Optimize task_thread_info()

Optimize task_thread_info()

[Chen, Kenneth W]

Kernel scheduler uses thread_info.cpu heavily, task_cpu(p) is use just
about in every hot scheduler functions.  Since thread_info structure
always has constant offset from task_struct, access to member variable
of thread_info can be as simple as adding a constant to task_struct
pointer.  Optimize ia64 task_thread_info() to use constant offset. This
saves a memory indirection whenever thread_info structure is used.

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>


--- linux-2.6.15/include/asm-ia64/thread_info.h.orig	2006-01-24 17:59:45.909971596 -0800
+++ linux-2.6.15/include/asm-ia64/thread_info.h	2006-01-24 18:31:05.396276697 -0800
@@ -57,9 +57,11 @@ struct thread_info {
 /* how to get the thread information struct from C */
 #define current_thread_info()	((struct thread_info *) ((char *) current + IA64_TASK_SIZE))
 #define alloc_thread_info(tsk)	((struct thread_info *) ((char *) (tsk) + IA64_TASK_SIZE))
+#define task_thread_info(p)	((struct thread_info *) ((char *) (p) + IA64_TASK_SIZE))
 #else
 #define current_thread_info()	((struct thread_info *) 0)
 #define alloc_thread_info(tsk)	((struct thread_info *) 0)
+#define task_thread_info(p)	((struct thread_info *) 0)
 #endif
 #define free_thread_info(ti)	/* nothing */
 
@@ -67,6 +69,18 @@ struct thread_info {
 #define alloc_task_struct()	((task_t *)__get_free_pages(GFP_KERNEL, KERNEL_STACK_SIZE_ORDER))
 #define free_task_struct(tsk)	free_pages((unsigned long) (tsk), KERNEL_STACK_SIZE_ORDER)
 
+#define __HAVE_THREAD_FUNCTIONS
+static inline void setup_thread_stack(struct task_struct *p, struct task_struct *org)
+{
+        *task_thread_info(p) = *task_thread_info(org);
+        task_thread_info(p)->task = p;
+}
+
+static inline unsigned long *end_of_stack(struct task_struct *p)
+{
+	return (unsigned long *) (task_thread_info(p) + 1);
+}
+
 #endif /* !__ASSEMBLY */
 
 /*

RE: Optimize task_thread_info()

[Chen, Kenneth W]

Chen, Kenneth wrote on Tuesday, January 24, 2006 6:34 PM
> Kernel scheduler uses thread_info.cpu heavily, task_cpu(p) is use just
> about in every hot scheduler functions.  Since thread_info structure
> always has constant offset from task_struct, access to member variable
> of thread_info can be as simple as adding a constant to task_struct
> pointer.  Optimize ia64 task_thread_info() to use constant offset. This
> saves a memory indirection whenever thread_info structure is used.

Nevermind, someone beat me to it by stunning 12 days!!

http://www.kernel.org/git/gitweb.cgi?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h«03591db110e8d195d381a68692eb37da981cdf

[PATCH] ia64: task_thread_info()

on ia64 thread_info is at the constant offset from task_struct and stack
is embedded into the same beast.  Set __HAVE_THREAD_FUNCTIONS, made
task_thread_info() just add a constant.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>