cairo/util/cairo-trace/trace.c

4481 lines
101 KiB
C
Raw Normal View History

/* cairo-trace - a utility to record and replay calls to the Cairo library.
*
* Copyright © 2008 Chris Wilson
*
* 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 3 of the License, 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. If not, see <http://www.gnu.org/licenses/>.
*/
#define _GNU_SOURCE
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <dlfcn.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <string.h>
#include <sys/types.h>
#include <unistd.h>
#include <errno.h>
#include <pthread.h>
#include <zlib.h>
#include <math.h>
#include <locale.h> /* for locale independent %f printing */
#include <ctype.h>
#include <assert.h>
#include <cairo.h>
#if CAIRO_HAS_FT_FONT
# include <cairo-ft.h>
#endif
#ifndef CAIRO_TRACE_OUTDIR
#define CAIRO_TRACE_OUTDIR "."
#endif
#if HAVE_BYTESWAP_H
# include <byteswap.h>
#endif
#ifndef bswap_16
# define bswap_16(p) \
(((((uint16_t)(p)) & 0x00ff) << 8) | \
(((uint16_t)(p)) >> 8));
#endif
#ifndef bswap_32
# define bswap_32(p) \
(((((uint32_t)(p)) & 0x000000ff) << 24) | \
((((uint32_t)(p)) & 0x0000ff00) << 8) | \
((((uint32_t)(p)) & 0x00ff0000) >> 8) | \
((((uint32_t)(p))) >> 24));
#endif
#include "lookup-symbol.h"
/* Reverse the bits in a byte with 7 operations (no 64-bit):
* Devised by Sean Anderson, July 13, 2001.
* Source: http://graphics.stanford.edu/~seander/bithacks.html#ReverseByteWith32Bits
*/
#define CAIRO_BITSWAP8(c) ((((c) * 0x0802LU & 0x22110LU) | ((c) * 0x8020LU & 0x88440LU)) * 0x10101LU >> 16)
#if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ > 4)
#define CAIRO_PRINTF_FORMAT(fmt_index, va_index) \
__attribute__((__format__(__printf__, fmt_index, va_index)))
#else
#define CAIRO_PRINTF_FORMAT(fmt_index, va_index)
#endif
/* XXX implement manual vprintf so that the user can control precision of
* printed numbers.
*/
static void *_dlhandle = RTLD_NEXT;
#define DLCALL(name, args...) ({ \
static typeof (&name) name##_real; \
if (name##_real == NULL) { \
name##_real = dlsym (_dlhandle, #name); \
if (name##_real == NULL && _dlhandle == RTLD_NEXT) { \
_dlhandle = dlopen ("libcairo.so", RTLD_LAZY); \
name##_real = dlsym (_dlhandle, #name); \
assert (name##_real != NULL); \
} \
} \
(*name##_real) (args); \
})
#define ARRAY_SIZE(a) (sizeof (a) / sizeof (a[0]))
#define ARRAY_LENGTH(a) ((int) ARRAY_SIZE(a))
#if SIZEOF_VOID_P == 4
#define PTR_SHIFT 2
#elif SIZEOF_VOID_P == 8
#define PTR_SHIFT 3
#else
#error Unexpected pointer size
#endif
#define BUCKET(b, ptr) (((unsigned long) (ptr) >> PTR_SHIFT) % ARRAY_LENGTH (b))
#if defined(__GNUC__) && (__GNUC__ > 2) && defined(__OPTIMIZE__)
#define _BOOLEAN_EXPR(expr) \
__extension__ ({ \
int _boolean_var_; \
if (expr) \
_boolean_var_ = 1; \
else \
_boolean_var_ = 0; \
_boolean_var_; \
})
#define LIKELY(expr) (__builtin_expect (_BOOLEAN_EXPR(expr), 1))
#define UNLIKELY(expr) (__builtin_expect (_BOOLEAN_EXPR(expr), 0))
#else
#define LIKELY(expr) (expr)
#define UNLIKELY(expr) (expr)
#endif
typedef struct _object Object;
typedef struct _type Type;
struct _object {
const void *addr;
Type *type;
unsigned long int token;
int width, height;
bool foreign;
bool defined;
int operand;
void *data;
void (*destroy)(void *);
Object *next, *prev;
};
struct _type {
const char *name;
enum operand_type {
NONE,
SURFACE,
CONTEXT,
FONT_FACE,
PATTERN,
SCALED_FONT,
_N_OP_TYPES
} op_type;
const char *op_code;
pthread_mutex_t mutex;
struct _bitmap {
unsigned long int min;
unsigned long int count;
unsigned int map[64];
struct _bitmap *next;
} map;
Object *objects[607];
Type *next;
};
static struct _type_table {
pthread_mutex_t mutex;
Type *op_types[_N_OP_TYPES];
} Types;
static FILE *logfile;
static bool _flush;
static bool _error;
static bool _line_info;
static bool _mark_dirty;
static const cairo_user_data_key_t destroy_key;
#if __GNUC__ >= 3
#define _emit_line_info() do { \
if (_line_info && _write_lock ()) { \
void *addr = __builtin_return_address(0); \
char caller[1024]; \
_trace_printf ("%% %s() called by %s\n", __FUNCTION__, \
lookup_symbol (caller, sizeof (caller), addr)); \
_write_unlock (); \
} \
} while (0)
#else
#define _emit_line_info()
#endif
static void
_type_release_token (Type *t, unsigned long int token)
{
struct _bitmap *b, **prev = NULL;
b = &t->map;
while (b != NULL) {
if (token < b->min + sizeof (b->map) * CHAR_BIT) {
unsigned int bit, elem;
token -= b->min;
elem = token / (sizeof (b->map[0]) * CHAR_BIT);
bit = token % (sizeof (b->map[0]) * CHAR_BIT);
b->map[elem] &= ~(1 << bit);
if (! --b->count && prev) {
*prev = b->next;
free (b);
}
return;
}
prev = &b->next;
b = b->next;
}
}
static unsigned long int
_type_next_token (Type *t)
{
struct _bitmap *b, *bb, **prev = NULL;
unsigned long int min = 0;
b = &t->map;
while (b != NULL) {
if (b->min != min)
break;
if (b->count < sizeof (b->map) * CHAR_BIT) {
unsigned int n, m, bit;
for (n = 0; n < ARRAY_SIZE (b->map); n++) {
if (b->map[n] == (unsigned int) -1)
continue;
for (m=0, bit=1; m<sizeof (b->map[0])*CHAR_BIT; m++, bit<<=1) {
if ((b->map[n] & bit) == 0) {
b->map[n] |= bit;
b->count++;
return n * sizeof (b->map[0])*CHAR_BIT + m + b->min;
}
}
}
}
min += sizeof (b->map) * CHAR_BIT;
prev = &b->next;
b = b->next;
}
bb = malloc (sizeof (struct _bitmap));
*prev = bb;
bb->next = b;
bb->min = min;
bb->count = 1;
bb->map[0] = 0x1;
memset (bb->map + 1, 0, sizeof (bb->map) - sizeof (bb->map[0]));
return min;
}
static void
_object_destroy (Object *obj)
{
int bucket;
pthread_mutex_lock (&obj->type->mutex);
bucket = BUCKET (obj->type->objects, obj->addr);
_type_release_token (obj->type, obj->token);
if (obj->prev != NULL)
obj->prev->next = obj->next;
else
obj->type->objects[bucket] = obj->next;
if (obj->next != NULL)
obj->next->prev = obj->prev;
pthread_mutex_unlock (&obj->type->mutex);
if (obj->data != NULL && obj->destroy != NULL)
obj->destroy (obj->data);
free (obj);
}
static void
_type_create (const char *typename,
enum operand_type op_type,
const char *op_code)
{
Type *t;
pthread_mutex_lock (&Types.mutex);
t = malloc (sizeof (Type));
t->name = typename;
t->op_type = op_type;
t->op_code = op_code;
pthread_mutex_init (&t->mutex, NULL);
t->map.min = 0;
t->map.count = 0;
memset (t->map.map, 0, sizeof (t->map.map));
t->map.next = NULL;
memset (t->objects, 0, sizeof (t->objects));
t->next = NULL;
Types.op_types[op_type] = t;
pthread_mutex_unlock (&Types.mutex);
}
static Type *
_get_type (enum operand_type type)
{
return Types.op_types[type];
}
static void
_type_destroy (Type *t)
{
int n;
struct _bitmap *b;
for (n = 0; n < ARRAY_LENGTH (t->objects); n++) {
Object *obj = t->objects[n];
while (obj != NULL) {
Object *next = obj->next;
_object_destroy (obj);
obj = next;
}
}
b = t->map.next;
while (b != NULL) {
struct _bitmap *next = b->next;
free (b);
b = next;
}
pthread_mutex_destroy (&t->mutex);
free (t);
}
static Object *
_type_get_object (Type *type, const void *ptr)
{
Object *obj;
int bucket = BUCKET (type->objects, ptr);
for (obj = type->objects[bucket]; obj != NULL; obj = obj->next) {
if (obj->addr == ptr) {
if (obj->prev != NULL) { /* mru */
obj->prev->next = obj->next;
if (obj->next != NULL)
obj->next->prev = obj->prev;
obj->prev = NULL;
type->objects[bucket]->prev = obj;
obj->next = type->objects[bucket];
type->objects[bucket] = obj;
}
return obj;
}
}
return NULL;
}
static Object *
_object_create (Type *type, const void *ptr)
{
Object *obj;
int bucket = BUCKET (type->objects, ptr);
obj = malloc (sizeof (Object));
obj->defined = false;
obj->foreign = false;
obj->operand = -1;
obj->type = type;
obj->addr = ptr;
obj->token = _type_next_token (type);
obj->data = NULL;
obj->destroy = NULL;
obj->prev = NULL;
obj->next = type->objects[bucket];
if (type->objects[bucket] != NULL)
type->objects[bucket]->prev = obj;
type->objects[bucket] = obj;
return obj;
}
static void __attribute__ ((constructor))
_init_trace (void)
{
pthread_mutex_init (&Types.mutex, NULL);
_type_create ("unclassed", NONE, "");
_type_create ("cairo_t", CONTEXT, "c");
_type_create ("cairo_font_face_t", FONT_FACE, "f");
_type_create ("cairo_pattern_t", PATTERN, "p");
_type_create ("cairo_scaled_font_t", SCALED_FONT, "sf");
_type_create ("cairo_surface_t", SURFACE, "s");
}
static void
_close_trace (void)
{
if (logfile != NULL) {
fclose (logfile);
logfile = NULL;
}
}
static void __attribute__ ((destructor))
_fini_trace (void)
{
int n;
_close_trace ();
for (n = 0; n < ARRAY_LENGTH (Types.op_types); n++) {
if (Types.op_types[n]) {
_type_destroy (Types.op_types[n]);
Types.op_types[n] = NULL;
}
}
pthread_mutex_destroy (&Types.mutex);
}
/* Format a double in a locale independent way and trim trailing
* zeros. Based on code from Alex Larson <alexl@redhat.com>.
* http://mail.gnome.org/archives/gtk-devel-list/2001-October/msg00087.html
*
* The code in the patch is copyright Red Hat, Inc under the LGPL.
*/
#define SIGNIFICANT_DIGITS_AFTER_DECIMAL 6
static void
_trace_dtostr (char *buffer, size_t size, double d)
{
struct lconv *locale_data;
const char *decimal_point;
int decimal_point_len;
char *p;
int decimal_len;
int num_zeros, decimal_digits;
/* Omit the minus sign from negative zero. */
if (d == 0.0)
d = 0.0;
locale_data = localeconv ();
decimal_point = locale_data->decimal_point;
decimal_point_len = strlen (decimal_point);
/* Using "%f" to print numbers less than 0.1 will result in
* reduced precision due to the default 6 digits after the
* decimal point.
*
* For numbers is < 0.1, we print with maximum precision and count
* the number of zeros between the decimal point and the first
* significant digit. We then print the number again with the
* number of decimal places that gives us the required number of
* significant digits. This ensures the number is correctly
* rounded.
*/
if (fabs (d) >= 0.1) {
snprintf (buffer, size, "%f", d);
} else {
snprintf (buffer, size, "%.18f", d);
p = buffer;
if (*p == '+' || *p == '-')
p++;
while (isdigit (*p))
p++;
if (strncmp (p, decimal_point, decimal_point_len) == 0)
p += decimal_point_len;
num_zeros = 0;
while (*p++ == '0')
num_zeros++;
decimal_digits = num_zeros + SIGNIFICANT_DIGITS_AFTER_DECIMAL;
if (decimal_digits < 18)
snprintf (buffer, size, "%.*f", decimal_digits, d);
}
p = buffer;
if (*p == '+' || *p == '-')
p++;
while (isdigit (*p))
p++;
if (strncmp (p, decimal_point, decimal_point_len) == 0) {
*p = '.';
decimal_len = strlen (p + decimal_point_len);
memmove (p + 1, p + decimal_point_len, decimal_len);
p[1 + decimal_len] = 0;
/* Remove trailing zeros and decimal point if possible. */
for (p = p + decimal_len; *p == '0'; p--)
*p = 0;
if (*p == '.') {
*p = 0;
p--;
}
}
}
enum {
LENGTH_MODIFIER_LONG = 0x100
};
/* Here's a limited reimplementation of printf. The reason for doing
* this is primarily to special case handling of doubles. We want
* locale independent formatting of doubles and we want to trim
* trailing zeros. This is handled by dtostr() above, and the code
* below handles everything else by calling snprintf() to do the
* formatting. This functionality is only for internal use and we
* only implement the formats we actually use.
*/
static void CAIRO_PRINTF_FORMAT(1, 0)
_trace_vprintf (const char *fmt, va_list ap)
{
#define SINGLE_FMT_BUFFER_SIZE 32
char buffer[512], single_fmt[SINGLE_FMT_BUFFER_SIZE];
int single_fmt_length;
char *p;
const char *f, *start;
int length_modifier, width;
cairo_bool_t var_width;
int ret_ignored;
f = fmt;
p = buffer;
while (*f != '\0') {
if (*f != '%') {
*p++ = *f++;
continue;
}
start = f;
f++;
if (*f == '0')
f++;
var_width = 0;
if (*f == '*') {
var_width = 1;
f++;
}
while (isdigit (*f))
f++;
length_modifier = 0;
if (*f == 'l') {
length_modifier = LENGTH_MODIFIER_LONG;
f++;
}
/* The only format strings exist in the cairo implementation
* itself. So there's an internal consistency problem if any
* of them is larger than our format buffer size. */
single_fmt_length = f - start + 1;
/* Reuse the format string for this conversion. */
memcpy (single_fmt, start, single_fmt_length);
single_fmt[single_fmt_length] = '\0';
/* Flush contents of buffer before snprintf()'ing into it. */
ret_ignored = fwrite (buffer, 1, p-buffer, logfile);
/* We group signed and unsigned together in this switch, the
* only thing that matters here is the size of the arguments,
* since we're just passing the data through to sprintf(). */
switch (*f | length_modifier) {
case '%':
buffer[0] = *f;
buffer[1] = 0;
break;
case 'd':
case 'u':
case 'o':
case 'x':
case 'X':
if (var_width) {
width = va_arg (ap, int);
snprintf (buffer, sizeof buffer,
single_fmt, width, va_arg (ap, int));
} else {
snprintf (buffer, sizeof buffer, single_fmt, va_arg (ap, int));
}
break;
case 'd' | LENGTH_MODIFIER_LONG:
case 'u' | LENGTH_MODIFIER_LONG:
case 'o' | LENGTH_MODIFIER_LONG:
case 'x' | LENGTH_MODIFIER_LONG:
case 'X' | LENGTH_MODIFIER_LONG:
if (var_width) {
width = va_arg (ap, int);
snprintf (buffer, sizeof buffer,
single_fmt, width, va_arg (ap, long int));
} else {
snprintf (buffer, sizeof buffer,
single_fmt, va_arg (ap, long int));
}
break;
case 's':
snprintf (buffer, sizeof buffer,
single_fmt, va_arg (ap, const char *));
break;
case 'f':
case 'g':
_trace_dtostr (buffer, sizeof buffer, va_arg (ap, double));
break;
case 'c':
buffer[0] = va_arg (ap, int);
buffer[1] = 0;
break;
default:
break;
}
p = buffer + strlen (buffer);
f++;
}
ret_ignored = fwrite (buffer, 1, p-buffer, logfile);
}
static void CAIRO_PRINTF_FORMAT(1, 2)
_trace_printf (const char *fmt, ...)
{
va_list ap;
va_start (ap, fmt);
_trace_vprintf (fmt, ap);
va_end (ap);
}
static void
get_prog_name (char *buf, int length)
{
char *slash;
FILE *file;
memset (buf, 0, length);
if (length == 0)
return;
file = fopen ("/proc/self/cmdline", "rb");
if (file != NULL) {
fgets (buf, length, file);
fclose (file);
} else {
char const *name = getenv ("CAIRO_TRACE_PROG_NAME");
if (name != NULL) {
strncpy (buf, name, length-1);
}
}
slash = strrchr (buf, '/');
if (slash != NULL) {
size_t len = strlen (slash+1);
memmove (buf, slash+1, len+1);
}
}
static void
_emit_header (void)
{
char name[4096] = "";
get_prog_name (name, sizeof (name));
_trace_printf ("%%!CairoScript - %s\n", name);
_trace_printf ("%%*** Warning CairoScript is still a new tracing format, and is subject to change.\n");
}
static bool
_init_logfile (void)
{
static bool initialized;
const char *filename;
const char *env;
if (initialized)
return logfile != NULL;
initialized = true;
env = getenv ("CAIRO_TRACE_FLUSH");
if (env != NULL)
_flush = atoi (env);
_line_info = true;
env = getenv ("CAIRO_TRACE_LINE_INFO");
if (env != NULL)
_line_info = atoi (env);
_mark_dirty = true;
env = getenv ("CAIRO_TRACE_MARK_DIRTY");
if (env != NULL)
_mark_dirty = atoi (env);
filename = getenv ("CAIRO_TRACE_FD");
if (filename != NULL) {
int fd = atoi (filename);
if (fd == -1)
return false;
logfile = fdopen (fd, "w");
if (logfile == NULL) {
fprintf (stderr, "Failed to open trace file descriptor '%s': %s\n",
filename, strerror (errno));
return false;
}
setenv ("CAIRO_TRACE_FD", "-1", 1);
goto done;
}
filename = getenv ("CAIRO_TRACE_OUTFILE_EXACT");
if (filename == NULL) {
char buf[4096], name[4096] = "";
filename = getenv ("CAIRO_TRACE_OUTDIR");
if (filename == NULL)
filename = CAIRO_TRACE_OUTDIR;
get_prog_name (name, sizeof (name));
if (*name == '\0')
strcpy (name, "cairo-trace.dat");
snprintf (buf, sizeof (buf), "%s/%s.%d.trace",
filename, name, getpid());
filename = buf;
setenv ("CAIRO_TRACE_FD", "-1", 1);
}
logfile = fopen (filename, "wb");
if (logfile == NULL) {
fprintf (stderr, "Failed to open trace file '%s': %s\n",
filename, strerror (errno));
return false;
}
fprintf (stderr, "cairo-trace: Recording cairo trace data to %s\n",
filename);
done:
atexit (_close_trace);
_emit_header ();
return true;
}
static bool
_write_lock (void)
{
if (_error)
return false;
if (! _init_logfile ())
return false;
#if HAVE_FLOCKFILE && HAVE_FUNLOCKFILE
flockfile (logfile);
#endif
return true;
}
static void
_write_unlock (void)
{
if (logfile == NULL)
return;
#if HAVE_FLOCKFILE && HAVE_FUNLOCKFILE
funlockfile (logfile);
#endif
if (_flush)
fflush (logfile);
}
static Object *
_type_object_create (enum operand_type op_type, const void *ptr)
{
Type *type;
Object *obj;
type = _get_type (op_type);
pthread_mutex_lock (&type->mutex);
obj = _object_create (type, ptr);
pthread_mutex_unlock (&type->mutex);
return obj;
}
static Object *
_get_object (enum operand_type op_type, const void *ptr)
{
Type *type;
Object *obj;
type = _get_type (op_type);
pthread_mutex_lock (&type->mutex);
obj = _type_get_object (type, ptr);
pthread_mutex_unlock (&type->mutex);
return obj;
}
static Object *current_object[2048]; /* XXX limit operand stack */
static int current_stack_depth;
static void
ensure_operands (int num_operands)
{
if (current_stack_depth < num_operands)
{
int n;
fprintf (stderr, "Operand stack underflow!\n");
for (n = 0; n < current_stack_depth; n++) {
Object *obj = current_object[n];
fprintf (stderr, " [%3d] = %s%ld\n",
n, obj->type->op_code, obj->token);
}
abort ();
}
}
static void
_consume_operand (void)
{
Object *obj;
ensure_operands (1);
obj = current_object[--current_stack_depth];
if (! obj->defined) {
_trace_printf ("dup /%s%ld exch def\n",
obj->type->op_code,
obj->token);
obj->defined = true;
}
obj->operand = -1;
}
static void
_exch_operands (void)
{
Object *tmp;
ensure_operands (2);
tmp = current_object[current_stack_depth-1];
tmp->operand--;
current_object[current_stack_depth-1] = current_object[current_stack_depth-2];
current_object[current_stack_depth-2] = tmp;
tmp = current_object[current_stack_depth-1];
tmp->operand++;
}
static bool
_pop_operands_to_object (Object *obj)
{
if (obj->operand == -1)
return false;
if (obj->operand == current_stack_depth - 2) {
_exch_operands ();
_trace_printf ("exch ");
return true;
}
while (current_stack_depth > obj->operand + 1) {
Object *c_obj;
ensure_operands (1);
c_obj = current_object[--current_stack_depth];
c_obj->operand = -1;
if (! c_obj->defined) {
_trace_printf ("/%s%ld exch def\n",
c_obj->type->op_code,
c_obj->token);
c_obj->defined = true;
} else {
_trace_printf ("pop %% %s%ld\n",
c_obj->type->op_code, c_obj->token);
}
}
return true;
}
static bool
_pop_operands_to (enum operand_type t, const void *ptr)
{
return _pop_operands_to_object (_get_object (t, ptr));
}
static bool
_is_current_object (Object *obj, int depth)
{
if (current_stack_depth <= depth)
return false;
return current_object[current_stack_depth-depth-1] == obj;
}
static bool
_is_current (enum operand_type type, const void *ptr, int depth)
{
return _is_current_object (_get_object (type, ptr), depth);
}
static void
_push_operand (enum operand_type t, const void *ptr)
{
Object *obj = _get_object (t, ptr);
if (current_stack_depth == ARRAY_LENGTH (current_object))
{
int n;
fprintf (stderr, "Operand stack overflow!\n");
for (n = 0; n < current_stack_depth; n++) {
obj = current_object[n];
fprintf (stderr, " [%3d] = %s%ld\n",
n, obj->type->op_code, obj->token);
}
abort ();
}
obj->operand = current_stack_depth;
current_object[current_stack_depth++] = obj;
}
static void
_object_remove (Object *obj)
{
if (obj->operand != -1) {
ensure_operands (1);
if (obj->operand == current_stack_depth - 1) {
_trace_printf ("pop %% %s%ld destroyed\n",
obj->type->op_code, obj->token);
} else if (obj->operand == current_stack_depth - 2) {
_exch_operands ();
_trace_printf ("exch pop %% %s%ld destroyed\n",
obj->type->op_code, obj->token);
} else {
int n;
_trace_printf ("%d -1 roll pop %% %s%ld destroyed\n",
current_stack_depth - obj->operand,
obj->type->op_code, obj->token);
for (n = obj->operand; n < current_stack_depth - 1; n++) {
current_object[n] = current_object[n+1];
current_object[n]->operand = n;
}
}
current_stack_depth--;
}
}
static void
_object_undef (void *ptr)
{
Object *obj = ptr;
if (_write_lock ()) {
_object_remove (obj);
if (obj->defined) {
_trace_printf ("/%s%ld undef\n",
obj->type->op_code, obj->token);
}
_write_unlock ();
}
_object_destroy (obj);
}
static long
_create_context_id (cairo_t *cr)
{
Object *obj;
obj = _get_object (CONTEXT, cr);
if (obj == NULL) {
obj = _type_object_create (CONTEXT, cr);
DLCALL (cairo_set_user_data,
cr, &destroy_key, obj, _object_undef);
}
return obj->token;
}
static long
_get_id (enum operand_type op_type, const void *ptr)
{
Object *obj;
obj = _get_object (op_type, ptr);
if (obj == NULL) {
if (logfile != NULL) {
_trace_printf ("%% Unknown object of type %s, trace is incomplete.",
_get_type (op_type)->name);
}
_error = true;
return -1;
}
return obj->token;
}
static bool
_has_id (enum operand_type op_type, const void *ptr)
{
return _get_object (op_type, ptr) != NULL;
}
static long
_get_context_id (cairo_t *cr)
{
return _get_id (CONTEXT, cr);
}
static long
_create_font_face_id (cairo_font_face_t *font_face)
{
Object *obj;
obj = _get_object (FONT_FACE, font_face);
if (obj == NULL) {
obj = _type_object_create (FONT_FACE, font_face);
DLCALL (cairo_font_face_set_user_data,
font_face, &destroy_key, obj, _object_undef);
}
return obj->token;
}
static long
_get_font_face_id (cairo_font_face_t *font_face)
{
return _get_id (FONT_FACE, font_face);
}
static bool
_has_font_face_id (cairo_font_face_t *font_face)
{
return _has_id (FONT_FACE, font_face);
}
static void
_emit_font_face_id (cairo_font_face_t *font_face)
{
Object *obj = _get_object (FONT_FACE, font_face);
if (obj == NULL) {
_trace_printf ("null ");
} else {
if (obj->defined) {
_trace_printf ("f%ld ", obj->token);
} else {
_trace_printf ("%d index ", current_stack_depth - obj->operand - 1);
}
}
}
static bool
_has_pattern_id (cairo_pattern_t *pattern)
{
return _has_id (PATTERN, pattern);
}
static long
_create_pattern_id (cairo_pattern_t *pattern)
{
Object *obj;
obj = _get_object (PATTERN, pattern);
if (obj == NULL) {
obj = _type_object_create (PATTERN, pattern);
DLCALL (cairo_pattern_set_user_data,
pattern, &destroy_key, obj, _object_undef);
}
return obj->token;
}
static long
_get_pattern_id (cairo_pattern_t *pattern)
{
return _get_id (PATTERN, pattern);
}
static void
_emit_pattern_id (cairo_pattern_t *pattern)
{
Object *obj = _get_object (PATTERN, pattern);
if (obj == NULL) {
_trace_printf ("null ");
} else {
if (obj->defined) {
_trace_printf ("p%ld ", obj->token);
} else {
_trace_printf ("%d index ",
current_stack_depth - obj->operand - 1);
}
}
}
static long
_create_scaled_font_id (cairo_scaled_font_t *font)
{
Object *obj;
obj = _get_object (SCALED_FONT, font);
if (obj == NULL) {
obj = _type_object_create (SCALED_FONT, font);
DLCALL (cairo_scaled_font_set_user_data,
font, &destroy_key, obj, _object_undef);
}
return obj->token;
}
static long
_get_scaled_font_id (const cairo_scaled_font_t *font)
{
return _get_id (SCALED_FONT, font);
}
static bool
_has_scaled_font_id (const cairo_scaled_font_t *font)
{
return _has_id (SCALED_FONT, font);
}
static bool
_has_surface_id (const cairo_surface_t *surface)
{
return _has_id (SURFACE, surface);
}
static long
_create_surface_id (cairo_surface_t *surface)
{
Object *obj;
obj = _get_object (SURFACE, surface);
if (obj == NULL) {
obj = _type_object_create (SURFACE, surface);
DLCALL (cairo_surface_set_user_data,
surface, &destroy_key, obj, _object_undef);
}
return obj->token;
}
static long
_get_surface_id (cairo_surface_t *surface)
{
return _get_id (SURFACE, surface);
}
static bool
_matrix_is_identity (const cairo_matrix_t *m)
{
return m->xx == 1. && m->yx == 0. &&
m->xy == 0. && m->yy == 1. &&
m->x0 == 0. && m->y0 == 0.;
}
#define BUFFER_SIZE 16384
struct _data_stream {
z_stream zlib_stream;
unsigned char zin_buf[BUFFER_SIZE];
unsigned char zout_buf[BUFFER_SIZE];
unsigned char four_tuple[4];
int base85_pending;
};
static void
_write_zlib_data_start (struct _data_stream *stream)
{
stream->zlib_stream.zalloc = Z_NULL;
stream->zlib_stream.zfree = Z_NULL;
stream->zlib_stream.opaque = Z_NULL;
deflateInit (&stream->zlib_stream, Z_DEFAULT_COMPRESSION);
stream->zlib_stream.next_in = stream->zin_buf;
stream->zlib_stream.avail_in = 0;
stream->zlib_stream.next_out = stream->zout_buf;
stream->zlib_stream.avail_out = BUFFER_SIZE;
}
static void
_write_base85_data_start (struct _data_stream *stream)
{
stream->base85_pending = 0;
}
static bool
_expand_four_tuple_to_five (unsigned char four_tuple[4],
unsigned char five_tuple[5])
{
uint32_t value;
int digit, i;
bool all_zero = true;
value = four_tuple[0] << 24 |
four_tuple[1] << 16 |
four_tuple[2] << 8 |
four_tuple[3] << 0;
for (i = 0; i < 5; i++) {
digit = value % 85;
if (digit != 0 && all_zero)
all_zero = false;
five_tuple[4-i] = digit + 33;
value = value / 85;
}
return all_zero;
}
static void
_write_base85_data (struct _data_stream *stream,
const unsigned char *data,
unsigned int length)
{
unsigned char five_tuple[5];
int ret;
while (length--) {
stream->four_tuple[stream->base85_pending++] = *data++;
if (stream->base85_pending == 4) {
if (_expand_four_tuple_to_five (stream->four_tuple, five_tuple))
ret = fwrite ("z", 1, 1, logfile);
else
ret = fwrite (five_tuple, 5, 1, logfile);
stream->base85_pending = 0;
}
}
}
static void
_write_zlib_data (struct _data_stream *stream, bool flush)
{
bool finished;
do {
int ret = deflate (&stream->zlib_stream, flush ? Z_FINISH : Z_NO_FLUSH);
if (flush || stream->zlib_stream.avail_out == 0) {
_write_base85_data (stream,
stream->zout_buf,
BUFFER_SIZE - stream->zlib_stream.avail_out);
stream->zlib_stream.next_out = stream->zout_buf;
stream->zlib_stream.avail_out = BUFFER_SIZE;
}
finished = true;
if (stream->zlib_stream.avail_in != 0)
finished = false;
if (flush && ret != Z_STREAM_END)
finished = false;
} while (! finished);
stream->zlib_stream.next_in = stream->zin_buf;
}
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
static void
_write_data_start (struct _data_stream *stream, uint32_t len)
{
_write_zlib_data_start (stream);
_write_base85_data_start (stream);
_trace_printf ("<|");
_write_base85_data (stream, (unsigned char *) &len, sizeof (len));
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
}
static void
_write_data (struct _data_stream *stream,
const void *data,
unsigned int length)
{
unsigned int count;
const unsigned char *p = data;
while (length) {
count = length;
if (count > BUFFER_SIZE - stream->zlib_stream.avail_in)
count = BUFFER_SIZE - stream->zlib_stream.avail_in;
memcpy (stream->zin_buf + stream->zlib_stream.avail_in, p, count);
p += count;
stream->zlib_stream.avail_in += count;
length -= count;
if (stream->zlib_stream.avail_in == BUFFER_SIZE)
_write_zlib_data (stream, false);
}
}
static void
_write_zlib_data_end (struct _data_stream *stream)
{
_write_zlib_data (stream, true);
deflateEnd (&stream->zlib_stream);
}
static void
_write_base85_data_end (struct _data_stream *stream)
{
unsigned char five_tuple[5];
int ret;
if (stream->base85_pending) {
memset (stream->four_tuple + stream->base85_pending,
0, 4 - stream->base85_pending);
_expand_four_tuple_to_five (stream->four_tuple, five_tuple);
ret = fwrite (five_tuple, stream->base85_pending+1, 1, logfile);
}
}
static void
_write_data_end (struct _data_stream *stream)
{
_write_zlib_data_end (stream);
_write_base85_data_end (stream);
_trace_printf ("~>");
}
static void
_emit_data (const void *data, unsigned int length)
{
struct _data_stream stream;
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_write_data_start (&stream, length);
_write_data (&stream, data, length);
_write_data_end (&stream);
}
static const char *
_format_to_string (cairo_format_t format)
{
#define f(name) case CAIRO_FORMAT_ ## name: return #name
switch (format) {
f(ARGB32);
f(RGB24);
f(A8);
f(A1);
}
#undef f
return "UNKNOWN_FORMAT";
}
static const char *
_status_to_string (cairo_status_t status)
{
#define f(name) case CAIRO_STATUS_ ## name: return "STATUS_" #name
switch (status) {
f(SUCCESS);
f(NO_MEMORY);
f(INVALID_RESTORE);
f(INVALID_POP_GROUP);
f(NO_CURRENT_POINT);
f(INVALID_MATRIX);
f(INVALID_STATUS);
f(NULL_POINTER);
f(INVALID_STRING);
f(INVALID_PATH_DATA);
f(READ_ERROR);
f(WRITE_ERROR);
f(SURFACE_FINISHED);
f(SURFACE_TYPE_MISMATCH);
f(PATTERN_TYPE_MISMATCH);
f(INVALID_CONTENT);
f(INVALID_FORMAT);
f(INVALID_VISUAL);
f(FILE_NOT_FOUND);
f(INVALID_DASH);
f(INVALID_DSC_COMMENT);
f(INVALID_INDEX);
f(CLIP_NOT_REPRESENTABLE);
f(TEMP_FILE_ERROR);
f(INVALID_STRIDE);
f(FONT_TYPE_MISMATCH);
f(USER_FONT_IMMUTABLE);
f(USER_FONT_ERROR);
f(NEGATIVE_COUNT);
f(INVALID_CLUSTERS);
f(INVALID_SLANT);
f(INVALID_WEIGHT);
f(INVALID_SIZE);
f(USER_FONT_NOT_IMPLEMENTED);
case CAIRO_STATUS_LAST_STATUS:
break;
}
return "UNKNOWN_STATUS";
#undef f
}
static void CAIRO_PRINTF_FORMAT(2, 3)
_emit_image (cairo_surface_t *image,
const char *info,
...)
{
int stride, row, width, height;
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
uint32_t len;
cairo_format_t format;
uint8_t row_stack[BUFFER_SIZE];
uint8_t *rowdata;
uint8_t *data;
struct _data_stream stream;
cairo_status_t status;
status = DLCALL (cairo_surface_status, image);
if (status) {
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf ("<< /status //%s >> image",
_status_to_string (status));
return;
}
width = DLCALL (cairo_image_surface_get_width, image);
height = DLCALL (cairo_image_surface_get_height, image);
stride = DLCALL (cairo_image_surface_get_stride, image);
format = DLCALL (cairo_image_surface_get_format, image);
data = DLCALL (cairo_image_surface_get_data, image);
_trace_printf ("dict\n"
" /width %d set\n"
" /height %d set\n"
" /format //%s set\n",
width, height,
_format_to_string (format));
if (info != NULL) {
va_list ap;
va_start (ap, info);
_trace_vprintf (info, ap);
va_end (ap);
}
if (DLCALL (cairo_version) >= CAIRO_VERSION_ENCODE (1, 9, 0)) {
const char *mime_types[] = {
CAIRO_MIME_TYPE_JPEG,
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
CAIRO_MIME_TYPE_JP2,
CAIRO_MIME_TYPE_PNG,
NULL
}, **mime_type;
for (mime_type = mime_types; *mime_type; mime_type++) {
const unsigned char *mime_data;
unsigned int mime_length;
DLCALL (cairo_surface_get_mime_data,
image, *mime_type, &mime_data, &mime_length);
if (mime_data != NULL) {
_trace_printf (" /mime-type (%s) set\n"
" /source <~",
*mime_type);
_write_base85_data_start (&stream);
_write_base85_data (&stream, mime_data, mime_length);
_write_base85_data_end (&stream);
_trace_printf ("~> set\n"
" image");
return;
}
}
}
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
switch (format) {
case CAIRO_FORMAT_A1: len = (width + 7)/8; break;
case CAIRO_FORMAT_A8: len = width; break;
case CAIRO_FORMAT_RGB24: len = 3*width; break;
default:
case CAIRO_FORMAT_ARGB32: len = 4*width; break;
}
_trace_printf (" /source ");
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_write_data_start (&stream, len * height);
#ifdef WORDS_BIGENDIAN
switch (format) {
case CAIRO_FORMAT_A1:
for (row = height; row--; ) {
_write_data (&stream, data, (width+7)/8);
data += stride;
}
break;
case CAIRO_FORMAT_A8:
for (row = height; row--; ) {
_write_data (&stream, data, width);
data += stride;
}
break;
case CAIRO_FORMAT_RGB24:
for (row = height; row--; ) {
int col;
rowdata = data;
for (col = width; col--; ) {
_write_data (&stream, rowdata, 3);
rowdata+=4;
}
data += stride;
}
break;
case CAIRO_FORMAT_ARGB32:
for (row = height; row--; ) {
_write_data (&stream, data, 4*width);
data += stride;
}
break;
default:
break;
}
#else
if (stride > ARRAY_LENGTH (row_stack)) {
rowdata = malloc (stride);
if (rowdata == NULL)
goto BAIL;
} else
rowdata = row_stack;
switch (format) {
case CAIRO_FORMAT_A1:
for (row = height; row--; ) {
int col;
for (col = 0; col < (width + 7)/8; col++)
rowdata[col] = CAIRO_BITSWAP8 (data[col]);
_write_data (&stream, rowdata, (width+7)/8);
data += stride;
}
break;
case CAIRO_FORMAT_A8:
for (row = height; row--; ) {
_write_data (&stream, rowdata, width);
data += stride;
}
break;
case CAIRO_FORMAT_RGB24:
for (row = height; row--; ) {
uint8_t *src = data;
int col;
for (col = 0; col < width; col++) {
rowdata[3*col+2] = *src++;
rowdata[3*col+1] = *src++;
rowdata[3*col+0] = *src++;
src++;
}
_write_data (&stream, rowdata, 3*width);
data += stride;
}
break;
case CAIRO_FORMAT_ARGB32:
for (row = height; row--; ) {
uint32_t *src = (uint32_t *) data;
uint32_t *dst = (uint32_t *) rowdata;
int col;
for (col = 0; col < width; col++)
dst[col] = bswap_32 (src[col]);
_write_data (&stream, rowdata, 4*width);
data += stride;
}
break;
default:
break;
}
if (rowdata != row_stack)
free (rowdata);
BAIL:
_write_data_end (&stream);
#endif
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" set\n image");
}
static void
_encode_string_literal (char *out, int max,
const char *utf8, int len)
{
char c;
const char *end;
*out++ = '(';
max--;
if (utf8 == NULL)
goto DONE;
if (len < 0)
len = strlen (utf8);
end = utf8 + len;
while (utf8 < end) {
if (max < 5)
break;
switch ((c = *utf8++)) {
case '\n':
*out++ = '\\';
*out++ = 'n';
max -= 2;
break;
case '\r':
*out++ = '\\';
*out++ = 'r';
max -= 2;
case '\t':
*out++ = '\\';
*out++ = 't';
max -= 2;
break;
case '\b':
*out++ = '\\';
*out++ = 'b';
max -= 2;
break;
case '\f':
*out++ = '\\';
*out++ = 'f';
max -= 2;
break;
case '\\':
case '(':
case ')':
*out++ = '\\';
*out++ = c;
max -= 2;
break;
default:
if (isprint (c) || isspace (c)) {
*out++ = c;
} else {
int octal = 0;
while (c) {
octal *= 10;
octal += c&7;
c /= 8;
}
octal = snprintf (out, max, "\\%03d", octal);
out += octal;
max -= octal;
}
break;
}
}
DONE:
*out++ = ')';
*out = '\0';
}
static void
_emit_string_literal (const char *utf8, int len)
{
char c;
const char *end;
if (utf8 == NULL) {
_trace_printf ("()");
return;
}
if (len < 0)
len = strlen (utf8);
end = utf8 + len;
_trace_printf ("(");
while (utf8 < end) {
switch ((c = *utf8++)) {
case '\n':
c = 'n';
goto ESCAPED_CHAR;
case '\r':
c = 'r';
goto ESCAPED_CHAR;
case '\t':
c = 't';
goto ESCAPED_CHAR;
case '\b':
c = 'b';
goto ESCAPED_CHAR;
case '\f':
c = 'f';
goto ESCAPED_CHAR;
case '\\':
case '(':
case ')':
ESCAPED_CHAR:
_trace_printf ("\\%c", c);
break;
default:
if (isprint (c) || isspace (c)) {
_trace_printf ("%c", c);
} else {
int octal = 0;
while (c) {
octal *= 10;
octal += c&7;
c /= 8;
}
_trace_printf ("\\%03d", octal);
}
break;
}
}
_trace_printf (")");
}
static void
_emit_current (Object *obj)
{
if (obj != NULL && ! _pop_operands_to_object (obj)) {
_trace_printf ("%s%ld\n", obj->type->op_code, obj->token);
_push_operand (obj->type->op_type, obj->addr);
}
}
static void
_emit_context (cairo_t *cr)
{
_emit_current (_get_object (CONTEXT, cr));
}
static void
_emit_font_face (cairo_font_face_t *font_face)
{
_emit_current (_get_object (FONT_FACE, font_face));
}
static void
_emit_pattern (cairo_pattern_t *pattern)
{
_emit_current (_get_object (PATTERN, pattern));
}
static void
_emit_scaled_font (cairo_scaled_font_t *scaled_font)
{
_emit_current (_get_object (SCALED_FONT, scaled_font));
}
static void
_emit_surface (cairo_surface_t *surface)
{
_emit_current (_get_object (SURFACE, surface));
}
static void CAIRO_PRINTF_FORMAT(2, 3)
_emit_cairo_op (cairo_t *cr, const char *fmt, ...)
{
va_list ap;
if (cr == NULL || ! _write_lock ())
return;
_emit_context (cr);
va_start (ap, fmt);
_trace_vprintf ( fmt, ap);
va_end (ap);
_write_unlock ();
}
cairo_t *
cairo_create (cairo_surface_t *target)
{
cairo_t *ret;
long surface_id;
long context_id;
ret = DLCALL (cairo_create, target);
context_id = _create_context_id (ret);
_emit_line_info ();
if (target != NULL && _write_lock ()) {
surface_id = _get_surface_id (target);
if (surface_id != -1) {
_get_object (SURFACE, target)->foreign = false;
/* we presume that we will continue to use the context */
if (_pop_operands_to (SURFACE, target)){
_consume_operand ();
} else {
_trace_printf ("s%ld ", surface_id);
}
_trace_printf ("context %% c%ld\n", context_id);
_push_operand (CONTEXT, ret);
}
_write_unlock ();
}
return ret;
}
void
cairo_save (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "save\n");
return DLCALL (cairo_save, cr);
}
void
cairo_restore (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "restore\n");
return DLCALL (cairo_restore, cr);
}
void
cairo_push_group (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "//COLOR_ALPHA push-group\n");
return DLCALL (cairo_push_group, cr);
}
static const char *
_content_to_string (cairo_content_t content)
{
switch (content) {
case CAIRO_CONTENT_ALPHA: return "ALPHA";
case CAIRO_CONTENT_COLOR: return "COLOR";
default:
case CAIRO_CONTENT_COLOR_ALPHA: return "COLOR_ALPHA";
}
}
void
cairo_push_group_with_content (cairo_t *cr, cairo_content_t content)
{
_emit_line_info ();
_emit_cairo_op (cr, "//%s push-group\n", _content_to_string (content));
return DLCALL (cairo_push_group_with_content, cr, content);
}
cairo_pattern_t *
cairo_pop_group (cairo_t *cr)
{
cairo_pattern_t *ret;
ret = DLCALL (cairo_pop_group, cr);
_emit_line_info ();
_emit_cairo_op (cr, "pop-group %% p%ld\n", _create_pattern_id (ret));
_push_operand (PATTERN, ret);
return ret;
}
void
cairo_pop_group_to_source (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "pop-group set-source\n");
return DLCALL (cairo_pop_group_to_source, cr);
}
static const char *
_operator_to_string (cairo_operator_t op)
{
#define f(name) case CAIRO_OPERATOR_ ## name: return #name
switch (op) {
f(OVER);
f(SOURCE);
f(CLEAR);
f(IN);
f(OUT);
f(ATOP);
f(DEST);
f(DEST_OVER);
f(DEST_IN);
f(DEST_OUT);
f(DEST_ATOP);
f(XOR);
f(ADD);
f(SATURATE);
f(MULTIPLY);
f(SCREEN);
f(OVERLAY);
f(DARKEN);
f(LIGHTEN);
case CAIRO_OPERATOR_COLOR_DODGE: return "DODGE";
case CAIRO_OPERATOR_COLOR_BURN: return "BURN";
f(HARD_LIGHT);
f(SOFT_LIGHT);
f(DIFFERENCE);
f(EXCLUSION);
f(HSL_HUE);
f(HSL_SATURATION);
f(HSL_COLOR);
f(HSL_LUMINOSITY);
}
#undef f
return "UNKNOWN_OPERATOR";
}
void
cairo_set_operator (cairo_t *cr, cairo_operator_t op)
{
_emit_line_info ();
_emit_cairo_op (cr, "//%s set-operator\n", _operator_to_string (op));
return DLCALL (cairo_set_operator, cr, op);
}
void
cairo_set_source_rgb (cairo_t *cr, double red, double green, double blue)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g %g %g set-source-rgb\n", red, green, blue);
return DLCALL (cairo_set_source_rgb, cr, red, green, blue);
}
void
cairo_set_source_rgba (cairo_t *cr, double red, double green, double blue, double alpha)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g %g %g %g set-source-rgba\n",
red, green, blue, alpha);
return DLCALL (cairo_set_source_rgba, cr, red, green, blue, alpha);
}
static void
_emit_source_image (cairo_surface_t *surface)
{
Object *obj;
cairo_surface_t *image;
cairo_t *cr;
obj = _get_object (SURFACE, surface);
if (obj == NULL)
return;
image = DLCALL (cairo_image_surface_create,
CAIRO_FORMAT_ARGB32,
obj->width,
obj->height);
cr = DLCALL (cairo_create, image);
DLCALL (cairo_set_source_surface, cr, surface, 0, 0);
DLCALL (cairo_paint, cr);
DLCALL (cairo_destroy, cr);
_emit_image (image, NULL);
_trace_printf (" set-source-image ");
DLCALL (cairo_surface_destroy, image);
obj->foreign = false;
}
static void
_emit_source_image_rectangle (cairo_surface_t *surface,
int x, int y,
int width, int height)
{
Object *obj;
cairo_surface_t *image;
cairo_t *cr;
obj = _get_object (SURFACE, surface);
if (obj == NULL)
return;
if (obj->foreign)
return _emit_source_image (surface);
image = DLCALL (cairo_image_surface_create,
CAIRO_FORMAT_ARGB32,
width,
height);
cr = DLCALL (cairo_create, image);
DLCALL (cairo_set_source_surface, cr, surface, x, y);
DLCALL (cairo_paint, cr);
DLCALL (cairo_destroy, cr);
_emit_image (image, NULL);
_trace_printf (" %d %d set-device-offset set-source-image ",
x, y);
DLCALL (cairo_surface_destroy, image);
}
void
cairo_set_source_surface (cairo_t *cr, cairo_surface_t *surface, double x, double y)
{
_emit_line_info ();
if (cr != NULL && surface != NULL && _write_lock ()) {
if (_is_current (SURFACE, surface, 0) &&
_is_current (CONTEXT, cr, 1))
{
_consume_operand ();
}
else if (_is_current (SURFACE, surface, 1) &&
_is_current (CONTEXT, cr, 0))
{
_trace_printf ("exch ");
_exch_operands ();
_consume_operand ();
} else {
_emit_context (cr);
_trace_printf ("s%ld ", _get_surface_id (surface));
}
if (_get_object (SURFACE, surface)->foreign)
_emit_source_image (surface);
_trace_printf ("pattern");
if (x != 0. || y != 0.)
_trace_printf (" %g %g translate", -x, -y);
_trace_printf (" set-source\n");
_write_unlock ();
}
return DLCALL (cairo_set_source_surface, cr, surface, x, y);
}
void
cairo_set_source (cairo_t *cr, cairo_pattern_t *source)
{
_emit_line_info ();
if (cr != NULL && source != NULL && _write_lock ()) {
Object *obj = _get_object (PATTERN, source);
bool need_context_and_pattern = true;
if (_is_current (PATTERN, source, 0) &&
_is_current (CONTEXT, cr, 1))
{
if (obj->defined) {
_consume_operand ();
need_context_and_pattern = false;
}
}
else if (_is_current (PATTERN, source, 1) &&
_is_current (CONTEXT, cr, 0))
{
if (obj->defined) {
_trace_printf ("exch ");
_exch_operands ();
_consume_operand ();
need_context_and_pattern = false;
}
}
if (need_context_and_pattern) {
_emit_context (cr);
_emit_pattern_id (source);
}
_trace_printf ("set-source\n");
_write_unlock ();
}
return DLCALL (cairo_set_source, cr, source);
}
cairo_pattern_t *
cairo_get_source (cairo_t *cr)
{
cairo_pattern_t *ret;
ret = DLCALL (cairo_get_source, cr);
if (! _has_pattern_id (ret)) {
_emit_cairo_op (cr, "/source get /p%ld exch def\n",
_create_pattern_id (ret));
_get_object (PATTERN, ret)->defined = true;
}
return ret;
}
void
cairo_set_tolerance (cairo_t *cr, double tolerance)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g set-tolerance\n", tolerance);
return DLCALL (cairo_set_tolerance, cr, tolerance);
}
static const char *
_antialias_to_string (cairo_antialias_t antialias)
{
#define f(name) case CAIRO_ANTIALIAS_ ## name: return "ANTIALIAS_" #name
switch (antialias) {
f(DEFAULT);
f(NONE);
f(GRAY);
f(SUBPIXEL);
};
#undef f
return "UNKNOWN_ANTIALIAS";
}
void
cairo_set_antialias (cairo_t *cr, cairo_antialias_t antialias)
{
_emit_line_info ();
_emit_cairo_op (cr,
"//%s set-antialias\n", _antialias_to_string (antialias));
return DLCALL (cairo_set_antialias, cr, antialias);
}
static const char *
_fill_rule_to_string (cairo_fill_rule_t rule)
{
#define f(name) case CAIRO_FILL_RULE_ ## name: return #name
switch (rule) {
f(WINDING);
f(EVEN_ODD);
};
#undef f
return "UNKNOWN_FILL_RULE";
}
void
cairo_set_fill_rule (cairo_t *cr, cairo_fill_rule_t fill_rule)
{
_emit_line_info ();
_emit_cairo_op (cr,
"//%s set-fill-rule\n", _fill_rule_to_string (fill_rule));
return DLCALL (cairo_set_fill_rule, cr, fill_rule);
}
void
cairo_set_line_width (cairo_t *cr, double width)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g set-line-width\n", width);
return DLCALL (cairo_set_line_width, cr, width);
}
static const char *
_line_cap_to_string (cairo_line_cap_t line_cap)
{
#define f(name) case CAIRO_LINE_CAP_ ## name: return "LINE_CAP_" #name
switch (line_cap) {
f(BUTT);
f(ROUND);
f(SQUARE);
};
#undef f
return "UNKNOWN_LINE_CAP";
}
void
cairo_set_line_cap (cairo_t *cr, cairo_line_cap_t line_cap)
{
_emit_line_info ();
_emit_cairo_op (cr, "//%s set-line-cap\n", _line_cap_to_string (line_cap));
return DLCALL (cairo_set_line_cap, cr, line_cap);
}
static const char *
_line_join_to_string (cairo_line_join_t line_join)
{
#define f(name) case CAIRO_LINE_JOIN_ ## name: return "LINE_JOIN_" #name
switch (line_join) {
f(MITER);
f(ROUND);
f(BEVEL);
};
#undef f
return "UNKNOWN_LINE_JOIN";
}
void
cairo_set_line_join (cairo_t *cr, cairo_line_join_t line_join)
{
_emit_line_info ();
_emit_cairo_op (cr,
"//%s set-line-join\n", _line_join_to_string (line_join));
return DLCALL (cairo_set_line_join, cr, line_join);
}
void
cairo_set_dash (cairo_t *cr, const double *dashes, int num_dashes, double offset)
{
_emit_line_info ();
if (cr != NULL && _write_lock ()) {
int n;
_emit_context (cr);
_trace_printf ("[");
for (n = 0; n < num_dashes; n++) {
if (n != 0)
_trace_printf (" ");
_trace_printf ("%g", dashes[n]);
}
_trace_printf ("] %g set-dash\n", offset);
_write_unlock ();
}
return DLCALL (cairo_set_dash, cr, dashes, num_dashes, offset);
}
void
cairo_set_miter_limit (cairo_t *cr, double limit)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g set-miter-limit\n", limit);
return DLCALL (cairo_set_miter_limit, cr, limit);
}
void
cairo_translate (cairo_t *cr, double tx, double ty)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g %g translate\n", tx, ty);
return DLCALL (cairo_translate, cr, tx, ty);
}
void
cairo_scale (cairo_t *cr, double sx, double sy)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g %g scale\n", sx, sy);
return DLCALL (cairo_scale, cr, sx, sy);
}
void
cairo_rotate (cairo_t *cr, double angle)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g rotate\n", angle);
return DLCALL (cairo_rotate, cr, angle);
}
void
cairo_transform (cairo_t *cr, const cairo_matrix_t *matrix)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g %g %g %g %g %g matrix transform\n",
matrix->xx, matrix->yx,
matrix->xy, matrix->yy,
matrix->x0, matrix->y0);
return DLCALL (cairo_transform, cr, matrix);
}
void
cairo_set_matrix (cairo_t *cr, const cairo_matrix_t *matrix)
{
_emit_line_info ();
if (_matrix_is_identity (matrix)) {
_emit_cairo_op (cr, "identity set-matrix\n");
} else {
_emit_cairo_op (cr, "%g %g %g %g %g %g matrix set-matrix\n",
matrix->xx, matrix->yx,
matrix->xy, matrix->yy,
matrix->x0, matrix->y0);
}
return DLCALL (cairo_set_matrix, cr, matrix);
}
cairo_surface_t *
cairo_get_target (cairo_t *cr)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_get_target, cr);
surface_id = _create_surface_id (ret);
if (cr != NULL && ! _has_surface_id (ret)) {
_emit_cairo_op (cr, "/target get /s%ld exch def\n", surface_id);
_get_object (SURFACE, ret)->defined = true;
}
return ret;
}
cairo_surface_t *
cairo_get_group_target (cairo_t *cr)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_get_group_target, cr);
surface_id = _create_surface_id (ret);
if (cr != NULL && ! _has_surface_id (ret)) {
_emit_cairo_op (cr, "/group-target get /s%ld exch def\n", surface_id);
_get_object (SURFACE, ret)->defined = true;
}
return ret;
}
void
cairo_identity_matrix (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "identity set-matrix\n");
return DLCALL (cairo_identity_matrix, cr);
}
void
cairo_new_path (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "n ");
return DLCALL (cairo_new_path, cr);
}
void
cairo_move_to (cairo_t *cr, double x, double y)
{
_emit_cairo_op (cr, "%g %g m ", x, y);
return DLCALL (cairo_move_to, cr, x, y);
}
void
cairo_new_sub_path (cairo_t *cr)
{
_emit_cairo_op (cr, "N ");
return DLCALL (cairo_new_sub_path, cr);
}
void
cairo_line_to (cairo_t *cr, double x, double y)
{
_emit_cairo_op (cr, "%g %g l ", x, y);
return DLCALL (cairo_line_to, cr, x, y);
}
void
cairo_curve_to (cairo_t *cr, double x1, double y1, double x2, double y2, double x3, double y3)
{
_emit_cairo_op (cr, "%g %g %g %g %g %g c ", x1, y1, x2, y2, x3, y3);
return DLCALL (cairo_curve_to, cr, x1, y1, x2, y2, x3, y3);
}
void
cairo_arc (cairo_t *cr, double xc, double yc, double radius, double angle1, double angle2)
{
_emit_cairo_op (cr, "%g %g %g %g %g arc\n", xc, yc, radius, angle1, angle2);
return DLCALL (cairo_arc, cr, xc, yc, radius, angle1, angle2);
}
void
cairo_arc_negative (cairo_t *cr, double xc, double yc, double radius, double angle1, double angle2)
{
_emit_cairo_op (cr, "%g %g %g %g %g arc-\n",
xc, yc, radius, angle1, angle2);
return DLCALL (cairo_arc_negative, cr, xc, yc, radius, angle1, angle2);
}
void
cairo_rel_move_to (cairo_t *cr, double dx, double dy)
{
_emit_cairo_op (cr, "%g %g M ", dx, dy);
return DLCALL (cairo_rel_move_to, cr, dx, dy);
}
void
cairo_rel_line_to (cairo_t *cr, double dx, double dy)
{
_emit_cairo_op (cr, "%g %g L ", dx, dy);
return DLCALL (cairo_rel_line_to, cr, dx, dy);
}
void
cairo_rel_curve_to (cairo_t *cr, double dx1, double dy1, double dx2, double dy2, double dx3, double dy3)
{
_emit_cairo_op (cr, "%g %g %g %g %g %g C ",
dx1, dy1, dx2, dy2, dx3, dy3);
return DLCALL (cairo_rel_curve_to, cr, dx1, dy1, dx2, dy2, dx3, dy3);
}
void
cairo_rectangle (cairo_t *cr, double x, double y, double width, double height)
{
_emit_cairo_op (cr, "%g %g %g %g rectangle\n", x, y, width, height);
return DLCALL (cairo_rectangle, cr, x, y, width, height);
}
void
cairo_close_path (cairo_t *cr)
{
_emit_cairo_op (cr, "h\n");
return DLCALL (cairo_close_path, cr);
}
void
cairo_paint (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "paint\n");
DLCALL (cairo_paint, cr);
}
void
cairo_paint_with_alpha (cairo_t *cr, double alpha)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g paint-with-alpha\n", alpha);
DLCALL (cairo_paint_with_alpha, cr, alpha);
}
void
cairo_mask (cairo_t *cr, cairo_pattern_t *pattern)
{
_emit_line_info ();
if (cr != NULL && pattern != NULL && _write_lock ()) {
Object *obj = _get_object (PATTERN, pattern);
bool need_context_and_pattern = true;
if (_is_current (PATTERN, pattern, 0) &&
_is_current (CONTEXT, cr, 1))
{
if (obj->defined) {
_consume_operand ();
need_context_and_pattern = false;
}
}
else if (_is_current (PATTERN, pattern, 1) &&
_is_current (CONTEXT, cr, 0))
{
if (obj->defined) {
_trace_printf ("exch ");
_exch_operands ();
_consume_operand ();
need_context_and_pattern = false;
}
}
if (need_context_and_pattern) {
_emit_context (cr);
_emit_pattern_id (pattern);
}
_trace_printf (" mask\n");
_write_unlock ();
}
DLCALL (cairo_mask, cr, pattern);
}
void
cairo_mask_surface (cairo_t *cr, cairo_surface_t *surface, double x, double y)
{
_emit_line_info ();
if (cr != NULL && surface != NULL && _write_lock ()) {
if (_is_current (SURFACE, surface, 0) &&
_is_current (CONTEXT, cr, 1))
{
_consume_operand ();
}
else if (_is_current (SURFACE, surface, 1) &&
_is_current (CONTEXT, cr, 0))
{
_trace_printf ("exch ");
_exch_operands ();
_consume_operand ();
} else {
_emit_context (cr);
_trace_printf ("s%ld ", _get_surface_id (surface));
}
_trace_printf ("pattern");
if (x != 0. || y != 0.)
_trace_printf (" %g %g translate", -x, -y);
_trace_printf (" mask\n");
_write_unlock ();
}
DLCALL (cairo_mask_surface, cr, surface, x, y);
}
void
cairo_stroke (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "stroke\n");
DLCALL (cairo_stroke, cr);
}
void
cairo_stroke_preserve (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "stroke+\n");
DLCALL (cairo_stroke_preserve, cr);
}
void
cairo_fill (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "fill\n");
DLCALL (cairo_fill, cr);
}
void
cairo_fill_preserve (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "fill+\n");
DLCALL (cairo_fill_preserve, cr);
}
void
cairo_copy_page (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "copy-page\n");
return DLCALL (cairo_copy_page, cr);
}
void
cairo_show_page (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "show-page\n");
return DLCALL (cairo_show_page, cr);
}
void
cairo_clip (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "clip\n");
DLCALL (cairo_clip, cr);
}
void
cairo_clip_preserve (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "clip+\n");
DLCALL (cairo_clip_preserve, cr);
}
void
cairo_reset_clip (cairo_t *cr)
{
_emit_line_info ();
_emit_cairo_op (cr, "reset-clip\n");
return DLCALL (cairo_reset_clip, cr);
}
static const char *
_slant_to_string (cairo_font_slant_t font_slant)
{
#define f(name) case CAIRO_FONT_SLANT_ ## name: return "SLANT_" #name
switch (font_slant) {
f(NORMAL);
f(ITALIC);
f(OBLIQUE);
};
#undef f
return "UNKNOWN_SLANT";
}
static const char *
_weight_to_string (cairo_font_weight_t font_weight)
{
#define f(name) case CAIRO_FONT_WEIGHT_ ## name: return "WEIGHT_" #name
switch (font_weight) {
f(NORMAL);
f(BOLD);
};
#undef f
return "UNKNOWN_WEIGHT";
}
void
cairo_select_font_face (cairo_t *cr, const char *family, cairo_font_slant_t slant, cairo_font_weight_t weight)
{
_emit_line_info ();
if (cr != NULL && _write_lock ()) {
_emit_context (cr);
_emit_string_literal (family, -1);
_trace_printf (" //%s //%s select-font-face\n",
_slant_to_string (slant),
_weight_to_string (weight));
_write_unlock ();
}
return DLCALL (cairo_select_font_face, cr, family, slant, weight);
}
cairo_font_face_t *
cairo_get_font_face (cairo_t *cr)
{
cairo_font_face_t *ret;
long font_face_id;
ret = DLCALL (cairo_get_font_face, cr);
font_face_id = _create_font_face_id (ret);
_emit_cairo_op (cr, "/font-face get %% f%ld\n", font_face_id);
_push_operand (FONT_FACE, ret);
return ret;
}
void
cairo_set_font_face (cairo_t *cr, cairo_font_face_t *font_face)
{
_emit_line_info ();
if (cr != NULL && font_face != NULL && _write_lock ()) {
if (_is_current (FONT_FACE, font_face, 0) &&
_is_current (CONTEXT, cr, 1))
{
_consume_operand ();
}
else if (_is_current (FONT_FACE, font_face, 1) &&
_is_current (CONTEXT, cr, 0))
{
_trace_printf ("exch ");
_exch_operands ();
_consume_operand ();
}
else
{
_emit_context (cr);
_emit_font_face_id (font_face);
}
_trace_printf ("set-font-face\n");
_write_unlock ();
}
return DLCALL (cairo_set_font_face, cr, font_face);
}
void
cairo_set_font_size (cairo_t *cr, double size)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g set-font-size\n", size);
return DLCALL (cairo_set_font_size, cr, size);
}
void
cairo_set_font_matrix (cairo_t *cr, const cairo_matrix_t *matrix)
{
_emit_line_info ();
_emit_cairo_op (cr, "%g %g %g %g %g %g matrix set-font-matrix\n",
matrix->xx, matrix->yx,
matrix->xy, matrix->yy,
matrix->x0, matrix->y0);
return DLCALL (cairo_set_font_matrix, cr, matrix);
}
static const char *
_subpixel_order_to_string (cairo_subpixel_order_t subpixel_order)
{
#define f(name) case CAIRO_SUBPIXEL_ORDER_ ## name: return "SUBPIXEL_ORDER_" #name
switch (subpixel_order) {
f(DEFAULT);
f(RGB);
f(BGR);
f(VRGB);
f(VBGR);
};
#undef f
return "UNKNOWN_SUBPIXEL_ORDER";
}
static const char *
_hint_style_to_string (cairo_hint_style_t hint_style)
{
#define f(name) case CAIRO_HINT_STYLE_ ## name: return "HINT_STYLE_" #name
switch (hint_style) {
f(DEFAULT);
f(NONE);
f(SLIGHT);
f(MEDIUM);
f(FULL);
};
#undef f
return "UNKNOWN_HINT_STYLE";
}
static const char *
_hint_metrics_to_string (cairo_hint_metrics_t hint_metrics)
{
#define f(name) case CAIRO_HINT_METRICS_ ## name: return "HINT_METRICS_" #name
switch (hint_metrics) {
f(DEFAULT);
f(OFF);
f(ON);
};
#undef f
return "UNKNOWN_HINT_METRICS";
}
static void
_emit_font_options (const cairo_font_options_t *options)
{
cairo_antialias_t antialias;
cairo_subpixel_order_t subpixel_order;
cairo_hint_style_t hint_style;
cairo_hint_metrics_t hint_metrics;
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf ("<<");
antialias = DLCALL (cairo_font_options_get_antialias, options);
if (antialias != CAIRO_ANTIALIAS_DEFAULT) {
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" /antialias //%s",
_antialias_to_string (antialias));
}
subpixel_order = DLCALL (cairo_font_options_get_subpixel_order, options);
if (subpixel_order != CAIRO_SUBPIXEL_ORDER_DEFAULT) {
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" /subpixel-order //%s",
_subpixel_order_to_string (subpixel_order));
}
hint_style = DLCALL (cairo_font_options_get_hint_style, options);
if (hint_style != CAIRO_HINT_STYLE_DEFAULT) {
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" /hint-style //%s",
_hint_style_to_string (hint_style));
}
hint_metrics = DLCALL (cairo_font_options_get_hint_metrics, options);
if (hint_style != CAIRO_HINT_METRICS_DEFAULT) {
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" /hint-metrics //%s",
_hint_metrics_to_string (hint_metrics));
}
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" >>");
}
void
cairo_set_font_options (cairo_t *cr, const cairo_font_options_t *options)
{
_emit_line_info ();
if (cr != NULL && options != NULL && _write_lock ()) {
_emit_context (cr);
_emit_font_options (options);
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf (" set-font-options\n");
_write_unlock ();
}
return DLCALL (cairo_set_font_options, cr, options);
}
cairo_scaled_font_t *
cairo_get_scaled_font (cairo_t *cr)
{
cairo_scaled_font_t *ret;
ret = DLCALL (cairo_get_scaled_font, cr);
if (cr != NULL && ! _has_scaled_font_id (ret)) {
_emit_cairo_op (cr, "/scaled-font get /sf%ld exch def\n",
_create_scaled_font_id (ret));
_get_object (SCALED_FONT, ret)->defined = true;
}
return ret;
}
void
cairo_set_scaled_font (cairo_t *cr, const cairo_scaled_font_t *scaled_font)
{
_emit_line_info ();
if (cr != NULL && scaled_font != NULL) {
if (_pop_operands_to (SCALED_FONT, scaled_font)) {
if (_is_current (CONTEXT, cr, 1)) {
if (_write_lock ()) {
_consume_operand ();
_trace_printf ("set-scaled-font\n");
_write_unlock ();
}
} else {
if (_get_object (CONTEXT, cr)->defined) {
if (_write_lock ()) {
_consume_operand ();
_trace_printf ("c%ld exch set-scaled-font pop\n",
_get_context_id (cr));
_write_unlock ();
}
} else {
_emit_cairo_op (cr, "sf%ld set-scaled-font\n",
_get_scaled_font_id (scaled_font));
}
}
} else {
_emit_cairo_op (cr, "sf%ld set-scaled-font\n",
_get_scaled_font_id (scaled_font));
}
}
return DLCALL (cairo_set_scaled_font, cr, scaled_font);
}
static void
_emit_matrix (const cairo_matrix_t *m)
{
if (_matrix_is_identity(m))
{
_trace_printf ("identity");
}
else
{
_trace_printf ("%g %g %g %g %g %g matrix",
m->xx, m->yx,
m->xy, m->yy,
m->x0, m->y0);
}
}
cairo_scaled_font_t *
cairo_scaled_font_create (cairo_font_face_t *font_face,
const cairo_matrix_t *font_matrix,
const cairo_matrix_t *ctm,
const cairo_font_options_t *options)
{
cairo_scaled_font_t *ret;
long scaled_font_id;
ret = DLCALL (cairo_scaled_font_create, font_face, font_matrix, ctm, options);
scaled_font_id = _create_scaled_font_id (ret);
_emit_line_info ();
if (font_face != NULL &&
font_matrix != NULL &&
ctm != NULL &&
options != NULL
&& _write_lock ())
{
if (_pop_operands_to (FONT_FACE, font_face))
_consume_operand ();
else
_trace_printf ("f%ld ", _get_font_face_id (font_face));
_emit_matrix (font_matrix);
_trace_printf (" ");
_emit_matrix (ctm);
_trace_printf (" ");
_emit_font_options (options);
if (_get_object (SCALED_FONT, ret)->defined) {
_trace_printf (" scaled-font pop %% sf%ld\n",
scaled_font_id);
} else {
_trace_printf (" scaled-font dup /sf%ld exch def\n",
scaled_font_id);
_push_operand (SCALED_FONT, ret);
_get_object (SCALED_FONT, ret)->defined = true;
}
_write_unlock ();
}
return ret;
}
void
cairo_show_text (cairo_t *cr, const char *utf8)
{
_emit_line_info ();
if (cr != NULL && _write_lock ()) {
_emit_context (cr);
_emit_string_literal (utf8, -1);
_trace_printf (" show-text\n");
_write_unlock ();
}
DLCALL (cairo_show_text, cr, utf8);
}
static void
_glyph_advance (cairo_scaled_font_t *font,
const cairo_glyph_t *glyph,
double *x, double *y)
{
cairo_text_extents_t extents;
DLCALL (cairo_scaled_font_glyph_extents, font, glyph, 1, &extents);
*x += extents.x_advance;
*y += extents.y_advance;
}
#define TOLERANCE 1e-5
static void
_emit_glyphs (cairo_scaled_font_t *font,
const cairo_glyph_t *glyphs,
int num_glyphs)
{
double x,y;
int n;
if (num_glyphs == 0) {
_trace_printf ("[]");
return;
}
for (n = 0; n < num_glyphs; n++) {
if (glyphs[n].index > 255)
break;
}
x = glyphs->x;
y = glyphs->y;
if (n < num_glyphs) { /* need full glyph range */
bool first;
_trace_printf ("[%g %g [", x, y);
first = true;
while (num_glyphs--) {
if (fabs (glyphs->x - x) > TOLERANCE ||
fabs (glyphs->y - y) > TOLERANCE)
{
x = glyphs->x;
y = glyphs->y;
_trace_printf ("] %g %g [", x, y);
first = true;
}
if (! first)
_trace_printf (" ");
_trace_printf ("%lu", glyphs->index);
first = false;
_glyph_advance (font, glyphs, &x, &y);
glyphs++;
}
_trace_printf ("]]");
} else {
struct _data_stream stream;
if (num_glyphs == 1) {
_trace_printf ("[%g %g <%02lx>]", x, y, glyphs->index);
} else {
_trace_printf ("[%g %g <~", x, y);
_write_base85_data_start (&stream);
while (num_glyphs--) {
unsigned char c;
if (fabs (glyphs->x - x) > TOLERANCE ||
fabs (glyphs->y - y) > TOLERANCE)
{
x = glyphs->x;
y = glyphs->y;
_write_base85_data_end (&stream);
_trace_printf ("~> %g %g <~", x, y);
_write_base85_data_start (&stream);
}
c = glyphs->index;
_write_base85_data (&stream, &c, 1);
_glyph_advance (font, glyphs, &x, &y);
glyphs++;
}
_write_base85_data_end (&stream);
_trace_printf ("~>]");
}
}
}
void
cairo_show_glyphs (cairo_t *cr, const cairo_glyph_t *glyphs, int num_glyphs)
{
_emit_line_info ();
if (cr != NULL && glyphs != NULL && _write_lock ()) {
cairo_scaled_font_t *font;
_emit_context (cr);
font = DLCALL (cairo_get_scaled_font, cr);
_emit_glyphs (font, glyphs, num_glyphs);
_trace_printf (" show-glyphs\n");
_write_unlock ();
}
DLCALL (cairo_show_glyphs, cr, glyphs, num_glyphs);
}
static const char *
_direction_to_string (cairo_bool_t backward)
{
const char *names[] = {
"FORWARD",
"BACKWARD"
};
return names[!!backward];
}
void
cairo_show_text_glyphs (cairo_t *cr,
const char *utf8,
int utf8_len,
const cairo_glyph_t *glyphs,
int num_glyphs,
const cairo_text_cluster_t *clusters,
int num_clusters,
cairo_text_cluster_flags_t backward)
{
cairo_scaled_font_t *font;
font = DLCALL (cairo_get_scaled_font, cr);
_emit_line_info ();
if (cr != NULL && glyphs != NULL && clusters != NULL && _write_lock ()) {
int n;
_emit_context (cr);
_emit_string_literal (utf8, utf8_len);
_emit_glyphs (font, glyphs, num_glyphs);
_trace_printf (" [");
for (n = 0; n < num_clusters; n++) {
_trace_printf (" %d %d",
clusters[n].num_bytes,
clusters[n].num_glyphs);
}
_trace_printf (" ] //%s show-text-glyphs\n",
_direction_to_string (backward));
_write_unlock ();
}
DLCALL (cairo_show_text_glyphs, cr,
utf8, utf8_len,
glyphs, num_glyphs,
clusters, num_clusters,
backward);
}
void
cairo_text_path (cairo_t *cr, const char *utf8)
{
_emit_line_info ();
if (cr != NULL && _write_lock ()) {
_emit_context (cr);
_emit_string_literal (utf8, -1);
_trace_printf (" text-path\n");
_write_unlock ();
}
return DLCALL (cairo_text_path, cr, utf8);
}
void
cairo_glyph_path (cairo_t *cr, const cairo_glyph_t *glyphs, int num_glyphs)
{
cairo_scaled_font_t *font;
font = DLCALL (cairo_get_scaled_font, cr);
_emit_line_info ();
if (cr != NULL && glyphs != NULL && _write_lock ()) {
_emit_context (cr);
_emit_glyphs (font, glyphs, num_glyphs);
_trace_printf (" glyph-path\n");
_write_unlock ();
}
return DLCALL (cairo_glyph_path, cr, glyphs, num_glyphs);
}
void
cairo_append_path (cairo_t *cr, const cairo_path_t *path)
{
/* XXX no support for named paths, so manually reconstruct */
int i;
cairo_path_data_t *p;
_emit_line_info ();
if (cr == NULL || path == NULL)
return DLCALL (cairo_append_path, cr, path);
for (i=0; i < path->num_data; i += path->data[i].header.length) {
p = &path->data[i];
switch (p->header.type) {
case CAIRO_PATH_MOVE_TO:
if (p->header.length >= 2)
cairo_move_to (cr, p[1].point.x, p[1].point.y);
break;
case CAIRO_PATH_LINE_TO:
if (p->header.length >= 2)
cairo_line_to (cr, p[1].point.x, p[1].point.y);
break;
case CAIRO_PATH_CURVE_TO:
if (p->header.length >= 4)
cairo_curve_to (cr,
p[1].point.x, p[1].point.y,
p[2].point.x, p[2].point.y,
p[3].point.x, p[3].point.y);
break;
case CAIRO_PATH_CLOSE_PATH:
if (p->header.length >= 1)
cairo_close_path (cr);
break;
default:
break;
}
}
}
cairo_surface_t *
cairo_image_surface_create (cairo_format_t format, int width, int height)
{
cairo_surface_t *ret;
long surface_id;
const char *format_str;
ret = DLCALL (cairo_image_surface_create, format, width, height);
surface_id = _create_surface_id (ret);
format_str = _format_to_string (format);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /width %d set\n"
" /height %d set\n"
" /format //%s set\n"
" image dup /s%ld exch def\n",
width, height, format_str, surface_id);
_get_object (SURFACE, ret)->width = width;
_get_object (SURFACE, ret)->height = height;
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_image_surface_create_for_data (unsigned char *data, cairo_format_t format, int width, int height, int stride)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_image_surface_create_for_data, data, format, width, height, stride);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
/* cairo_image_surface_create_for_data() is both used to supply
* foreign pixel data to cairo and in order to read pixels back.
* Defer grabbing the pixel contents until we have to, but only for
* "large" images, for small images the overhead of embedding pixels
* is negligible.
*
* Choose 32x32 as that captures most icons which thanks to GdkPixbuf
* are frequently reloaded.
*/
if (width * height < 32*32) {
_emit_image (ret, NULL);
_trace_printf (" dup /s%ld exch def\n",
surface_id);
} else {
_trace_printf ("dict\n"
" /width %d set\n"
" /height %d set\n"
" /format //%s set\n"
" image dup /s%ld exch def\n",
width, height,
_format_to_string (format),
surface_id);
_get_object (SURFACE, ret)->foreign = true;
}
_get_object (SURFACE, ret)->width = width;
_get_object (SURFACE, ret)->height = height;
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_surface_create_similar (cairo_surface_t *other,
cairo_content_t content,
int width, int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_surface_create_similar, other, content, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (other != NULL && _write_lock ()) {
Object *obj;
obj = _get_object (SURFACE, other);
if (obj->defined)
_trace_printf ("s%ld ", obj->token);
else if (current_stack_depth == obj->operand + 1)
_trace_printf ("dup ");
else
_trace_printf ("%d index ", current_stack_depth - obj->operand - 1);
_trace_printf ("%d %d //%s similar %% s%ld\n",
width,
height,
_content_to_string (content),
surface_id);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
static void CAIRO_PRINTF_FORMAT(2, 3)
_emit_surface_op (cairo_surface_t *surface, const char *fmt, ...)
{
va_list ap;
if (surface == NULL || ! _write_lock ())
return;
_emit_surface (surface);
va_start (ap, fmt);
_trace_vprintf ( fmt, ap);
va_end (ap);
_write_unlock ();
}
void
cairo_surface_finish (cairo_surface_t *surface)
{
_emit_line_info ();
return DLCALL (cairo_surface_finish, surface);
}
void
cairo_surface_flush (cairo_surface_t *surface)
{
_emit_line_info ();
return DLCALL (cairo_surface_flush, surface);
}
void
cairo_surface_mark_dirty (cairo_surface_t *surface)
{
_emit_line_info ();
if (surface != NULL && _write_lock ()) {
if (_mark_dirty) {
_emit_surface (surface);
_trace_printf ("%% mark-dirty\n");
_emit_source_image (surface);
} else
_trace_printf ("%% s%ld mark-dirty\n", _get_surface_id (surface));
_write_unlock ();
}
return DLCALL (cairo_surface_mark_dirty, surface);
}
void
cairo_surface_mark_dirty_rectangle (cairo_surface_t *surface,
int x, int y, int width, int height)
{
_emit_line_info ();
if (surface != NULL && _write_lock ()) {
if (_mark_dirty) {
_emit_surface (surface);
_trace_printf ("%% %d %d %d %d mark-dirty-rectangle\n",
x, y, width, height);
_emit_source_image_rectangle (surface, x,y, width, height);
} else
_trace_printf ("%% s%ld %d %d %d %d mark-dirty-rectangle\n",
_get_surface_id (surface), x, y, width, height);
_write_unlock ();
}
return DLCALL (cairo_surface_mark_dirty_rectangle, surface, x, y, width, height);
}
void
cairo_surface_set_device_offset (cairo_surface_t *surface, double x_offset, double y_offset)
{
_emit_line_info ();
_emit_surface_op (surface, "%g %g set-device-offset\n",
x_offset, y_offset);
return DLCALL (cairo_surface_set_device_offset, surface, x_offset, y_offset);
}
void
cairo_surface_set_fallback_resolution (cairo_surface_t *surface, double x_pixels_per_inch, double y_pixels_per_inch)
{
_emit_line_info ();
_emit_surface_op (surface, "%g %g set-fallback-resolution\n",
x_pixels_per_inch, y_pixels_per_inch);
return DLCALL (cairo_surface_set_fallback_resolution, surface, x_pixels_per_inch, y_pixels_per_inch);
}
void
cairo_surface_copy_page (cairo_surface_t *surface)
{
_emit_line_info ();
_emit_surface_op (surface, "copy-page\n");
return DLCALL (cairo_surface_copy_page, surface);
}
void
cairo_surface_show_page (cairo_surface_t *surface)
{
_emit_line_info ();
_emit_surface_op (surface, "show-page\n");
return DLCALL (cairo_surface_show_page, surface);
}
cairo_status_t
cairo_surface_set_mime_data (cairo_surface_t *surface,
const char *mime_type,
const unsigned char *data,
unsigned int length,
cairo_destroy_func_t destroy,
void *closure)
{
_emit_line_info ();
if (surface != NULL && _write_lock ()) {
_emit_surface (surface);
_emit_string_literal (mime_type, -1);
_trace_printf (" ");
_emit_data (data, length);
_trace_printf (" /deflate filter set-mime-data\n");
_write_unlock ();
}
return DLCALL (cairo_surface_set_mime_data,
surface,
mime_type,
data, length,
destroy,
closure);
}
#if CAIRO_HAS_PNG_FUNCTIONS
cairo_status_t
cairo_surface_write_to_png (cairo_surface_t *surface, const char *filename)
{
_emit_line_info ();
if (surface != NULL && _write_lock ()) {
_trace_printf ("%% s%ld ", _get_surface_id (surface));
_emit_string_literal (filename, -1);
_trace_printf (" write-to-png\n");
_write_unlock ();
}
return DLCALL (cairo_surface_write_to_png, surface, filename);
}
cairo_status_t
cairo_surface_write_to_png_stream (cairo_surface_t *surface,
cairo_write_func_t write_func,
void *data)
{
_emit_line_info ();
if (surface != NULL && _write_lock ()) {
char symbol[1024];
_trace_printf ("%% s%ld ", _get_surface_id (surface));
_emit_string_literal (lookup_symbol (symbol, sizeof (symbol),
write_func),
-1);
_trace_printf (" write-to-png-stream\n");
_write_unlock ();
}
return DLCALL (cairo_surface_write_to_png_stream,
surface, write_func, data);
}
#endif
static void CAIRO_PRINTF_FORMAT(2, 3)
_emit_pattern_op (cairo_pattern_t *pattern, const char *fmt, ...)
{
va_list ap;
if (pattern == NULL || ! _write_lock ())
return;
_emit_pattern (pattern);
va_start (ap, fmt);
_trace_vprintf (fmt, ap);
va_end (ap);
_write_unlock ();
}
cairo_pattern_t *
cairo_pattern_create_rgb (double red, double green, double blue)
{
cairo_pattern_t *ret;
long pattern_id;
ret = DLCALL (cairo_pattern_create_rgb, red, green, blue);
pattern_id = _create_pattern_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("/p%ld %g %g %g rgb def\n",
pattern_id, red, green, blue);
_get_object (PATTERN, ret)->defined = true;
_write_unlock ();
}
return ret;
}
cairo_pattern_t *
cairo_pattern_create_rgba (double red, double green, double blue, double alpha)
{
cairo_pattern_t *ret;
long pattern_id;
ret = DLCALL (cairo_pattern_create_rgba, red, green, blue, alpha);
pattern_id = _create_pattern_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("/p%ld %g %g %g %g rgba def\n",
pattern_id, red, green, blue, alpha);
_get_object (PATTERN, ret)->defined = true;
_write_unlock ();
}
return ret;
}
cairo_pattern_t *
cairo_pattern_create_for_surface (cairo_surface_t *surface)
{
cairo_pattern_t *ret;
long pattern_id;
long surface_id;
ret = DLCALL (cairo_pattern_create_for_surface, surface);
pattern_id = _create_pattern_id (ret);
_emit_line_info ();
if (surface != NULL && _write_lock ()) {
surface_id = _get_surface_id (surface);
if (_pop_operands_to (SURFACE, surface)) {
_consume_operand ();
} else {
_trace_printf ("s%ld ", surface_id);
}
if (_get_object (SURFACE, surface)->foreign)
_emit_source_image (surface);
_trace_printf ("pattern %% p%ld\n", pattern_id);
_push_operand (PATTERN, ret);
_write_unlock ();
}
return ret;
}
cairo_pattern_t *
cairo_pattern_create_linear (double x0, double y0, double x1, double y1)
{
cairo_pattern_t *ret;
long pattern_id;
ret = DLCALL (cairo_pattern_create_linear, x0, y0, x1, y1);
pattern_id = _create_pattern_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("%g %g %g %g linear %% p%ld\n",
x0, y0, x1, y1, pattern_id);
_push_operand (PATTERN, ret);
_write_unlock ();
}
return ret;
}
cairo_pattern_t *
cairo_pattern_create_radial (double cx0, double cy0, double radius0, double cx1, double cy1, double radius1)
{
cairo_pattern_t *ret;
long pattern_id;
ret = DLCALL (cairo_pattern_create_radial,
cx0, cy0, radius0,
cx1, cy1, radius1);
pattern_id = _create_pattern_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("%g %g %g %g %g %g radial %% p%ld\n",
cx0, cy0, radius0, cx1, cy1, radius1,
pattern_id);
_push_operand (PATTERN, ret);
_write_unlock ();
}
return ret;
}
void
cairo_pattern_add_color_stop_rgb (cairo_pattern_t *pattern, double offset, double red, double green, double blue)
{
_emit_line_info ();
_emit_pattern_op (pattern,
"%g %g %g %g 1 add-color-stop\n",
offset, red, green, blue);
return DLCALL (cairo_pattern_add_color_stop_rgb, pattern, offset, red, green, blue);
}
void
cairo_pattern_add_color_stop_rgba (cairo_pattern_t *pattern, double offset, double red, double green, double blue, double alpha)
{
_emit_line_info ();
_emit_pattern_op (pattern,
"%g %g %g %g %g add-color-stop\n",
offset, red, green, blue, alpha);
return DLCALL (cairo_pattern_add_color_stop_rgba, pattern, offset, red, green, blue, alpha);
}
void
cairo_pattern_set_matrix (cairo_pattern_t *pattern, const cairo_matrix_t *matrix)
{
_emit_line_info ();
if (_matrix_is_identity (matrix)) {
_emit_pattern_op (pattern, "identity set-matrix\n");
} else {
_emit_pattern_op (pattern,
"%g %g %g %g %g %g matrix set-matrix\n",
matrix->xx, matrix->yx,
matrix->xy, matrix->yy,
matrix->x0, matrix->y0);
}
return DLCALL (cairo_pattern_set_matrix, pattern, matrix);
}
static const char *
_filter_to_string (cairo_filter_t filter)
{
#define f(name) case CAIRO_FILTER_ ## name: return "FILTER_" #name
switch (filter) {
f(FAST);
f(GOOD);
f(BEST);
f(NEAREST);
f(BILINEAR);
f(GAUSSIAN);
};
#undef f
return "UNKNOWN_FILTER";
}
void
cairo_pattern_set_filter (cairo_pattern_t *pattern, cairo_filter_t filter)
{
_emit_line_info ();
_emit_pattern_op (pattern, "//%s set-filter\n", _filter_to_string (filter));
return DLCALL (cairo_pattern_set_filter, pattern, filter);
}
static const char *
_extend_to_string (cairo_extend_t extend)
{
#define f(name) case CAIRO_EXTEND_ ## name: return "EXTEND_" #name
switch (extend) {
f(NONE);
f(REPEAT);
f(REFLECT);
f(PAD);
};
#undef f
return "UNKNOWN_EXTEND";
}
void
cairo_pattern_set_extend (cairo_pattern_t *pattern, cairo_extend_t extend)
{
_emit_line_info ();
_emit_pattern_op (pattern, "//%s set-extend\n", _extend_to_string (extend));
return DLCALL (cairo_pattern_set_extend, pattern, extend);
}
#if CAIRO_HAS_FT_FONT
#if CAIRO_HAS_FC_FONT
cairo_font_face_t *
cairo_ft_font_face_create_for_pattern (FcPattern *pattern)
{
cairo_font_face_t *ret;
long font_face_id;
ret = DLCALL (cairo_ft_font_face_create_for_pattern, pattern);
font_face_id = _create_font_face_id (ret);
_emit_line_info ();
if (pattern != NULL && _write_lock ()) {
Object *obj;
FcChar8 *parsed;
obj = _get_object (FONT_FACE, ret);
if (obj->operand != -1)
_object_remove (obj);
parsed = DLCALL (FcNameUnparse, pattern);
_trace_printf ("dict\n"
" /type 42 set\n"
" /pattern ");
_emit_string_literal ((char *) parsed, -1);
_trace_printf (" set\n"
" font %% f%ld\n",
font_face_id);
_push_operand (FONT_FACE, ret);
_write_unlock ();
free (parsed);
}
return ret;
}
#endif /* CAIRO_HAS_FC_FONT*/
typedef struct _ft_face_data {
unsigned long index;
unsigned long size;
void *data;
} FtFaceData;
static void
_ft_face_data_destroy (void *arg)
{
FtFaceData *data = arg;
free (data->data);
free (data);
}
cairo_font_face_t *
cairo_ft_font_face_create_for_ft_face (FT_Face face, int load_flags)
{
cairo_font_face_t *ret;
Object *obj;
FtFaceData *data;
long font_face_id;
ret = DLCALL (cairo_ft_font_face_create_for_ft_face, face, load_flags);
font_face_id = _create_font_face_id (ret);
if (face == NULL)
return ret;
obj = _get_object (NONE, face);
data = obj->data;
if (data == NULL)
return ret;
_emit_line_info ();
if (_write_lock ()) {
obj = _get_object (FONT_FACE, ret);
if (obj->operand != -1)
_object_remove (obj);
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_trace_printf ("<< /type 42 /source ");
_emit_data (data->data, data->size);
_trace_printf (" /index %lu /flags %d >> font %% f%ld\n",
data->index, load_flags, font_face_id);
_push_operand (FONT_FACE, ret);
_write_unlock ();
}
return ret;
}
static bool
_ft_read_file (FtFaceData *data, const char *path)
{
char buf[8192];
FILE *file;
file = fopen (path, "rb");
if (file != NULL) {
size_t ret;
unsigned long int allocated = sizeof (buf);
data->data = malloc (allocated);
do {
ret = fread (buf, 1, sizeof (buf), file);
if (ret == 0)
break;
memcpy ((char *) data->data + data->size, buf, ret);
data->size += ret;
if (ret != sizeof (buf))
break;
if (data->size == allocated) {
allocated *= 2;
data->data = realloc (data->data, allocated);
}
} while (true);
fclose (file);
}
return file != NULL;
}
FT_Error
FT_New_Face (FT_Library library, const char *pathname, FT_Long index, FT_Face *face)
{
FT_Error ret;
ret = DLCALL (FT_New_Face, library, pathname, index, face);
if (ret == 0) {
Object *obj = _type_object_create (NONE, *face);
FtFaceData *data = malloc (sizeof (FtFaceData));
data->index = index;
data->size = 0;
data->data = NULL;
_ft_read_file (data, pathname);
obj->data = data;
obj->destroy = _ft_face_data_destroy;
}
return ret;
}
FT_Error
FT_New_Memory_Face (FT_Library library, const FT_Byte *mem, FT_Long size, FT_Long index, FT_Face *face)
{
FT_Error ret;
ret = DLCALL (FT_New_Memory_Face, library, mem, size, index, face);
if (ret == 0) {
Object *obj = _type_object_create (NONE, *face);
FtFaceData *data = malloc (sizeof (FtFaceData));
data->index = index;
data->size = size;
data->data = malloc (size);
memcpy (data->data, mem, size);
obj->data = data;
obj->destroy = _ft_face_data_destroy;
}
return ret;
}
/* XXX
* FT_New_Memory_Face() and FT_New_Face() appear to wrap FT_Open_Face() so we
* get a redundant call to FT_Open_Face() from those paths (no PLT hiding
* within FT, naughty library!) but we do not intercept a direct call to
* FT_Open_Face(). So far this has not caused any issues, but it will one
* day...
*/
#if 0
FT_Error
FT_Open_Face (FT_Library library, const FT_Open_Args *args, FT_Long index, FT_Face *face)
{
FT_Error ret;
ret = DLCALL (FT_Open_Face, library, args, index, face);
if (args->flags & FT_OPEN_MEMORY)
fprintf (stderr, "FT_Open_Face (mem=%p, %ld, %ld) = %p\n",
args->memory_base, args->memory_size,
index, *face);
else if (args->flags & FT_OPEN_STREAM)
fprintf (stderr, "FT_Open_Face (stream, %ld) = %p\n",
index, *face);
else if (args->flags & FT_OPEN_PATHNAME)
fprintf (stderr, "FT_Open_Face (path=%s, %ld) = %p\n",
args->pathname, index, *face);
return ret;
}
#endif
FT_Error
FT_Done_Face (FT_Face face)
{
_object_destroy (_get_object (NONE, face));
return DLCALL (FT_Done_Face, face);
}
#endif
static void
_surface_object_set_size (cairo_surface_t *surface, int width, int height)
{
Object *obj;
obj = _get_object (SURFACE, surface);
obj->width = width;
obj->height = height;
}
static void
_surface_object_set_size_from_surface (cairo_surface_t *surface)
{
_surface_object_set_size (surface,
DLCALL (cairo_image_surface_get_width, surface),
DLCALL (cairo_image_surface_get_height, surface));
}
#if CAIRO_HAS_PS_SURFACE
#include<cairo-ps.h>
cairo_surface_t *
cairo_ps_surface_create (const char *filename, double width_in_points, double height_in_points)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_ps_surface_create, filename, width_in_points, height_in_points);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /PS set\n"
" /filename ");
_emit_string_literal (filename, -1);
_trace_printf (" set\n"
" /width %g set\n"
" /height %g set\n"
" surface %% s%ld\n",
width_in_points,
height_in_points,
surface_id);
_surface_object_set_size (ret, width_in_points, height_in_points);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_ps_surface_create_for_stream (cairo_write_func_t write_func, void *closure, double width_in_points, double height_in_points)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_ps_surface_create_for_stream, write_func, closure, width_in_points, height_in_points);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /PS set\n"
" /width %g set\n"
" /height %g set\n"
" surface %% s%ld\n",
width_in_points,
height_in_points,
surface_id);
_surface_object_set_size (ret, width_in_points, height_in_points);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
void
cairo_ps_surface_set_size (cairo_surface_t *surface, double width_in_points, double height_in_points)
{
_emit_line_info ();
return DLCALL (cairo_ps_surface_set_size, surface, width_in_points, height_in_points);
}
#endif
#if CAIRO_HAS_PDF_SURFACE
#include <cairo-pdf.h>
cairo_surface_t *
cairo_pdf_surface_create (const char *filename, double width_in_points, double height_in_points)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_pdf_surface_create, filename, width_in_points, height_in_points);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /PDF set\n"
" /filename ");
_emit_string_literal (filename, -1);
_trace_printf (" set\n"
" /width %g set\n"
" /height %g set\n"
" surface %% s%ld\n",
width_in_points,
height_in_points,
surface_id);
_surface_object_set_size (ret, width_in_points, height_in_points);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_pdf_surface_create_for_stream (cairo_write_func_t write_func, void *closure, double width_in_points, double height_in_points)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_pdf_surface_create_for_stream, write_func, closure, width_in_points, height_in_points);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /PDF set\n"
" /width %g set\n"
" /height %g set\n"
" surface %% s%ld\n",
width_in_points,
height_in_points,
surface_id);
_surface_object_set_size (ret, width_in_points, height_in_points);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
void
cairo_pdf_surface_set_size (cairo_surface_t *surface, double width_in_points, double height_in_points)
{
_emit_line_info ();
return DLCALL (cairo_pdf_surface_set_size, surface, width_in_points, height_in_points);
}
#endif
#if CAIRO_HAS_SVG_SURFACE
#include <cairo-svg.h>
cairo_surface_t *
cairo_svg_surface_create (const char *filename, double width, double height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_svg_surface_create, filename, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /SVG set\n"
" /filename ");
_emit_string_literal (filename, -1);
_trace_printf (" set\n"
" /width %g set\n"
" /height %g set\n"
" surface %% s%ld\n",
width,
height,
surface_id);
_surface_object_set_size (ret, width, height);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_svg_surface_create_for_stream (cairo_write_func_t write_func, void *closure, double width, double height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_svg_surface_create_for_stream, write_func, closure, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /SVG set\n"
" /width %g set\n"
" /height %g set\n"
" surface %% s%ld\n",
width,
height,
surface_id);
_surface_object_set_size (ret, width, height);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#endif
#if CAIRO_HAS_PNG_FUNCTIONS
cairo_surface_t *
cairo_image_surface_create_from_png (const char *filename)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_image_surface_create_from_png, filename);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
char filename_string[4096];
_encode_string_literal (filename_string, sizeof (filename_string),
filename, -1);
_emit_image (ret, " /filename %s set\n", filename_string);
_trace_printf (" dup /s%ld exch def\n",
surface_id);
_surface_object_set_size_from_surface (ret);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_image_surface_create_from_png_stream (cairo_read_func_t read_func, void *closure)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_image_surface_create_from_png_stream, read_func, closure);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_emit_image (ret, NULL);
_trace_printf (" dup /s%ld exch def\n",
surface_id);
_surface_object_set_size_from_surface (ret);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#endif
static const char *
_content_from_surface (cairo_surface_t *surface)
{
return _content_to_string (DLCALL (cairo_surface_get_content, surface));
}
#if CAIRO_HAS_XLIB_SURFACE
#include <cairo-xlib.h>
cairo_surface_t *
cairo_xlib_surface_create (Display *dpy,
Drawable drawable,
Visual *visual,
int width, int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_xlib_surface_create,
dpy, drawable, visual, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /xlib set\n"
" /drawable 16!%lx set\n"
" /content //%s set\n"
" /width %d set\n"
" /height %d set\n"
" surface dup /s%ld exch def\n",
drawable,
_content_from_surface (ret),
width,
height,
surface_id);
_get_object (SURFACE, ret)->defined = true;
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->foreign = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_xlib_surface_create_for_bitmap (Display *dpy,
Pixmap bitmap,
Screen *screen,
int width, int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_xlib_surface_create_for_bitmap,
dpy, bitmap, screen, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /xlib set\n"
" /drawable 16!%lx set\n"
" /content //%s set\n"
" /width %d set\n"
" /height %d set\n"
" /depth 1 set\n"
" surface dup /s%ld exch def\n",
bitmap,
_content_from_surface (ret),
width,
height,
surface_id);
_get_object (SURFACE, ret)->defined = true;
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->foreign = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#if CAIRO_HAS_XLIB_XRENDER_SURFACE
#include <cairo-xlib-xrender.h>
cairo_surface_t *
cairo_xlib_surface_create_with_xrender_format (Display *dpy,
Drawable drawable,
Screen *screen,
XRenderPictFormat *format,
int width, int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_xlib_surface_create_with_xrender_format,
dpy, drawable, screen, format, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /xrender set\n"
" /drawable 16!%lx set\n"
" /content //%s set\n"
" /width %d set\n"
" /height %d set\n"
" /depth %d set\n"
" surface dup /s%ld exch def\n",
drawable,
_content_from_surface (ret),
width,
height,
format->depth,
surface_id);
_get_object (SURFACE, ret)->defined = true;
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->foreign = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#endif
#endif
#if CAIRO_HAS_SCRIPT_SURFACE
#include <cairo-script.h>
cairo_surface_t *
cairo_script_surface_create (cairo_script_context_t *ctx,
cairo_content_t content,
double width,
double height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_script_surface_create, ctx, content, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /script set\n"
" /content %s set\n"
" /width %g set\n"
" /height %g set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
width, height,
surface_id);
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_script_surface_create_for_target (cairo_script_context_t *ctx,
cairo_surface_t *target)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_script_surface_create_for_target, ctx, target);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /script set\n"
" surface dup /s%ld exch def\n",
surface_id);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#endif
#if CAIRO_HAS_TEST_SURFACES
#include <test-fallback-surface.h>
cairo_surface_t *
_cairo_test_fallback_surface_create (cairo_content_t content,
int width,
int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (_cairo_test_fallback_surface_create, content, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /test-fallback set\n"
" /content //%s set\n"
" /width %d set\n"
" /height %d set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
width, height,
surface_id);
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#include <test-paginated-surface.h>
cairo_surface_t *
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_cairo_test_paginated_surface_create (cairo_surface_t *surface)
{
cairo_surface_t *ret;
long surface_id;
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
ret = DLCALL (_cairo_test_paginated_surface_create, surface);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
/* XXX store initial data? */
_trace_printf ("dict\n"
" /type /test-paginated set\n"
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
" /target s%ld set\n"
" surface dup /s%ld exch def\n",
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
_get_surface_id (surface),
surface_id);
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#include <test-null-surface.h>
cairo_surface_t *
_cairo_test_null_surface_create (cairo_content_t content)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (_cairo_test_null_surface_create, content);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /test-null set\n"
" /content //%s set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
surface_id);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#endif
cairo_surface_t *
cairo_meta_surface_create (cairo_content_t content,
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
const cairo_rectangle_t *extents)
{
cairo_surface_t *ret;
long surface_id;
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
ret = DLCALL (cairo_meta_surface_create, content, extents);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
Remove clip handling from generic surface layer. Handling clip as part of the surface state, as opposed to being part of the operation state, is cumbersome and a hindrance to providing true proxy surface support. For example, the clip must be copied from the surface onto the fallback image, but this was forgotten causing undue hassle in each backend. Another example is the contortion the meta surface endures to ensure the clip is correctly recorded. By contrast passing the clip along with the operation is quite simple and enables us to write generic handlers for providing surface wrappers. (And in the future, we should be able to write more esoteric wrappers, e.g. automatic 2x FSAA, trivially.) In brief, instead of the surface automatically applying the clip before calling the backend, the backend can call into a generic helper to apply clipping. For raster surfaces, clip regions are handled automatically as part of the composite interface. For vector surfaces, a clip helper is introduced to replay and callback into an intersect_clip_path() function as necessary. Whilst this is not primarily a performance related change (the change should just move the computation of the clip from the moment it is applied by the user to the moment it is required by the backend), it is important to track any potential regression: ppc: Speedups ======== image-rgba evolution-20090607-0 1026085.22 0.18% -> 672972.07 0.77%: 1.52x speedup ▌ image-rgba evolution-20090618-0 680579.98 0.12% -> 573237.66 0.16%: 1.19x speedup ▎ image-rgba swfdec-fill-rate-4xaa-0 460296.92 0.36% -> 407464.63 0.42%: 1.13x speedup ▏ image-rgba swfdec-fill-rate-2xaa-0 128431.95 0.47% -> 115051.86 0.42%: 1.12x speedup ▏ Slowdowns ========= image-rgba firefox-periodic-table-0 56837.61 0.78% -> 66055.17 3.20%: 1.09x slowdown ▏
2009-07-23 15:32:13 +01:00
if (extents) {
_trace_printf ("dict\n"
" /type /meta set\n"
" /content //%s set\n"
" /extents [%f %f %f %f] set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
extents->x, extents->y,
extents->width, extents->height,
surface_id);
_surface_object_set_size (ret, extents->width, extents->height);
} else {
_trace_printf ("dict\n"
" /type /meta set\n"
" /content //%s set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
surface_id);
}
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_status_t
cairo_meta_surface_replay (cairo_surface_t *meta, cairo_surface_t *target)
{
cairo_status_t
ret = DLCALL (cairo_meta_surface_replay, meta, target);
_emit_line_info ();
if (_write_lock ()) {
_emit_surface (target);
_emit_surface (meta);
_trace_printf ("replay");
_consume_operand ();
}
return ret;
}
#if CAIRO_HAS_VG_SURFACE
#include <cairo-vg.h>
cairo_surface_t *
cairo_vg_surface_create (cairo_vg_context_t *context,
cairo_content_t content,
int width, int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_vg_surface_create, context, content, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
_trace_printf ("dict\n"
" /type /vg set\n"
" /content //%s set\n"
" /width %d set\n"
" /height %d set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
width, height,
surface_id);
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
cairo_surface_t *
cairo_vg_surface_create_for_image (cairo_vg_context_t *context,
VGImage image,
VGImageFormat format,
int width, int height)
{
cairo_surface_t *ret;
long surface_id;
ret = DLCALL (cairo_vg_surface_create_for_image,
context, image, format, width, height);
surface_id = _create_surface_id (ret);
_emit_line_info ();
if (_write_lock ()) {
cairo_content_t content;
content = DLCALL (cairo_surface_get_content, ret);
_trace_printf ("dict\n"
" /type /vg set\n"
" /content //%s set\n"
" /width %d set\n"
" /height %d set\n"
" surface dup /s%ld exch def\n",
_content_to_string (content),
width, height,
surface_id);
_surface_object_set_size (ret, width, height);
_get_object (SURFACE, ret)->defined = true;
_push_operand (SURFACE, ret);
_write_unlock ();
}
return ret;
}
#endif