From 6e5dc42b5add25c94ce0e95da87122f91b4bfdb3 Mon Sep 17 00:00:00 2001 From: Grant Grundler Date: Sun, 10 Sep 2006 12:57:55 -0700 Subject: [PATCH] [PARISC] Further updates to timer_interrupt() This version (relative to the current tree): o eliminates "while (ticks_elapsed)" loop. It's not needed. o drop "ticks_elapsed" completely from timer_interrupt(). o Estimates elapsed cycles (based on HZ) to see which kind of math we want to use to calculate "cycles_remainder". o Fixes a bug where we would loose a tick if we decided we wanted to skip one interrupt. Signed-off-by: Grant Grundler Signed-off-by: Kyle McMartin --- arch/parisc/kernel/time.c | 127 +++++++++++++++++++------------------- 1 file changed, 65 insertions(+), 62 deletions(-) diff --git a/arch/parisc/kernel/time.c b/arch/parisc/kernel/time.c index c43e847a4b..9d642d820f 100644 --- a/arch/parisc/kernel/time.c +++ b/arch/parisc/kernel/time.c @@ -43,12 +43,11 @@ irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) unsigned long now; unsigned long next_tick; unsigned long cycles_elapsed; - unsigned long cycles_remainder; - unsigned long ticks_elapsed = 1; /* at least one elapsed */ - int cpu = smp_processor_id(); + unsigned long cycles_remainder; + unsigned int cpu = smp_processor_id(); /* gcc can optimize for "read-only" case with a local clocktick */ - unsigned long local_ct = clocktick; + unsigned long cpt = clocktick; profile_tick(CPU_PROFILING, regs); @@ -63,28 +62,16 @@ irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) cycles_elapsed = now - next_tick; - /* Determine how much time elapsed. */ - if (now < next_tick) { - /* Scenario 2: CR16 wrapped after clock tick. - * 1's complement will give us the "elapse cycles". - * - * This "cr16 wrapped" cruft is primarily for 32-bit kernels. - * So think "unsigned long is u32" when reading the code. - * And yes, of course 64-bit will someday wrap, but only - * every 198841 days on a 1GHz machine. + if ((cycles_elapsed >> 5) < cpt) { + /* use "cheap" math (add/subtract) instead + * of the more expensive div/mul method */ - cycles_elapsed = ~cycles_elapsed; /* off by one cycle - don't care */ - } - - if (likely(cycles_elapsed < local_ct)) { - /* ticks_elapsed = 1 -- We already assumed one tick elapsed. */ cycles_remainder = cycles_elapsed; + while (cycles_remainder > cpt) { + cycles_remainder -= cpt; + } } else { - /* more than one tick elapsed. Do "expensive" math. */ - ticks_elapsed += cycles_elapsed / local_ct; - - /* Faster version of "remainder = elapsed % clocktick" */ - cycles_remainder = cycles_elapsed - (ticks_elapsed * local_ct); + cycles_remainder = cycles_elapsed % cpt; } /* Can we differentiate between "early CR16" (aka Scenario 1) and @@ -94,51 +81,65 @@ irqreturn_t timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) * cycles after the IT fires. But it's arbitrary how much time passes * before we call it "late". I've picked one second. */ - if (ticks_elapsed > HZ) { +/* aproximate HZ with shifts. Intended math is "(elapsed/clocktick) > HZ" */ +#if HZ == 1000 + if (cycles_elapsed > (cpt << 10) ) +#elif HZ == 250 + if (cycles_elapsed > (cpt << 8) ) +#elif HZ == 100 + if (cycles_elapsed > (cpt << 7) ) +#else +#warn WTF is HZ set to anyway? + if (cycles_elapsed > (HZ * cpt) ) +#endif + { /* Scenario 3: very long delay? bad in any case */ printk (KERN_CRIT "timer_interrupt(CPU %d): delayed!" - " ticks %ld cycles %lX rem %lX" + " cycles %lX rem %lX " " next/now %lX/%lX\n", cpu, - ticks_elapsed, cycles_elapsed, cycles_remainder, + cycles_elapsed, cycles_remainder, next_tick, now ); } + /* convert from "division remainder" to "remainder of clock tick" */ + cycles_remainder = cpt - cycles_remainder; /* Determine when (in CR16 cycles) next IT interrupt will fire. * We want IT to fire modulo clocktick even if we miss/skip some. * But those interrupts don't in fact get delivered that regularly. */ - next_tick = now + (local_ct - cycles_remainder); + next_tick = now + cycles_remainder; + + cpu_data[cpu].it_value = next_tick; /* Skip one clocktick on purpose if we are likely to miss next_tick. - * We'll catch what we missed on the tick after that. - * We should never need 0x1000 cycles to read CR16, calc the - * new next_tick, then write CR16 back. */ - if (!((local_ct - cycles_remainder) >> 12)) - next_tick += local_ct; + * We want to avoid the new next_tick being less than CR16. + * If that happened, itimer wouldn't fire until CR16 wrapped. + * We'll catch the tick we missed on the tick after that. + */ + if (!(cycles_remainder >> 13)) + next_tick += cpt; /* Program the IT when to deliver the next interrupt. */ /* Only bottom 32-bits of next_tick are written to cr16. */ - cpu_data[cpu].it_value = next_tick; mtctl(next_tick, 16); - /* Now that we are done mucking with unreliable delivery of interrupts, - * go do system house keeping. + + /* Done mucking with unreliable delivery of interrupts. + * Go do system house keeping. */ - while (ticks_elapsed--) { #ifdef CONFIG_SMP - smp_do_timer(regs); + smp_do_timer(regs); #else - update_process_times(user_mode(regs)); + update_process_times(user_mode(regs)); #endif - if (cpu == 0) { - write_seqlock(&xtime_lock); - do_timer(1); - write_sequnlock(&xtime_lock); - } + if (cpu == 0) { + write_seqlock(&xtime_lock); + do_timer(regs); + write_sequnlock(&xtime_lock); } - + /* check soft power switch status */ if (cpu == 0 && !atomic_read(&power_tasklet.count)) tasklet_schedule(&power_tasklet); @@ -164,14 +165,12 @@ unsigned long profile_pc(struct pt_regs *regs) EXPORT_SYMBOL(profile_pc); -/*** converted from ia64 ***/ /* * Return the number of micro-seconds that elapsed since the last * update to wall time (aka xtime). The xtime_lock * must be at least read-locked when calling this routine. */ -static inline unsigned long -gettimeoffset (void) +static inline unsigned long gettimeoffset (void) { #ifndef CONFIG_SMP /* @@ -185,36 +184,40 @@ gettimeoffset (void) unsigned long elapsed_cycles; unsigned long usec; unsigned long cpuid = smp_processor_id(); - unsigned long local_ct = clocktick; + unsigned long cpt = clocktick; next_tick = cpu_data[cpuid].it_value; now = mfctl(16); /* Read the hardware interval timer. */ - prev_tick = next_tick - local_ct; + prev_tick = next_tick - cpt; /* Assume Scenario 1: "now" is later than prev_tick. */ elapsed_cycles = now - prev_tick; - if (now < prev_tick) { - /* Scenario 2: CR16 wrapped! - * ones complement is off-by-one. Don't care. - */ - elapsed_cycles = ~elapsed_cycles; - } - - if (elapsed_cycles > (HZ * local_ct)) { +/* aproximate HZ with shifts. Intended math is "(elapsed/clocktick) > HZ" */ +#if HZ == 1000 + if (elapsed_cycles > (cpt << 10) ) +#elif HZ == 250 + if (elapsed_cycles > (cpt << 8) ) +#elif HZ == 100 + if (elapsed_cycles > (cpt << 7) ) +#else +#warn WTF is HZ set to anyway? + if (elapsed_cycles > (HZ * cpt) ) +#endif + { /* Scenario 3: clock ticks are missing. */ - printk (KERN_CRIT "gettimeoffset(CPU %d): missing ticks!" - "cycles %lX prev/now/next %lX/%lX/%lX clock %lX\n", - cpuid, - elapsed_cycles, prev_tick, now, next_tick, local_ct); + printk (KERN_CRIT "gettimeoffset(CPU %ld): missing %ld ticks!" + " cycles %lX prev/now/next %lX/%lX/%lX clock %lX\n", + cpuid, elapsed_cycles / cpt, + elapsed_cycles, prev_tick, now, next_tick, cpt); } /* FIXME: Can we improve the precision? Not with PAGE0. */ usec = (elapsed_cycles * 10000) / PAGE0->mem_10msec; /* add in "lost" jiffies */ - usec += local_ct * (jiffies - wall_jiffies); + usec += cpt * (jiffies - wall_jiffies); return usec; #else return 0; -- 2.39.5