maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/CodeGenCXX/attr-target-mv-member-funcs...

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

221 lines
11 KiB
C++
Raw Normal View History

[OpaquePtrs][Clang] Add -no-opaque-pointers to tests (NFC) This adds -no-opaque-pointers to clang tests whose output will change when opaque pointers are enabled by default. This is intended to be part of the migration approach described in https://discourse.llvm.org/t/enabling-opaque-pointers-by-default/61322/9. The patch has been produced by replacing %clang_cc1 with %clang_cc1 -no-opaque-pointers for tests that fail with opaque pointers enabled. Worth noting that this doesn't cover all tests, there's a remaining ~40 tests not using %clang_cc1 that will need a followup change. Differential Revision: https://reviews.llvm.org/D123115
2022-04-07 18:03:55 +08:00
// RUN: %clang_cc1 -no-opaque-pointers -std=c++11 -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=LINUX
// RUN: %clang_cc1 -no-opaque-pointers -std=c++11 -triple x86_64-windows-pc -emit-llvm %s -o - | FileCheck %s --check-prefix=WINDOWS
Implement Attribute Target MultiVersioning GCC's attribute 'target', in addition to being an optimization hint, also allows function multiversioning. We currently have the former implemented, this is the latter's implementation. This works by enabling functions with the same name/signature to coexist, so that they can all be emitted. Multiversion state is stored in the FunctionDecl itself, and SemaDecl manages the definitions. Note that it ends up having to permit redefinition of functions so that they can all be emitted. Additionally, all versions of the function must be emitted, so this also manages that. Note that this includes some additional rules that GCC does not, since defining something as a MultiVersion function after a usage has been made illegal. The only 'history rewriting' that happens is if a function is emitted before it has been converted to a multiversion'ed function, at which point its name needs to be changed. Function templates and virtual functions are NOT yet supported (not supported in GCC either). Additionally, constructors/destructors are disallowed, but the former is planned. llvm-svn: 322028
2018-01-09 05:34:17 +08:00
struct S {
int __attribute__((target("sse4.2"))) foo(int) { return 0; }
int __attribute__((target("arch=sandybridge"))) foo(int);
int __attribute__((target("arch=ivybridge"))) foo(int) { return 1; }
int __attribute__((target("default"))) foo(int) { return 2; }
S &__attribute__((target("arch=ivybridge"))) operator=(const S &) {
return *this;
}
S &__attribute__((target("default"))) operator=(const S &) {
return *this;
}
};
struct ConvertTo {
__attribute__((target("arch=ivybridge"))) operator S() const {
return S{};
}
__attribute__((target("default"))) operator S() const {
return S{};
}
};
int bar() {
S s;
S s2;
s2 = s;
ConvertTo C;
s2 = static_cast<S>(C);
return s.foo(0);
}
struct S2 {
int __attribute__((target("sse4.2"))) foo(int);
int __attribute__((target("arch=sandybridge"))) foo(int);
int __attribute__((target("arch=ivybridge"))) foo(int);
int __attribute__((target("default"))) foo(int);
};
int bar2() {
S2 s;
return s.foo(0);
}
int __attribute__((target("sse4.2"))) S2::foo(int) { return 0; }
int __attribute__((target("arch=ivybridge"))) S2::foo(int) { return 1; }
int __attribute__((target("default"))) S2::foo(int) { return 2; }
template<typename T>
struct templ {
int __attribute__((target("sse4.2"))) foo(int) { return 0; }
int __attribute__((target("arch=sandybridge"))) foo(int);
int __attribute__((target("arch=ivybridge"))) foo(int) { return 1; }
int __attribute__((target("default"))) foo(int) { return 2; }
};
int templ_use() {
templ<int> a;
templ<double> b;
return a.foo(1) + b.foo(2);
}
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: @_ZN1SaSERKS_.ifunc = weak_odr ifunc %struct.S* (%struct.S*, %struct.S*), %struct.S* (%struct.S*, %struct.S*)* ()* @_ZN1SaSERKS_.resolver
// LINUX: @_ZNK9ConvertTocv1SEv.ifunc = weak_odr ifunc void (%struct.ConvertTo*), void (%struct.ConvertTo*)* ()* @_ZNK9ConvertTocv1SEv.resolver
// LINUX: @_ZN1S3fooEi.ifunc = weak_odr ifunc i32 (%struct.S*, i32), i32 (%struct.S*, i32)* ()* @_ZN1S3fooEi.resolver
// LINUX: @_ZN2S23fooEi.ifunc = weak_odr ifunc i32 (%struct.S2*, i32), i32 (%struct.S2*, i32)* ()* @_ZN2S23fooEi.resolver
Implement Attribute Target MultiVersioning GCC's attribute 'target', in addition to being an optimization hint, also allows function multiversioning. We currently have the former implemented, this is the latter's implementation. This works by enabling functions with the same name/signature to coexist, so that they can all be emitted. Multiversion state is stored in the FunctionDecl itself, and SemaDecl manages the definitions. Note that it ends up having to permit redefinition of functions so that they can all be emitted. Additionally, all versions of the function must be emitted, so this also manages that. Note that this includes some additional rules that GCC does not, since defining something as a MultiVersion function after a usage has been made illegal. The only 'history rewriting' that happens is if a function is emitted before it has been converted to a multiversion'ed function, at which point its name needs to be changed. Function templates and virtual functions are NOT yet supported (not supported in GCC either). Additionally, constructors/destructors are disallowed, but the former is planned. llvm-svn: 322028
2018-01-09 05:34:17 +08:00
// Templates:
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: @_ZN5templIiE3fooEi.ifunc = weak_odr ifunc i32 (%struct.templ*, i32), i32 (%struct.templ*, i32)* ()* @_ZN5templIiE3fooEi.resolver
// LINUX: @_ZN5templIdE3fooEi.ifunc = weak_odr ifunc i32 (%struct.templ.0*, i32), i32 (%struct.templ.0*, i32)* ()* @_ZN5templIdE3fooEi.resolver
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[test] Add {{.*}} to make ELF tests immune to dso_local/dso_preemptable/(none) differences For a default visibility external linkage definition, dso_local is set for ELF -fno-pic/-fpie and COFF and Mach-O. Since default clang -cc1 for ELF is similar to -fpic ("PIC Level" is not set), this nuance causes unneeded binary format differences. To make emitted IR similar, ELF -cc1 -fpic will default to -fno-semantic-interposition, which sets dso_local for default visibility external linkage definitions. To make this flip smooth and enable future (dso_local as definition default), this patch replaces (function) `define ` with `define{{.*}} `, (variable/constant/alias) `= ` with `={{.*}} `, or inserts appropriate `{{.*}} `.
2020-12-31 16:27:11 +08:00
// LINUX: define{{.*}} i32 @_Z3barv()
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: %s = alloca %struct.S, align 1
// LINUX: %s2 = alloca %struct.S, align 1
// LINUX: %C = alloca %struct.ConvertTo, align 1
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: call noundef nonnull align 1 dereferenceable(1) %struct.S* @_ZN1SaSERKS_.ifunc(%struct.S* {{[^,]*}} %s2
[CodeGen] Apply 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments. * Adds 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments * Gates 'nonnull' on -f(no-)delete-null-pointer-checks * Introduces this-nonnull.cpp and microsoft-abi-this-nullable.cpp tests to explicitly test the behavior of this change * Refactors hundreds of over-constrained clang tests to permit these attributes, where needed * Updates Clang12 patch notes mentioning this change Reviewed-by: rsmith, jdoerfert Differential Revision: https://reviews.llvm.org/D17993
2020-11-17 07:04:55 +08:00
// LINUX: call void @_ZNK9ConvertTocv1SEv.ifunc(%struct.ConvertTo* {{[^,]*}} %C)
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: call noundef nonnull align 1 dereferenceable(1) %struct.S* @_ZN1SaSERKS_.ifunc(%struct.S* {{[^,]*}} %s2
// LINUX: call noundef i32 @_ZN1S3fooEi.ifunc(%struct.S* {{[^,]*}} %s, i32 noundef 0)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define dso_local noundef i32 @"?bar@@YAHXZ"()
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: %s = alloca %struct.S, align 1
// WINDOWS: %s2 = alloca %struct.S, align 1
// WINDOWS: %C = alloca %struct.ConvertTo, align 1
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: call noundef nonnull align 1 dereferenceable(1) %struct.S* @"??4S@@QEAAAEAU0@AEBU0@@Z.resolver"(%struct.S* {{[^,]*}} %s2
[CodeGen] Apply 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments. * Adds 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments * Gates 'nonnull' on -f(no-)delete-null-pointer-checks * Introduces this-nonnull.cpp and microsoft-abi-this-nullable.cpp tests to explicitly test the behavior of this change * Refactors hundreds of over-constrained clang tests to permit these attributes, where needed * Updates Clang12 patch notes mentioning this change Reviewed-by: rsmith, jdoerfert Differential Revision: https://reviews.llvm.org/D17993
2020-11-17 07:04:55 +08:00
// WINDOWS: call void @"??BConvertTo@@QEBA?AUS@@XZ.resolver"(%struct.ConvertTo* {{[^,]*}} %C
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: call noundef nonnull align 1 dereferenceable(1) %struct.S* @"??4S@@QEAAAEAU0@AEBU0@@Z.resolver"(%struct.S* {{[^,]*}} %s2
// WINDOWS: call noundef i32 @"?foo@S@@QEAAHH@Z.resolver"(%struct.S* {{[^,]*}} %s, i32 noundef 0)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: define weak_odr %struct.S* (%struct.S*, %struct.S*)* @_ZN1SaSERKS_.resolver() comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: ret %struct.S* (%struct.S*, %struct.S*)* @_ZN1SaSERKS_.arch_ivybridge
// LINUX: ret %struct.S* (%struct.S*, %struct.S*)* @_ZN1SaSERKS_
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// WINDOWS: define weak_odr dso_local %struct.S* @"??4S@@QEAAAEAU0@AEBU0@@Z.resolver"(%struct.S* %0, %struct.S* %1)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: call %struct.S* @"??4S@@QEAAAEAU0@AEBU0@@Z.arch_ivybridge"
// WINDOWS: call %struct.S* @"??4S@@QEAAAEAU0@AEBU0@@Z"
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: define weak_odr void (%struct.ConvertTo*)* @_ZNK9ConvertTocv1SEv.resolver() comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: ret void (%struct.ConvertTo*)* @_ZNK9ConvertTocv1SEv.arch_ivybridge
// LINUX: ret void (%struct.ConvertTo*)* @_ZNK9ConvertTocv1SEv
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// WINDOWS: define weak_odr dso_local void @"??BConvertTo@@QEBA?AUS@@XZ.resolver"(%struct.ConvertTo* %0, %struct.S* %1)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: call void @"??BConvertTo@@QEBA?AUS@@XZ.arch_ivybridge"
// WINDOWS: call void @"??BConvertTo@@QEBA?AUS@@XZ"
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: define weak_odr i32 (%struct.S*, i32)* @_ZN1S3fooEi.resolver() comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: ret i32 (%struct.S*, i32)* @_ZN1S3fooEi.arch_sandybridge
// LINUX: ret i32 (%struct.S*, i32)* @_ZN1S3fooEi.arch_ivybridge
// LINUX: ret i32 (%struct.S*, i32)* @_ZN1S3fooEi.sse4.2
// LINUX: ret i32 (%struct.S*, i32)* @_ZN1S3fooEi
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// WINDOWS: define weak_odr dso_local i32 @"?foo@S@@QEAAHH@Z.resolver"(%struct.S* %0, i32 %1)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: call i32 @"?foo@S@@QEAAHH@Z.arch_sandybridge"
// WINDOWS: call i32 @"?foo@S@@QEAAHH@Z.arch_ivybridge"
// WINDOWS: call i32 @"?foo@S@@QEAAHH@Z.sse4.2"
// WINDOWS: call i32 @"?foo@S@@QEAAHH@Z"
[test] Add {{.*}} to make ELF tests immune to dso_local/dso_preemptable/(none) differences For a default visibility external linkage definition, dso_local is set for ELF -fno-pic/-fpie and COFF and Mach-O. Since default clang -cc1 for ELF is similar to -fpic ("PIC Level" is not set), this nuance causes unneeded binary format differences. To make emitted IR similar, ELF -cc1 -fpic will default to -fno-semantic-interposition, which sets dso_local for default visibility external linkage definitions. To make this flip smooth and enable future (dso_local as definition default), this patch replaces (function) `define ` with `define{{.*}} `, (variable/constant/alias) `= ` with `={{.*}} `, or inserts appropriate `{{.*}} `.
2020-12-31 16:27:11 +08:00
// LINUX: define{{.*}} i32 @_Z4bar2v()
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: call noundef i32 @_ZN2S23fooEi.ifunc
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define dso_local noundef i32 @"?bar2@@YAHXZ"()
// WINDOWS: call noundef i32 @"?foo@S2@@QEAAHH@Z.resolver"
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: define weak_odr i32 (%struct.S2*, i32)* @_ZN2S23fooEi.resolver() comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: ret i32 (%struct.S2*, i32)* @_ZN2S23fooEi.arch_sandybridge
// LINUX: ret i32 (%struct.S2*, i32)* @_ZN2S23fooEi.arch_ivybridge
// LINUX: ret i32 (%struct.S2*, i32)* @_ZN2S23fooEi.sse4.2
// LINUX: ret i32 (%struct.S2*, i32)* @_ZN2S23fooEi
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// WINDOWS: define weak_odr dso_local i32 @"?foo@S2@@QEAAHH@Z.resolver"(%struct.S2* %0, i32 %1)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: call i32 @"?foo@S2@@QEAAHH@Z.arch_sandybridge"
// WINDOWS: call i32 @"?foo@S2@@QEAAHH@Z.arch_ivybridge"
// WINDOWS: call i32 @"?foo@S2@@QEAAHH@Z.sse4.2"
// WINDOWS: call i32 @"?foo@S2@@QEAAHH@Z"
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: define{{.*}} i32 @_ZN2S23fooEi.sse4.2(%struct.S2* {{[^,]*}} %this, i32 noundef %0)
// LINUX: define{{.*}} i32 @_ZN2S23fooEi.arch_ivybridge(%struct.S2* {{[^,]*}} %this, i32 noundef %0)
// LINUX: define{{.*}} i32 @_ZN2S23fooEi(%struct.S2* {{[^,]*}} %this, i32 noundef %0)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define dso_local noundef i32 @"?foo@S2@@QEAAHH@Z.sse4.2"(%struct.S2* {{[^,]*}} %this, i32 noundef %0)
// WINDOWS: define dso_local noundef i32 @"?foo@S2@@QEAAHH@Z.arch_ivybridge"(%struct.S2* {{[^,]*}} %this, i32 noundef %0)
// WINDOWS: define dso_local noundef i32 @"?foo@S2@@QEAAHH@Z"(%struct.S2* {{[^,]*}} %this, i32 noundef %0)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[test] Add {{.*}} to make ELF tests immune to dso_local/dso_preemptable/(none) differences For a default visibility external linkage definition, dso_local is set for ELF -fno-pic/-fpie and COFF and Mach-O. Since default clang -cc1 for ELF is similar to -fpic ("PIC Level" is not set), this nuance causes unneeded binary format differences. To make emitted IR similar, ELF -cc1 -fpic will default to -fno-semantic-interposition, which sets dso_local for default visibility external linkage definitions. To make this flip smooth and enable future (dso_local as definition default), this patch replaces (function) `define ` with `define{{.*}} `, (variable/constant/alias) `= ` with `={{.*}} `, or inserts appropriate `{{.*}} `.
2020-12-31 16:27:11 +08:00
// LINUX: define{{.*}} i32 @_Z9templ_usev()
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: call noundef i32 @_ZN5templIiE3fooEi.ifunc
// LINUX: call noundef i32 @_ZN5templIdE3fooEi.ifunc
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define dso_local noundef i32 @"?templ_use@@YAHXZ"()
// WINDOWS: call noundef i32 @"?foo@?$templ@H@@QEAAHH@Z.resolver"
// WINDOWS: call noundef i32 @"?foo@?$templ@N@@QEAAHH@Z.resolver"
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: define weak_odr i32 (%struct.templ*, i32)* @_ZN5templIiE3fooEi.resolver() comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: ret i32 (%struct.templ*, i32)* @_ZN5templIiE3fooEi.arch_sandybridge
// LINUX: ret i32 (%struct.templ*, i32)* @_ZN5templIiE3fooEi.arch_ivybridge
// LINUX: ret i32 (%struct.templ*, i32)* @_ZN5templIiE3fooEi.sse4.2
// LINUX: ret i32 (%struct.templ*, i32)* @_ZN5templIiE3fooEi
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// WINDOWS: define weak_odr dso_local i32 @"?foo@?$templ@H@@QEAAHH@Z.resolver"(%struct.templ* %0, i32 %1)
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: call i32 @"?foo@?$templ@H@@QEAAHH@Z.arch_sandybridge"
// WINDOWS: call i32 @"?foo@?$templ@H@@QEAAHH@Z.arch_ivybridge"
// WINDOWS: call i32 @"?foo@?$templ@H@@QEAAHH@Z.sse4.2"
// WINDOWS: call i32 @"?foo@?$templ@H@@QEAAHH@Z"
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// LINUX: define weak_odr i32 (%struct.templ.0*, i32)* @_ZN5templIdE3fooEi.resolver() comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// LINUX: ret i32 (%struct.templ.0*, i32)* @_ZN5templIdE3fooEi.arch_sandybridge
// LINUX: ret i32 (%struct.templ.0*, i32)* @_ZN5templIdE3fooEi.arch_ivybridge
// LINUX: ret i32 (%struct.templ.0*, i32)* @_ZN5templIdE3fooEi.sse4.2
// LINUX: ret i32 (%struct.templ.0*, i32)* @_ZN5templIdE3fooEi
[CodeGen] Add alias for cpu_dispatch function with IFunc & Fix resolver linkage type Multi-versioned functions defined by cpu_dispatch and implemented with IFunc can not be called outside the translation units where they are defined due to lack of symbols. This patch add function aliases for these functions and thus make them visible outside. Differential Revision: https://reviews.llvm.org/D67058 Patch by Senran Zhang llvm-svn: 371586
2019-09-11 09:54:48 +08:00
// WINDOWS: define weak_odr dso_local i32 @"?foo@?$templ@N@@QEAAHH@Z.resolver"(%struct.templ.0* %0, i32 %1) comdat
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
// WINDOWS: call i32 @"?foo@?$templ@N@@QEAAHH@Z.arch_sandybridge"
// WINDOWS: call i32 @"?foo@?$templ@N@@QEAAHH@Z.arch_ivybridge"
// WINDOWS: call i32 @"?foo@?$templ@N@@QEAAHH@Z.sse4.2"
// WINDOWS: call i32 @"?foo@?$templ@N@@QEAAHH@Z"
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: define linkonce_odr noundef i32 @_ZN1S3fooEi.sse4.2(%struct.S* {{[^,]*}} %this, i32 noundef %0)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
// LINUX: ret i32 0
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@S@@QEAAHH@Z.sse4.2"(%struct.S* {{[^,]*}} %this, i32 noundef %0)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
// WINDOWS: ret i32 0
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: declare noundef i32 @_ZN1S3fooEi.arch_sandybridge(%struct.S* {{[^,]*}}, i32 noundef)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: declare dso_local noundef i32 @"?foo@S@@QEAAHH@Z.arch_sandybridge"(%struct.S* {{[^,]*}}, i32 noundef)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: define linkonce_odr noundef i32 @_ZN1S3fooEi.arch_ivybridge(%struct.S* {{[^,]*}} %this, i32 noundef %0)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
// LINUX: ret i32 1
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@S@@QEAAHH@Z.arch_ivybridge"(%struct.S* {{[^,]*}} %this, i32 noundef %0)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
// WINDOWS: ret i32 1
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: define linkonce_odr noundef i32 @_ZN1S3fooEi(%struct.S* {{[^,]*}} %this, i32 noundef %0)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
// LINUX: ret i32 2
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@S@@QEAAHH@Z"(%struct.S* {{[^,]*}} %this, i32 noundef %0)
Revert "Multiversioning- Ensure all MV functions are emitted." This reverts commit 65df29f9318ac13a633c0ce13b2b0bccf06e79ca. AS suggested by @rsmith here: https://reviews.llvm.org/rL345839 I'm reverting this and solving the initial problem in a different way. llvm-svn: 348595
2018-12-07 22:56:50 +08:00
// WINDOWS: ret i32 2
[Clang/Test]: Rename enable_noundef_analysis to disable-noundef-analysis and turn it off by default Turning on `enable_noundef_analysis` flag allows better codegen by removing freeze instructions. I modified clang by renaming `enable_noundef_analysis` flag to `disable-noundef-analysis` and turning it off by default. Test updates are made as a separate patch: D108453 Reviewed By: eugenis Differential Revision: https://reviews.llvm.org/D105169
2022-01-16 17:53:11 +08:00
// LINUX: define linkonce_odr noundef i32 @_ZN5templIiE3fooEi.sse4.2
// LINUX: declare noundef i32 @_ZN5templIiE3fooEi.arch_sandybridge
// LINUX: define linkonce_odr noundef i32 @_ZN5templIiE3fooEi.arch_ivybridge
// LINUX: define linkonce_odr noundef i32 @_ZN5templIiE3fooEi
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@?$templ@H@@QEAAHH@Z.sse4.2"
// WINDOWS: declare dso_local noundef i32 @"?foo@?$templ@H@@QEAAHH@Z.arch_sandybridge"
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@?$templ@H@@QEAAHH@Z.arch_ivybridge"
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@?$templ@H@@QEAAHH@Z"
// LINUX: define linkonce_odr noundef i32 @_ZN5templIdE3fooEi.sse4.2
// LINUX: declare noundef i32 @_ZN5templIdE3fooEi.arch_sandybridge
// LINUX: define linkonce_odr noundef i32 @_ZN5templIdE3fooEi.arch_ivybridge
// LINUX: define linkonce_odr noundef i32 @_ZN5templIdE3fooEi
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@?$templ@N@@QEAAHH@Z.sse4.2"
// WINDOWS: declare dso_local noundef i32 @"?foo@?$templ@N@@QEAAHH@Z.arch_sandybridge"
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@?$templ@N@@QEAAHH@Z.arch_ivybridge"
// WINDOWS: define linkonce_odr dso_local noundef i32 @"?foo@?$templ@N@@QEAAHH@Z"