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/noalias.ll

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

785 lines
31 KiB
LLVM
Raw Normal View History

[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
; 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=7 -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=7 -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_CGSCC_OPM,NOT_CGSCC_NPM,NOT_TUNIT_OPM,IS__TUNIT____,IS________NPM,IS__TUNIT_NPM
[Attributor] Partially disable three tests to unblock the windows bot The windows bot reported a crash [0] which seems to not happen on other platforms. We disable the old pass manager cgscc runs while this is under investigation. [0] http://lab.llvm.org:8011/builders/llvm-clang-x86_64-expensive-checks-win/builds/23151
2020-04-21 10:11:02 +08:00
; TODO: The old pass manager cgscc run is disabled as it causes a crash on windows which is under investigation: http://lab.llvm.org:8011/builders/llvm-clang-x86_64-expensive-checks-win/builds/23151
[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
; 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] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
; TEST 1 - negative.
; void *G;
; void *foo(){
; void *V = malloc(4);
; G = V;
; return V;
; }
@G = external global i8*
define i8* @foo() {
[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 {{[^@]+}}@foo()
; CHECK-NEXT: [[TMP1:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: store i8* [[TMP1]], i8** @G, align 8
; CHECK-NEXT: ret i8* [[TMP1]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%1 = tail call noalias i8* @malloc(i64 4)
store i8* %1, i8** @G, align 8
ret i8* %1
}
declare noalias i8* @malloc(i64)
; TEST 2
; call noalias function in return instruction.
define i8* @return_noalias(){
[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 {{[^@]+}}@return_noalias()
; CHECK-NEXT: [[TMP1:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: ret i8* [[TMP1]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%1 = tail call noalias i8* @malloc(i64 4)
ret i8* %1
}
[Attributor] Deduce "no-capture" argument attribute Add the no-capture argument attribute deduction to the Attributor fixpoint framework. The new string attributed "no-capture-maybe-returned" is introduced to allow deduction of no-capture through functions that "capture" an argument but only by "returning" it. It is only used by the Attributor for testing. Differential Revision: https://reviews.llvm.org/D59922 llvm-svn: 370817
2019-09-04 04:37:24 +08:00
define void @nocapture(i8* %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 {{[^@]+}}@nocapture
; CHECK-SAME: (i8* nocapture nofree readnone [[A:%.*]])
; CHECK-NEXT: ret void
;
[Attributor] Deduce "no-capture" argument attribute Add the no-capture argument attribute deduction to the Attributor fixpoint framework. The new string attributed "no-capture-maybe-returned" is introduced to allow deduction of no-capture through functions that "capture" an argument but only by "returning" it. It is only used by the Attributor for testing. Differential Revision: https://reviews.llvm.org/D59922 llvm-svn: 370817
2019-09-04 04:37:24 +08:00
ret void
}
define i8* @return_noalias_looks_like_capture(){
[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 {{[^@]+}}@return_noalias_looks_like_capture()
; CHECK-NEXT: [[TMP1:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: ret i8* [[TMP1]]
;
[Attributor] Deduce "no-capture" argument attribute Add the no-capture argument attribute deduction to the Attributor fixpoint framework. The new string attributed "no-capture-maybe-returned" is introduced to allow deduction of no-capture through functions that "capture" an argument but only by "returning" it. It is only used by the Attributor for testing. Differential Revision: https://reviews.llvm.org/D59922 llvm-svn: 370817
2019-09-04 04:37:24 +08:00
%1 = tail call noalias i8* @malloc(i64 4)
call void @nocapture(i8* %1)
ret i8* %1
}
[Attributor] Simple casts preserve no-alias property This is a minimal but important advancement over the existing code. A cast with an operand that is only used in the cast retains the no-alias property of the operand.
2020-01-26 10:21:41 +08:00
define i16* @return_noalias_casted(){
[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 {{[^@]+}}@return_noalias_casted()
; CHECK-NEXT: [[TMP1:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: [[C:%.*]] = bitcast i8* [[TMP1]] to i16*
; CHECK-NEXT: ret i16* [[C]]
;
[Attributor] Simple casts preserve no-alias property This is a minimal but important advancement over the existing code. A cast with an operand that is only used in the cast retains the no-alias property of the operand.
2020-01-26 10:21:41 +08:00
%1 = tail call noalias i8* @malloc(i64 4)
%c = bitcast i8* %1 to i16*
ret i16* %c
}
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
declare i8* @alias()
; TEST 3
define i8* @call_alias(){
[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 {{[^@]+}}@call_alias()
; CHECK-NEXT: [[TMP1:%.*]] = tail call i8* @alias()
; CHECK-NEXT: ret i8* [[TMP1]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%1 = tail call i8* @alias()
ret i8* %1
}
; TEST 4
; void *baz();
; void *foo(int a);
;
; void *bar() {
; foo(0);
; return baz();
; }
;
; void *foo(int a) {
; if (a)
; bar();
; return malloc(4);
; }
define i8* @bar() nounwind uwtable {
[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 {{[^@]+}}@bar()
; CHECK-NEXT: [[TMP1:%.*]] = tail call i8* (...) @baz()
; CHECK-NEXT: ret i8* [[TMP1]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%1 = tail call i8* (...) @baz()
ret i8* %1
}
IR: print value numbers for unnamed function arguments For consistency with normal instructions and clarity when reading IR, it's best to print the %0, %1, ... names of function arguments in definitions. Also modifies the parser to accept IR in that form for obvious reasons. llvm-svn: 367755
2019-08-03 22:28:34 +08:00
define i8* @foo1(i32 %0) nounwind uwtable {
[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 {{[^@]+}}@foo1
; CHECK-SAME: (i32 [[TMP0:%.*]])
; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP0]], 0
; CHECK-NEXT: br i1 [[TMP2]], label [[TMP5:%.*]], label [[TMP3:%.*]]
; CHECK: 3:
; CHECK-NEXT: [[TMP4:%.*]] = tail call i8* (...) @baz()
; CHECK-NEXT: br label [[TMP5]]
; CHECK: 5:
; CHECK-NEXT: [[TMP6:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: ret i8* [[TMP6]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%2 = icmp eq i32 %0, 0
br i1 %2, label %5, label %3
3: ; preds = %1
%4 = tail call i8* (...) @baz()
br label %5
5: ; preds = %1, %3
%6 = tail call noalias i8* @malloc(i64 4)
ret i8* %6
}
declare i8* @baz(...) nounwind uwtable
; TEST 5
; Returning global pointer. Should not be noalias.
define i8** @getter() {
[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 {{[^@]+}}@getter()
; CHECK-NEXT: ret i8** @G
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
ret i8** @G
}
; Returning global pointer. Should not be noalias.
define i8** @calle1(){
[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 {{[^@]+}}@calle1()
; CHECK-NEXT: ret i8** @G
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%1 = call i8** @getter()
ret i8** %1
}
; TEST 6
declare noalias i8* @strdup(i8* nocapture) nounwind
define i8* @test6() nounwind uwtable ssp {
[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 {{[^@]+}}@test6()
; CHECK-NEXT: [[X:%.*]] = alloca [2 x i8], align 1
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i8], [2 x i8]* [[X]], i64 0, i64 0
; CHECK-NEXT: store i8 97, i8* [[ARRAYIDX]], align 1
; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds [2 x i8], [2 x i8]* [[X]], i64 0, i64 1
; CHECK-NEXT: store i8 0, i8* [[ARRAYIDX1]], align 1
; CHECK-NEXT: [[CALL:%.*]] = call noalias i8* @strdup(i8* nonnull dereferenceable(2) [[ARRAYIDX]])
; CHECK-NEXT: ret i8* [[CALL]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%x = alloca [2 x i8], align 1
%arrayidx = getelementptr inbounds [2 x i8], [2 x i8]* %x, i64 0, i64 0
store i8 97, i8* %arrayidx, align 1
%arrayidx1 = getelementptr inbounds [2 x i8], [2 x i8]* %x, i64 0, i64 1
store i8 0, i8* %arrayidx1, align 1
%call = call noalias i8* @strdup(i8* %arrayidx) nounwind
ret i8* %call
}
; TEST 7
define i8* @test7() nounwind {
[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 {{[^@]+}}@test7()
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.*]] = call noalias i8* @malloc(i64 4)
; CHECK-NEXT: [[TOBOOL:%.*]] = icmp eq i8* [[A]], null
; CHECK-NEXT: br i1 [[TOBOOL]], label [[RETURN:%.*]], label [[IF_END:%.*]]
; CHECK: if.end:
; CHECK-NEXT: store i8 7, i8* [[A]]
; CHECK-NEXT: br label [[RETURN]]
; CHECK: return:
; CHECK-NEXT: [[RETVAL_0:%.*]] = phi i8* [ [[A]], [[IF_END]] ], [ null, [[ENTRY:%.*]] ]
; CHECK-NEXT: ret i8* [[RETVAL_0]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
entry:
%A = call noalias i8* @malloc(i64 4) nounwind
%tobool = icmp eq i8* %A, null
br i1 %tobool, label %return, label %if.end
if.end:
store i8 7, i8* %A
br label %return
return:
%retval.0 = phi i8* [ %A, %if.end ], [ null, %entry ]
ret i8* %retval.0
}
; TEST 8
IR: print value numbers for unnamed function arguments For consistency with normal instructions and clarity when reading IR, it's best to print the %0, %1, ... names of function arguments in definitions. Also modifies the parser to accept IR in that form for obvious reasons. llvm-svn: 367755
2019-08-03 22:28:34 +08:00
define i8* @test8(i32* %0) nounwind uwtable {
[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 {{[^@]+}}@test8
; CHECK-SAME: (i32* [[TMP0:%.*]])
; CHECK-NEXT: [[TMP2:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: [[TMP3:%.*]] = icmp ne i32* [[TMP0]], null
; CHECK-NEXT: br i1 [[TMP3]], label [[TMP4:%.*]], label [[TMP5:%.*]]
; CHECK: 4:
; CHECK-NEXT: store i8 10, i8* [[TMP2]]
; CHECK-NEXT: br label [[TMP5]]
; CHECK: 5:
; CHECK-NEXT: ret i8* [[TMP2]]
;
[Attributor] NoAlias on return values. Porting function return value attribute noalias to attributor. This will be followed with a patch for callsite and function argumets. Reviewers: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D63067 llvm-svn: 366728
2019-07-23 03:36:27 +08:00
%2 = tail call noalias i8* @malloc(i64 4)
%3 = icmp ne i32* %0, null
br i1 %3, label %4, label %5
4: ; preds = %1
store i8 10, i8* %2
br label %5
5: ; preds = %1, %4
ret i8* %2
}
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
; TEST 9
; Simple Argument Test
[Attributor] Modularize AANoAliasCallSiteArgument to simplify extensions This patch modularizes the way we check for no-alias call site arguments by putting the existing logic into helper functions. The reasoning was not changed but special cases for readonly/readnone were added.
2020-01-28 12:24:32 +08:00
declare void @use_i8(i8* nocapture)
[Attributor] Function signature rewrite infrastructure As part of the Attributor manifest we want to change the signature of functions. This patch introduces a fairly generic interface to do so. As a first, very simple, use case, we remove unused arguments. A second use case, pointer privatization, will be committed with this patch as well. A lot of the code and ideas are taken from argument promotion and we run all argument promotion tests through this framework as well. Reviewed By: uenoku Differential Revision: https://reviews.llvm.org/D68765
2019-10-10 14:39:16 +08:00
define internal void @test9a(i8* %a, i8* %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
; CHECK-LABEL: define {{[^@]+}}@test9a()
; CHECK-NEXT: call void @use_i8(i8* noalias align 536870912 null)
; CHECK-NEXT: ret void
;
[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 @use_i8(i8* null)
[Attributor] Function signature rewrite infrastructure As part of the Attributor manifest we want to change the signature of functions. This patch introduces a fairly generic interface to do so. As a first, very simple, use case, we remove unused arguments. A second use case, pointer privatization, will be committed with this patch as well. A lot of the code and ideas are taken from argument promotion and we run all argument promotion tests through this framework as well. Reviewed By: uenoku Differential Revision: https://reviews.llvm.org/D68765
2019-10-10 14:39:16 +08:00
ret void
}
define internal void @test9b(i8* %a, i8* %b) {
[Attributor] Modularize AANoAliasCallSiteArgument to simplify extensions This patch modularizes the way we check for no-alias call site arguments by putting the existing logic into helper functions. The reasoning was not changed but special cases for readonly/readnone were added.
2020-01-28 12:24:32 +08:00
; FIXME: %b should be noalias
[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 {{[^@]+}}@test9b
; CHECK-SAME: (i8* noalias nocapture [[A:%.*]], i8* nocapture [[B:%.*]])
; CHECK-NEXT: call void @use_i8(i8* noalias nocapture [[A]])
; CHECK-NEXT: call void @use_i8(i8* nocapture [[B]])
; CHECK-NEXT: ret void
;
[Attributor] Function signature rewrite infrastructure As part of the Attributor manifest we want to change the signature of functions. This patch introduces a fairly generic interface to do so. As a first, very simple, use case, we remove unused arguments. A second use case, pointer privatization, will be committed with this patch as well. A lot of the code and ideas are taken from argument promotion and we run all argument promotion tests through this framework as well. Reviewed By: uenoku Differential Revision: https://reviews.llvm.org/D68765
2019-10-10 14:39:16 +08:00
call void @use_i8(i8* %a)
call void @use_i8(i8* %b)
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
ret void
}
[Attributor] Modularize AANoAliasCallSiteArgument to simplify extensions This patch modularizes the way we check for no-alias call site arguments by putting the existing logic into helper functions. The reasoning was not changed but special cases for readonly/readnone were added.
2020-01-28 12:24:32 +08:00
define internal void @test9c(i8* %a, i8* %b, i8* %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 {{[^@]+}}@test9c
; CHECK-SAME: (i8* noalias nocapture [[A:%.*]], i8* nocapture [[B:%.*]], i8* nocapture [[C:%.*]])
; CHECK-NEXT: call void @use_i8(i8* noalias nocapture [[A]])
; CHECK-NEXT: call void @use_i8(i8* nocapture [[B]])
; CHECK-NEXT: call void @use_i8(i8* nocapture [[C]])
; CHECK-NEXT: ret void
;
[Attributor] Modularize AANoAliasCallSiteArgument to simplify extensions This patch modularizes the way we check for no-alias call site arguments by putting the existing logic into helper functions. The reasoning was not changed but special cases for readonly/readnone were added.
2020-01-28 12:24:32 +08:00
call void @use_i8(i8* %a)
call void @use_i8(i8* %b)
call void @use_i8(i8* %c)
ret void
}
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
define void @test9_helper(i8* %a, i8* %b) {
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_CGSCC_NPM-LABEL: define {{[^@]+}}@test9_helper
; NOT_CGSCC_NPM-SAME: (i8* nocapture [[A:%.*]], i8* nocapture [[B:%.*]])
; NOT_CGSCC_NPM-NEXT: tail call void @test9a()
; NOT_CGSCC_NPM-NEXT: tail call void @test9a()
; NOT_CGSCC_NPM-NEXT: tail call void @test9b(i8* noalias nocapture [[A]], i8* nocapture [[B]])
; NOT_CGSCC_NPM-NEXT: tail call void @test9b(i8* noalias nocapture [[B]], i8* noalias nocapture [[A]])
; NOT_CGSCC_NPM-NEXT: tail call void @test9c(i8* noalias nocapture [[A]], i8* nocapture [[B]], i8* nocapture [[B]])
; NOT_CGSCC_NPM-NEXT: tail call void @test9c(i8* noalias nocapture [[B]], i8* noalias nocapture [[A]], i8* noalias nocapture [[A]])
; NOT_CGSCC_NPM-NEXT: ret void
[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 {{[^@]+}}@test9_helper
; IS__CGSCC____-SAME: (i8* nocapture [[A:%.*]], i8* nocapture [[B:%.*]])
; IS__CGSCC____-NEXT: tail call void @test9a()
; IS__CGSCC____-NEXT: tail call void @test9a()
; IS__CGSCC____-NEXT: tail call void @test9b(i8* noalias [[A]], i8* [[B]])
; IS__CGSCC____-NEXT: tail call void @test9b(i8* noalias [[B]], i8* noalias [[A]])
; IS__CGSCC____-NEXT: tail call void @test9c(i8* noalias [[A]], i8* [[B]], i8* [[B]])
; IS__CGSCC____-NEXT: tail call void @test9c(i8* noalias nocapture [[B]], i8* noalias [[A]], i8* noalias nocapture [[A]])
; IS__CGSCC____-NEXT: ret void
;
[Attributor] Function signature rewrite infrastructure As part of the Attributor manifest we want to change the signature of functions. This patch introduces a fairly generic interface to do so. As a first, very simple, use case, we remove unused arguments. A second use case, pointer privatization, will be committed with this patch as well. A lot of the code and ideas are taken from argument promotion and we run all argument promotion tests through this framework as well. Reviewed By: uenoku Differential Revision: https://reviews.llvm.org/D68765
2019-10-10 14:39:16 +08:00
tail call void @test9a(i8* noalias %a, i8* %b)
tail call void @test9a(i8* noalias %b, i8* noalias %a)
tail call void @test9b(i8* noalias %a, i8* %b)
tail call void @test9b(i8* noalias %b, i8* noalias %a)
[Attributor] Modularize AANoAliasCallSiteArgument to simplify extensions This patch modularizes the way we check for no-alias call site arguments by putting the existing logic into helper functions. The reasoning was not changed but special cases for readonly/readnone were added.
2020-01-28 12:24:32 +08:00
tail call void @test9c(i8* noalias %a, i8* %b, i8* %b)
tail call void @test9c(i8* noalias %b, i8* noalias %a, i8* noalias %a)
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
ret void
}
; TEST 10
; Simple CallSite Test
[Attributor] Implement AANoAliasCallSiteArgument initialization Summary: This patch adds an appropriate `initialize` method for `AANoAliasCallSiteArgument`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66927 llvm-svn: 370456
2019-08-30 18:00:32 +08:00
declare void @test10_helper_1(i8* %a)
define void @test10_helper_2(i8* noalias %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 {{[^@]+}}@test10_helper_2
; CHECK-SAME: (i8* noalias [[A:%.*]])
; CHECK-NEXT: tail call void @test10_helper_1(i8* [[A]])
; CHECK-NEXT: 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
tail call void @test10_helper_1(i8* %a)
[Attributor] Implement AANoAliasCallSiteArgument initialization Summary: This patch adds an appropriate `initialize` method for `AANoAliasCallSiteArgument`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66927 llvm-svn: 370456
2019-08-30 18:00:32 +08:00
ret void
}
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
define void @test10(i8* noalias %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 {{[^@]+}}@test10
; CHECK-SAME: (i8* noalias [[A:%.*]])
; CHECK-NEXT: tail call void @test10_helper_1(i8* [[A]])
; CHECK-NEXT: tail call void @test10_helper_2(i8* noalias [[A]])
; CHECK-NEXT: ret void
;
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
; FIXME: missing noalias
[Attributor] Implement AANoAliasCallSiteArgument initialization Summary: This patch adds an appropriate `initialize` method for `AANoAliasCallSiteArgument`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D66927 llvm-svn: 370456
2019-08-30 18:00:32 +08:00
tail call void @test10_helper_1(i8* %a)
tail call void @test10_helper_2(i8* %a)
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
ret void
}
; TEST 11
; CallSite Test
declare void @test11_helper(i8* %a, i8 *%b)
define void @test11(i8* noalias %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 {{[^@]+}}@test11
; CHECK-SAME: (i8* noalias [[A:%.*]])
; CHECK-NEXT: tail call void @test11_helper(i8* [[A]], i8* [[A]])
; CHECK-NEXT: ret void
;
[Attributor] Deduce "noalias" attribute Summary: This patch adds very basic deduction for noalias. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Tags: LLVM Differential Revision: https://reviews.llvm.org/D66207 llvm-svn: 370295
2019-08-29 13:52:00 +08:00
tail call void @test11_helper(i8* %a, i8* %a)
ret void
}
[Attributor] Implement "noalias" callsite argument deduction Summary: Now, `nocapture` is deduced in Attributor therefore, this patch introduces deduction for `noalias` callsite argument using `nocapture`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67286 llvm-svn: 371590
2019-09-11 15:00:33 +08:00
; TEST 12
; CallSite Argument
declare void @use_nocapture(i8* nocapture)
declare void @use(i8*)
define void @test12_1() {
[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________OPM-LABEL: define {{[^@]+}}@test12_1()
; IS________OPM-NEXT: [[A:%.*]] = alloca i8, align 4
; IS________OPM-NEXT: [[B:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS________OPM-NEXT: tail call void @use_nocapture(i8* noalias nonnull align 4 dereferenceable(1) [[A]])
; IS________OPM-NEXT: tail call void @use_nocapture(i8* noalias nonnull align 4 dereferenceable(1) [[A]])
; IS________OPM-NEXT: tail call void @use_nocapture(i8* noalias [[B]])
; IS________OPM-NEXT: tail call void @use_nocapture(i8* noalias [[B]])
; IS________OPM-NEXT: ret void
;
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_TUNIT_OPM-LABEL: define {{[^@]+}}@test12_1()
; NOT_TUNIT_OPM-NEXT: [[A:%.*]] = alloca i8, align 4
; NOT_TUNIT_OPM-NEXT: [[B:%.*]] = tail call noalias i8* @malloc(i64 4)
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* noalias nonnull align 4 dereferenceable(1) [[A]])
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* noalias nonnull align 4 dereferenceable(1) [[A]])
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* noalias nocapture [[B]])
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* noalias nocapture [[B]])
; NOT_TUNIT_OPM-NEXT: ret void
[Attributor] Implement "noalias" callsite argument deduction Summary: Now, `nocapture` is deduced in Attributor therefore, this patch introduces deduction for `noalias` callsite argument using `nocapture`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67286 llvm-svn: 371590
2019-09-11 15:00:33 +08:00
;
%A = alloca i8, align 4
%B = tail call noalias i8* @malloc(i64 4)
tail call void @use_nocapture(i8* %A)
tail call void @use_nocapture(i8* %A)
tail call void @use_nocapture(i8* %B)
tail call void @use_nocapture(i8* %B)
ret void
}
define void @test12_2(){
[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________OPM-LABEL: define {{[^@]+}}@test12_2()
; IS________OPM-NEXT: [[A:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS________OPM-NEXT: tail call void @use_nocapture(i8* noalias nocapture [[A]])
; IS________OPM-NEXT: tail call void @use_nocapture(i8* noalias nocapture [[A]])
; IS________OPM-NEXT: tail call void @use(i8* [[A]])
; IS________OPM-NEXT: tail call void @use_nocapture(i8* [[A]])
; IS________OPM-NEXT: ret void
[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][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_TUNIT_OPM-LABEL: define {{[^@]+}}@test12_2()
; NOT_TUNIT_OPM-NEXT: [[A:%.*]] = tail call noalias i8* @malloc(i64 4)
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* noalias nocapture [[A]])
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* noalias nocapture [[A]])
; NOT_TUNIT_OPM-NEXT: tail call void @use(i8* [[A]])
; NOT_TUNIT_OPM-NEXT: tail call void @use_nocapture(i8* nocapture [[A]])
; NOT_TUNIT_OPM-NEXT: ret void
[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
;
; FIXME: This should be @use_nocapture(i8* noalias [[A]])
; FIXME: This should be @use_nocapture(i8* noalias nocapture [[A]])
[Attributor] Implement "noalias" callsite argument deduction Summary: Now, `nocapture` is deduced in Attributor therefore, this patch introduces deduction for `noalias` callsite argument using `nocapture`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67286 llvm-svn: 371590
2019-09-11 15:00:33 +08:00
%A = tail call noalias i8* @malloc(i64 4)
tail call void @use_nocapture(i8* %A)
tail call void @use_nocapture(i8* %A)
tail call void @use(i8* %A)
tail call void @use_nocapture(i8* %A)
ret void
}
declare void @two_args(i8* nocapture , i8* nocapture)
define void @test12_3(){
[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________OPM-LABEL: define {{[^@]+}}@test12_3()
; IS________OPM-NEXT: [[A:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS________OPM-NEXT: tail call void @two_args(i8* [[A]], i8* [[A]])
; IS________OPM-NEXT: ret void
;
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_TUNIT_OPM-LABEL: define {{[^@]+}}@test12_3()
; NOT_TUNIT_OPM-NEXT: [[A:%.*]] = tail call noalias i8* @malloc(i64 4)
; NOT_TUNIT_OPM-NEXT: tail call void @two_args(i8* nocapture [[A]], i8* nocapture [[A]])
; NOT_TUNIT_OPM-NEXT: ret void
[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] Implement "noalias" callsite argument deduction Summary: Now, `nocapture` is deduced in Attributor therefore, this patch introduces deduction for `noalias` callsite argument using `nocapture`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67286 llvm-svn: 371590
2019-09-11 15:00:33 +08:00
%A = tail call noalias i8* @malloc(i64 4)
tail call void @two_args(i8* %A, i8* %A)
ret void
}
define void @test12_4(){
[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________OPM-LABEL: define {{[^@]+}}@test12_4()
; IS________OPM-NEXT: [[A:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS________OPM-NEXT: [[B:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS________OPM-NEXT: [[A_0:%.*]] = getelementptr i8, i8* [[A]], i64 0
; IS________OPM-NEXT: [[A_1:%.*]] = getelementptr i8, i8* [[A]], i64 1
; IS________OPM-NEXT: [[B_0:%.*]] = getelementptr i8, i8* [[B]], i64 0
; IS________OPM-NEXT: tail call void @two_args(i8* [[A]], i8* [[B]])
; IS________OPM-NEXT: tail call void @two_args(i8* [[A]], i8* [[A_0]])
; IS________OPM-NEXT: tail call void @two_args(i8* [[A]], i8* [[A_1]])
; IS________OPM-NEXT: tail call void @two_args(i8* [[A_0]], i8* [[B_0]])
; IS________OPM-NEXT: ret void
;
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_TUNIT_OPM-LABEL: define {{[^@]+}}@test12_4()
; NOT_TUNIT_OPM-NEXT: [[A:%.*]] = tail call noalias i8* @malloc(i64 4)
; NOT_TUNIT_OPM-NEXT: [[B:%.*]] = tail call noalias i8* @malloc(i64 4)
; NOT_TUNIT_OPM-NEXT: [[A_0:%.*]] = getelementptr i8, i8* [[A]], i64 0
; NOT_TUNIT_OPM-NEXT: [[A_1:%.*]] = getelementptr i8, i8* [[A]], i64 1
; NOT_TUNIT_OPM-NEXT: [[B_0:%.*]] = getelementptr i8, i8* [[B]], i64 0
; NOT_TUNIT_OPM-NEXT: tail call void @two_args(i8* noalias nocapture [[A]], i8* noalias nocapture [[B]])
; NOT_TUNIT_OPM-NEXT: tail call void @two_args(i8* nocapture [[A]], i8* nocapture [[A_0]])
; NOT_TUNIT_OPM-NEXT: tail call void @two_args(i8* nocapture [[A]], i8* nocapture [[A_1]])
; NOT_TUNIT_OPM-NEXT: tail call void @two_args(i8* nocapture [[A_0]], i8* nocapture [[B_0]])
; NOT_TUNIT_OPM-NEXT: ret void
[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] Implement "noalias" callsite argument deduction Summary: Now, `nocapture` is deduced in Attributor therefore, this patch introduces deduction for `noalias` callsite argument using `nocapture`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67286 llvm-svn: 371590
2019-09-11 15:00:33 +08:00
%A = tail call noalias i8* @malloc(i64 4)
%B = tail call noalias i8* @malloc(i64 4)
%A_0 = getelementptr i8, i8* %A, i64 0
%A_1 = getelementptr i8, i8* %A, i64 1
%B_0 = getelementptr i8, i8* %B, i64 0
tail call void @two_args(i8* %A, i8* %B)
tail call void @two_args(i8* %A, i8* %A_0)
tail call void @two_args(i8* %A, i8* %A_1)
[Attributor] Add "free"-based heap2stack deduction Summary: If there is a unique free of the allocated that has to be reached from the malloc, we can apply the heap-2-stack transformation even if the pointer escapes. Reviewers: hfinkel, sstefan1, uenoku Subscribers: hiraditya, bollu, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D68958
2019-10-15 06:29:05 +08:00
; FIXME: This should be @two_args(i8* noalias nocapture %A_0, i8* noalias nocapture %B_0)
[Attributor] Implement "noalias" callsite argument deduction Summary: Now, `nocapture` is deduced in Attributor therefore, this patch introduces deduction for `noalias` callsite argument using `nocapture`. Reviewers: jdoerfert, sstefan1 Reviewed By: jdoerfert Subscribers: lebedev.ri, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D67286 llvm-svn: 371590
2019-09-11 15:00:33 +08:00
tail call void @two_args(i8* %A_0, i8* %B_0)
ret void
}
[Attributor] Simple casts preserve no-alias property This is a minimal but important advancement over the existing code. A cast with an operand that is only used in the cast retains the no-alias property of the operand.
2020-01-26 10:21:41 +08:00
; TEST 13
define void @use_i8_internal(i8* %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 {{[^@]+}}@use_i8_internal
; CHECK-SAME: (i8* nocapture [[A:%.*]])
; CHECK-NEXT: call void @use_i8(i8* nocapture [[A]])
; CHECK-NEXT: ret void
;
[Attributor] Simple casts preserve no-alias property This is a minimal but important advancement over the existing code. A cast with an operand that is only used in the cast retains the no-alias property of the operand.
2020-01-26 10:21:41 +08:00
call void @use_i8(i8* %a)
ret void
}
define void @test13_use_noalias(){
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; CHECK-LABEL: define {{[^@]+}}@test13_use_noalias()
; CHECK-NEXT: [[M1:%.*]] = tail call noalias i8* @malloc(i64 4)
; CHECK-NEXT: [[C1:%.*]] = bitcast i8* [[M1]] to i16*
; CHECK-NEXT: [[C2:%.*]] = bitcast i16* [[C1]] to i8*
; CHECK-NEXT: call void @use_i8_internal(i8* noalias nocapture [[C2]])
; CHECK-NEXT: ret void
[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 {{[^@]+}}@test13_use_noalias()
; IS__CGSCC_OPM-NEXT: [[M1:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS__CGSCC_OPM-NEXT: [[C1:%.*]] = bitcast i8* [[M1]] to i16*
; IS__CGSCC_OPM-NEXT: [[C2:%.*]] = bitcast i16* [[C1]] to i8*
; IS__CGSCC_OPM-NEXT: call void @use_i8_internal(i8* noalias [[C2]])
; IS__CGSCC_OPM-NEXT: ret void
[Attributor] Simple casts preserve no-alias property This is a minimal but important advancement over the existing code. A cast with an operand that is only used in the cast retains the no-alias property of the operand.
2020-01-26 10:21:41 +08:00
%m1 = tail call noalias i8* @malloc(i64 4)
%c1 = bitcast i8* %m1 to i16*
%c2 = bitcast i16* %c1 to i8*
call void @use_i8_internal(i8* %c2)
ret void
}
define void @test13_use_alias(){
[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________OPM-LABEL: define {{[^@]+}}@test13_use_alias()
; IS________OPM-NEXT: [[M1:%.*]] = tail call noalias i8* @malloc(i64 4)
; IS________OPM-NEXT: [[C1:%.*]] = bitcast i8* [[M1]] to i16*
; IS________OPM-NEXT: [[C2A:%.*]] = bitcast i16* [[C1]] to i8*
; IS________OPM-NEXT: [[C2B:%.*]] = bitcast i16* [[C1]] to i8*
; IS________OPM-NEXT: call void @use_i8_internal(i8* [[C2A]])
; IS________OPM-NEXT: call void @use_i8_internal(i8* [[C2B]])
; IS________OPM-NEXT: ret void
;
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_TUNIT_OPM-LABEL: define {{[^@]+}}@test13_use_alias()
; NOT_TUNIT_OPM-NEXT: [[M1:%.*]] = tail call noalias i8* @malloc(i64 4)
; NOT_TUNIT_OPM-NEXT: [[C1:%.*]] = bitcast i8* [[M1]] to i16*
; NOT_TUNIT_OPM-NEXT: [[C2A:%.*]] = bitcast i16* [[C1]] to i8*
; NOT_TUNIT_OPM-NEXT: [[C2B:%.*]] = bitcast i16* [[C1]] to i8*
; NOT_TUNIT_OPM-NEXT: call void @use_i8_internal(i8* nocapture [[C2A]])
; NOT_TUNIT_OPM-NEXT: call void @use_i8_internal(i8* nocapture [[C2B]])
; NOT_TUNIT_OPM-NEXT: ret void
[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] Simple casts preserve no-alias property This is a minimal but important advancement over the existing code. A cast with an operand that is only used in the cast retains the no-alias property of the operand.
2020-01-26 10:21:41 +08:00
%m1 = tail call noalias i8* @malloc(i64 4)
%c1 = bitcast i8* %m1 to i16*
%c2a = bitcast i16* %c1 to i8*
%c2b = bitcast i16* %c1 to i8*
call void @use_i8_internal(i8* %c2a)
call void @use_i8_internal(i8* %c2b)
ret void
}
[Attributor][FIX] Carefully strip casts in AANoAlias We can strip casts in AANoAlias but that might cause us to end up with a non-pointer type. We do properly handle that case now.
2020-02-15 09:23:21 +08:00
; TEST 14 i2p casts
define internal i32 @p2i(i32* %arg) {
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_CGSCC_NPM-LABEL: define {{[^@]+}}@p2i
; NOT_CGSCC_NPM-SAME: (i32* noalias nofree readnone [[ARG:%.*]])
; NOT_CGSCC_NPM-NEXT: [[P2I:%.*]] = ptrtoint i32* [[ARG]] to i32
; NOT_CGSCC_NPM-NEXT: ret i32 [[P2I]]
[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 {{[^@]+}}@p2i
; IS__CGSCC____-SAME: (i32* nofree readnone [[ARG:%.*]])
; IS__CGSCC____-NEXT: [[P2I:%.*]] = ptrtoint i32* [[ARG]] to i32
; IS__CGSCC____-NEXT: ret i32 [[P2I]]
;
[Attributor][FIX] Carefully strip casts in AANoAlias We can strip casts in AANoAlias but that might cause us to end up with a non-pointer type. We do properly handle that case now.
2020-02-15 09:23:21 +08:00
%p2i = ptrtoint i32* %arg to i32
ret i32 %p2i
}
define i32 @i2p(i32* %arg) {
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_CGSCC_NPM-LABEL: define {{[^@]+}}@i2p
; NOT_CGSCC_NPM-SAME: (i32* nofree readonly [[ARG:%.*]])
; NOT_CGSCC_NPM-NEXT: [[C:%.*]] = call i32 @p2i(i32* noalias nofree readnone [[ARG]])
; NOT_CGSCC_NPM-NEXT: [[I2P:%.*]] = inttoptr i32 [[C]] to i8*
; NOT_CGSCC_NPM-NEXT: [[BC:%.*]] = bitcast i8* [[I2P]] to i32*
; NOT_CGSCC_NPM-NEXT: [[CALL:%.*]] = call i32 @ret(i32* nofree readonly align 4 [[BC]])
; NOT_CGSCC_NPM-NEXT: ret i32 [[CALL]]
[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 {{[^@]+}}@i2p
; IS__CGSCC____-SAME: (i32* nofree readonly [[ARG:%.*]])
; IS__CGSCC____-NEXT: [[C:%.*]] = call i32 @p2i(i32* noalias nofree readnone [[ARG]])
; IS__CGSCC____-NEXT: [[I2P:%.*]] = inttoptr i32 [[C]] to i8*
; IS__CGSCC____-NEXT: [[BC:%.*]] = bitcast i8* [[I2P]] to i32*
; IS__CGSCC____-NEXT: [[CALL:%.*]] = call i32 @ret(i32* nofree nonnull readonly align 4 dereferenceable(4) [[BC]])
; IS__CGSCC____-NEXT: ret i32 [[CALL]]
;
[Attributor][FIX] Carefully strip casts in AANoAlias We can strip casts in AANoAlias but that might cause us to end up with a non-pointer type. We do properly handle that case now.
2020-02-15 09:23:21 +08:00
%c = call i32 @p2i(i32* %arg)
%i2p = inttoptr i32 %c to i8*
%bc = bitcast i8* %i2p to i32*
%call = call i32 @ret(i32* %bc)
ret i32 %call
}
define internal i32 @ret(i32* %arg) {
[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 {{[^@]+}}@ret
; CHECK-SAME: (i32* nocapture nofree nonnull readonly align 4 dereferenceable(4) [[ARG:%.*]])
; CHECK-NEXT: [[L:%.*]] = load i32, i32* [[ARG]], align 4
; CHECK-NEXT: ret i32 [[L]]
;
[Attributor][FIX] Carefully strip casts in AANoAlias We can strip casts in AANoAlias but that might cause us to end up with a non-pointer type. We do properly handle that case now.
2020-02-15 09:23:21 +08:00
%l = load i32, i32* %arg
ret i32 %l
}
[Attributor] Improve noalias preservation using reachability Resolution for below fixme: (ii) Check whether the value is captured in the scope using AANoCapture. FIXME: This is conservative though, it is better to look at CFG and check only uses possibly executed before this callsite. Propagates caller argument's noalias attribute to callee. Reviewed by: jdoerfert, uenoku Reviewers: jdoerfert, sstefan1, uenoku Subscribers: uenoku, sstefan1, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D71617
2020-03-13 23:39:08 +08:00
; Test to propagate noalias where value is assumed to be no-capture in all the
; uses possibly executed before this callsite.
; IR referred from musl/src/strtod.c file
%struct._IO_FILE = type { i32, i8*, i8*, i32 (%struct._IO_FILE*)*, i8*, i8*, i8*, i8*, i32 (%struct._IO_FILE*, i8*, i32)*, i32 (%struct._IO_FILE*, i8*, i32)*, i64 (%struct._IO_FILE*, i64, i32)*, i8*, i32, %struct._IO_FILE*, %struct._IO_FILE*, i32, i32, i32, i16, i8, i8, i32, i32, i8*, i64, i8*, i8*, i8*, [4 x i8], i64, i64, %struct._IO_FILE*, %struct._IO_FILE*, %struct.__locale_struct*, [4 x i8] }
%struct.__locale_struct = type { [6 x %struct.__locale_map*] }
%struct.__locale_map = type opaque
; Function Attrs: nounwind optsize
define internal fastcc double @strtox(i8* %s, i8** %p, i32 %prec) unnamed_addr {
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_CGSCC_NPM-LABEL: define {{[^@]+}}@strtox
; NOT_CGSCC_NPM-SAME: (i8* noalias [[S:%.*]]) unnamed_addr
; NOT_CGSCC_NPM-NEXT: entry:
; NOT_CGSCC_NPM-NEXT: [[F:%.*]] = alloca [[STRUCT__IO_FILE:%.*]], align 8
; NOT_CGSCC_NPM-NEXT: [[TMP0:%.*]] = bitcast %struct._IO_FILE* [[F]] to i8*
; NOT_CGSCC_NPM-NEXT: call void @llvm.lifetime.start.p0i8(i64 144, i8* nonnull align 8 dereferenceable(240) [[TMP0]])
; NOT_CGSCC_NPM-NEXT: [[CALL:%.*]] = call i32 bitcast (i32 (...)* @sh_fromstring to i32 (%struct._IO_FILE*, i8*)*)(%struct._IO_FILE* nonnull align 8 dereferenceable(240) [[F]], i8* [[S]])
; NOT_CGSCC_NPM-NEXT: call void @__shlim(%struct._IO_FILE* nonnull align 8 dereferenceable(240) [[F]], i64 0)
; NOT_CGSCC_NPM-NEXT: [[CALL1:%.*]] = call double @__floatscan(%struct._IO_FILE* nonnull align 8 dereferenceable(240) [[F]], i32 1, i32 1)
; NOT_CGSCC_NPM-NEXT: call void @llvm.lifetime.end.p0i8(i64 144, i8* nonnull align 8 dereferenceable(240) [[TMP0]])
; NOT_CGSCC_NPM-NEXT: ret double [[CALL1]]
[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 {{[^@]+}}@strtox
; IS__CGSCC____-SAME: (i8* [[S:%.*]]) unnamed_addr
; IS__CGSCC____-NEXT: entry:
; IS__CGSCC____-NEXT: [[F:%.*]] = alloca [[STRUCT__IO_FILE:%.*]], align 8
; IS__CGSCC____-NEXT: [[TMP0:%.*]] = bitcast %struct._IO_FILE* [[F]] to i8*
; IS__CGSCC____-NEXT: call void @llvm.lifetime.start.p0i8(i64 144, i8* nonnull align 8 dereferenceable(240) [[TMP0]])
; IS__CGSCC____-NEXT: [[CALL:%.*]] = call i32 bitcast (i32 (...)* @sh_fromstring to i32 (%struct._IO_FILE*, i8*)*)(%struct._IO_FILE* nonnull align 8 dereferenceable(240) [[F]], i8* [[S]])
; IS__CGSCC____-NEXT: call void @__shlim(%struct._IO_FILE* nonnull align 8 dereferenceable(240) [[F]], i64 0)
; IS__CGSCC____-NEXT: [[CALL1:%.*]] = call double @__floatscan(%struct._IO_FILE* nonnull align 8 dereferenceable(240) [[F]], i32 1, i32 1)
; IS__CGSCC____-NEXT: call void @llvm.lifetime.end.p0i8(i64 144, i8* nonnull align 8 dereferenceable(240) [[TMP0]])
; IS__CGSCC____-NEXT: ret double [[CALL1]]
;
[Attributor] Improve noalias preservation using reachability Resolution for below fixme: (ii) Check whether the value is captured in the scope using AANoCapture. FIXME: This is conservative though, it is better to look at CFG and check only uses possibly executed before this callsite. Propagates caller argument's noalias attribute to callee. Reviewed by: jdoerfert, uenoku Reviewers: jdoerfert, sstefan1, uenoku Subscribers: uenoku, sstefan1, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D71617
2020-03-13 23:39:08 +08:00
entry:
%f = alloca %struct._IO_FILE, align 8
%0 = bitcast %struct._IO_FILE* %f to i8*
call void @llvm.lifetime.start.p0i8(i64 144, i8* nonnull %0)
%call = call i32 bitcast (i32 (...)* @sh_fromstring to i32 (%struct._IO_FILE*, i8*)*)(%struct._IO_FILE* nonnull %f, i8* %s)
call void @__shlim(%struct._IO_FILE* nonnull %f, i64 0)
%call1 = call double @__floatscan(%struct._IO_FILE* nonnull %f, i32 %prec, i32 1)
call void @llvm.lifetime.end.p0i8(i64 144, i8* nonnull %0)
ret double %call1
}
; Function Attrs: nounwind optsize
define dso_local double @strtod(i8* noalias %s, i8** noalias %p) {
[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 {{[^@]+}}@strtod
; CHECK-SAME: (i8* noalias [[S:%.*]], i8** noalias nocapture nofree readnone [[P:%.*]])
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CALL:%.*]] = tail call fastcc double @strtox(i8* noalias [[S]])
; CHECK-NEXT: ret double [[CALL]]
;
[Attributor] Improve noalias preservation using reachability Resolution for below fixme: (ii) Check whether the value is captured in the scope using AANoCapture. FIXME: This is conservative though, it is better to look at CFG and check only uses possibly executed before this callsite. Propagates caller argument's noalias attribute to callee. Reviewed by: jdoerfert, uenoku Reviewers: jdoerfert, sstefan1, uenoku Subscribers: uenoku, sstefan1, hiraditya, llvm-commits Differential Revision: https://reviews.llvm.org/D71617
2020-03-13 23:39:08 +08:00
entry:
%call = tail call fastcc double @strtox(i8* %s, i8** %p, i32 1)
ret double %call
}
; Function Attrs: argmemonly nounwind willreturn
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture)
; Function Attrs: optsize
declare dso_local i32 @sh_fromstring(...) local_unnamed_addr
; Function Attrs: optsize
declare dso_local void @__shlim(%struct._IO_FILE*, i64) local_unnamed_addr
; Function Attrs: optsize
declare dso_local double @__floatscan(%struct._IO_FILE*, i32, i32) local_unnamed_addr
; Function Attrs: argmemonly nounwind willreturn
declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture)
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
; Test 15
[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
; propagate noalias to some callsite arguments that there is no possibly reachable capture before it
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
@alias_of_p = external global i32*
define void @make_alias(i32* %p) {
[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 {{[^@]+}}@make_alias
; CHECK-SAME: (i32* nofree writeonly [[P:%.*]])
; CHECK-NEXT: store i32* [[P]], i32** @alias_of_p, align 8
; CHECK-NEXT: ret void
;
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
store i32* %p, i32** @alias_of_p
ret void
}
define void @only_store(i32* %p) {
[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 {{[^@]+}}@only_store
; CHECK-SAME: (i32* nocapture nofree nonnull writeonly align 4 dereferenceable(4) [[P:%.*]])
; CHECK-NEXT: store i32 0, i32* [[P]], align 4
; CHECK-NEXT: ret void
;
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
store i32 0, i32* %p
ret void
}
define void @test15_caller(i32* noalias %p, i32 %c) {
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_CGSCC_NPM-LABEL: define {{[^@]+}}@test15_caller
; NOT_CGSCC_NPM-SAME: (i32* noalias nofree writeonly [[P:%.*]], i32 [[C:%.*]])
; NOT_CGSCC_NPM-NEXT: [[TOBOOL:%.*]] = icmp eq i32 [[C]], 0
; NOT_CGSCC_NPM-NEXT: br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
; NOT_CGSCC_NPM: if.then:
; NOT_CGSCC_NPM-NEXT: tail call void @only_store(i32* noalias nocapture nofree writeonly align 4 [[P]])
; NOT_CGSCC_NPM-NEXT: br label [[IF_END]]
; NOT_CGSCC_NPM: if.end:
; NOT_CGSCC_NPM-NEXT: tail call void @make_alias(i32* nofree writeonly [[P]])
; NOT_CGSCC_NPM-NEXT: ret void
[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 {{[^@]+}}@test15_caller
; IS__CGSCC____-SAME: (i32* noalias nofree writeonly [[P:%.*]], i32 [[C:%.*]])
; IS__CGSCC____-NEXT: [[TOBOOL:%.*]] = icmp eq i32 [[C]], 0
; IS__CGSCC____-NEXT: br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
; IS__CGSCC____: if.then:
; IS__CGSCC____-NEXT: tail call void @only_store(i32* noalias nocapture nofree nonnull writeonly align 4 dereferenceable(4) [[P]])
; IS__CGSCC____-NEXT: br label [[IF_END]]
; IS__CGSCC____: if.end:
; IS__CGSCC____-NEXT: tail call void @make_alias(i32* nofree writeonly [[P]])
; IS__CGSCC____-NEXT: ret void
;
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
%tobool = icmp eq i32 %c, 0
br i1 %tobool, label %if.end, label %if.then
if.then:
tail call void @only_store(i32* %p)
br label %if.end
if.end:
tail call void @make_alias(i32* %p)
ret void
}
; Test 16
;
; __attribute__((noinline)) static void test16_sub(int * restrict p, int c1, int c2) {
; if (c1) {
; only_store(p);
; make_alias(p);
; }
; if (!c2) {
; only_store(p);
; }
; }
; void test16_caller(int * restrict p, int c) {
; test16_sub(p, c, 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
;
; FIXME: this should be tail @only_store(i32* noalias %p)
; when test16_caller is called, c1 always equals to c2. (Note that linkage is internal)
; Therefore, only one of the two conditions of if statementes will be fulfilled.
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
define internal void @test16_sub(i32* noalias %p, i32 %c1, i32 %c2) {
[Attributor][NFC] Strip check lines not used while 3 tests are partially disable The three tests modified by this commit have been partially disabled (one run line is commented out). As a consequence subsequent updates will have weird effects on the check lines. This is a commit to avoid such effects by making the check lines match the three remaining run lines.
2020-04-23 07:06:02 +08:00
; NOT_CGSCC_NPM-LABEL: define {{[^@]+}}@test16_sub
; NOT_CGSCC_NPM-SAME: (i32* noalias nofree writeonly [[P:%.*]], i32 [[C1:%.*]], i32 [[C2:%.*]])
; NOT_CGSCC_NPM-NEXT: [[TOBOOL:%.*]] = icmp eq i32 [[C1]], 0
; NOT_CGSCC_NPM-NEXT: br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
; NOT_CGSCC_NPM: if.then:
; NOT_CGSCC_NPM-NEXT: tail call void @only_store(i32* noalias nocapture nofree writeonly align 4 [[P]])
; NOT_CGSCC_NPM-NEXT: tail call void @make_alias(i32* nofree writeonly align 4 [[P]])
; NOT_CGSCC_NPM-NEXT: br label [[IF_END]]
; NOT_CGSCC_NPM: if.end:
; NOT_CGSCC_NPM-NEXT: [[TOBOOL1:%.*]] = icmp eq i32 [[C2]], 0
; NOT_CGSCC_NPM-NEXT: br i1 [[TOBOOL1]], label [[IF_THEN2:%.*]], label [[IF_END3:%.*]]
; NOT_CGSCC_NPM: if.then2:
; NOT_CGSCC_NPM-NEXT: tail call void @only_store(i32* nofree writeonly align 4 [[P]])
; NOT_CGSCC_NPM-NEXT: br label [[IF_END3]]
; NOT_CGSCC_NPM: if.end3:
; NOT_CGSCC_NPM-NEXT: ret void
[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 {{[^@]+}}@test16_sub
; IS__CGSCC____-SAME: (i32* noalias nofree writeonly [[P:%.*]], i32 [[C1:%.*]], i32 [[C2:%.*]])
; IS__CGSCC____-NEXT: [[TOBOOL:%.*]] = icmp eq i32 [[C1]], 0
; IS__CGSCC____-NEXT: br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
; IS__CGSCC____: if.then:
; IS__CGSCC____-NEXT: tail call void @only_store(i32* noalias nocapture nofree nonnull writeonly align 4 dereferenceable(4) [[P]])
; IS__CGSCC____-NEXT: tail call void @make_alias(i32* nofree nonnull writeonly align 4 dereferenceable(4) [[P]])
; IS__CGSCC____-NEXT: br label [[IF_END]]
; IS__CGSCC____: if.end:
; IS__CGSCC____-NEXT: [[TOBOOL1:%.*]] = icmp eq i32 [[C2]], 0
; IS__CGSCC____-NEXT: br i1 [[TOBOOL1]], label [[IF_THEN2:%.*]], label [[IF_END3:%.*]]
; IS__CGSCC____: if.then2:
; IS__CGSCC____-NEXT: tail call void @only_store(i32* nofree nonnull writeonly align 4 dereferenceable(4) [[P]])
; IS__CGSCC____-NEXT: br label [[IF_END3]]
; IS__CGSCC____: if.end3:
; IS__CGSCC____-NEXT: ret void
;
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
%tobool = icmp eq i32 %c1, 0
br i1 %tobool, label %if.end, label %if.then
if.then:
[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
tail call void @only_store(i32* %p)
tail call void @make_alias(i32* %p)
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
br label %if.end
if.end:
%tobool1 = icmp eq i32 %c2, 0
br i1 %tobool1, label %if.then2, label %if.end3
if.then2:
tail call void @only_store(i32* %p)
br label %if.end3
if.end3:
ret void
}
define void @test16_caller(i32* %p, i32 %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 {{[^@]+}}@test16_caller
; CHECK-SAME: (i32* nofree writeonly [[P:%.*]], i32 [[C:%.*]])
; CHECK-NEXT: tail call void @test16_sub(i32* noalias nofree writeonly [[P]], i32 [[C]], i32 [[C]])
; CHECK-NEXT: ret void
;
[Attributor] AAReachability : use isPotentiallyReachable in isKnownReachable `isKnownReachable` had only interface (always returns true). Changed it to call `isPotentiallyReachable`. This change enables deductions of other Abstract Attributes depending on AAReachability to use reachability information obtained from CFG, and it can make them stronger. Reviewed By: jdoerfert Differential Revision: https://reviews.llvm.org/D76210
2020-04-05 08:15:26 +08:00
tail call void @test16_sub(i32* %p, i32 %c, i32 %c)
ret void
}