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/value-simplify.ll

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

502 lines
16 KiB
LLVM
Raw Normal View History

[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --scrub-attributes
[Attributor][NFC] Remove obsolete option from tests Since D76871 it is sufficient to run `opt -atributor` or `-attributor-cgscc`.
2020-04-18 10:09:16 +08:00
; RUN: opt -attributor -attributor-manifest-internal -attributor-max-iterations-verify -attributor-annotate-decl-cs -attributor-max-iterations=4 -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_CGSCC_NPM,NOT_CGSCC_OPM,NOT_TUNIT_NPM,IS__TUNIT____,IS________OPM,IS__TUNIT_OPM
; RUN: opt -aa-pipeline=basic-aa -passes=attributor -attributor-manifest-internal -attributor-max-iterations-verify -attributor-annotate-decl-cs -attributor-max-iterations=4 -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_CGSCC_OPM,NOT_CGSCC_NPM,NOT_TUNIT_OPM,IS__TUNIT____,IS________NPM,IS__TUNIT_NPM
; RUN: opt -attributor-cgscc -attributor-manifest-internal -attributor-annotate-decl-cs -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_TUNIT_NPM,NOT_TUNIT_OPM,NOT_CGSCC_NPM,IS__CGSCC____,IS________OPM,IS__CGSCC_OPM
; RUN: opt -aa-pipeline=basic-aa -passes=attributor-cgscc -attributor-manifest-internal -attributor-annotate-decl-cs -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_TUNIT_NPM,NOT_TUNIT_OPM,NOT_CGSCC_OPM,IS__CGSCC____,IS________NPM,IS__CGSCC_NPM
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
declare void @f(i32)
Reland [X86] Codegen for preallocated See https://reviews.llvm.org/D74651 for the preallocated IR constructs and LangRef changes. In X86TargetLowering::LowerCall(), if a call is preallocated, record each argument's offset from the stack pointer and the total stack adjustment. Associate the call Value with an integer index. Store the info in X86MachineFunctionInfo with the integer index as the key. This adds two new target independent ISDOpcodes and two new target dependent Opcodes corresponding to @llvm.call.preallocated.{setup,arg}. The setup ISelDAG node takes in a chain and outputs a chain and a SrcValue of the preallocated call Value. It is lowered to a target dependent node with the SrcValue replaced with the integer index key by looking in X86MachineFunctionInfo. In X86TargetLowering::EmitInstrWithCustomInserter() this is lowered to an %esp adjustment, the exact amount determined by looking in X86MachineFunctionInfo with the integer index key. The arg ISelDAG node takes in a chain, a SrcValue of the preallocated call Value, and the arg index int constant. It produces a chain and the pointer fo the arg. It is lowered to a target dependent node with the SrcValue replaced with the integer index key by looking in X86MachineFunctionInfo. In X86TargetLowering::EmitInstrWithCustomInserter() this is lowered to a lea of the stack pointer plus an offset determined by looking in X86MachineFunctionInfo with the integer index key. Force any function containing a preallocated call to use the frame pointer. Does not yet handle a setup without a call, or a conditional call. Does not yet handle musttail. That requires a LangRef change first. Tried to look at all references to inalloca and see if they apply to preallocated. I've made preallocated versions of tests testing inalloca whenever possible and when they make sense (e.g. not alloca related, inalloca edge cases). Aside from the tests added here, I checked that this codegen produces correct code for something like ``` struct A { A(); A(A&&); ~A(); }; void bar() { foo(foo(foo(foo(foo(A(), 4), 5), 6), 7), 8); } ``` by replacing the inalloca version of the .ll file with the appropriate preallocated code. Running the executable produces the same results as using the current inalloca implementation. Reverted due to unexpectedly passing tests, added REQUIRES: asserts for reland. Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D77689
2020-03-17 03:32:36 +08:00
declare token @llvm.call.preallocated.setup(i32)
declare i8* @llvm.call.preallocated.arg(token, i32)
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
; Test1: Replace argument with constant
define internal void @test1(i32 %a) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@test1()
; CHECK-NEXT: tail call void @f(i32 1)
; CHECK-NEXT: ret void
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
tail call void @f(i32 %a)
ret void
}
define void @test1_helper() {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@test1_helper()
; CHECK-NEXT: tail call void @test1()
; CHECK-NEXT: ret void
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
tail call void @test1(i32 1)
ret void
}
; TEST 2 : Simplify return value
define i32 @return0() {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@return0()
; CHECK-NEXT: ret i32 0
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
ret i32 0
}
define i32 @return1() {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@return1()
; CHECK-NEXT: ret i32 1
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
ret i32 1
}
define i32 @test2_1(i1 %c) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@test2_1
; CHECK-SAME: (i1 [[C:%.*]])
; CHECK-NEXT: br i1 [[C]], label [[IF_TRUE:%.*]], label [[IF_FALSE:%.*]]
; CHECK: if.true:
; CHECK-NEXT: [[RET0:%.*]] = add i32 0, 1
; CHECK-NEXT: br label [[END:%.*]]
; CHECK: if.false:
; CHECK-NEXT: br label [[END]]
; CHECK: end:
; CHECK-NEXT: [[RET:%.*]] = phi i32 [ [[RET0]], [[IF_TRUE]] ], [ 1, [[IF_FALSE]] ]
; CHECK-NEXT: ret i32 1
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
br i1 %c, label %if.true, label %if.false
if.true:
%call = tail call i32 @return0()
%ret0 = add i32 %call, 1
br label %end
if.false:
%ret1 = tail call i32 @return1()
br label %end
end:
%ret = phi i32 [ %ret0, %if.true ], [ %ret1, %if.false ]
[Attributor] AAValueConstantRange: Value range analysis using constant range Summary: This patch introduces `AAValueConstantRange`, which answers a possible range for integer value in a specific program point. One of the motivations is propagating existing `range` metadata. (I think we need to change the situation that `range` metadata cannot be put to Argument). The state is a tuple of `ConstantRange` and it is initialized to (known, assumed) = ([-∞, +∞], empty). Currently, AAValueConstantRange is created in `getAssumedConstant` method when `AAValueSimplify` returns `nullptr`(worst state). Supported - BinaryOperator(add, sub, ...) - CmpInst(icmp eq, ...) - !range metadata `AAValueConstantRange` is not intended to extend to polyhedral range value analysis. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: phosek, davezarzycki, baziotis, hiraditya, javed.absar, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71620
2020-01-15 14:25:52 +08:00
ret i32 1
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
}
define i32 @test2_2(i1 %c) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@test2_2
; CHECK-SAME: (i1 [[C:%.*]])
; CHECK-NEXT: ret i32 1
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
%ret = tail call i32 @test2_1(i1 %c)
ret i32 %ret
}
declare void @use(i32)
define void @test3(i1 %c) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@test3
; CHECK-SAME: (i1 [[C:%.*]])
; CHECK-NEXT: br i1 [[C]], label [[IF_TRUE:%.*]], label [[IF_FALSE:%.*]]
; CHECK: if.true:
; CHECK-NEXT: br label [[END:%.*]]
; CHECK: if.false:
; CHECK-NEXT: br label [[END]]
; CHECK: end:
; CHECK-NEXT: [[R:%.*]] = phi i32 [ 1, [[IF_TRUE]] ], [ 1, [[IF_FALSE]] ]
; CHECK-NEXT: tail call void @use(i32 1)
; CHECK-NEXT: ret void
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
br i1 %c, label %if.true, label %if.false
if.true:
br label %end
if.false:
%ret1 = tail call i32 @return1()
br label %end
end:
%r = phi i32 [ 1, %if.true ], [ %ret1, %if.false ]
tail call void @use(i32 %r)
ret void
}
define void @test-select-phi(i1 %c) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-LABEL: define {{[^@]+}}@test-select-phi
; CHECK-SAME: (i1 [[C:%.*]])
; CHECK-NEXT: tail call void @use(i32 1)
; CHECK-NEXT: [[SELECT_NOT_SAME:%.*]] = select i1 [[C]], i32 1, i32 0
; CHECK-NEXT: tail call void @use(i32 [[SELECT_NOT_SAME]])
; CHECK-NEXT: br i1 [[C]], label [[IF_TRUE:%.*]], label [[IF_FALSE:%.*]]
; CHECK: if-true:
; CHECK-NEXT: br label [[END:%.*]]
; CHECK: if-false:
; CHECK-NEXT: br label [[END]]
; CHECK: end:
; CHECK-NEXT: [[PHI_SAME:%.*]] = phi i32 [ 1, [[IF_TRUE]] ], [ 1, [[IF_FALSE]] ]
; CHECK-NEXT: [[PHI_NOT_SAME:%.*]] = phi i32 [ 0, [[IF_TRUE]] ], [ 1, [[IF_FALSE]] ]
; CHECK-NEXT: [[PHI_SAME_PROP:%.*]] = phi i32 [ 1, [[IF_TRUE]] ], [ 1, [[IF_FALSE]] ]
; CHECK-NEXT: [[PHI_SAME_UNDEF:%.*]] = phi i32 [ 1, [[IF_TRUE]] ], [ undef, [[IF_FALSE]] ]
; CHECK-NEXT: [[SELECT_NOT_SAME_UNDEF:%.*]] = select i1 [[C]], i32 [[PHI_NOT_SAME]], i32 undef
; CHECK-NEXT: tail call void @use(i32 1)
; CHECK-NEXT: tail call void @use(i32 [[PHI_NOT_SAME]])
; CHECK-NEXT: tail call void @use(i32 1)
; CHECK-NEXT: tail call void @use(i32 1)
; CHECK-NEXT: tail call void @use(i32 [[SELECT_NOT_SAME_UNDEF]])
; CHECK-NEXT: ret void
;
[Attributor] ValueSimplify Abstract Attribute Summary: This patch introduces initial `AAValueSimplify` which simplifies a value in a context. example - (for function returned) If all the return values are the same and constant, then we can replace callsite returned with the constant. - If an internal function takes the same value(constant) as an argument in the callsite, then we can replace the argument with that constant. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66967 llvm-svn: 371291
2019-09-07 15:03:05 +08:00
%select-same = select i1 %c, i32 1, i32 1
tail call void @use(i32 %select-same)
%select-not-same = select i1 %c, i32 1, i32 0
tail call void @use(i32 %select-not-same)
br i1 %c, label %if-true, label %if-false
if-true:
br label %end
if-false:
br label %end
end:
%phi-same = phi i32 [ 1, %if-true ], [ 1, %if-false ]
%phi-not-same = phi i32 [ 0, %if-true ], [ 1, %if-false ]
%phi-same-prop = phi i32 [ 1, %if-true ], [ %select-same, %if-false ]
%phi-same-undef = phi i32 [ 1, %if-true ], [ undef, %if-false ]
%select-not-same-undef = select i1 %c, i32 %phi-not-same, i32 undef
tail call void @use(i32 %phi-same)
tail call void @use(i32 %phi-not-same)
tail call void @use(i32 %phi-same-prop)
tail call void @use(i32 %phi-same-undef)
tail call void @use(i32 %select-not-same-undef)
ret void
}
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
define i32 @ipccp1(i32 %a) {
; CHECK-LABEL: define {{[^@]+}}@ipccp1
[Attributor] AAValueConstantRange: Value range analysis using constant range Summary: This patch introduces `AAValueConstantRange`, which answers a possible range for integer value in a specific program point. One of the motivations is propagating existing `range` metadata. (I think we need to change the situation that `range` metadata cannot be put to Argument). The state is a tuple of `ConstantRange` and it is initialized to (known, assumed) = ([-∞, +∞], empty). Currently, AAValueConstantRange is created in `getAssumedConstant` method when `AAValueSimplify` returns `nullptr`(worst state). Supported - BinaryOperator(add, sub, ...) - CmpInst(icmp eq, ...) - !range metadata `AAValueConstantRange` is not intended to extend to polyhedral range value analysis. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: phosek, davezarzycki, baziotis, hiraditya, javed.absar, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D71620
2020-01-15 14:25:52 +08:00
; CHECK-SAME: (i32 returned [[A:%.*]])
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
; CHECK-NEXT: br i1 true, label [[T:%.*]], label [[F:%.*]]
; CHECK: t:
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; CHECK-NEXT: ret i32 [[A]]
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
; CHECK: f:
; CHECK-NEXT: unreachable
;
br i1 true, label %t, label %f
t:
ret i32 %a
f:
%r = call i32 @ipccp1(i32 5)
ret i32 %r
}
define internal i1 @ipccp2i(i1 %a) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC____-LABEL: define {{[^@]+}}@ipccp2i()
; IS__CGSCC____-NEXT: br label [[T:%.*]]
; IS__CGSCC____: t:
; IS__CGSCC____-NEXT: ret i1 true
; IS__CGSCC____: f:
; IS__CGSCC____-NEXT: unreachable
;
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
br i1 %a, label %t, label %f
t:
ret i1 %a
f:
%r = call i1 @ipccp2i(i1 false)
ret i1 %r
}
define i1 @ipccp2() {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__TUNIT____-LABEL: define {{[^@]+}}@ipccp2()
; IS__TUNIT____-NEXT: ret i1 true
;
; IS__CGSCC____-LABEL: define {{[^@]+}}@ipccp2()
; IS__CGSCC____-NEXT: [[R:%.*]] = call i1 @ipccp2i()
; IS__CGSCC____-NEXT: ret i1 [[R]]
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
;
%r = call i1 @ipccp2i(i1 true)
ret i1 %r
}
[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
define internal i1 @ipccp2ib(i1 %a) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC____-LABEL: define {{[^@]+}}@ipccp2ib()
; IS__CGSCC____-NEXT: br label [[T:%.*]]
; IS__CGSCC____: t:
; IS__CGSCC____-NEXT: ret i1 true
; IS__CGSCC____: f:
; IS__CGSCC____-NEXT: unreachable
;
[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
br i1 %a, label %t, label %f
t:
ret i1 true
f:
%r = call i1 @ipccp2ib(i1 false)
ret i1 %r
}
define i1 @ipccp2b() {
[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
; CHECK-LABEL: define {{[^@]+}}@ipccp2b()
[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
; CHECK-NEXT: ret i1 true
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
;
[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
%r = call i1 @ipccp2ib(i1 true)
ret i1 %r
}
define internal i32 @ipccp3i(i32 %a) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC____-LABEL: define {{[^@]+}}@ipccp3i()
; IS__CGSCC____-NEXT: br label [[T:%.*]]
; IS__CGSCC____: t:
; IS__CGSCC____-NEXT: ret i32 7
; IS__CGSCC____: f:
; IS__CGSCC____-NEXT: unreachable
;
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
%c = icmp eq i32 %a, 7
br i1 %c, label %t, label %f
t:
ret i32 %a
f:
%r = call i32 @ipccp3i(i32 5)
ret i32 %r
}
define i32 @ipccp3() {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__TUNIT____-LABEL: define {{[^@]+}}@ipccp3()
; IS__TUNIT____-NEXT: ret i32 7
;
; IS__CGSCC____-LABEL: define {{[^@]+}}@ipccp3()
; IS__CGSCC____-NEXT: [[R:%.*]] = call i32 @ipccp3i()
; IS__CGSCC____-NEXT: ret i32 [[R]]
;
[Attributor] Make liveness "edge-based" Summary: If control is transferred to a successor is the key question when it comes to liveness. The new implementation puts that question in the focus and thereby providing a clean way to assume certain CFG edges are dead or instructions will not transfer control. Reviewers: sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69605
2019-10-30 00:47:47 +08:00
%r = call i32 @ipccp3i(i32 7)
ret i32 %r
}
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
; Do not touch complicated arguments (for now)
%struct.X = type { i8* }
define internal i32* @test_inalloca(i32* inalloca %a) {
[Attributor] Use existing helpers to determine IR facts We now use getPointerDereferenceableBytes to determine `nonnull` and `dereferenceable` facts from the IR. We also use getPointerAlignment in AAAlign for the same reason. The latter can interfere with callbacks so we do restrict it to non-function-pointers for now.
2020-05-11 02:16:49 +08:00
; CHECK-LABEL: define {{[^@]+}}@test_inalloca
; CHECK-SAME: (i32* inalloca noalias nofree returned writeonly align 536870912 "no-capture-maybe-returned" [[A:%.*]])
; CHECK-NEXT: ret i32* [[A]]
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
;
ret i32* %a
}
define i32* @complicated_args_inalloca() {
; CHECK-LABEL: define {{[^@]+}}@complicated_args_inalloca()
[Attributor][NFC] Proactively ask for `nocapure` on call site arguments This minimizes test noise later on and is in line with other attributes we derive proactively.
2020-05-04 08:08:30 +08:00
; CHECK-NEXT: [[CALL:%.*]] = call i32* @test_inalloca(i32* noalias nocapture nofree writeonly align 536870912 null)
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
; CHECK-NEXT: ret i32* [[CALL]]
;
%call = call i32* @test_inalloca(i32* null)
ret i32* %call
}
Reland [X86] Codegen for preallocated See https://reviews.llvm.org/D74651 for the preallocated IR constructs and LangRef changes. In X86TargetLowering::LowerCall(), if a call is preallocated, record each argument's offset from the stack pointer and the total stack adjustment. Associate the call Value with an integer index. Store the info in X86MachineFunctionInfo with the integer index as the key. This adds two new target independent ISDOpcodes and two new target dependent Opcodes corresponding to @llvm.call.preallocated.{setup,arg}. The setup ISelDAG node takes in a chain and outputs a chain and a SrcValue of the preallocated call Value. It is lowered to a target dependent node with the SrcValue replaced with the integer index key by looking in X86MachineFunctionInfo. In X86TargetLowering::EmitInstrWithCustomInserter() this is lowered to an %esp adjustment, the exact amount determined by looking in X86MachineFunctionInfo with the integer index key. The arg ISelDAG node takes in a chain, a SrcValue of the preallocated call Value, and the arg index int constant. It produces a chain and the pointer fo the arg. It is lowered to a target dependent node with the SrcValue replaced with the integer index key by looking in X86MachineFunctionInfo. In X86TargetLowering::EmitInstrWithCustomInserter() this is lowered to a lea of the stack pointer plus an offset determined by looking in X86MachineFunctionInfo with the integer index key. Force any function containing a preallocated call to use the frame pointer. Does not yet handle a setup without a call, or a conditional call. Does not yet handle musttail. That requires a LangRef change first. Tried to look at all references to inalloca and see if they apply to preallocated. I've made preallocated versions of tests testing inalloca whenever possible and when they make sense (e.g. not alloca related, inalloca edge cases). Aside from the tests added here, I checked that this codegen produces correct code for something like ``` struct A { A(); A(A&&); ~A(); }; void bar() { foo(foo(foo(foo(foo(A(), 4), 5), 6), 7), 8); } ``` by replacing the inalloca version of the .ll file with the appropriate preallocated code. Running the executable produces the same results as using the current inalloca implementation. Reverted due to unexpectedly passing tests, added REQUIRES: asserts for reland. Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D77689
2020-03-17 03:32:36 +08:00
define internal i32* @test_preallocated(i32* preallocated(i32) %a) {
; CHECK-LABEL: define {{[^@]+}}@test_preallocated
; CHECK-SAME: (i32* noalias nofree returned writeonly preallocated(i32) align 536870912 "no-capture-maybe-returned" [[A:%.*]])
; CHECK-NEXT: ret i32* [[A]]
;
ret i32* %a
}
define i32* @complicated_args_preallocated() {
; CHECK-LABEL: define {{[^@]+}}@complicated_args_preallocated()
; CHECK-NEXT: [[C:%.*]] = call token @llvm.call.preallocated.setup(i32 1)
; CHECK-NEXT: [[CALL:%.*]] = call i32* @test_preallocated(i32* noalias nocapture nofree writeonly preallocated(i32) align 536870912 null)
; CHECK-NEXT: ret i32* [[CALL]]
;
%c = call token @llvm.call.preallocated.setup(i32 1)
%call = call i32* @test_preallocated(i32* preallocated(i32) null) ["preallocated"(token %c)]
ret i32* %call
}
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
define internal void @test_sret(%struct.X* sret %a, %struct.X** %b) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
;
[Attributor] Use existing helpers to determine IR facts We now use getPointerDereferenceableBytes to determine `nonnull` and `dereferenceable` facts from the IR. We also use getPointerAlignment in AAAlign for the same reason. The latter can interfere with callbacks so we do restrict it to non-function-pointers for now.
2020-05-11 02:16:49 +08:00
; CHECK-LABEL: define {{[^@]+}}@test_sret
; CHECK-SAME: (%struct.X* noalias nofree nonnull sret writeonly align 536870912 dereferenceable(8) [[A:%.*]], %struct.X** nocapture nofree nonnull writeonly align 8 dereferenceable(8) [[B:%.*]])
; CHECK-NEXT: store %struct.X* [[A]], %struct.X** [[B]], align 8
; CHECK-NEXT: ret void
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
;
store %struct.X* %a, %struct.X** %b
ret void
}
[Attributor] Derive better alignment for accessed pointers Use DL & ABI information for better alignment deduction, e.g., if a type is accessed and the ABI specifies an alignment requirement for such an access we can use it. This is based on a patch by @lebedev.ri and inspired by getBaseAlign in Loads.cpp. Depends on D76673. Reviewed By: lebedev.ri Differential Revision: https://reviews.llvm.org/D76674
2020-04-02 10:46:03 +08:00
; FIXME: Alignment and dereferenceability are not propagated to the argument
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
define void @complicated_args_sret(%struct.X** %b) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
;
; IS__TUNIT____-LABEL: define {{[^@]+}}@complicated_args_sret
; IS__TUNIT____-SAME: (%struct.X** nocapture nofree writeonly [[B:%.*]])
[Attributor][NFC] Proactively ask for `nocapure` on call site arguments This minimizes test noise later on and is in line with other attributes we derive proactively.
2020-05-04 08:08:30 +08:00
; IS__TUNIT____-NEXT: call void @test_sret(%struct.X* noalias nocapture nofree writeonly align 536870912 null, %struct.X** nocapture nofree writeonly align 8 [[B]])
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__TUNIT____-NEXT: ret void
;
; IS__CGSCC____-LABEL: define {{[^@]+}}@complicated_args_sret
; IS__CGSCC____-SAME: (%struct.X** nocapture nofree nonnull writeonly align 8 dereferenceable(8) [[B:%.*]])
[Attributor] Use existing helpers to determine IR facts We now use getPointerDereferenceableBytes to determine `nonnull` and `dereferenceable` facts from the IR. We also use getPointerAlignment in AAAlign for the same reason. The latter can interfere with callbacks so we do restrict it to non-function-pointers for now.
2020-05-11 02:16:49 +08:00
; IS__CGSCC____-NEXT: call void @test_sret(%struct.X* noalias nocapture nofree nonnull writeonly align 536870912 dereferenceable(8) null, %struct.X** nocapture nofree nonnull writeonly align 8 dereferenceable(8) [[B]])
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC____-NEXT: ret void
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
;
call void @test_sret(%struct.X* null, %struct.X** %b)
ret void
}
define internal %struct.X* @test_nest(%struct.X* nest %a) {
[Attributor] Use existing helpers to determine IR facts We now use getPointerDereferenceableBytes to determine `nonnull` and `dereferenceable` facts from the IR. We also use getPointerAlignment in AAAlign for the same reason. The latter can interfere with callbacks so we do restrict it to non-function-pointers for now.
2020-05-11 02:16:49 +08:00
; CHECK-LABEL: define {{[^@]+}}@test_nest
; CHECK-SAME: (%struct.X* nest noalias nofree readnone returned align 536870912 "no-capture-maybe-returned" [[A:%.*]])
; CHECK-NEXT: ret %struct.X* [[A]]
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
;
ret %struct.X* %a
}
define %struct.X* @complicated_args_nest() {
; CHECK-LABEL: define {{[^@]+}}@complicated_args_nest()
[Attributor][NFC] Proactively ask for `nocapure` on call site arguments This minimizes test noise later on and is in line with other attributes we derive proactively.
2020-05-04 08:08:30 +08:00
; CHECK-NEXT: [[CALL:%.*]] = call %struct.X* @test_nest(%struct.X* noalias nocapture nofree readnone align 536870912 null)
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
; CHECK-NEXT: ret %struct.X* [[CALL]]
;
%call = call %struct.X* @test_nest(%struct.X* null)
ret %struct.X* %call
}
@S = external global %struct.X
define internal void @test_byval(%struct.X* byval %a) {
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC_OPM-LABEL: define {{[^@]+}}@test_byval
; IS__CGSCC_OPM-SAME: (%struct.X* noalias nocapture nofree nonnull writeonly byval align 8 dereferenceable(8) [[A:%.*]])
; IS__CGSCC_OPM-NEXT: [[G0:%.*]] = getelementptr [[STRUCT_X:%.*]], %struct.X* [[A]], i32 0, i32 0
; IS__CGSCC_OPM-NEXT: store i8* null, i8** [[G0]], align 8
; IS__CGSCC_OPM-NEXT: ret void
;
; IS__CGSCC_NPM-LABEL: define {{[^@]+}}@test_byval
; IS__CGSCC_NPM-SAME: (i8* nocapture nofree readnone [[TMP0:%.*]])
AllocaInst should store Align instead of MaybeAlign. Along the lines of D77454 and D79968. Unlike loads and stores, the default alignment is getPrefTypeAlign, to match the existing handling in various places, including SelectionDAG and InstCombine. Differential Revision: https://reviews.llvm.org/D80044
2020-05-16 04:23:14 +08:00
; IS__CGSCC_NPM-NEXT: [[A_PRIV:%.*]] = alloca [[STRUCT_X:%.*]], align 8
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC_NPM-NEXT: [[A_PRIV_CAST:%.*]] = bitcast %struct.X* [[A_PRIV]] to i8**
[Attributor] Inititialize "value attributes" w/ must-be-executed-context info Attributes that only depend on the value (=bit pattern) can be initialized from uses in the must-be-executed-context (MBEC). We did use `AAComposeTwoGenericDeduction` and `AAFromMustBeExecutedContext` before to do this for some positions of these attributes but not for all. This was fairly complicated and also problematic as we did run it in every `updateImpl` call even though we only use known information. The new implementation removes `AAComposeTwoGenericDeduction`* and `AAFromMustBeExecutedContext` in favor of a simple interface `AddInformation::fromMBEContext(...)` which we call from the `initialize` methods of the "value attribute" `Impl` classes, e.g. `AANonNullImpl:initialize`. There can be two types of test changes: 1) Artifacts were we miss some information that was known before a global fixpoint was reached and therefore available in an update but not at the beginning. 2) Deduction for values we did not derive via the MBEC before or which were not found as the `AAFromMustBeExecutedContext::updateImpl` was never invoked. * An improved version of AAComposeTwoGenericDeduction can be found in D78718. Once we find a new use case that implementation will be able to handle "generic" AAs better. --- Single run of the Attributor module and then CGSCC pass (oldPM) for SPASS/clause.c (~10k LLVM-IR loc): Before: ``` calls to allocation functions: 468428 (328952/s) temporary memory allocations: 77480 (54410/s) peak heap memory consumption: 32.71MB peak RSS (including heaptrack overhead): 122.46MB total memory leaked: 269.10KB ``` After: ``` calls to allocation functions: 554720 (351310/s) temporary memory allocations: 101650 (64376/s) peak heap memory consumption: 28.46MB peak RSS (including heaptrack overhead): 116.75MB total memory leaked: 269.10KB ``` Difference: ``` calls to allocation functions: 86292 (556722/s) temporary memory allocations: 24170 (155935/s) peak heap memory consumption: -4.25MB peak RSS (including heaptrack overhead): 0B total memory leaked: 0B ``` Reviewed By: uenoku Differential Revision: https://reviews.llvm.org/D78719
2020-04-23 06:49:58 +08:00
; IS__CGSCC_NPM-NEXT: store i8* [[TMP0]], i8** [[A_PRIV_CAST]], align 8
[Attributor] Unify testing (=updates,prefixes,run configurations,...) When the Attributor was created the test update scripts were not well suited to deal with the challenges of IR attribute checking. This partially improved. Since then we also added three additional configurations that need testing; in total we now have the following four: { TUNIT, CGSCC } x { old pass manager (OPM), new pass manager (NPM) } Finally, the number of developers and tests grew rapidly (partially due to the addition of ArgumentPromotion and IPConstantProp tests), which resulted in tests only being run in some configurations, different prefixes being used, and different "styles" of checks being used. Due to the above reasons I believed we needed to take another look at the test update scripts. While we started to use them, via UTC_ARGS: --enable/disable, the other problems remained. To improve the testing situation for *all* configurations, to simplify future updates to the test, and to help identify subtle effects of future changes, we now use the test update scripts for (almost) all Attributor tests. An exhaustive prefix list minimizes the number of check lines and makes it easy to identify and compare configurations. Tests have been adjusted in the process but we tried to keep their intend unchanged. Reviewed By: sstefan1 Differential Revision: https://reviews.llvm.org/D76588
2020-04-09 08:04:57 +08:00
; IS__CGSCC_NPM-NEXT: [[G0:%.*]] = getelementptr [[STRUCT_X]], %struct.X* [[A_PRIV]], i32 0, i32 0
; IS__CGSCC_NPM-NEXT: store i8* null, i8** [[G0]], align 8
; IS__CGSCC_NPM-NEXT: ret void
;
[Attributor][FIX] Make value simplification aware of "complicated" attributes We cannot simply replace arguments that carry attributes like `nest`, `inalloca`, `sret`, and `byval`. Except for the last one, which we can replace if it is not written, we bail for now.
2019-11-02 02:42:54 +08:00
%g0 = getelementptr %struct.X, %struct.X* %a, i32 0, i32 0
store i8* null, i8** %g0
ret void
}
define void @complicated_args_byval() {
; CHECK-LABEL: define {{[^@]+}}@complicated_args_byval()
; CHECK-NEXT: ret void
;
call void @test_byval(%struct.X* @S)
ret void
}
[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
define internal i8*@test_byval2(%struct.X* byval %a) {
[Attributor] Generalize `getAssumedConstantInt` interface We are often interested in an assumed constant and sometimes it has to be an integer constant. Before we only looked for the latter, now we can ask for either.
2020-02-17 06:45:28 +08:00
; CHECK-LABEL: define {{[^@]+}}@test_byval2()
[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-NEXT: [[G0:%.*]] = getelementptr [[STRUCT_X:%.*]], %struct.X* @S, i32 0, i32 0
; CHECK-NEXT: [[L:%.*]] = load i8*, i8** [[G0]], align 8
; CHECK-NEXT: ret i8* [[L]]
;
%g0 = getelementptr %struct.X, %struct.X* %a, i32 0, i32 0
%l = load i8*, i8** %g0
ret i8* %l
}
define i8* @complicated_args_byval2() {
; CHECK-LABEL: define {{[^@]+}}@complicated_args_byval2()
[Attributor] Generalize `getAssumedConstantInt` interface We are often interested in an assumed constant and sometimes it has to be an integer constant. Before we only looked for the latter, now we can ask for either.
2020-02-17 06:45:28 +08:00
; CHECK-NEXT: [[C:%.*]] = call i8* @test_byval2()
[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-NEXT: ret i8* [[C]]
;
%c = call i8* @test_byval2(%struct.X* @S)
ret i8* %c
}
[Attributor][FIX] Return CHANGED once a pessimistic fixpoint is reached.
2020-01-01 03:25:47 +08:00
define void @fixpoint_changed(i32* %p) {
; CHECK-LABEL: define {{[^@]+}}@fixpoint_changed
; CHECK-SAME: (i32* nocapture nofree writeonly [[P:%.*]])
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_COND:%.*]]
; CHECK: for.cond:
; CHECK-NEXT: [[J_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[SW_EPILOG:%.*]] ]
; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[J_0]], 30
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
; CHECK: for.body:
; CHECK-NEXT: switch i32 [[J_0]], label [[SW_EPILOG]] [
; CHECK-NEXT: i32 1, label [[SW_BB:%.*]]
; CHECK-NEXT: ]
; CHECK: sw.bb:
; CHECK-NEXT: br label [[SW_EPILOG]]
; CHECK: sw.epilog:
; CHECK-NEXT: [[X_0:%.*]] = phi i32 [ 255, [[FOR_BODY]] ], [ 253, [[SW_BB]] ]
StoreInst should store Align, not MaybeAlign This is D77454, except for stores. All the infrastructure work was done for loads, so the remaining changes necessary are relatively small. Differential Revision: https://reviews.llvm.org/D79968
2020-05-15 05:48:10 +08:00
; CHECK-NEXT: store i32 [[X_0]], i32* [[P]], align 4
[Attributor][FIX] Return CHANGED once a pessimistic fixpoint is reached.
2020-01-01 03:25:47 +08:00
; CHECK-NEXT: [[INC]] = add nsw i32 [[J_0]], 1
; CHECK-NEXT: br label [[FOR_COND]]
; CHECK: for.end:
; CHECK-NEXT: ret void
;
entry:
br label %for.cond
for.cond:
%j.0 = phi i32 [ 0, %entry ], [ %inc, %sw.epilog ]
%cmp = icmp slt i32 %j.0, 30
br i1 %cmp, label %for.body, label %for.end
for.body:
switch i32 %j.0, label %sw.epilog [
i32 1, label %sw.bb
]
sw.bb:
br label %sw.epilog
sw.epilog:
%x.0 = phi i32 [ 255, %for.body ], [ 253, %sw.bb ]
store i32 %x.0, i32* %p
%inc = add nsw i32 %j.0, 1
br label %for.cond
for.end:
ret void
}
[Attributor] Manifest simplified (return) values properly If we simplify a function return value we have to modify the return instructions.
2020-02-15 10:11:22 +08:00
; Check we merge undef and a constant properly.
; FIXME fold the addition and return the constant.
define i8 @caller0() {
; CHECK-LABEL: define {{[^@]+}}@caller0()
; CHECK-NEXT: [[C:%.*]] = call i8 @callee()
; CHECK-NEXT: ret i8 [[C]]
;
%c = call i8 @callee(i8 undef)
ret i8 %c
}
define i8 @caller1() {
; CHECK-LABEL: define {{[^@]+}}@caller1()
; CHECK-NEXT: [[C:%.*]] = call i8 @callee()
; CHECK-NEXT: ret i8 [[C]]
;
%c = call i8 @callee(i8 undef)
ret i8 %c
}
define i8 @caller2() {
; CHECK-LABEL: define {{[^@]+}}@caller2()
; CHECK-NEXT: [[C:%.*]] = call i8 @callee()
; CHECK-NEXT: ret i8 [[C]]
;
%c = call i8 @callee(i8 undef)
ret i8 %c
}
define i8 @caller_middle() {
; CHECK-LABEL: define {{[^@]+}}@caller_middle()
; CHECK-NEXT: [[C:%.*]] = call i8 @callee()
; CHECK-NEXT: ret i8 [[C]]
;
%c = call i8 @callee(i8 42)
ret i8 %c
}
define i8 @caller3() {
; CHECK-LABEL: define {{[^@]+}}@caller3()
; CHECK-NEXT: [[C:%.*]] = call i8 @callee()
; CHECK-NEXT: ret i8 [[C]]
;
%c = call i8 @callee(i8 undef)
ret i8 %c
}
define i8 @caller4() {
; CHECK-LABEL: define {{[^@]+}}@caller4()
; CHECK-NEXT: [[C:%.*]] = call i8 @callee()
; CHECK-NEXT: ret i8 [[C]]
;
%c = call i8 @callee(i8 undef)
ret i8 %c
}
define internal i8 @callee(i8 %a) {
; CHECK-LABEL: define {{[^@]+}}@callee()
; CHECK-NEXT: [[C:%.*]] = add i8 42, 7
; CHECK-NEXT: ret i8 [[C]]
;
%c = add i8 %a, 7
ret i8 %c
}