summaryrefslogtreecommitdiffstats
path: root/firmware/profile.c
blob: 6700eca404afb3ceeb43e32df5eef9816aeb2ad2 (plain)
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
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
/***************************************************************************
 *             __________               __   ___.
 *   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
 *   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
 *   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
 *   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
 *                     \/            \/     \/    \/            \/
 * $Id$
 *
 * Profiling routines counts ticks and calls to each profiled function.
 * 
 * Copyright (C) 2005 by Brandon Low
 *
 * 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
 * of the License, or (at your option) any later version.
 *
 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
 * KIND, either express or implied.
 *
 ****************************************************************************
 *
 * profile_func_enter() based on mcount found in gmon.c:
 *
 ***************************************************************************
 * Copyright (c) 1991, 1998 The Regents of the University of California.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. [rescinded 22 July 1999]
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 * @(#)gmon.c 5.3 (Berkeley) 5/22/91
 */

#include <file.h>
#include <sprintf.h>
#include <system.h>
#include <string.h>
#include <timer.h>
#include <sys/types.h>
#include "profile.h"

/* PFD is Profiled Function Data */

/* Indices are shorts which means that we use 4k of RAM */
#define INDEX_BITS 11    /* What is a reasonable size for this? */
#define INDEX_SIZE 2048  /* 2 ^ INDEX_BITS */
#define INDEX_MASK 0x7FF /* lower INDEX_BITS 1 */

/*
 * In the current setup (pfd has 4 longs and 2 shorts) this uses 20k of RAM
 * for profiling, and allows for profiling sections of code with up-to
 * 1024 function caller->callee pairs
 */
#define NUMPFDS 512

struct pfd_struct {
    void *self_pc;
    unsigned long count;
    unsigned long time;
    unsigned short link;
    struct pfd_struct *caller;
};

/* Possible states of profiling */
#define PROF_ON 0x00
#define PROF_BUSY 0x01
#define PROF_ERROR 0x02
#define PROF_OFF 0x03
/* Masks for thread switches */
#define PROF_OFF_THREAD 0x10
#define PROF_ON_THREAD 0x0F

static unsigned short profiling = PROF_OFF;
static size_t recursion_level;
static unsigned short indices[INDEX_SIZE];
static struct pfd_struct pfds[NUMPFDS];
/* This holds a pointer to the last pfd effected for time tracking */
static struct pfd_struct *last_pfd;
/* These are used to track the time when we've lost the CPU so it doesn't count
 * against any of the profiled functions */
static int profiling_thread = -1;

/* internal function prototypes */
static void profile_timer_tick(void);
static void profile_timer_unregister(void);

static void write_function_recursive(int fd, struct pfd_struct *pfd, int depth);

/* Be careful to use the right one for the size of your variable */
#ifdef CPU_COLDFIRE
#define ADDQI_L(_var,_value) \
    asm ("addq.l %[value],%[var];" \
         : [var] "+g" (_var) \
         : [value] "I" (_value) )
#else
#define ADDQI_L(var, value) var += value
#endif

void profile_thread_stopped(int current_thread) {
    if (current_thread == profiling_thread) {
        /* If profiling is busy or idle */
        if (profiling < PROF_ERROR) {
            /* Unregister the timer so that other threads aren't interrupted */
            timer_unregister();
        }
        /* Make sure we don't waste time profiling when we're running the
         * wrong thread */
        profiling |= PROF_OFF_THREAD;
    }
}

void profile_thread_started(int current_thread) {
    if (current_thread == profiling_thread) {
        /* Now we are allowed to profile again */
        profiling &= PROF_ON_THREAD;
        /* if profiling was busy or idle */
        if (profiling < PROF_ERROR) {
            /* After we de-mask, if profiling is active, reactivate the timer */
            timer_register(0, profile_timer_unregister, 
                    TIMER_FREQ/10000, profile_timer_tick IF_COP(, 0 ) );
        }
    }
}

static void profile_timer_tick(void) {
    if (!profiling) {
        register struct pfd_struct *my_last_pfd = last_pfd;
        if (my_last_pfd) {
            ADDQI_L(my_last_pfd->time,1);
        }
    }
}

static void profile_timer_unregister(void) {
    profiling = PROF_ERROR;
    profstop();
}

/* This function clears the links on top level linkers, and clears the needed
 * parts of memory in the index array */
void profstart(int current_thread) {
    recursion_level = 0;
    profiling_thread = current_thread;
    last_pfd = (struct pfd_struct*)0;
    pfds[0].link = 0;
    pfds[0].self_pc = 0;
    memset(indices,0,INDEX_SIZE * sizeof(unsigned short));
    timer_register(
        0, profile_timer_unregister, TIMER_FREQ/10000, profile_timer_tick IF_COP(, 0 ) );
    profiling = PROF_ON;
}

static void write_function_recursive(int fd, struct pfd_struct *pfd, int depth){
    unsigned short link = pfd->link;
    fdprintf(fd,"0x%08lX\t%08ld\t%08ld\t%04d\n", (size_t)pfd->self_pc, 
            pfd->count, pfd->time, depth);
    if (link > 0 && link < NUMPFDS) { 
        write_function_recursive(fd, &pfds[link], ++depth);
    }
}

void profstop() {
    int profiling_exit = profiling;
    int fd = 0;
    int i;
    unsigned short current_index;
    timer_unregister();
    profiling = PROF_OFF;
    fd = open("/profile.out", O_WRONLY|O_CREAT|O_TRUNC);
    if (profiling_exit == PROF_ERROR) {
        fdprintf(fd,"Profiling exited with an error.\n");
        fdprintf(fd,"Overflow or timer stolen most likely.\n");
    }
    fdprintf(fd,"PROFILE_THREAD\tPFDS_USED\n");
    fdprintf(fd,"%08d\t%08d\n", profiling_thread,
        pfds[0].link);
    fdprintf(fd,"FUNCTION_PC\tCALL_COUNT\tTICKS\t\tDEPTH\n");
    for (i = 0; i < INDEX_SIZE; i++) {
        current_index = indices[i];
        if (current_index != 0) { 
            write_function_recursive(fd, &pfds[current_index], 0);
        }
    }
    fdprintf(fd,"DEBUG PROFILE DATA FOLLOWS\n");
    fdprintf(fd,"INDEX\tLOCATION\tSELF_PC\t\tCOUNT\t\tTIME\t\tLINK\tCALLER_IDX\n");
    for (i = 0; i < NUMPFDS; i++) {
        struct pfd_struct *my_last_pfd = &pfds[i];
        if (my_last_pfd->self_pc != 0) {
            fdprintf(fd,
                    "%04d\t0x%08lX\t0x%08lX\t0x%08lX\t0x%08lX\t%04d\t0x%08lX\n",
                    i, (size_t)my_last_pfd, (size_t)my_last_pfd->self_pc,
                    my_last_pfd->count, my_last_pfd->time, my_last_pfd->link,
                    (size_t)my_last_pfd->caller );
        }
    }
    fdprintf(fd,"INDEX_ADDRESS=INDEX\n");
    for (i=0; i < INDEX_SIZE; i++) {
        fdprintf(fd,"%08lX=%04d\n",(size_t)&indices[i],indices[i]);
    }
    close(fd);
}

void __cyg_profile_func_exit(void *self_pc, void *call_site) {
    (void)call_site;
    (void)self_pc;
    /* When we started timing, we set the time to the tick at that time
     * less the time already used in function */
    if (profiling) {
        return;
    }
    profiling = PROF_BUSY;
    {
        register unsigned short my_recursion_level = recursion_level;
        if (my_recursion_level) {
            my_recursion_level--;
            recursion_level = my_recursion_level;
        } else {
            /* This shouldn't be necessary, maybe exit could be called first */
            register struct pfd_struct *my_last_pfd = last_pfd;
            if (my_last_pfd) {
                last_pfd = my_last_pfd->caller;
            }
        }
    }
    profiling = PROF_ON;
}

#define ALLOCATE_PFD(temp) \
    temp = ++pfds[0].link;\
    if (temp >= NUMPFDS) goto overflow; \
    pfd = &pfds[temp];\
    pfd->self_pc = self_pc; pfd->count = 1; pfd->time = 0

void __cyg_profile_func_enter(void *self_pc, void *from_pc) {
    struct pfd_struct *pfd;
    struct pfd_struct *prev_pfd;
    unsigned short *pfd_index_pointer;
    unsigned short pfd_index;

    /* check that we are profiling and that we aren't recursively invoked
     * this is equivalent to 'if (profiling != PROF_ON)' but it's faster */
    if (profiling) {
        return;
    }
    profiling = PROF_BUSY;
    /* A check that the PC is in the code range here wouldn't hurt, but this is
     * logically guaranteed to be a valid address unless the constants are
     * breaking the rules.  */
    pfd_index_pointer = &indices[((size_t)from_pc)&INDEX_MASK];
    pfd_index = *pfd_index_pointer;
    if (pfd_index == 0) {
        /* new caller, allocate new storage */
        ALLOCATE_PFD(pfd_index);
        pfd->link = 0;
        *pfd_index_pointer = pfd_index;
        goto done;
    }
    pfd = &pfds[pfd_index];
    if (pfd->self_pc == self_pc) {
        /* only / most recent function called by this caller, usual case */
        /* increment count, start timing and exit */
        goto found;
    }
    /* collision, bad for performance, look down the list of functions called by
     * colliding PCs */
    for (; /* goto done */; ) {
        pfd_index = pfd->link;
        if (pfd_index == 0) {
            /* no more previously called functions, allocate a new one */
            ALLOCATE_PFD(pfd_index);
            /* this function becomes the new head, link to the old head */
            pfd->link = *pfd_index_pointer;
            /* and set the index to point to this function */
            *pfd_index_pointer = pfd_index;
            /* start timing and exit */
            goto done;
        }
        /* move along the chain */
        prev_pfd = pfd;
        pfd = &pfds[pfd_index];
        if (pfd->self_pc == self_pc) {
            /* found ourself */
            /* Remove me from my old spot */
            prev_pfd->link = pfd->link;
            /* Link to the old head */
            pfd->link = *pfd_index_pointer;
            /* Make me head */
            *pfd_index_pointer = pfd_index;
            /* increment count, start timing and exit */
            goto found;
        }

    }
/* We've found a pfd, increment it */
found:
    ADDQI_L(pfd->count,1);
/* We've (found or created) and updated our pfd, save it and start timing */
done:
    {
        register struct pfd_struct *my_last_pfd = last_pfd;
        if (pfd != my_last_pfd) {
            /* If we are not recursing */
            pfd->caller = my_last_pfd;
            last_pfd = pfd;
        } else {
            ADDQI_L(recursion_level,1);
        }
    }
    /* Start timing this function */
    profiling = PROF_ON;
    return;  /* normal return restores saved registers */

overflow:
    /* this is the same as 'profiling = PROF_ERROR' */
    profiling = PROF_ERROR;
    return;
}