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

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

336 lines
14 KiB
C
Raw Normal View History

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
// RUN: %clang_cc1 -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=LINUX
// RUN: %clang_cc1 -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
int __attribute__((target("sse4.2"))) foo(void) { return 0; }
int __attribute__((target("arch=sandybridge"))) foo(void);
int __attribute__((target("arch=ivybridge"))) foo(void) {return 1;}
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
int __attribute__((target("arch=goldmont"))) foo(void) {return 3;}
int __attribute__((target("arch=goldmont-plus"))) foo(void) {return 4;}
int __attribute__((target("arch=tremont"))) foo(void) {return 5;}
int __attribute__((target("arch=icelake-client"))) foo(void) {return 6;}
int __attribute__((target("arch=icelake-server"))) foo(void) {return 7;}
[X86] Assign a feature priority to 'tigerlake' so it won't assert when used with function multiversioning Also test cooperlake since it was also just added to function multiversioning when it was enabled for __builtin_cpu_is.
2020-06-22 04:24:43 +08:00
int __attribute__((target("arch=cooperlake"))) foo(void) {return 8;}
int __attribute__((target("arch=tigerlake"))) foo(void) {return 9;}
[X86] Support -march=sapphirerapids Support -march=sapphirerapids for x86. Compare with Icelake Server, it includes 14 more new features. They are amxtile, amxint8, amxbf16, avx512bf16, avx512vp2intersect, cldemote, enqcmd, movdir64b, movdiri, ptwrite, serialize, shstk, tsxldtrk, waitpkg. Reviewed By: craig.topper Differential Revision: https://reviews.llvm.org/D86503
2020-08-25 12:27:02 +08:00
int __attribute__((target("arch=sapphirerapids"))) foo(void) {return 10;}
[X86] Add a stub for Intel's alderlake. No scheduling, no autodetection.
2020-10-25 00:00:33 +08:00
int __attribute__((target("arch=alderlake"))) foo(void) {return 11;}
[X86] Support -march=rocketlake Reviewed By: skan, craig.topper, MaskRay Differential Revision: https://reviews.llvm.org/D100085
2021-04-13 09:24:34 +08:00
int __attribute__((target("arch=rocketlake"))) foo(void) {return 12;}
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
int __attribute__((target("default"))) foo(void) { return 2; }
int bar() {
return foo();
}
inline int __attribute__((target("sse4.2"))) foo_inline(void) { return 0; }
inline int __attribute__((target("arch=sandybridge"))) foo_inline(void);
inline int __attribute__((target("arch=ivybridge"))) foo_inline(void) {return 1;}
inline int __attribute__((target("default"))) foo_inline(void) { return 2; }
int bar2() {
return foo_inline();
}
inline __attribute__((target("default"))) void foo_decls(void);
inline __attribute__((target("sse4.2"))) void foo_decls(void);
void bar3() {
foo_decls();
}
inline __attribute__((target("default"))) void foo_decls(void) {}
inline __attribute__((target("sse4.2"))) void foo_decls(void) {}
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
inline __attribute__((target("default"))) void foo_multi(int i, double d) {}
inline __attribute__((target("avx,sse4.2"))) void foo_multi(int i, double d) {}
inline __attribute__((target("sse4.2,fma4"))) void foo_multi(int i, double d) {}
inline __attribute__((target("arch=ivybridge,fma4,sse4.2"))) void foo_multi(int i, double d) {}
[CodeGen] Fix a crash on mangling multiversioned functions `multiVersionSortPriority` expects features to have no prefix. We currently carry them around in the format "+${feature}". llvm-svn: 322618
2018-01-17 12:46:04 +08:00
void bar4() {
Implement Function Multiversioning for Non-ELF Systems. Similar to how ICC handles CPU-Dispatch on Windows, this patch uses the resolver function directly to forward the call to the proper function. This is not nearly as efficient as IFuncs of course, but is still quite useful for large functions specifically developed for certain processors. This is unfortunately still limited to x86, since it depends on __builtin_cpu_supports and __builtin_cpu_is, which are x86 builtins. The naming for the resolver/forwarding function for cpu-dispatch was taken from ICC's implementation, which uses the unmodified name for this (no mangling additions). This is possible, since cpu-dispatch uses '.A' for the 'default' version. In 'target' multiversioning, this function keeps the '.resolver' extension in order to keep the default function keeping the default mangling. Change-Id: I4731555a39be26c7ad59a2d8fda6fa1a50f73284 Differential Revision: https://reviews.llvm.org/D53586 llvm-svn: 345298
2018-10-26 02:57:19 +08:00
foo_multi(1, 5.0);
[CodeGen] Fix a crash on mangling multiversioned functions `multiVersionSortPriority` expects features to have no prefix. We currently carry them around in the format "+${feature}". llvm-svn: 322618
2018-01-17 12:46:04 +08:00
}
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
int fwd_decl_default(void);
int __attribute__((target("default"))) fwd_decl_default(void) { return 2; }
int fwd_decl_avx(void);
int __attribute__((target("avx"))) fwd_decl_avx(void) { return 2; }
int __attribute__((target("default"))) fwd_decl_avx(void) { return 2; }
void bar5() {
fwd_decl_default();
fwd_decl_avx();
}
Fix Target Multiversioning renaming. The initial implementation only did 'first declaration renaming' when a default version came after. This is insufficient in cases where a default does not exist, so this patch makes sure that we do the renaming in all cases. This renaming is necessary because we emit the first declaration before knowing that it IS a target multiversion function, which would change its name. The second declaration (the one that caused the multiversioning) then needs to make sure that the first one has its name changed to be consistent with the resolver usage.
2020-03-09 23:06:58 +08:00
int __attribute__((target("avx"))) changed_to_mv(void) { return 0;}
int __attribute__((target("fma4"))) changed_to_mv(void) { return 1;}
Permit attribute 'used' with 'target' multiversioning. This adds infrastructure for a multiversioning whitelist, plus adds 'used' to the allowed list with 'target'. The behavior here mirrors the implementation in GCC, where 'used' only applies to the single declaration and doesn't apply to the ifunc or resolver. This is not being applied to cpu_dispatch and cpu_specific, since the rules are more complicated for cpu_specific, which emits multiple symbols. Additionally, the author isn't currently aware of uses in the wild of this combination, but is aware of a number of target+used combinations.
2020-03-10 01:15:45 +08:00
__attribute__((target("default"), used)) inline void foo_used(int i, double d) {}
__attribute__((target("avx,sse4.2"))) inline void foo_used(int i, double d) {}
__attribute__((target("default"))) inline void foo_used2(int i, double d) {}
__attribute__((target("avx,sse4.2"), used)) inline void foo_used2(int i, double d) {}
Ensure target-multiversioning emits deferred declarations As reported in PR50025, sometimes we would end up not emitting functions needed by inline multiversioned variants. This is because we typically use the 'deferred decl' mechanism to emit these. However, the variants are emitted after that typically happens. This fixes that by ensuring we re-run deferred decls after this happens. Also, the multiversion emission is done recursively to ensure that MV functions that require other MV functions to be emitted get emitted.
2021-04-20 22:35:57 +08:00
// PR50025:
static void must_be_emitted(void) {}
inline __attribute__((target("default"))) void pr50025(void) { must_be_emitted(); }
void calls_pr50025() { pr50025(); }
// Also need to make sure we get other multiversion functions.
inline __attribute__((target("default"))) void pr50025b(void) { must_be_emitted(); }
inline __attribute__((target("default"))) void pr50025c(void) { pr50025b(); }
void calls_pr50025c() { pr50025c(); }
Change some addUsedGlobal to addUsedOrCompilerUsedGlobal An global value in the `llvm.used` list does not have GC root semantics on ELF targets. This will be changed in a subsequent backend patch. Change some `llvm.used` in the ELF code path to use `llvm.compiler.used` to prevent undesired GC root semantics. Change one extern "C" alias (due to `__attribute__((used))` in extern "C") to use `llvm.compiler.used` on all targets. GNU ld has a rule "`__start_/__stop_` references from a live input section retain the associated C identifier name sections", which LLD may drop entirely (currently refined to exclude SHF_LINK_ORDER/SHF_GROUP) in a future release (the rule makes it clumsy to GC metadata sections; D96914 added a way to try the potential future behavior). For `llvm.used` global values defined in a C identifier name section, keep using `llvm.used` so that the future LLD change will not affect them. rnk kindly categorized the changes: ``` ObjC/blocks: this wants GC root semantics, since ObjC mainly runs on Mac. MS C++ ABI stuff: wants GC root semantics, no change OpenMP: unsure, but GC root semantics probably don't hurt CodeGenModule: affected in this patch to *not* use GC root semantics so that __attribute__((used)) behavior remains the same on ELF, plus two other minor use cases that don't want GC semantics Coverage: Probably want GC root semantics CGExpr.cpp: refers to LTO, wants GC root CGDeclCXX.cpp: one is MS ABI specific, so yes GC root, one is some other C++ init functionality, which should form GC roots (C++ initializers can have side effects and must run) CGDecl.cpp: Changed in this patch for __attribute__((used)) ``` Differential Revision: https://reviews.llvm.org/D97446
2021-02-27 02:42:07 +08:00
// LINUX: @llvm.compiler.used = appending global [2 x i8*] [i8* bitcast (void (i32, double)* @foo_used to i8*), i8* bitcast (void (i32, double)* @foo_used2.avx_sse4.2 to i8*)], section "llvm.metadata"
Permit attribute 'used' with 'target' multiversioning. This adds infrastructure for a multiversioning whitelist, plus adds 'used' to the allowed list with 'target'. The behavior here mirrors the implementation in GCC, where 'used' only applies to the single declaration and doesn't apply to the ifunc or resolver. This is not being applied to cpu_dispatch and cpu_specific, since the rules are more complicated for cpu_specific, which emits multiple symbols. Additionally, the author isn't currently aware of uses in the wild of this combination, but is aware of a number of target+used combinations.
2020-03-10 01:15:45 +08:00
// WINDOWS: @llvm.used = appending global [2 x i8*] [i8* bitcast (void (i32, double)* @foo_used to i8*), i8* bitcast (void (i32, double)* @foo_used2.avx_sse4.2 to i8*)], section "llvm.metadata"
Revert "[CodeGenModule] Assume dso_local for -fpic -fno-semantic-interposition" This reverts commit 789a46f2d742e11edaade28cb59a0f4d2a1d770e. Accidentally committed.
2020-02-04 02:09:39 +08:00
// LINUX: @foo.ifunc = weak_odr ifunc i32 (), i32 ()* ()* @foo.resolver
// LINUX: @foo_inline.ifunc = weak_odr ifunc i32 (), i32 ()* ()* @foo_inline.resolver
// LINUX: @foo_decls.ifunc = weak_odr ifunc void (), void ()* ()* @foo_decls.resolver
// LINUX: @foo_multi.ifunc = weak_odr ifunc void (i32, double), void (i32, double)* ()* @foo_multi.resolver
// LINUX: @fwd_decl_default.ifunc = weak_odr ifunc i32 (), i32 ()* ()* @fwd_decl_default.resolver
// LINUX: @fwd_decl_avx.ifunc = weak_odr ifunc i32 (), i32 ()* ()* @fwd_decl_avx.resolver
[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 @foo.sse4.2()
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 0
[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 @foo.arch_ivybridge()
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 1
[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 @foo.arch_goldmont()
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
// LINUX: ret i32 3
[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 @foo.arch_goldmont-plus()
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
// LINUX: ret i32 4
[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 @foo.arch_tremont()
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
// LINUX: ret i32 5
[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 @foo.arch_icelake-client()
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
// LINUX: ret i32 6
[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 @foo.arch_icelake-server()
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
// LINUX: ret i32 7
[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 @foo.arch_cooperlake()
[X86] Assign a feature priority to 'tigerlake' so it won't assert when used with function multiversioning Also test cooperlake since it was also just added to function multiversioning when it was enabled for __builtin_cpu_is.
2020-06-22 04:24:43 +08:00
// LINUX: ret i32 8
[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 @foo.arch_tigerlake()
[X86] Assign a feature priority to 'tigerlake' so it won't assert when used with function multiversioning Also test cooperlake since it was also just added to function multiversioning when it was enabled for __builtin_cpu_is.
2020-06-22 04:24:43 +08:00
// LINUX: ret i32 9
[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 @foo.arch_sapphirerapids()
[X86] Support -march=sapphirerapids Support -march=sapphirerapids for x86. Compare with Icelake Server, it includes 14 more new features. They are amxtile, amxint8, amxbf16, avx512bf16, avx512vp2intersect, cldemote, enqcmd, movdir64b, movdiri, ptwrite, serialize, shstk, tsxldtrk, waitpkg. Reviewed By: craig.topper Differential Revision: https://reviews.llvm.org/D86503
2020-08-25 12:27:02 +08:00
// LINUX: ret i32 10
[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 @foo.arch_alderlake()
[X86] Add a stub for Intel's alderlake. No scheduling, no autodetection.
2020-10-25 00:00:33 +08:00
// LINUX: ret i32 11
[X86] Support -march=rocketlake Reviewed By: skan, craig.topper, MaskRay Differential Revision: https://reviews.llvm.org/D100085
2021-04-13 09:24:34 +08:00
// LINUX: define{{.*}} i32 @foo.arch_rocketlake()
// LINUX: ret i32 12
[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 @foo()
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 2
[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 @bar()
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: call i32 @foo.ifunc()
// WINDOWS: define dso_local i32 @foo.sse4.2()
// WINDOWS: ret i32 0
// WINDOWS: define dso_local i32 @foo.arch_ivybridge()
// WINDOWS: ret i32 1
[X86] Assign a feature to tremont, goldmont, goldmont-plus, icelake-client, and icelake for target multiversioning priority. Without this these CPUs all caused the compiler to assert when used for multiversioning.
2020-06-10 06:25:46 +08:00
// WINDOWS: define dso_local i32 @foo.arch_goldmont()
// WINDOWS: ret i32 3
// WINDOWS: define dso_local i32 @foo.arch_goldmont-plus()
// WINDOWS: ret i32 4
// WINDOWS: define dso_local i32 @foo.arch_tremont()
// WINDOWS: ret i32 5
// WINDOWS: define dso_local i32 @foo.arch_icelake-client()
// WINDOWS: ret i32 6
// WINDOWS: define dso_local i32 @foo.arch_icelake-server()
// WINDOWS: ret i32 7
[NFC][X86] add missing tests in clang/test/CodeGen/attr-target-mv.c Reviewed By: pengfei Differential Revision: https://reviews.llvm.org/D106849
2021-07-29 13:27:40 +08:00
// WINDOWS: define dso_local i32 @foo.arch_cooperlake()
// WINDOWS: ret i32 8
// WINDOWS: define dso_local i32 @foo.arch_tigerlake()
// WINDOWS: ret i32 9
// WINDOWS: define dso_local i32 @foo.arch_sapphirerapids()
// WINDOWS: ret i32 10
// WINDOWS: define dso_local i32 @foo.arch_alderlake()
// WINDOWS: ret i32 11
// WINDOWS: define dso_local i32 @foo.arch_rocketlake()
// WINDOWS: ret i32 12
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: define dso_local i32 @foo()
// WINDOWS: ret i32 2
// WINDOWS: define dso_local i32 @bar()
// WINDOWS: call i32 @foo.resolver()
[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 ()* @foo.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: call void @__cpu_indicator_init()
// LINUX: ret i32 ()* @foo.arch_sandybridge
// LINUX: ret i32 ()* @foo.arch_ivybridge
// LINUX: ret i32 ()* @foo.sse4.2
// LINUX: ret i32 ()* @foo
[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.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
// WINDOWS: call void @__cpu_indicator_init()
// WINDOWS: call i32 @foo.arch_sandybridge
// WINDOWS: call i32 @foo.arch_ivybridge
// WINDOWS: call i32 @foo.sse4.2
// WINDOWS: call i32 @foo
[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 @bar2()
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: call i32 @foo_inline.ifunc()
// WINDOWS: define dso_local i32 @bar2()
// WINDOWS: call i32 @foo_inline.resolver()
[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 ()* @foo_inline.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: call void @__cpu_indicator_init()
// LINUX: ret i32 ()* @foo_inline.arch_sandybridge
// LINUX: ret i32 ()* @foo_inline.arch_ivybridge
// LINUX: ret i32 ()* @foo_inline.sse4.2
// LINUX: ret i32 ()* @foo_inline
[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_inline.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
// WINDOWS: call void @__cpu_indicator_init()
// WINDOWS: call i32 @foo_inline.arch_sandybridge
// WINDOWS: call i32 @foo_inline.arch_ivybridge
// WINDOWS: call i32 @foo_inline.sse4.2
// WINDOWS: call i32 @foo_inline
[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{{.*}} void @bar3()
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: call void @foo_decls.ifunc()
// WINDOWS: define dso_local void @bar3()
// WINDOWS: call void @foo_decls.resolver()
[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 ()* @foo_decls.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 ()* @foo_decls.sse4.2
// LINUX: ret void ()* @foo_decls
[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 @foo_decls.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
// WINDOWS: call void @foo_decls.sse4.2
Fix spelling of WINDOWS in a test Change-Id: I232515655359f14308e1c5509c4b7db96d1fafcb llvm-svn: 348598
2018-12-07 23:06:43 +08:00
// WINDOWS: call void @foo_decls
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{{.*}} void @bar4()
[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 void @foo_multi.ifunc(i32 noundef 1, double noundef 5.{{[0+e]*}})
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: define dso_local void @bar4()
[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 void @foo_multi.resolver(i32 noundef 1, double noundef 5.{{[0+e]*}})
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 void (i32, double)* @foo_multi.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: and i32 %{{.*}}, 4352
// LINUX: icmp eq i32 %{{.*}}, 4352
// LINUX: ret void (i32, double)* @foo_multi.fma4_sse4.2
// LINUX: icmp eq i32 %{{.*}}, 12
// LINUX: and i32 %{{.*}}, 4352
// LINUX: icmp eq i32 %{{.*}}, 4352
// LINUX: ret void (i32, double)* @foo_multi.arch_ivybridge_fma4_sse4.2
// LINUX: and i32 %{{.*}}, 768
// LINUX: icmp eq i32 %{{.*}}, 768
// LINUX: ret void (i32, double)* @foo_multi.avx_sse4.2
// LINUX: ret void (i32, double)* @foo_multi
[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 @foo_multi.resolver(i32 %0, double %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: and i32 %{{.*}}, 4352
// WINDOWS: icmp eq i32 %{{.*}}, 4352
// WINDOWS: call void @foo_multi.fma4_sse4.2(i32 %0, double %1)
// WINDOWS-NEXT: ret void
// WINDOWS: icmp eq i32 %{{.*}}, 12
// WINDOWS: and i32 %{{.*}}, 4352
// WINDOWS: icmp eq i32 %{{.*}}, 4352
// WINDOWS: call void @foo_multi.arch_ivybridge_fma4_sse4.2(i32 %0, double %1)
// WINDOWS-NEXT: ret void
// WINDOWS: and i32 %{{.*}}, 768
// WINDOWS: icmp eq i32 %{{.*}}, 768
// WINDOWS: call void @foo_multi.avx_sse4.2(i32 %0, double %1)
// WINDOWS-NEXT: ret void
// WINDOWS: call void @foo_multi(i32 %0, double %1)
// WINDOWS-NEXT: ret void
[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 @fwd_decl_default()
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: ret i32 2
[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 @fwd_decl_avx.avx()
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: ret i32 2
[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 @fwd_decl_avx()
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: ret i32 2
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
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// WINDOWS: define dso_local i32 @fwd_decl_default()
// WINDOWS: ret i32 2
// WINDOWS: define dso_local i32 @fwd_decl_avx.avx()
// WINDOWS: ret i32 2
// WINDOWS: define dso_local i32 @fwd_decl_avx()
// WINDOWS: ret i32 2
[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{{.*}} void @bar5()
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: call i32 @fwd_decl_default.ifunc()
// LINUX: call i32 @fwd_decl_avx.ifunc()
// WINDOWS: define dso_local void @bar5()
// WINDOWS: call i32 @fwd_decl_default.resolver()
// WINDOWS: call i32 @fwd_decl_avx.resolver()
[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 ()* @fwd_decl_default.resolver() comdat
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: call void @__cpu_indicator_init()
// LINUX: ret i32 ()* @fwd_decl_default
[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 ()* @fwd_decl_avx.resolver() comdat
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: call void @__cpu_indicator_init()
// LINUX: ret i32 ()* @fwd_decl_avx.avx
// LINUX: ret i32 ()* @fwd_decl_avx
[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 @fwd_decl_default.resolver() comdat
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// WINDOWS: call void @__cpu_indicator_init()
// WINDOWS: call i32 @fwd_decl_default
[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 @fwd_decl_avx.resolver() comdat
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// WINDOWS: call void @__cpu_indicator_init()
// WINDOWS: call i32 @fwd_decl_avx.avx
// WINDOWS: call i32 @fwd_decl_avx
[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 @changed_to_mv.avx()
// LINUX: define{{.*}} i32 @changed_to_mv.fma4()
Fix Target Multiversioning renaming. The initial implementation only did 'first declaration renaming' when a default version came after. This is insufficient in cases where a default does not exist, so this patch makes sure that we do the renaming in all cases. This renaming is necessary because we emit the first declaration before knowing that it IS a target multiversion function, which would change its name. The second declaration (the one that caused the multiversioning) then needs to make sure that the first one has its name changed to be consistent with the resolver usage.
2020-03-09 23:06:58 +08:00
// WINDOWS: define dso_local i32 @changed_to_mv.avx()
// WINDOWS: define dso_local i32 @changed_to_mv.fma4()
[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 void @foo_used(i32 noundef %{{.*}}, double noundef %{{.*}})
Permit attribute 'used' with 'target' multiversioning. This adds infrastructure for a multiversioning whitelist, plus adds 'used' to the allowed list with 'target'. The behavior here mirrors the implementation in GCC, where 'used' only applies to the single declaration and doesn't apply to the ifunc or resolver. This is not being applied to cpu_dispatch and cpu_specific, since the rules are more complicated for cpu_specific, which emits multiple symbols. Additionally, the author isn't currently aware of uses in the wild of this combination, but is aware of a number of target+used combinations.
2020-03-10 01:15:45 +08:00
// LINUX-NOT: @foo_used.avx_sse4.2(
// LINUX-NOT: @foo_used2(
[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 void @foo_used2.avx_sse4.2(i32 noundef %{{.*}}, double noundef %{{.*}})
Permit attribute 'used' with 'target' multiversioning. This adds infrastructure for a multiversioning whitelist, plus adds 'used' to the allowed list with 'target'. The behavior here mirrors the implementation in GCC, where 'used' only applies to the single declaration and doesn't apply to the ifunc or resolver. This is not being applied to cpu_dispatch and cpu_specific, since the rules are more complicated for cpu_specific, which emits multiple symbols. Additionally, the author isn't currently aware of uses in the wild of this combination, but is aware of a number of target+used combinations.
2020-03-10 01:15:45 +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 linkonce_odr dso_local void @foo_used(i32 noundef %{{.*}}, double noundef %{{.*}})
Permit attribute 'used' with 'target' multiversioning. This adds infrastructure for a multiversioning whitelist, plus adds 'used' to the allowed list with 'target'. The behavior here mirrors the implementation in GCC, where 'used' only applies to the single declaration and doesn't apply to the ifunc or resolver. This is not being applied to cpu_dispatch and cpu_specific, since the rules are more complicated for cpu_specific, which emits multiple symbols. Additionally, the author isn't currently aware of uses in the wild of this combination, but is aware of a number of target+used combinations.
2020-03-10 01:15:45 +08:00
// WINDOWS-NOT: @foo_used.avx_sse4.2(
// WINDOWS-NOT: @foo_used2(
[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 void @foo_used2.avx_sse4.2(i32 noundef %{{.*}}, double noundef %{{.*}})
Permit attribute 'used' with 'target' multiversioning. This adds infrastructure for a multiversioning whitelist, plus adds 'used' to the allowed list with 'target'. The behavior here mirrors the implementation in GCC, where 'used' only applies to the single declaration and doesn't apply to the ifunc or resolver. This is not being applied to cpu_dispatch and cpu_specific, since the rules are more complicated for cpu_specific, which emits multiple symbols. Additionally, the author isn't currently aware of uses in the wild of this combination, but is aware of a number of target+used combinations.
2020-03-10 01:15:45 +08:00
Correct 'target' default behavior on redecl, allow forward declaration. Declarations without the attribute were disallowed because it would be ambiguous which 'target' it was supposed to be on. For example: void ___attribute__((target("v1"))) foo(); void foo(); // Redecl of above, or fwd decl of below? void ___attribute__((target("v2"))) foo(); However, a first declaration doesn't have that problem, and erroring prevents it from working in cases where the forward declaration is useful. Additionally, a forward declaration of target==default wouldn't properly cause multiversioning, so this patch fixes that. The patch was not split since the 'default' fix would require implementing the same check for that case, followed by undoing the same change for the fwd-decl implementation. Change-Id: I66f2c5bc2477bcd3f7544b9c16c83ece257077b0 llvm-svn: 347805
2018-11-29 04:58:43 +08:00
// LINUX: declare i32 @foo.arch_sandybridge()
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: declare dso_local i32 @foo.arch_sandybridge()
Revert "[CodeGenModule] Assume dso_local for -fpic -fno-semantic-interposition" This reverts commit 789a46f2d742e11edaade28cb59a0f4d2a1d770e. Accidentally committed.
2020-02-04 02:09:39 +08:00
// LINUX: define linkonce i32 @foo_inline.sse4.2()
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 0
// WINDOWS: define linkonce_odr dso_local i32 @foo_inline.sse4.2()
// WINDOWS: ret i32 0
// LINUX: declare i32 @foo_inline.arch_sandybridge()
// WINDOWS: declare dso_local i32 @foo_inline.arch_sandybridge()
Revert "[CodeGenModule] Assume dso_local for -fpic -fno-semantic-interposition" This reverts commit 789a46f2d742e11edaade28cb59a0f4d2a1d770e. Accidentally committed.
2020-02-04 02:09:39 +08:00
// LINUX: define linkonce i32 @foo_inline.arch_ivybridge()
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
Revert "[CodeGenModule] Assume dso_local for -fpic -fno-semantic-interposition" This reverts commit 789a46f2d742e11edaade28cb59a0f4d2a1d770e. Accidentally committed.
2020-02-04 02:09:39 +08:00
// LINUX: define linkonce i32 @foo_inline()
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
// WINDOWS: define linkonce_odr dso_local i32 @foo_inline.arch_ivybridge()
// WINDOWS: ret i32 1
// WINDOWS: define linkonce_odr dso_local i32 @foo_inline()
// WINDOWS: ret i32 2
Revert "[CodeGenModule] Assume dso_local for -fpic -fno-semantic-interposition" This reverts commit 789a46f2d742e11edaade28cb59a0f4d2a1d770e. Accidentally committed.
2020-02-04 02:09:39 +08:00
// LINUX: define linkonce void @foo_decls()
// LINUX: define linkonce void @foo_decls.sse4.2()
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: define linkonce_odr dso_local void @foo_decls()
// WINDOWS: define linkonce_odr dso_local void @foo_decls.sse4.2()
[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 void @foo_multi(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
// LINUX: define linkonce void @foo_multi.avx_sse4.2(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
// LINUX: define linkonce void @foo_multi.fma4_sse4.2(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
// LINUX: define linkonce void @foo_multi.arch_ivybridge_fma4_sse4.2(i32 noundef %{{[^,]+}}, double 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: define linkonce_odr dso_local void @foo_multi(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
// WINDOWS: define linkonce_odr dso_local void @foo_multi.avx_sse4.2(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
// WINDOWS: define linkonce_odr dso_local void @foo_multi.fma4_sse4.2(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
// WINDOWS: define linkonce_odr dso_local void @foo_multi.arch_ivybridge_fma4_sse4.2(i32 noundef %{{[^,]+}}, double noundef %{{[^\)]+}})
Ensure target-multiversioning emits deferred declarations As reported in PR50025, sometimes we would end up not emitting functions needed by inline multiversioned variants. This is because we typically use the 'deferred decl' mechanism to emit these. However, the variants are emitted after that typically happens. This fixes that by ensuring we re-run deferred decls after this happens. Also, the multiversion emission is done recursively to ensure that MV functions that require other MV functions to be emitted get emitted.
2021-04-20 22:35:57 +08:00
// Ensure that we emit the 'static' function here.
// LINUX: define linkonce void @pr50025()
// LINUX: call void @must_be_emitted
// LINUX: define internal void @must_be_emitted()
// WINDOWS: define linkonce_odr dso_local void @pr50025()
// WINDOWS: call void @must_be_emitted
// WINDOWS: define internal void @must_be_emitted()
// LINUX: define linkonce void @pr50025c()
// LINUX: define linkonce void @pr50025b()
// WINDOWS: define linkonce_odr dso_local void @pr50025c()
// WINDOWS: define linkonce_odr dso_local void @pr50025b()