Loading...
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 | /* * Copyright (C) 1993-1996 Bas Laarhoven. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; see the file COPYING. If not, write to the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. * * $Source: /homes/cvs/ftape-stacked/ftape/lowlevel/ftape-calibr.c,v $ * $Revision: 1.2 $ * $Date: 1997/10/05 19:18:08 $ * * GP calibration routine for processor speed dependent * functions. */ #include <linux/config.h> #include <linux/errno.h> #include <linux/jiffies.h> #include <asm/system.h> #include <asm/io.h> #if defined(__alpha__) # include <asm/hwrpb.h> #elif defined(__x86_64__) # include <asm/msr.h> # include <asm/timex.h> #elif defined(__i386__) # include <linux/timex.h> #endif #include <linux/ftape.h> #include "../lowlevel/ftape-tracing.h" #include "../lowlevel/ftape-calibr.h" #include "../lowlevel/fdc-io.h" #undef DEBUG #if !defined(__alpha__) && !defined(__i386__) && !defined(__x86_64__) # error Ftape is not implemented for this architecture! #endif #if defined(__alpha__) || defined(__x86_64__) static unsigned long ps_per_cycle = 0; #endif static spinlock_t calibr_lock; /* * Note: On Intel PCs, the clock ticks at 100 Hz (HZ==100) which is * too slow for certain timeouts (and that clock doesn't even tick * when interrupts are disabled). For that reason, the 8254 timer is * used directly to implement fine-grained timeouts. However, on * Alpha PCs, the 8254 is *not* used to implement the clock tick * (which is 1024 Hz, normally) and the 8254 timer runs at some * "random" frequency (it seems to run at 18Hz, but it's not safe to * rely on this value). Instead, we use the Alpha's "rpcc" * instruction to read cycle counts. As this is a 32 bit counter, * it will overflow only once per 30 seconds (on a 200MHz machine), * which is plenty. */ unsigned int ftape_timestamp(void) { #if defined(__alpha__) unsigned long r; asm volatile ("rpcc %0" : "=r" (r)); return r; #elif defined(__x86_64__) unsigned long r; rdtscl(r); return r; #elif defined(__i386__) /* * Note that there is some time between counter underflowing and jiffies * increasing, so the code below won't always give correct output. * -Vojtech */ unsigned long flags; __u16 lo; __u16 hi; spin_lock_irqsave(&calibr_lock, flags); outb_p(0x00, 0x43); /* latch the count ASAP */ lo = inb_p(0x40); /* read the latched count */ lo |= inb(0x40) << 8; hi = jiffies; spin_unlock_irqrestore(&calibr_lock, flags); return ((hi + 1) * (unsigned int) LATCH) - lo; /* downcounter ! */ #endif } static unsigned int short_ftape_timestamp(void) { #if defined(__alpha__) || defined(__x86_64__) return ftape_timestamp(); #elif defined(__i386__) unsigned int count; unsigned long flags; spin_lock_irqsave(&calibr_lock, flags); outb_p(0x00, 0x43); /* latch the count ASAP */ count = inb_p(0x40); /* read the latched count */ count |= inb(0x40) << 8; spin_unlock_irqrestore(&calibr_lock, flags); return (LATCH - count); /* normal: downcounter */ #endif } static unsigned int diff(unsigned int t0, unsigned int t1) { #if defined(__alpha__) || defined(__x86_64__) return (t1 - t0); #elif defined(__i386__) /* * This is tricky: to work for both short and full ftape_timestamps * we'll have to discriminate between these. * If it _looks_ like short stamps with wrapping around we'll * asume it are. This will generate a small error if it really * was a (very large) delta from full ftape_timestamps. */ return (t1 <= t0 && t0 <= LATCH) ? t1 + LATCH - t0 : t1 - t0; #endif } static unsigned int usecs(unsigned int count) { #if defined(__alpha__) || defined(__x86_64__) return (ps_per_cycle * count) / 1000000UL; #elif defined(__i386__) return (10000 * count) / ((CLOCK_TICK_RATE + 50) / 100); #endif } unsigned int ftape_timediff(unsigned int t0, unsigned int t1) { /* * Calculate difference in usec for ftape_timestamp results t0 & t1. * Note that on the i386 platform with short time-stamps, the * maximum allowed timespan is 1/HZ or we'll lose ticks! */ return usecs(diff(t0, t1)); } /* To get an indication of the I/O performance, * measure the duration of the inb() function. */ static void time_inb(void) { int i; int t0, t1; unsigned long flags; int status; TRACE_FUN(ft_t_any); spin_lock_irqsave(&calibr_lock, flags); t0 = short_ftape_timestamp(); for (i = 0; i < 1000; ++i) { status = inb(fdc.msr); } t1 = short_ftape_timestamp(); spin_unlock_irqrestore(&calibr_lock, flags); TRACE(ft_t_info, "inb() duration: %d nsec", ftape_timediff(t0, t1)); TRACE_EXIT; } static void init_clock(void) { TRACE_FUN(ft_t_any); #if defined(__x86_64__) ps_per_cycle = 1000000000UL / cpu_khz; #elif defined(__alpha__) extern struct hwrpb_struct *hwrpb; ps_per_cycle = (1000*1000*1000*1000UL) / hwrpb->cycle_freq; #endif TRACE_EXIT; } /* * Input: function taking int count as parameter. * pointers to calculated calibration variables. */ void ftape_calibrate(char *name, void (*fun) (unsigned int), unsigned int *calibr_count, unsigned int *calibr_time) { static int first_time = 1; int i; unsigned int tc = 0; unsigned int count; unsigned int time; #if defined(__i386__) unsigned int old_tc = 0; unsigned int old_count = 1; unsigned int old_time = 1; #endif TRACE_FUN(ft_t_flow); if (first_time) { /* get idea of I/O performance */ init_clock(); time_inb(); first_time = 0; } /* value of timeout must be set so that on very slow systems * it will give a time less than one jiffy, and on * very fast systems it'll give reasonable precision. */ count = 40; for (i = 0; i < 15; ++i) { unsigned int t0; unsigned int t1; unsigned int once; unsigned int multiple; unsigned long flags; *calibr_count = *calibr_time = count; /* set TC to 1 */ spin_lock_irqsave(&calibr_lock, flags); fun(0); /* dummy, get code into cache */ t0 = short_ftape_timestamp(); fun(0); /* overhead + one test */ t1 = short_ftape_timestamp(); once = diff(t0, t1); t0 = short_ftape_timestamp(); fun(count); /* overhead + count tests */ t1 = short_ftape_timestamp(); multiple = diff(t0, t1); spin_unlock_irqrestore(&calibr_lock, flags); time = ftape_timediff(0, multiple - once); tc = (1000 * time) / (count - 1); TRACE(ft_t_any, "once:%3d us,%6d times:%6d us, TC:%5d ns", usecs(once), count - 1, usecs(multiple), tc); #if defined(__alpha__) || defined(__x86_64__) /* * Increase the calibration count exponentially until the * calibration time exceeds 100 ms. */ if (time >= 100*1000) { break; } #elif defined(__i386__) /* * increase the count until the resulting time nears 2/HZ, * then the tc will drop sharply because we lose LATCH counts. */ if (tc <= old_tc / 2) { time = old_time; count = old_count; break; } old_tc = tc; old_count = count; old_time = time; #endif count *= 2; } *calibr_count = count - 1; *calibr_time = time; TRACE(ft_t_info, "TC for `%s()' = %d nsec (at %d counts)", name, (1000 * *calibr_time) / *calibr_count, *calibr_count); TRACE_EXIT; } |