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 <stuart.menefy@st.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

8 files changed, 48 insertions, 42 deletions

diff --git a/arch/sh/include/asm/fpu.h b/arch/sh/include/asm/fpu.h
index bfd78e19de1..d7709c06fac 100644
--- a/arch/sh/include/asm/fpu.h
+++ b/arch/sh/include/asm/fpu.h
@@ -18,17 +18,14 @@ static inline void grab_fpu(struct pt_regs *regs)
 
 struct task_struct;
 
-extern void save_fpu(struct task_struct *__tsk, struct pt_regs *regs);
+extern void save_fpu(struct task_struct *__tsk);
 void fpu_state_restore(struct pt_regs *regs);
 #else
 
+#define save_fpu(tsk)		do { } while (0)
 #define release_fpu(regs)	do { } while (0)
 #define grab_fpu(regs)		do { } while (0)
 
-static inline void save_fpu(struct task_struct *tsk, struct pt_regs *regs)
-{
-	clear_tsk_thread_flag(tsk, TIF_USEDFPU);
-}
 #endif
 
 struct user_regset;
@@ -40,21 +37,28 @@ extern int fpregs_get(struct task_struct *target,
 		      unsigned int pos, unsigned int count,
 		      void *kbuf, void __user *ubuf);
 
+static inline void __unlazy_fpu(struct task_struct *tsk, struct pt_regs *regs)
+{
+	if (task_thread_info(tsk)->status & TS_USEDFPU) {
+		task_thread_info(tsk)->status &= ~TS_USEDFPU;
+		save_fpu(tsk);
+		release_fpu(regs);
+	} else
+		tsk->fpu_counter = 0;
+}
+
 static inline void unlazy_fpu(struct task_struct *tsk, struct pt_regs *regs)
 {
 	preempt_disable();
-	if (test_tsk_thread_flag(tsk, TIF_USEDFPU))
-		save_fpu(tsk, regs);
-	else
-		tsk->fpu_counter = 0;
+	__unlazy_fpu(tsk, regs);
 	preempt_enable();
 }
 
 static inline void clear_fpu(struct task_struct *tsk, struct pt_regs *regs)
 {
 	preempt_disable();
-	if (test_tsk_thread_flag(tsk, TIF_USEDFPU)) {
-		clear_tsk_thread_flag(tsk, TIF_USEDFPU);
+	if (task_thread_info(tsk)->status & TS_USEDFPU) {
+		task_thread_info(tsk)->status &= ~TS_USEDFPU;
 		release_fpu(regs);
 	}
 	preempt_enable();
diff --git a/arch/sh/include/asm/processor_32.h b/arch/sh/include/asm/processor_32.h
index 9a8714945dc..1f3d6fab660 100644
--- a/arch/sh/include/asm/processor_32.h
+++ b/arch/sh/include/asm/processor_32.h
@@ -56,6 +56,7 @@ asmlinkage void __init sh_cpu_init(void);
 #define SR_DSP		0x00001000
 #define SR_IMASK	0x000000f0
 #define SR_FD		0x00008000
+#define SR_MD		0x40000000
 
 /*
  * DSP structure and data
@@ -136,7 +137,7 @@ struct mm_struct;
 extern void release_thread(struct task_struct *);
 
 /* Prepare to copy thread state - unlazy all lazy status */
-#define prepare_to_copy(tsk)	do { } while (0)
+void prepare_to_copy(struct task_struct *tsk);
 
 /*
  * create a kernel thread without removing it from tasklists
diff --git a/arch/sh/include/asm/thread_info.h b/arch/sh/include/asm/thread_info.h
index 23eeed89467..1f3d927e226 100644
--- a/arch/sh/include/asm/thread_info.h
+++ b/arch/sh/include/asm/thread_info.h
@@ -51,6 +51,7 @@ struct thread_info {
 	.task		= &tsk,			\
 	.exec_domain	= &default_exec_domain,	\
 	.flags		= 0,			\
+	.status		= 0,			\
 	.cpu		= 0,			\
 	.preempt_count	= INIT_PREEMPT_COUNT,	\
 	.addr_limit	= KERNEL_DS,		\
@@ -117,7 +118,6 @@ extern void free_thread_info(struct thread_info *ti);
 #define TIF_SECCOMP		6	/* secure computing */
 #define TIF_NOTIFY_RESUME	7	/* callback before returning to user */
 #define TIF_SYSCALL_TRACEPOINT	8	/* for ftrace syscall instrumentation */
-#define TIF_USEDFPU		16	/* FPU was used by this task this quantum (SMP) */
 #define TIF_POLLING_NRFLAG	17	/* true if poll_idle() is polling TIF_NEED_RESCHED */
 #define TIF_MEMDIE		18
 #define TIF_FREEZE		19	/* Freezing for suspend */
@@ -130,7 +130,6 @@ extern void free_thread_info(struct thread_info *ti);
 #define _TIF_SECCOMP		(1 << TIF_SECCOMP)
 #define _TIF_NOTIFY_RESUME	(1 << TIF_NOTIFY_RESUME)
 #define _TIF_SYSCALL_TRACEPOINT	(1 << TIF_SYSCALL_TRACEPOINT)
-#define _TIF_USEDFPU		(1 << TIF_USEDFPU)
 #define _TIF_POLLING_NRFLAG	(1 << TIF_POLLING_NRFLAG)
 #define _TIF_FREEZE		(1 << TIF_FREEZE)
 
@@ -163,6 +162,7 @@ extern void free_thread_info(struct thread_info *ti);
  * have to worry about atomic accesses.
  */
 #define TS_RESTORE_SIGMASK	0x0001	/* restore signal mask in do_signal() */
+#define TS_USEDFPU		0x0002	/* FPU used by this task this quantum */
 
 #ifndef __ASSEMBLY__
 #define HAVE_SET_RESTORE_SIGMASK	1
diff --git a/arch/sh/kernel/cpu/init.c b/arch/sh/kernel/cpu/init.c
index 580d58b94cc..ad9dfff9427 100644
--- a/arch/sh/kernel/cpu/init.c
+++ b/arch/sh/kernel/cpu/init.c
@@ -311,12 +311,12 @@ asmlinkage void __init sh_cpu_init(void)
 	if (fpu_disabled) {
 		printk("FPU Disabled\n");
 		current_cpu_data.flags &= ~CPU_HAS_FPU;
-		disable_fpu();
 	}
 
 	/* FPU initialization */
+	disable_fpu();
 	if ((current_cpu_data.flags & CPU_HAS_FPU)) {
-		clear_thread_flag(TIF_USEDFPU);
+		current_thread_info()->status &= ~TS_USEDFPU;
 		clear_used_math();
 	}
 
diff --git a/arch/sh/kernel/cpu/sh2a/fpu.c b/arch/sh/kernel/cpu/sh2a/fpu.c
index 6df2fb98eb3..13817ee49d5 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;
 }
diff --git a/arch/sh/kernel/cpu/sh4/fpu.c b/arch/sh/kernel/cpu/sh4/fpu.c
index d79226fa59d..e97857aec8a 100644
--- a/arch/sh/kernel/cpu/sh4/fpu.c
+++ b/arch/sh/kernel/cpu/sh4/fpu.c
@@ -41,13 +41,11 @@ static unsigned int fpu_exception_flags;
 
 /*
  * 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)
+void 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"
@@ -92,7 +90,6 @@ void save_fpu(struct task_struct *tsk, struct pt_regs *regs)
 		      :"memory");
 
 	disable_fpu();
-	release_fpu(regs);
 }
 
 static void restore_fpu(struct task_struct *tsk)
@@ -285,7 +282,6 @@ static int ieee_fpe_handler(struct pt_regs *regs)
 		/* fcnvsd */
 		struct task_struct *tsk = current;
 
-		save_fpu(tsk, regs);
 		if ((tsk->thread.fpu.hard.fpscr & FPSCR_CAUSE_ERROR))
 			/* FPU error */
 			denormal_to_double(&tsk->thread.fpu.hard,
@@ -462,7 +458,7 @@ BUILD_TRAP_HANDLER(fpu_error)
 	struct task_struct *tsk = current;
 	TRAP_HANDLER_DECL;
 
-	save_fpu(tsk, regs);
+	__unlazy_fpu(tsk, regs);
 	fpu_exception_flags = 0;
 	if (ieee_fpe_handler(regs)) {
 		tsk->thread.fpu.hard.fpscr &=
@@ -473,7 +469,7 @@ BUILD_TRAP_HANDLER(fpu_error)
 		tsk->thread.fpu.hard.fpscr |= (fpu_exception_flags >> 10);
 		grab_fpu(regs);
 		restore_fpu(tsk);
-		set_tsk_thread_flag(tsk, TIF_USEDFPU);
+		task_thread_info(tsk)->status |= TS_USEDFPU;
 		if ((((tsk->thread.fpu.hard.fpscr & FPSCR_ENABLE_MASK) >> 7) &
 		     (fpu_exception_flags >> 2)) == 0) {
 			return;
@@ -502,7 +498,7 @@ void fpu_state_restore(struct pt_regs *regs)
 		fpu_init();
 		set_used_math();
 	}
-	set_tsk_thread_flag(tsk, TIF_USEDFPU);
+	task_thread_info(tsk)->status |= TS_USEDFPU;
 	tsk->fpu_counter++;
 }
 
diff --git a/arch/sh/kernel/process_32.c b/arch/sh/kernel/process_32.c
index 7733f5fa6bb..d721f9297c0 100644
--- a/arch/sh/kernel/process_32.c
+++ b/arch/sh/kernel/process_32.c
@@ -134,7 +134,10 @@ int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
 	regs.regs[5] = (unsigned long)fn;
 
 	regs.pc = (unsigned long)kernel_thread_helper;
-	regs.sr = (1 << 30);
+	regs.sr = SR_MD;
+#if defined(CONFIG_SH_FPU)
+	regs.sr |= SR_FD;
+#endif
 
 	/* Ok, create the new process.. */
 	pid = do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0,
@@ -189,6 +192,15 @@ int dump_fpu(struct pt_regs *regs, elf_fpregset_t *fpu)
 }
 EXPORT_SYMBOL(dump_fpu);
 
+/*
+ * This gets called before we allocate a new thread and copy
+ * the current task into it.
+ */
+void prepare_to_copy(struct task_struct *tsk)
+{
+	unlazy_fpu(tsk, task_pt_regs(tsk));
+}
+
 asmlinkage void ret_from_fork(void);
 
 int copy_thread(unsigned long clone_flags, unsigned long usp,
@@ -197,16 +209,10 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
 {
 	struct thread_info *ti = task_thread_info(p);
 	struct pt_regs *childregs;
-#if defined(CONFIG_SH_FPU) || defined(CONFIG_SH_DSP)
+#if defined(CONFIG_SH_DSP)
 	struct task_struct *tsk = current;
 #endif
 
-#if defined(CONFIG_SH_FPU)
-	unlazy_fpu(tsk, regs);
-	p->thread.fpu = tsk->thread.fpu;
-	copy_to_stopped_child_used_math(p);
-#endif
-
 #if defined(CONFIG_SH_DSP)
 	if (is_dsp_enabled(tsk)) {
 		/* We can use the __save_dsp or just copy the struct:
@@ -226,6 +232,8 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
 	} else {
 		childregs->regs[15] = (unsigned long)childregs;
 		ti->addr_limit = KERNEL_DS;
+		ti->status &= ~TS_USEDFPU;
+		p->fpu_counter = 0;
 	}
 
 	if (clone_flags & CLONE_SETTLS)
diff --git a/arch/sh/math-emu/math.c b/arch/sh/math-emu/math.c
index ac2d7abd256..d6c15cae091 100644
--- a/arch/sh/math-emu/math.c
+++ b/arch/sh/math-emu/math.c
@@ -558,7 +558,7 @@ static int ieee_fpe_handler(struct pt_regs *regs)
 					    (finsn >> 8) & 0xf);
 			tsk->thread.fpu.hard.fpscr &=
 				~(FPSCR_CAUSE_MASK | FPSCR_FLAG_MASK);
-			set_tsk_thread_flag(tsk, TIF_USEDFPU);
+			task_thread_info(tsk)->status |= TS_USEDFPU;
 		} else {
 			info.si_signo = SIGFPE;
 			info.si_errno = 0;
@@ -619,10 +619,10 @@ int do_fpu_inst(unsigned short inst, struct pt_regs *regs)
 	struct task_struct *tsk = current;
 	struct sh_fpu_soft_struct *fpu = &(tsk->thread.fpu.soft);
 
-	if (!test_tsk_thread_flag(tsk, TIF_USEDFPU)) {
+	if (!(task_thread_info(tsk)->status & TS_USEDFPU)) {
 		/* initialize once. */
 		fpu_init(fpu);
-		set_tsk_thread_flag(tsk, TIF_USEDFPU);
+		task_thread_info(tsk)->status |= TS_USEDFPU;
 	}
 
 	return fpu_emulate(inst, fpu, regs);