscribers: hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D66155 llvm-svn: 369470
2019-08-21 07:16:57 +08:00
; SCC test
;
; char a1 __attribute__((aligned(8)));
; char a2 __attribute__((aligned(16)));
;
; char* f1(char* a ){
; return a?a:f2(&a1);
; }
; char* f2(char* a){
; return a?f1(a):f3(&a2);
; }
;
; char* f3(char* a){
; return a?&a1: f1(&a2);
; }
@a1 = common global i8 0, align 8
@a2 = common global i8 0, align 16
define internal i8* @f1(i8* readnone %0) local_unnamed_addr #0 {
; ATTRIBUTOR: define internal i8* @f1(i8* readnone %0)
%2 = icmp eq i8* %0, null
br i1 %2, label %3, label %5
; <label>:3: ; preds = %1
; ATTRIBUTOR: %4 = tail call i8* undef(i8* nonnull align 8 @a1)
%4 = tail call i8* @f2(i8* nonnull @a1)
br label %5
; <label>:5: ; preds = %1, %3
%6 = phi i8* [ %4, %3 ], [ %0, %1 ]
ret i8* %6
}
define internal i8* @f2(i8* readnone %0) local_unnamed_addr #0 {
; ATTRIBUTOR: define internal i8* @f2(i8* readnone %0)
%2 = icmp eq i8* %0, null
br i1 %2, label %5, label %3
; <label>:3: ; preds = %1
; ATTRIBUTOR: %4 = tail call i8* undef(i8* nonnull align 8 %0)
%4 = tail call i8* @f1(i8* nonnull %0)
br label %7
; <label>:5: ; preds = %1
; ATTRIBUTOR: %6 = tail call i8* undef(i8* nonnull align 16 @a2)
%6 = tail call i8* @f3(i8* nonnull @a2)
br label %7
; <label>:7: ; preds = %5, %3
%8 = phi i8* [ %4, %3 ], [ %6, %5 ]
ret i8* %8
}
define internal i8* @f3(i8* readnone %0) local_unnamed_addr #0 {
; ATTRIBUTOR: define internal i8* @f3(i8* readnone %0)
%2 = icmp eq i8* %0, null
br i1 %2, label %3, label %5
; <label>:3: ; preds = %1
; ATTRIBUTOR: %4 = tail call i8* undef(i8* nonnull align 16 @a2)
%4 = tail call i8* @f1(i8* nonnull @a2)
br label %5
; <label>:5: ; preds = %1, %3
%6 = phi i8* [ %4, %3 ], [ @a1, %1 ]
ret i8* %6
}
[Attributor] Use existing unreachable instead of introducing new ones So far we split the unreachable off and placed a new one, this is not necessary. llvm-svn: 369575
2019-08-22 04:56:41 +08:00
[Attributor][FIX] Remove duplicate, half-broken functionality The changeXXXAfterManifest functions are better suited to deal with changes so we should prefer them. These functions also recursively delete dead instructions which is why we see test changes.
2020-01-12 14:17:08 +08:00
declare void @sink() nofree nosync nounwind willreturn
[Attributor] Use existing unreachable instead of introducing new ones So far we split the unreachable off and placed a new one, this is not necessary. llvm-svn: 369575
2019-08-22 04:56:41 +08:00
define void @test_unreachable() {
; CHECK: define void @test_unreachable()
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
; CHECK-NEXT: call void @sink()
[Attributor] Use existing unreachable instead of introducing new ones So far we split the unreachable off and placed a new one, this is not necessary. llvm-svn: 369575
2019-08-22 04:56:41 +08:00
; CHECK-NEXT: call void @test_unreachable()
; CHECK-NEXT: unreachable
; CHECK-NEXT: }
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
call void @sink()
[Attributor] Use existing unreachable instead of introducing new ones So far we split the unreachable off and placed a new one, this is not necessary. llvm-svn: 369575
2019-08-22 04:56:41 +08:00
call void @test_unreachable()
unreachable
}
[Attributor] Look at internal functions only on-demand Summary: Instead of building attributes for internal functions which we do not update as long as we assume they are dead, we now do not create attributes until we assume the internal function to be live. This improves the number of required iterations, as well as the number of required updates, in real code. On our tests, the results are mixed. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66914 llvm-svn: 370924
2019-09-05 00:35:20 +08:00
define linkonce_odr void @non_exact1() {
call void @non_dead_a0()
call void @non_dead_a1()
call void @non_dead_a2()
call void @non_dead_a3()
call void @non_dead_a4()
call void @non_dead_a5()
call void @non_dead_a6()
call void @non_dead_a7()
call void @non_dead_a8()
call void @non_dead_a9()
call void @non_dead_a10()
call void @non_dead_a11()
call void @non_dead_a12()
call void @non_dead_a13()
call void @non_dead_a14()
call void @non_dead_a15()
call void @middle()
ret void
}
define internal void @middle() {
bb0:
call void @non_dead_b0()
call void @non_dead_b1()
call void @non_dead_b2()
call void @non_dead_b3()
br label %bb1
bb1:
call void @non_dead_b4()
call void @non_dead_b5()
call void @non_dead_b6()
call void @non_dead_b7()
br label %bb2
bb2:
call void @non_dead_b8()
call void @non_dead_b9()
call void @non_dead_b10()
call void @non_dead_b11()
br label %bb3
bb3:
call void @non_dead_b12()
call void @non_dead_b13()
call void @non_dead_b14()
call void @non_dead_b15()
br label %bb4
bb4:
call void @non_exact2()
ret void
}
define linkonce_odr void @non_exact2() {
call void @non_dead_c0()
call void @non_dead_c1()
call void @non_dead_c2()
call void @non_dead_c3()
call void @non_dead_c4()
call void @non_dead_c5()
call void @non_dead_c6()
call void @non_dead_c7()
call void @non_dead_c8()
call void @non_dead_c9()
call void @non_dead_c10()
call void @non_dead_c11()
call void @non_dead_c12()
call void @non_dead_c13()
call void @non_dead_c14()
call void @non_dead_c15()
call void @non_exact3()
ret void
}
define linkonce_odr void @non_exact3() {
call void @non_dead_d0()
call void @non_dead_d1()
call void @non_dead_d2()
call void @non_dead_d3()
call void @non_dead_d4()
call void @non_dead_d5()
call void @non_dead_d6()
call void @non_dead_d7()
call void @non_dead_d8()
call void @non_dead_d9()
call void @non_dead_d10()
call void @non_dead_d11()
call void @non_dead_d12()
call void @non_dead_d13()
call void @non_dead_d14()
call void @non_dead_d15()
%nr = call i32 @foo_noreturn()
call void @dead_e1()
ret void
}
; CHECK: define linkonce_odr void @non_exact3() {
; CHECK-NEXT: call void @non_dead_d0()
; CHECK-NEXT: call void @non_dead_d1()
; CHECK-NEXT: call void @non_dead_d2()
; CHECK-NEXT: call void @non_dead_d3()
; CHECK-NEXT: call void @non_dead_d4()
; CHECK-NEXT: call void @non_dead_d5()
; CHECK-NEXT: call void @non_dead_d6()
; CHECK-NEXT: call void @non_dead_d7()
; CHECK-NEXT: call void @non_dead_d8()
; CHECK-NEXT: call void @non_dead_d9()
; CHECK-NEXT: call void @non_dead_d10()
; CHECK-NEXT: call void @non_dead_d11()
; CHECK-NEXT: call void @non_dead_d12()
; CHECK-NEXT: call void @non_dead_d13()
; CHECK-NEXT: call void @non_dead_d14()
; CHECK-NEXT: call void @non_dead_d15()
; CHECK-NEXT: %nr = call i32 @foo_noreturn()
; CHECK-NEXT: unreachable
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
define internal void @non_dead_a0() {
call void @sink()
ret void
}
define internal void @non_dead_a1() {
call void @sink()
ret void
}
define internal void @non_dead_a2() {
call void @sink()
ret void
}
define internal void @non_dead_a3() {
call void @sink()
ret void
}
define internal void @non_dead_a4() {
call void @sink()
ret void
}
define internal void @non_dead_a5() {
call void @sink()
ret void
}
define internal void @non_dead_a6() {
call void @sink()
ret void
}
define internal void @non_dead_a7() {
call void @sink()
ret void
}
define internal void @non_dead_a8() {
call void @sink()
ret void
}
define internal void @non_dead_a9() {
call void @sink()
ret void
}
define internal void @non_dead_a10() {
call void @sink()
ret void
}
define internal void @non_dead_a11() {
call void @sink()
ret void
}
define internal void @non_dead_a12() {
call void @sink()
ret void
}
define internal void @non_dead_a13() {
call void @sink()
ret void
}
define internal void @non_dead_a14() {
call void @sink()
ret void
}
define internal void @non_dead_a15() {
call void @sink()
ret void
}
define internal void @non_dead_b0() {
call void @sink()
ret void
}
define internal void @non_dead_b1() {
call void @sink()
ret void
}
define internal void @non_dead_b2() {
call void @sink()
ret void
}
define internal void @non_dead_b3() {
call void @sink()
ret void
}
define internal void @non_dead_b4() {
call void @sink()
ret void
}
define internal void @non_dead_b5() {
call void @sink()
ret void
}
define internal void @non_dead_b6() {
call void @sink()
ret void
}
define internal void @non_dead_b7() {
call void @sink()
ret void
}
define internal void @non_dead_b8() {
call void @sink()
ret void
}
define internal void @non_dead_b9() {
call void @sink()
ret void
}
define internal void @non_dead_b10() {
call void @sink()
ret void
}
define internal void @non_dead_b11() {
call void @sink()
ret void
}
define internal void @non_dead_b12() {
call void @sink()
ret void
}
define internal void @non_dead_b13() {
call void @sink()
ret void
}
define internal void @non_dead_b14() {
call void @sink()
ret void
}
define internal void @non_dead_b15() {
call void @sink()
ret void
}
define internal void @non_dead_c0() {
call void @sink()
ret void
}
define internal void @non_dead_c1() {
call void @sink()
ret void
}
define internal void @non_dead_c2() {
call void @sink()
ret void
}
define internal void @non_dead_c3() {
call void @sink()
ret void
}
define internal void @non_dead_c4() {
call void @sink()
ret void
}
define internal void @non_dead_c5() {
call void @sink()
ret void
}
define internal void @non_dead_c6() {
call void @sink()
ret void
}
define internal void @non_dead_c7() {
call void @sink()
ret void
}
define internal void @non_dead_c8() {
call void @sink()
ret void
}
define internal void @non_dead_c9() {
call void @sink()
ret void
}
define internal void @non_dead_c10() {
call void @sink()
ret void
}
define internal void @non_dead_c11() {
call void @sink()
ret void
}
define internal void @non_dead_c12() {
call void @sink()
ret void
}
define internal void @non_dead_c13() {
call void @sink()
ret void
}
define internal void @non_dead_c14() {
call void @sink()
ret void
}
define internal void @non_dead_c15() {
call void @sink()
ret void
}
define internal void @non_dead_d0() {
call void @sink()
ret void
}
define internal void @non_dead_d1() {
call void @sink()
ret void
}
define internal void @non_dead_d2() {
call void @sink()
ret void
}
define internal void @non_dead_d3() {
call void @sink()
ret void
}
define internal void @non_dead_d4() {
call void @sink()
ret void
}
define internal void @non_dead_d5() {
call void @sink()
ret void
}
define internal void @non_dead_d6() {
call void @sink()
ret void
}
define internal void @non_dead_d7() {
call void @sink()
ret void
}
define internal void @non_dead_d8() {
call void @sink()
ret void
}
define internal void @non_dead_d9() {
call void @sink()
ret void
}
define internal void @non_dead_d10() {
call void @sink()
ret void
}
define internal void @non_dead_d11() {
call void @sink()
ret void
}
define internal void @non_dead_d12() {
call void @sink()
ret void
}
define internal void @non_dead_d13() {
call void @sink()
ret void
}
define internal void @non_dead_d14() {
call void @sink()
ret void
}
define internal void @non_dead_d15() {
call void @sink()
ret void
}
[Attributor] Look at internal functions only on-demand Summary: Instead of building attributes for internal functions which we do not update as long as we assume they are dead, we now do not create attributes until we assume the internal function to be live. This improves the number of required iterations, as well as the number of required updates, in real code. On our tests, the results are mixed. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66914 llvm-svn: 370924
2019-09-05 00:35:20 +08:00
define internal void @dead_e0() { call void @dead_e1() ret void }
define internal void @dead_e1() { call void @dead_e2() ret void }
define internal void @dead_e2() { ret void }
; CHECK: define internal void @non_dead_a0()
; CHECK: define internal void @non_dead_a1()
; CHECK: define internal void @non_dead_a2()
; CHECK: define internal void @non_dead_a3()
; CHECK: define internal void @non_dead_a4()
; CHECK: define internal void @non_dead_a5()
; CHECK: define internal void @non_dead_a6()
; CHECK: define internal void @non_dead_a7()
; CHECK: define internal void @non_dead_a8()
; CHECK: define internal void @non_dead_a9()
; CHECK: define internal void @non_dead_a10()
; CHECK: define internal void @non_dead_a11()
; CHECK: define internal void @non_dead_a12()
; CHECK: define internal void @non_dead_a13()
; CHECK: define internal void @non_dead_a14()
; CHECK: define internal void @non_dead_a15()
; CHECK: define internal void @non_dead_b0()
; CHECK: define internal void @non_dead_b1()
; CHECK: define internal void @non_dead_b2()
; CHECK: define internal void @non_dead_b3()
; CHECK: define internal void @non_dead_b4()
; CHECK: define internal void @non_dead_b5()
; CHECK: define internal void @non_dead_b6()
; CHECK: define internal void @non_dead_b7()
; CHECK: define internal void @non_dead_b8()
; CHECK: define internal void @non_dead_b9()
; CHECK: define internal void @non_dead_b10()
; CHECK: define internal void @non_dead_b11()
; CHECK: define internal void @non_dead_b12()
; CHECK: define internal void @non_dead_b13()
; CHECK: define internal void @non_dead_b14()
; CHECK: define internal void @non_dead_b15()
; CHECK: define internal void @non_dead_c0()
; CHECK: define internal void @non_dead_c1()
; CHECK: define internal void @non_dead_c2()
; CHECK: define internal void @non_dead_c3()
; CHECK: define internal void @non_dead_c4()
; CHECK: define internal void @non_dead_c5()
; CHECK: define internal void @non_dead_c6()
; CHECK: define internal void @non_dead_c7()
; CHECK: define internal void @non_dead_c8()
; CHECK: define internal void @non_dead_c9()
; CHECK: define internal void @non_dead_c10()
; CHECK: define internal void @non_dead_c11()
; CHECK: define internal void @non_dead_c12()
; CHECK: define internal void @non_dead_c13()
; CHECK: define internal void @non_dead_c14()
; CHECK: define internal void @non_dead_c15()
; CHECK: define internal void @non_dead_d0()
; CHECK: define internal void @non_dead_d1()
; CHECK: define internal void @non_dead_d2()
; CHECK: define internal void @non_dead_d3()
; CHECK: define internal void @non_dead_d4()
; CHECK: define internal void @non_dead_d5()
; CHECK: define internal void @non_dead_d6()
; CHECK: define internal void @non_dead_d7()
; CHECK: define internal void @non_dead_d8()
; CHECK: define internal void @non_dead_d9()
; CHECK: define internal void @non_dead_d10()
; CHECK: define internal void @non_dead_d11()
; CHECK: define internal void @non_dead_d12()
; CHECK: define internal void @non_dead_d13()
; CHECK: define internal void @non_dead_d14()
; Verify we actually deduce information for these functions.
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
; MODULE: Function Attrs: nofree nosync nounwind willreturn
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE-NEXT: define internal void @non_dead_d15()
; MODULE-NOT: define internal void @dead_e
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
; CGSCC: Function Attrs: nofree nosync nounwind willreturn
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; CGSCC-NEXT: define internal void @non_dead_d15()
[Attributor][Fix] Make sure we do not delete live code Summary: Liveness needs to mark edges, not blocks as dead. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67191 llvm-svn: 370975
2019-09-05 04:34:52 +08:00
declare void @blowup() noreturn
define void @live_with_dead_entry() personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
; CHECK: define void @live_with_dead_entry(
; CHECK-NEXT: entry:
; CHECK-NEXT: invoke void @blowup()
; CHECK-NEXT: to label %live_with_dead_entry.dead unwind label %lpad
; CHECK: lpad: ; preds = %entry
; CHECK-NEXT: %0 = landingpad { i8*, i32 }
; CHECK-NEXT: catch i8* null
; CHECK-NEXT: br label %live_with_dead_entry
; CHECK: live_with_dead_entry.dead: ; preds = %entry
; CHECK-NEXT: unreachable
; CHECK: live_with_dead_entry: ; preds = %lpad
; CHECK-NEXT: ret void
entry:
invoke void @blowup() to label %live_with_dead_entry unwind label %lpad
lpad:
%0 = landingpad { i8*, i32 } catch i8* null
br label %live_with_dead_entry
live_with_dead_entry:
ret void
}
define void @live_with_dead_entry_lp() personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
; CHECK: define void @live_with_dead_entry_lp(
; CHECK-NEXT: entry:
; CHECK-NEXT: invoke void @blowup()
[Attributor][FIX] Carefully change invokes to calls (after manifest) Before we manually inserted unreachable early but that could lead to broken PHI nodes. Now we use the existing late modification functionality.
2020-01-08 06:01:57 +08:00
; CHECK-NEXT: to label %[[LIVE_WITH_DEAD_ENTRY_DEAD1:.*]] unwind label %[[LP1:.*]]
; CHECK: [[LP1]]: ; preds = %entry
[Attributor][Fix] Make sure we do not delete live code Summary: Liveness needs to mark edges, not blocks as dead. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67191 llvm-svn: 370975
2019-09-05 04:34:52 +08:00
; CHECK-NEXT: %lp = landingpad { i8*, i32 }
; CHECK-NEXT: catch i8* null
; CHECK-NEXT: invoke void @blowup()
[Attributor][FIX] Carefully change invokes to calls (after manifest) Before we manually inserted unreachable early but that could lead to broken PHI nodes. Now we use the existing late modification functionality.
2020-01-08 06:01:57 +08:00
; CHECK-NEXT: to label %[[LIVE_WITH_DEAD_ENTRY_DEAD2:.*]] unwind label %[[LP2:.*]]
; CHECK: [[LP2]]: ; preds = %lp1
[Attributor][Fix] Make sure we do not delete live code Summary: Liveness needs to mark edges, not blocks as dead. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67191 llvm-svn: 370975
2019-09-05 04:34:52 +08:00
; CHECK-NEXT: %0 = landingpad { i8*, i32 }
; CHECK-NEXT: catch i8* null
; CHECK-NEXT: br label %live_with_dead_entry
[Attributor][FIX] Carefully change invokes to calls (after manifest) Before we manually inserted unreachable early but that could lead to broken PHI nodes. Now we use the existing late modification functionality.
2020-01-08 06:01:57 +08:00
; CHECK: [[LIVE_WITH_DEAD_ENTRY_DEAD1]]:
[Attributor][Fix] Make sure we do not delete live code Summary: Liveness needs to mark edges, not blocks as dead. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67191 llvm-svn: 370975
2019-09-05 04:34:52 +08:00
; CHECK-NEXT: unreachable
[Attributor][FIX] Carefully change invokes to calls (after manifest) Before we manually inserted unreachable early but that could lead to broken PHI nodes. Now we use the existing late modification functionality.
2020-01-08 06:01:57 +08:00
; CHECK: [[LIVE_WITH_DEAD_ENTRY_DEAD2]]:
[Attributor][Fix] Make sure we do not delete live code Summary: Liveness needs to mark edges, not blocks as dead. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67191 llvm-svn: 370975
2019-09-05 04:34:52 +08:00
; CHECK-NEXT: unreachable
; CHECK: live_with_dead_entry: ; preds = %lp2
; CHECK-NEXT: ret void
entry:
invoke void @blowup() to label %live_with_dead_entry unwind label %lp1
lp1:
%lp = landingpad { i8*, i32 } catch i8* null
invoke void @blowup() to label %live_with_dead_entry unwind label %lp2
lp2:
%0 = landingpad { i8*, i32 } catch i8* null
br label %live_with_dead_entry
live_with_dead_entry:
ret void
}
[Attributor] Liveness for values Summary: This patch introduces liveness (AAIsDead) for all positions, thus for all kinds of values. For now, we say an instruction is dead if it would be removed assuming all users are dead. A call site return is different as we just look at the users. If all call site returns have been eliminated, the return values can return undef instead of their original value, eliminating uses. We try to recursively delete dead instructions now and we introduce a simple check interface for use-traversal. This is the idea tried out in D68626 but implemented in the right way. Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68925
2019-10-13 16:08:18 +08:00
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE: define internal void @useless_arg_sink()
; CGSCC: define internal void @useless_arg_sink(i32*{{.*}} %a)
[Attributor] Liveness for values Summary: This patch introduces liveness (AAIsDead) for all positions, thus for all kinds of values. For now, we say an instruction is dead if it would be removed assuming all users are dead. A call site return is different as we just look at the users. If all call site returns have been eliminated, the return values can return undef instead of their original value, eliminating uses. We try to recursively delete dead instructions now and we introduce a simple check interface for use-traversal. This is the idea tried out in D68626 but implemented in the right way. Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68925
2019-10-13 16:08:18 +08:00
define internal void @useless_arg_sink(i32* %a) {
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
call void @sink()
[Attributor] Liveness for values Summary: This patch introduces liveness (AAIsDead) for all positions, thus for all kinds of values. For now, we say an instruction is dead if it would be removed assuming all users are dead. A call site return is different as we just look at the users. If all call site returns have been eliminated, the return values can return undef instead of their original value, eliminating uses. We try to recursively delete dead instructions now and we introduce a simple check interface for use-traversal. This is the idea tried out in D68626 but implemented in the right way. Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68925
2019-10-13 16:08:18 +08:00
ret void
}
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE: define internal void @useless_arg_almost_sink()
; CGSCC: define internal void @useless_arg_almost_sink(i32*{{.*}} %a)
[Attributor] Liveness for values Summary: This patch introduces liveness (AAIsDead) for all positions, thus for all kinds of values. For now, we say an instruction is dead if it would be removed assuming all users are dead. A call site return is different as we just look at the users. If all call site returns have been eliminated, the return values can return undef instead of their original value, eliminating uses. We try to recursively delete dead instructions now and we introduce a simple check interface for use-traversal. This is the idea tried out in D68626 but implemented in the right way. Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68925
2019-10-13 16:08:18 +08:00
define internal void @useless_arg_almost_sink(i32* %a) {
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE: call void @useless_arg_sink()
[Attributor] Use assumed information to determine side-effects We relied on wouldInstructionBeTriviallyDead before but that functions does not take assumed information, especially for calls, into account. The replacement, AAIsDead::isAssumeSideEffectFree, does. This change makes AAIsDeadCallSiteReturn more complex as we can have a dead call or only dead users. The test have been modified to include a side effect where there was none in order to keep the coverage. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D73311
2020-01-12 14:34:38 +08:00
; CGSCC: call void @useless_arg_sink(i32* noalias nocapture nofree readnone %a)
[Attributor] Liveness for values Summary: This patch introduces liveness (AAIsDead) for all positions, thus for all kinds of values. For now, we say an instruction is dead if it would be removed assuming all users are dead. A call site return is different as we just look at the users. If all call site returns have been eliminated, the return values can return undef instead of their original value, eliminating uses. We try to recursively delete dead instructions now and we introduce a simple check interface for use-traversal. This is the idea tried out in D68626 but implemented in the right way. Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68925
2019-10-13 16:08:18 +08:00
call void @useless_arg_sink(i32* %a)
ret void
}
; Check we do not annotate the function interface of this weak function.
; CHECK: define weak_odr void @useless_arg_ext(i32* %a)
define weak_odr void @useless_arg_ext(i32* %a) {
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; MODULE: call void @useless_arg_almost_sink()
; CGSCC: call void @useless_arg_almost_sink(i32* noalias nofree readnone %a)
[Attributor] Liveness for values Summary: This patch introduces liveness (AAIsDead) for all positions, thus for all kinds of values. For now, we say an instruction is dead if it would be removed assuming all users are dead. A call site return is different as we just look at the users. If all call site returns have been eliminated, the return values can return undef instead of their original value, eliminating uses. We try to recursively delete dead instructions now and we introduce a simple check interface for use-traversal. This is the idea tried out in D68626 but implemented in the right way. Reviewers: uenoku, sstefan1 Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68925
2019-10-13 16:08:18 +08:00
call void @useless_arg_almost_sink(i32* %a)
ret void
}
; CHECK: define internal void @useless_arg_ext_int(i32* %a)
define internal void @useless_arg_ext_int(i32* %a) {
; CHECK: call void @useless_arg_ext(i32* %a)
call void @useless_arg_ext(i32* %a)
ret void
}
define void @useless_arg_ext_int_ext(i32* %a) {
; CHECK: call void @useless_arg_ext_int(i32* %a)
call void @useless_arg_ext_int(i32* %a)
ret void
}
[Attributor][FIX] Handle the default case of a switch In D69605 only the "cases" of a switch were handled but if none matched we did not make the default case life. This is fixed now and properly tested (with code from IPConstantProp/user-with-multiple-uses.ll).
2019-11-02 02:57:49 +08:00
[Utils] Allow "on-the-fly" argument changes for update_test_check scripts Update test scripts were limited because they performed a single action on the entire file and if that action was controlled by arguments, like the one introduced in D68819, there was no record of it. This patch introduces the capability of changing the arguments passed to the script "on-the-fly" while processing a test file. In addition, an "on/off" switch was added so that processing can be disabled for parts of the file where the content is simply copied. The last extension is a record of the invocation arguments in the auto generated NOTE. These arguments are also picked up in a subsequent invocation, allowing updates with special options enabled without user interaction. To change the arguments the string `UTC_ARGS:` has to be present in a line, followed by "additional command line arguments". That is everything that follows `UTC_ARGS:` will be added to a growing list of "command line arguments" which is reparsed after every update. Reviewed By: arichardson Differential Revision: https://reviews.llvm.org/D69701
2019-11-01 11:45:17 +08:00
; UTC_ARGS: --enable
[Attributor][FIX] Handle the default case of a switch In D69605 only the "cases" of a switch were handled but if none matched we did not make the default case life. This is fixed now and properly tested (with code from IPConstantProp/user-with-multiple-uses.ll).
2019-11-02 02:57:49 +08:00
; FIXME: We should fold terminators.
define internal i32 @switch_default(i64 %i) nounwind {
[Attributor] Ignore uses if a value is simplified If we have a replacement for a value, via AAValueSimplify, the original value will lose all its uses. Thus, as long as a value is simplified we can skip the uses in checkForAllUses, given that these uses are transitive uses for the simplified version and will therefore affect the simplified version as necessary. Since this allowed us to remove calls without side-effects and a known return value, we need to make sure not to eliminate `musttail` calls. Those we keep around, or later remove the entire `musttail` call chain.
2020-01-24 07:12:56 +08:00
; MODULE-LABEL: define {{[^@]+}}@switch_default()
; MODULE-NEXT: entry:
; MODULE-NEXT: switch i64 0, label [[SW_DEFAULT:%.*]] [
; MODULE-NEXT: i64 3, label [[RETURN:%.*]]
; MODULE-NEXT: i64 10, label [[RETURN]]
; MODULE-NEXT: ]
; MODULE: sw.default:
; MODULE-NEXT: call void @sink()
; MODULE-NEXT: ret i32 undef
; MODULE: return:
; MODULE-NEXT: unreachable
;
; CGSCC-LABEL: define {{[^@]+}}@switch_default
; CGSCC-SAME: (i64 [[I:%.*]])
; CGSCC-NEXT: entry:
; CGSCC-NEXT: switch i64 0, label [[SW_DEFAULT:%.*]] [
; CGSCC-NEXT: i64 3, label [[RETURN:%.*]]
; CGSCC-NEXT: i64 10, label [[RETURN]]
; CGSCC-NEXT: ]
; CGSCC: sw.default:
; CGSCC-NEXT: call void @sink()
; CGSCC-NEXT: ret i32 123
; CGSCC: return:
; CGSCC-NEXT: unreachable
[Attributor][FIX] Handle the default case of a switch In D69605 only the "cases" of a switch were handled but if none matched we did not make the default case life. This is fixed now and properly tested (with code from IPConstantProp/user-with-multiple-uses.ll).
2019-11-02 02:57:49 +08:00
;
entry:
switch i64 %i, label %sw.default [
i64 3, label %return
i64 10, label %return
]
sw.default:
[Attributor][FIX] Remove duplicate, half-broken functionality The changeXXXAfterManifest functions are better suited to deal with changes so we should prefer them. These functions also recursively delete dead instructions which is why we see test changes.
2020-01-12 14:17:08 +08:00
call void @sink()
[Attributor][FIX] Handle the default case of a switch In D69605 only the "cases" of a switch were handled but if none matched we did not make the default case life. This is fixed now and properly tested (with code from IPConstantProp/user-with-multiple-uses.ll).
2019-11-02 02:57:49 +08:00
ret i32 123
return:
ret i32 0
}
define i32 @switch_default_caller() {
[Attributor] Ignore uses if a value is simplified If we have a replacement for a value, via AAValueSimplify, the original value will lose all its uses. Thus, as long as a value is simplified we can skip the uses in checkForAllUses, given that these uses are transitive uses for the simplified version and will therefore affect the simplified version as necessary. Since this allowed us to remove calls without side-effects and a known return value, we need to make sure not to eliminate `musttail` calls. Those we keep around, or later remove the entire `musttail` call chain.
2020-01-24 07:12:56 +08:00
; MODULE-LABEL: define {{[^@]+}}@switch_default_caller()
; MODULE-NEXT: [[CALL2:%.*]] = tail call i32 @switch_default()
; MODULE-NEXT: ret i32 123
;
[Attributor][FIX] Remove duplicate, half-broken functionality The changeXXXAfterManifest functions are better suited to deal with changes so we should prefer them. These functions also recursively delete dead instructions which is why we see test changes.
2020-01-12 14:17:08 +08:00
; CGSCC-LABEL: define {{[^@]+}}@switch_default_caller()
; CGSCC-NEXT: [[CALL2:%.*]] = tail call i32 @switch_default(i64 0)
; CGSCC-NEXT: ret i32 123
[Attributor][FIX] Handle the default case of a switch In D69605 only the "cases" of a switch were handled but if none matched we did not make the default case life. This is fixed now and properly tested (with code from IPConstantProp/user-with-multiple-uses.ll).
2019-11-02 02:57:49 +08:00
;
%call2 = tail call i32 @switch_default(i64 0)
ret i32 %call2
}
[Attributor][FIX] Remove duplicate, half-broken functionality The changeXXXAfterManifest functions are better suited to deal with changes so we should prefer them. These functions also recursively delete dead instructions which is why we see test changes.
2020-01-12 14:17:08 +08:00
define internal i32 @switch_default_dead(i64 %i) nounwind {
entry:
switch i64 %i, label %sw.default [
i64 3, label %return
i64 10, label %return
]
sw.default:
ret i32 123
return:
ret i32 0
}
define i32 @switch_default_dead_caller() {
; CHECK-LABEL: define {{[^@]+}}@switch_default_dead_caller()
; CHECK-NEXT: ret i32 123
;
%call2 = tail call i32 @switch_default_dead(i64 0)
ret i32 %call2
}
[Utils] Allow "on-the-fly" argument changes for update_test_check scripts Update test scripts were limited because they performed a single action on the entire file and if that action was controlled by arguments, like the one introduced in D68819, there was no record of it. This patch introduces the capability of changing the arguments passed to the script "on-the-fly" while processing a test file. In addition, an "on/off" switch was added so that processing can be disabled for parts of the file where the content is simply copied. The last extension is a record of the invocation arguments in the auto generated NOTE. These arguments are also picked up in a subsequent invocation, allowing updates with special options enabled without user interaction. To change the arguments the string `UTC_ARGS:` has to be present in a line, followed by "additional command line arguments". That is everything that follows `UTC_ARGS:` will be added to a growing list of "command line arguments" which is reparsed after every update. Reviewed By: arichardson Differential Revision: https://reviews.llvm.org/D69701
2019-11-01 11:45:17 +08:00
; UTC_ARGS: --disable
[Attributor] Ignore BlockAddress users in call site traversal BlockAddress users will not "call" the function so they do not qualify as call sites in the first place. When we delete a function with BlockAddress users we need to first remove the body so they are properly discarded.
2019-11-02 11:35:18 +08:00
; Allow blockaddress users
[Attributor] Add an Attributor CGSCC pass and run it In addition to the module pass, this patch introduces a CGSCC pass that runs the Attributor on a strongly connected component of the call graph (both old and new PM). The Attributor was always design to be used on a subset of functions which makes this patch mostly mechanical. The one change is that we give up `norecurse` deduction in the module pass in favor of doing it during the CGSCC pass. This makes the interfaces simpler but can be revisited if needed. Reviewed By: hfinkel Differential Revision: https://reviews.llvm.org/D70767
2019-11-27 14:30:12 +08:00
; ALL_BUT_OLD_CGSCCC-NOT @dead_with_blockaddress_users
[Attributor] Ignore BlockAddress users in call site traversal BlockAddress users will not "call" the function so they do not qualify as call sites in the first place. When we delete a function with BlockAddress users we need to first remove the body so they are properly discarded.
2019-11-02 11:35:18 +08:00
define internal void @dead_with_blockaddress_users(i32* nocapture %pc) nounwind readonly {
entry:
br label %indirectgoto
lab0: ; preds = %indirectgoto
%indvar.next = add i32 %indvar, 1 ; <i32> [#uses=1]
br label %indirectgoto
end: ; preds = %indirectgoto
ret void
indirectgoto: ; preds = %lab0, %entry
%indvar = phi i32 [ %indvar.next, %lab0 ], [ 0, %entry ] ; <i32> [#uses=2]
%pc.addr.0 = getelementptr i32, i32* %pc, i32 %indvar ; <i32*> [#uses=1]
%tmp1.pn = load i32, i32* %pc.addr.0 ; <i32> [#uses=1]
%indirect.goto.dest.in = getelementptr inbounds [2 x i8*], [2 x i8*]* @dead_with_blockaddress_users.l, i32 0, i32 %tmp1.pn ; <i8**> [#uses=1]
%indirect.goto.dest = load i8*, i8** %indirect.goto.dest.in ; <i8*> [#uses=1]
indirectbr i8* %indirect.goto.dest, [label %lab0, label %end]
}