[CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting
rL131311 added `asm()` support for builtin functions, but `asm()` for builtins with
specialized emitting (e.g. memcpy, various math functions) still do not work.
This patch makes these functions work for `asm()` and `#pragma redefine_extname`.
glibc uses `asm()` to redirect internal libc function calls to hidden aliases.
Limitation: such a function is a builtin in clang, but will not be recognized as
a libcall in optimization passes because Clang does not annotate the renamed
function as a libcall. In GCC -O1 or above, `abs` can be optimized out but we can't.
Additionally, we cannot redirect `__builtin_sin` to `real_sin` in the following example:
double sin(double x) asm("real_sin");
double f(double d) { return __builtin_sin(d); }
---
According to @rsmith, the following three statements cannot be simultaneously true:
(1) The frontend function foo has known, builtin semantics X.
(2) The symbol foo has known, builtin semantics X.
(3) It's not correct to lower a call to the frontend function foo to the symbol foo.
People do want (1) (if it is profitable to expand a memcpy, do it).
This also means that people do not want to add -fno-builtin-memcpy.
People do want (3): that is why they use asm("__GI_memcpy") in the first place.
So unfortunately we make a compromise by not refuting (2) (see the limitation above).
For most libcalls, there is a small loss because compilers don't synthesize them.
For the few glibc cares about, it uses `asm("memcpy = __GI_memcpy");` to make
the assembly level redirection.
(Changing function names (e.g. `__memcpy`) is a hit to ergonomics which is not acceptable).
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D88712
2020-10-16 06:11:45 +08:00
|
|
|
// RUN: %clang_cc1 -triple=i686-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefixes=CHECK,LINUX
|
|
|
|
|
// RUN: %clang_cc1 -triple=i686-apple-darwin9 -emit-llvm %s -o - | FileCheck %s --check-prefixes=CHECK,DARWIN
|
2011-05-14 05:12:10 +08:00
|
|
|
|
|
|
|
|
char *strerror(int) asm("alias");
|
2011-11-21 05:05:04 +08:00
|
|
|
int x __asm("foo");
|
2011-05-14 05:12:10 +08:00
|
|
|
|
2011-11-21 05:05:04 +08:00
|
|
|
int *test(void) {
|
|
|
|
|
static int y __asm("bar");
|
|
|
|
|
strerror(-1);
|
|
|
|
|
return &y;
|
2011-05-14 05:12:10 +08:00
|
|
|
}
|
|
|
|
|
|
2011-11-21 05:05:04 +08:00
|
|
|
// LINUX: @bar = internal global i32 0
|
2020-03-10 03:25:24 +08:00
|
|
|
// LINUX: @foo = global i32 0
|
2011-05-14 05:12:10 +08:00
|
|
|
// LINUX: declare i8* @alias(i32)
|
2011-11-21 05:05:04 +08:00
|
|
|
|
|
|
|
|
// DARWIN: @"\01bar" = internal global i32 0
|
2020-03-10 03:25:24 +08:00
|
|
|
// DARWIN: @"\01foo" = global i32 0
|
2011-05-14 05:12:10 +08:00
|
|
|
// DARWIN: declare i8* @"\01alias"(i32)
|
[CGBuiltin] Respect asm labels and redefine_extname for builtins with specialized emitting
rL131311 added `asm()` support for builtin functions, but `asm()` for builtins with
specialized emitting (e.g. memcpy, various math functions) still do not work.
This patch makes these functions work for `asm()` and `#pragma redefine_extname`.
glibc uses `asm()` to redirect internal libc function calls to hidden aliases.
Limitation: such a function is a builtin in clang, but will not be recognized as
a libcall in optimization passes because Clang does not annotate the renamed
function as a libcall. In GCC -O1 or above, `abs` can be optimized out but we can't.
Additionally, we cannot redirect `__builtin_sin` to `real_sin` in the following example:
double sin(double x) asm("real_sin");
double f(double d) { return __builtin_sin(d); }
---
According to @rsmith, the following three statements cannot be simultaneously true:
(1) The frontend function foo has known, builtin semantics X.
(2) The symbol foo has known, builtin semantics X.
(3) It's not correct to lower a call to the frontend function foo to the symbol foo.
People do want (1) (if it is profitable to expand a memcpy, do it).
This also means that people do not want to add -fno-builtin-memcpy.
People do want (3): that is why they use asm("__GI_memcpy") in the first place.
So unfortunately we make a compromise by not refuting (2) (see the limitation above).
For most libcalls, there is a small loss because compilers don't synthesize them.
For the few glibc cares about, it uses `asm("memcpy = __GI_memcpy");` to make
the assembly level redirection.
(Changing function names (e.g. `__memcpy`) is a hit to ergonomics which is not acceptable).
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D88712
2020-10-16 06:11:45 +08:00
|
|
|
|
|
|
|
|
extern void *memcpy(void *__restrict, const void *__restrict, unsigned long);
|
|
|
|
|
extern __typeof(memcpy) memcpy asm("__GI_memcpy");
|
|
|
|
|
void test_memcpy(void *dst, void *src, unsigned long n) {
|
|
|
|
|
memcpy(dst, src, n);
|
|
|
|
|
}
|
|
|
|
|
// CHECK-LABEL: @test_memcpy(
|
|
|
|
|
// LINUX: call i8* @__GI_memcpy(
|
|
|
|
|
// LINUX: declare i8* @__GI_memcpy(i8*, i8*, i32)
|
|
|
|
|
// DARWIN: call i8* @"\01__GI_memcpy"(
|
|
|
|
|
// DARWIN: declare i8* @"\01__GI_memcpy"(i8*, i8*, i32)
|
|
|
|
|
|
|
|
|
|
long lrint(double x) asm("__GI_lrint");
|
|
|
|
|
long test_lrint(double x) {
|
|
|
|
|
return lrint(x);
|
|
|
|
|
}
|
|
|
|
|
// CHECK-LABEL: @test_lrint(
|
|
|
|
|
// LINUX: call i32 @__GI_lrint(
|
|
|
|
|
// LINUX: declare i32 @__GI_lrint(double)
|
|
|
|
|
// DARWIN: call i32 @"\01__GI_lrint"(
|
|
|
|
|
// DARWIN: declare i32 @"\01__GI_lrint"(double)
|
|
|
|
|
|
|
|
|
|
/// NOTE: GCC can optimize out abs in -O1 or above. Clang does not
|
|
|
|
|
/// communicate the mapping to the backend so the libcall cannot be eliminated.
|
|
|
|
|
int abs(int x) asm("__GI_abs");
|
|
|
|
|
long test_abs(int x) {
|
|
|
|
|
return abs(x);
|
|
|
|
|
}
|
|
|
|
|
// CHECK-LABEL: @test_abs(
|
|
|
|
|
// LINUX: call i32 @__GI_abs(
|
|
|
|
|
|
|
|
|
|
/// FIXME: test_sin should call real_sin instead.
|
|
|
|
|
double sin(double x) asm("real_sin");
|
|
|
|
|
double test_sin(double d) { return __builtin_sin(d); }
|
|
|
|
|
// CHECK-LABEL: @test_sin(
|
|
|
|
|
// LINUX: call double @llvm.sin.f64(
|