From d3ea9fa0a563620fe9f416f94bb8927c64390917 Mon Sep 17 00:00:00 2001 From: Stuart Menefy Date: Fri, 25 Sep 2009 18:25:10 +0100 Subject: sh: Minor optimisations to FPU handling A number of small optimisations to FPU handling, in particular: - move the task USEDFPU flag from the thread_info flags field (which is accessed asynchronously to the thread) to a new status field, which is only accessed by the thread itself. This allows locking to be removed in most cases, or can be reduced to a preempt_lock(). This mimics the i386 behaviour. - move the modification of regs->sr and thread_info->status flags out of save_fpu() to __unlazy_fpu(). This gives the compiler a better chance to optimise things, as well as making save_fpu() symmetrical with restore_fpu() and init_fpu(). - implement prepare_to_copy(), so that when creating a thread, we can unlazy the FPU prior to copying the thread data structures. Also make sure that the FPU is disabled while in the kernel, in particular while booting, and for newly created kernel threads, In a very artificial benchmark, the execution time for 2500000 context switches was reduced from 50 to 45 seconds. Signed-off-by: Stuart Menefy Signed-off-by: Paul Mundt --- arch/sh/kernel/cpu/sh2a/fpu.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) (limited to 'arch/sh/kernel/cpu/sh2a/fpu.c') diff --git a/arch/sh/kernel/cpu/sh2a/fpu.c b/arch/sh/kernel/cpu/sh2a/fpu.c index 6df2fb98eb30..13817ee49d52 100644 --- a/arch/sh/kernel/cpu/sh2a/fpu.c +++ b/arch/sh/kernel/cpu/sh2a/fpu.c @@ -25,14 +25,12 @@ /* * Save FPU registers onto task structure. - * Assume called with FPU enabled (SR.FD=0). */ void -save_fpu(struct task_struct *tsk, struct pt_regs *regs) +save_fpu(struct task_struct *tsk) { unsigned long dummy; - clear_tsk_thread_flag(tsk, TIF_USEDFPU); enable_fpu(); asm volatile("sts.l fpul, @-%0\n\t" "sts.l fpscr, @-%0\n\t" @@ -60,7 +58,6 @@ save_fpu(struct task_struct *tsk, struct pt_regs *regs) : "memory"); disable_fpu(); - release_fpu(regs); } static void @@ -598,13 +595,13 @@ BUILD_TRAP_HANDLER(fpu_error) struct task_struct *tsk = current; TRAP_HANDLER_DECL; - save_fpu(tsk, regs); + __unlazy_fpu(tsk, regs); if (ieee_fpe_handler(regs)) { tsk->thread.fpu.hard.fpscr &= ~(FPSCR_CAUSE_MASK | FPSCR_FLAG_MASK); grab_fpu(regs); restore_fpu(tsk); - set_tsk_thread_flag(tsk, TIF_USEDFPU); + task_thread_info(tsk)->status |= TS_USEDFPU; return; } @@ -630,5 +627,5 @@ BUILD_TRAP_HANDLER(fpu_state_restore) fpu_init(); set_used_math(); } - set_tsk_thread_flag(tsk, TIF_USEDFPU); + task_thread_info(tsk)->status |= TS_USEDFPU; } -- cgit From 0f09e197a39c081fa8c2752ee65919cb6cba963a Mon Sep 17 00:00:00 2001 From: Paul Mundt Date: Tue, 24 Nov 2009 17:56:17 +0900 Subject: sh: Apply the sleazy FPU changes for SH-2A FPU as well. This plugs in the fpu_counter manipulation for the SH-2A side also. Signed-off-by: Paul Mundt --- arch/sh/kernel/cpu/sh2a/fpu.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'arch/sh/kernel/cpu/sh2a/fpu.c') diff --git a/arch/sh/kernel/cpu/sh2a/fpu.c b/arch/sh/kernel/cpu/sh2a/fpu.c index 13817ee49d52..d395ce5740e7 100644 --- a/arch/sh/kernel/cpu/sh2a/fpu.c +++ b/arch/sh/kernel/cpu/sh2a/fpu.c @@ -608,18 +608,18 @@ BUILD_TRAP_HANDLER(fpu_error) force_sig(SIGFPE, tsk); } -BUILD_TRAP_HANDLER(fpu_state_restore) +void fpu_state_restore(struct pt_regs *regs) { struct task_struct *tsk = current; - TRAP_HANDLER_DECL; grab_fpu(regs); - if (!user_mode(regs)) { + if (unlikely(!user_mode(regs))) { printk(KERN_ERR "BUG: FPU is used in kernel mode.\n"); + BUG(); return; } - if (used_math()) { + if (likely(used_math())) { /* Using the FPU again. */ restore_fpu(tsk); } else { @@ -628,4 +628,12 @@ BUILD_TRAP_HANDLER(fpu_state_restore) set_used_math(); } task_thread_info(tsk)->status |= TS_USEDFPU; + tsk->fpu_counter++; +} + +BUILD_TRAP_HANDLER(fpu_state_restore) +{ + TRAP_HANDLER_DECL; + + fpu_state_restore(regs); } -- cgit