Skip to contentSkip to conversation

sLeAZY FPU feature - x86_64 support

PATCH Voluspa 2 participants 2 messages Jul 2, 2006

lore
Share & tools
lore

Voluspa

author
#0 lore raw

You have a very strange 2.6.17 kernel there. The include/linux/sched.h is so incompatible that the patching (with fuzz) places "unsigned char fpu_counter;" in a totally unrelated struct, and not in "struct task_struct {".

Here's a working rebase of that part - sorry about mangling by this webmail client... Btw, the whole thing has no measurable effect on real world stuff like rendering through blender - on my machine, at least.

diff -Nur linux-2.6.17-git19-original/include/linux/sched.h linux-2.6.17-git19- sleazyfpu/include/linux/sched.h
--- linux-2.6.17-git19-original/include/linux/sched.h 2006-07-02 01:17:
36.000000000 +0200
+++ linux-2.6.17-git19-sleazyfpu/include/linux/sched.h 2006-07-02 01:10:
42.000000000 +0200
@@ -926,6 +926,16 @@

         * cache last used pipe for splice
         */
        struct pipe_inode_info *splice_pipe;
+
+       /*
+       * fpu_counter contains the number of consecutive context switches
+       * that the FPU is used. If this is over a threshold, the lazy fpu
+       * saving becomes unlazy to save the trap. This is an unsigned char
+       * so that after 256 times the counter wraps and the behavior turns
+       * lazy again; this to deal with bursty apps that only use FPU for
+       * a short time
+       */
+       unsigned char fpu_counter;

};

static inline pid_t process_group(struct task_struct *tsk)

Mvh
Mats Johannesson

#1 lore raw
Quoted 1 line · Voluspa

You have a very strange 2.6.17 kernel there.

oh sorry it's a 2.6.17-mm4 kernel... I should have mentioned that..

Quoted 1 line