i386: use C code for current_thread_info()
PATCH 1/1 i386: use C code for current_thread_info()
Using C code for current_thread_info() lets the compiler optimize it. With gcc 4.0.2, kernel is smaller:
text data bss dec hex filename
3645212 555556 312024 4512792 44dc18 2.6.17-rc6-nb-post/vmlinux 3647276 555556 312024 4514856 44e428 2.6.17-rc6-nb/vmlinux
-------
-2064
include/asm-i386/thread_info.h +4 -6
@@ -84,17 +84,15 @@ struct thread_info {#define init_stack (init_thread_union.stack) +/* how to get the current stack pointer from C */ +register unsigned long current_stack_pointer asm("esp") __attribute_used__; +/* how to get the thread information struct from C */ static inline struct thread_info *current_thread_info (void ) { -struct thread_info *ti; - __asm__("andl %%esp,%0; " :"=r" (ti) :"0" (~(THREAD_SIZE -1 ))); -return ti; +return (struct thread_info *)(current_stack_pointer & ~(THREAD_SIZE -1 )); } -/* how to get the current stack pointer from C */ -register unsigned long current_stack_pointer asm("esp") __attribute_used__; -/* thread information allocation */ #ifdef CONFIG_DEBUG_STACK_USAGE #define alloc_thread_info (tsk) \
Signature · 1 line
Chuck
Quoted 2 lines · Chuck Ebbert
Using C code for
current_thread_info()lets the compiler optimize it.
Ok, me likee. I just worry that this might break some older gcc version. Have you checked with gcc-3.2 or something?
We've had that "current_stack_pointer" thing for a long while, but we only use it in the irq-stack path (where it's really very incidental: the whole use of the "previous_esp" thing is only for the oops tracing. Maybe there was some reason (like gcc generating bad code) for not using it earlier?
(Admittedly, a more likely reason is probably just nobody looking at it, but it sounds like we should check it out anyway, just in case)
Linus
Quoted 10 lines
Using C code for
current_thread_info()lets the compiler optimize it. With gcc 4.0.2, kernel is smaller:text data bss dec hex filename
3645212 555556 312024 4512792 44dc18 2.6.17-rc6-nb-post/vmlinux 3647276 555556 312024 4514856 44e428 2.6.17-rc6-nb/vmlinux-------
-2064
If possible, can you or someone post the results for x86_64?
Jan Engelhardt
Quoted 3 lines
If possible, can you or someone post the results for
x86_64?
Ignore this request - 'was a asm-i386 change ^_^
Quoted 10 lines · Jan Engelhardt
Quoted 8 lines
Using C code for
current_thread_info()lets the compiler optimize it. With gcc 4.0.2, kernel is smaller:text data bss dec hex filename
3645212 555556 312024 4512792 44dc18 2.6.17-rc6-nb-post/vmlinux 3647276 555556 312024 4514856 44e428 2.6.17-rc6-nb/vmlinux-------
-2064If possible, can you or someone post the results for
x86_64?
Patch is for i386, x86_64's current_thread_info() is already C.
Signature · 5 lines
Cheers, Alistair. Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 10 lines · Alistair John Strachan
Quoted 9 lines · Jan Engelhardt
Quoted 8 lines
Using C code for
current_thread_info()lets the compiler optimize it. With gcc 4.0.2, kernel is smaller:text data bss dec hex filename
3645212 555556 312024 4512792 44dc18 2.6.17-rc6-nb-post/vmlinux 3647276 555556 312024 4514856 44e428 2.6.17-rc6-nb/vmlinux-------
-2064
Quoted 3 lines
Quoted 1 line
If possible, can you or someone post the results for
x86_64?Patch is for i386, x86_64's
current_thread_info()is already C.
Actually read_pda() is inline assembly. But so far gcc/binutils don't support %gs references in the way the kernel needs it directly, so it has to stay this way.
-Andi