sergiotarxz/libvips
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

add lots of instrumentation

Browse Source 
also svg output
This commit is contained in:
John Cupitt 2013-11-26 14:57:56 +00:00
parent 3a195d4d5c
commit eb3ad6d64f
20 changed files with 167 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
libvips/arithmetic/arithmetic.c
View File

@ -478,6 +478,8 @@ vips_arithmetic_gen( VipsRegion *or,
p[i] = NULL;
q = (VipsPel *) VIPS_REGION_ADDR( or, r->left, r->top );
VIPS_GATE_START( "vips_arithmetic_gen: work" );
for( y = 0; y < r->height; y++ ) {
class->process_line( arithmetic, q, p, r->width );
@ -486,6 +488,8 @@ vips_arithmetic_gen( VipsRegion *or,
q += VIPS_REGION_LSKIP( or );
}
VIPS_GATE_STOP( "vips_arithmetic_gen: work" );
return( 0 );
}

4
libvips/colour/XYZ2Lab.c
View File

@ -122,10 +122,6 @@ vips_XYZ2Lab_line( VipsColour *colour, VipsPel *out, VipsPel **in, int width )
(void) g_once( &once, table_init, NULL );
g_mutex_lock( vips__global_lock );
VIPS_OPERATION( colour )->pixels += width;
g_mutex_unlock( vips__global_lock );
for( x = 0; x < width; x++ ) {
float nX, nY, nZ;
int i;

4
libvips/colour/colour.c
View File

@ -259,6 +259,8 @@ vips_colour_gen( VipsRegion *or,
if( vips_region_prepare( ir[i], r ) )
return( -1 );
VIPS_GATE_START( "vips_colour_gen: work" );
for( y = 0; y < r->height; y++ ) {
for( i = 0; ir[i]; i++ )
p[i] = VIPS_REGION_ADDR( ir[i], r->left, r->top + y );
@ -268,6 +270,8 @@ vips_colour_gen( VipsRegion *or,
class->process_line( colour, q, p, r->width );
}
VIPS_GATE_STOP( "vips_colour_gen: work" );
return( 0 );
}

4
libvips/conversion/bandary.c
View File

@ -100,6 +100,8 @@ vips_bandary_gen( VipsRegion *or, void *seq, void *a, void *b, gboolean *stop )
p[i] = NULL;
q = VIPS_REGION_ADDR( or, r->left, r->top );
VIPS_GATE_START( "vips_bandary_gen: work" );
for( y = 0; y < r->height; y++ ) {
class->process_line( bandary, q, p, r->width );
@ -108,6 +110,8 @@ vips_bandary_gen( VipsRegion *or, void *seq, void *a, void *b, gboolean *stop )
q += VIPS_REGION_LSKIP( or );
}
VIPS_GATE_STOP( "vips_bandary_gen: work" );
return( 0 );
}

4
libvips/conversion/cast.c
View File

@ -339,6 +339,8 @@ vips_cast_gen( VipsRegion *or, void *vseq, void *a, void *b,
if( vips_region_prepare( ir, r ) )
return( -1 );
VIPS_GATE_START( "vips_cast_gen: work" );
for( y = to; y < bo; y++ ) {
VipsPel *in = VIPS_REGION_ADDR( ir, le, y );
VipsPel *out = VIPS_REGION_ADDR( or, le, y );
@ -419,6 +421,8 @@ vips_cast_gen( VipsRegion *or, void *vseq, void *a, void *b,
}
}
VIPS_GATE_STOP( "vips_cast_gen: work" );
return( 0 );
}

20
libvips/conversion/embed.c
View File

@ -126,7 +126,8 @@ vips_embed_find_edge( VipsEmbed *embed, VipsRect *r, int i, VipsRect *out )
* right edge. If we're strictly up/down/left/right of the image, we
* can trim.
*/
if( i == 0 || i == 2 ) {
if( i == 0 ||
i == 2 ) {
VipsRect extend;
/* Above or below.
@ -136,7 +137,8 @@ vips_embed_find_edge( VipsEmbed *embed, VipsRect *r, int i, VipsRect *out )
extend.height = embed->height;
vips_rect_intersectrect( out, &extend, out );
}
if( i == 1 || i == 3 ) {
if( i == 1 ||
i == 3 ) {
VipsRect extend;
/* Left or right.
@ -176,6 +178,8 @@ vips_embed_paint_edge( VipsEmbed *embed,
VipsPel *q;
int y;
VIPS_GATE_START( "vips_embed_paint_edge: work" );
/* Pixels left to paint.
*/
todo = *r;
@ -209,6 +213,8 @@ vips_embed_paint_edge( VipsEmbed *embed,
memcpy( q, p, bs * todo.width );
}
}
VIPS_GATE_STOP( "vips_embed_paint_edge: work" );
}
static int
@ -258,19 +264,29 @@ vips_embed_gen( VipsRegion *or, void *seq, void *a, void *b, gboolean *stop )
switch( embed->extend ) {
case VIPS_EXTEND_BLACK:
case VIPS_EXTEND_WHITE:
VIPS_GATE_START( "vips_embed_gen: work1" );
/* Paint the borders a solid value.
*/
for( i = 0; i < 8; i++ )
vips_region_paint( or, &embed->border[i],
embed->extend == 0 ? 0 : 255 );
VIPS_GATE_STOP( "vips_embed_gen: work1" );
break;
case VIPS_EXTEND_BACKGROUND:
VIPS_GATE_START( "vips_embed_gen: work2" );
/* Paint the borders a solid value.
*/
for( i = 0; i < 8; i++ )
vips_region_paint_pel( or, &embed->border[i],
embed->ink );
VIPS_GATE_STOP( "vips_embed_gen: work2" );
break;
case VIPS_EXTEND_COPY:

8
libvips/conversion/sequential.c
View File

@ -129,8 +129,12 @@ vips_sequential_generate( VipsRegion *or,
"request for line %d, height %d",
r->top, r->height );
VIPS_GATE_START( "vips_sequential_generate: wait" );
g_mutex_lock( sequential->lock );
VIPS_GATE_STOP( "vips_sequential_generate: wait" );
VIPS_DEBUG_MSG_GREEN( "thread %p has lock ...\n", g_thread_self() );
/* If we've seen an error, everything must stop.
@ -167,6 +171,8 @@ vips_sequential_generate( VipsRegion *or,
VIPS_DEBUG_MSG_GREEN( "thread %p stalling for up to %gs ...\n",
g_thread_self(), STALL_TIME );
VIPS_GATE_START( "vips_sequential_generate: wait" );
/* Exit the loop on timeout or condition passes. We have to
* be wary of spurious wakeups.
*/
@ -181,6 +187,8 @@ vips_sequential_generate( VipsRegion *or,
break;
#endif
VIPS_GATE_STOP( "vips_sequential_generate: wait" );
VIPS_DEBUG_MSG_GREEN( "thread %p awake again ...\n",
g_thread_self() );
}

21
libvips/conversion/tilecache.c
View File

@ -598,8 +598,12 @@ vips_tile_cache_gen( VipsRegion *or,
GSList *work;
GSList *p;
VIPS_GATE_START( "vips_tile_cache_gen: wait1" );
g_mutex_lock( cache->lock );
VIPS_GATE_STOP( "vips_tile_cache_gen: wait1" );
VIPS_DEBUG_MSG( "vips_tile_cache_gen: "
"left = %d, top = %d, width = %d, height = %d\n",
r->left, r->top, r->width, r->height );
@ -661,9 +665,16 @@ vips_tile_cache_gen( VipsRegion *or,
&tile->pos,
tile->pos.left, tile->pos.top );
if( cache->threaded )
if( cache->threaded ) {
VIPS_GATE_START( "vips_tile_cache_gen: "
"wait2" );
g_mutex_lock( cache->lock );
VIPS_GATE_STOP( "vips_tile_cache_gen: "
"wait2" );
}
/* If there was an error calculating this
* tile, just warn and carry on.
*
@ -713,8 +724,12 @@ vips_tile_cache_gen( VipsRegion *or,
VIPS_DEBUG_MSG( "vips_tile_cache_gen: waiting\n" );
VIPS_GATE_START( "vips_tile_cache_gen: wait3" );
g_cond_wait( cache->new_tile, cache->lock );
VIPS_GATE_STOP( "vips_tile_cache_gen: wait3" );
VIPS_DEBUG_MSG( "vips_tile_cache_gen: awake!\n" );
}
}
@ -861,8 +876,12 @@ vips_line_cache_gen( VipsRegion *or,
{
VipsBlockCache *block_cache = (VipsBlockCache *) b;
VIPS_GATE_START( "vips_line_cache_gen: wait" );
g_mutex_lock( block_cache->lock );
VIPS_GATE_STOP( "vips_line_cache_gen: wait" );
/* We size up the cache to the largest request.
*/
if( or->valid.height >

12
libvips/convolution/im_conv.c
View File

@ -710,6 +710,8 @@ conv_gen( REGION *or, void *vseq, void *a, void *b )
if( im_prepare( ir, &s ) )
return( -1 );
VIPS_GATE_START( "conv_gen: work" );
/* Fill offset array. Only do this if the bpl has changed since the
* previous im_prepare().
*/
@ -774,6 +776,8 @@ conv_gen( REGION *or, void *vseq, void *a, void *b )
}
}
VIPS_GATE_STOP( "conv_gen: work" );
return( 0 );
}
@ -880,6 +884,8 @@ conv3x3_gen( REGION *or, void *vseq, void *a, void *b )
if( im_prepare( ir, &s ) )
return( -1 );
VIPS_GATE_START( "conv3x3_gen: work" );
for( y = to; y < bo; y++ ) {
switch( in->BandFmt ) {
case IM_BANDFMT_UCHAR:
@ -927,6 +933,8 @@ conv3x3_gen( REGION *or, void *vseq, void *a, void *b )
}
}
VIPS_GATE_STOP( "conv3x3_gen: work" );
return( 0 );
}
@ -958,6 +966,8 @@ convvec_gen( REGION *or, void *vseq, void *a, void *b )
if( im_prepare( ir, &s ) )
return( -1 );
VIPS_GATE_START( "convvec_gen: work" );
for( j = 0; j < conv->n_pass; j++ )
vips_executor_set_program( &convolve[j],
conv->pass[j].vector, sz );
@ -1006,6 +1016,8 @@ convvec_gen( REGION *or, void *vseq, void *a, void *b )
#endif /*DEBUG_PIXELS*/
}
VIPS_GATE_STOP( "convvec_gen: work" );
return( 0 );
}

4
libvips/convolution/sharpen.c
View File

@ -138,6 +138,8 @@ vips_sharpen_generate( VipsRegion *or,
vips_region_prepare( in[1], r ) )
return( -1 );
VIPS_GATE_START( "vips_sharpen_generate: work" );
for( y = 0; y < r->height; y++ ) {
short *p1 = (short *)
VIPS_REGION_ADDR( in[0], r->left, r->top + y );
@ -179,6 +181,8 @@ vips_sharpen_generate( VipsRegion *or,
}
}
VIPS_GATE_STOP( "vips_sharpen_generate: work" );
return( 0 );
}

4
libvips/foreign/jpeg2vips.c
View File

@ -888,6 +888,8 @@ read_jpeg_generate( VipsRegion *or,
printf( "read_jpeg_generate: line %d, %d rows\n",
r->top, r->height );
#endif /*DEBUG*/
VIPS_GATE_START( "read_jpeg_generate: work" );
/* We're inside a tilecache where tiles are the full image width, so
* this should always be true.
@ -926,6 +928,8 @@ read_jpeg_generate( VipsRegion *or,
}
}
VIPS_GATE_STOP( "read_jpeg_generate: work" );
return( 0 );
}

2
libvips/include/vips/operation.h
View File

@ -74,7 +74,7 @@ typedef struct _VipsOperation {
guint hash;
gboolean found_hash;
/* Pixels calculated ... handy for measuring over calculation.
/* Pixels calculated ... handy for measuring over-calculation.
*/
int pixels;

6
libvips/iofuncs/image.c
View File

@ -960,9 +960,15 @@ vips_image_real_invalidate( VipsImage *image )
{
VIPS_DEBUG_MSG( "vips_image_real_invalidate: %p\n", image );
VIPS_GATE_START( "vips_image_real_invalidate: wait" );
g_mutex_lock( image->sslock );
VIPS_GATE_STOP( "vips_image_real_invalidate: wait" );
(void) vips_slist_map2( image->regions,
(VipsSListMap2Fn) vips_region_invalidate, NULL, NULL );
g_mutex_unlock( image->sslock );
}

33
libvips/iofuncs/region.c
View File

@ -203,9 +203,15 @@ vips__region_start( VipsRegion *region )
VipsImage *image = region->im;
if( !region->seq && image->start_fn ) {
VIPS_GATE_START( "vips__region_start: wait" );
g_mutex_lock( image->sslock );
VIPS_GATE_STOP( "vips__region_start: wait" );
region->seq = image->start_fn( image,
image->client1, image->client2 );
g_mutex_unlock( image->sslock );
if( !region->seq ) {
@ -229,9 +235,15 @@ vips__region_stop( VipsRegion *region )
if( region->seq && image->stop_fn ) {
int result;
VIPS_GATE_START( "vips__region_stop: wait" );
g_mutex_lock( image->sslock );
VIPS_GATE_STOP( "vips__region_stop: wait" );
result = image->stop_fn( region->seq,
image->client1, image->client2 );
g_mutex_unlock( image->sslock );
/* stop function can return an error, but we have nothing we
@ -279,9 +291,16 @@ vips_region_dispose( GObject *gobject )
/* Detach from image.
*/
VIPS_GATE_START( "vips_region_dispose: wait" );
g_mutex_lock( image->sslock );
VIPS_GATE_STOP( "vips_region_dispose: wait" );
image->regions = g_slist_remove( image->regions, region );
g_mutex_unlock( image->sslock );
region->im = NULL;
g_object_unref( image );
@ -363,8 +382,12 @@ vips__region_take_ownership( VipsRegion *region )
/* Lock so that there's a memory barrier with the thread doing the
* vips__region_no_ownership() before us.
*/
VIPS_GATE_START( "vips__region_take_ownership: wait" );
g_mutex_lock( region->im->sslock );
VIPS_GATE_STOP( "vips__region_take_ownership: wait" );
if( region->thread != g_thread_self() ) {
g_assert( region->thread == NULL );
@ -399,8 +422,12 @@ vips__region_check_ownership( VipsRegion *region )
void
vips__region_no_ownership( VipsRegion *region )
{
VIPS_GATE_START( "vips__region_no_ownership: wait" );
g_mutex_lock( region->im->sslock );
VIPS_GATE_STOP( "vips__region_no_ownership: wait" );
vips__region_check_ownership( region );
region->thread = NULL;
@ -425,8 +452,14 @@ vips_region_build( VipsObject *object )
/* We're usually inside the ss lock anyway. But be safe ...
*/
VIPS_GATE_START( "vips_region_build: wait" );
g_mutex_lock( image->sslock );
VIPS_GATE_STOP( "vips_region_build: wait" );
image->regions = g_slist_prepend( image->regions, region );
g_mutex_unlock( image->sslock );
return( 0 );

4
libvips/iofuncs/semaphore.c
View File

@ -116,6 +116,8 @@ vips_semaphore_downn( VipsSemaphore *s, int n )
{
int value_after_op;
VIPS_GATE_START( "vips_semaphore_downn: wait" );
g_mutex_lock( s->mutex );
while( s->v < n )
@ -130,6 +132,8 @@ vips_semaphore_downn( VipsSemaphore *s, int n )
s->name, n, value_after_op );
#endif /*DEBUG_IO*/
VIPS_GATE_STOP( "vips_semaphore_downn: wait" );
return( value_after_op );
}

2
libvips/iofuncs/sink.c
View File

@ -234,7 +234,7 @@ sink_init( Sink *sink,
sink->b = b;
if( !(sink->t = vips_image_new()) ||
im_copy( sink->sink_base.im, sink->t ) ) {
vips_image_write( sink->sink_base.im, sink->t ) ) {
sink_free( sink );
return( -1 );
}

8
libvips/iofuncs/sinkdisc.c
View File

@ -169,8 +169,10 @@ wbuffer_write( WriteBuffer *wbuffer )
wbuffer->region->bpl * wbuffer->area.height, wbuffer );
VIPS_GATE_START( "wbuffer_write: " );
wbuffer->write_errno = write->write_fn( wbuffer->region,
&wbuffer->area, write->a );
VIPS_GATE_STOP( "wbuffer_write: " );
}
@ -184,18 +186,14 @@ wbuffer_write_thread( void *data )
for(;;) {
/* Wait to be told to write.
*/
VIPS_GATE_START( "wbuffer_write_thread: wait go" );
vips_semaphore_down( &wbuffer->go );
VIPS_GATE_STOP( "wbuffer_write_thread: wait go" );
if( wbuffer->kill )
break;
/* Now block until the last worker finishes on this buffer.
*/
VIPS_GATE_START( "wbuffer_write_thread: wait tile" );
vips_semaphore_downn( &wbuffer->nwrite, 0 );
VIPS_GATE_STOP( "wbuffer_write_thread: wait tile" );
wbuffer_write( wbuffer );
@ -513,7 +511,7 @@ vips_sink_disc( VipsImage *im, VipsRegionWrite write_fn, void *a )
* started (if the allocate failed), and in any case, we don't care if
* the final write went through or not.
*/
if( !result )
if( !result )
vips_semaphore_down( &write.buf->done );
vips_image_posteval( im );

6
libvips/iofuncs/threadpool.c
View File

@ -546,11 +546,11 @@ vips_thread_work_unit( VipsThread *thr )
if( thr->error )
return;
VIPS_GATE_START( "vips_thread_work_unit: allocate-wait" );
VIPS_GATE_START( "vips_thread_work_unit: wait" );
g_mutex_lock( pool->allocate_lock );
VIPS_GATE_STOP( "vips_thread_work_unit: allocate-wait" );
VIPS_GATE_STOP( "vips_thread_work_unit: wait" );
/* Has another worker signaled stop while we've been working?
*/
@ -577,12 +577,10 @@ vips_thread_work_unit( VipsThread *thr )
/* Process a work unit.
*/
VIPS_GATE_START( "vips_thread_work_unit: work" );
if( pool->work( thr->state, pool->a ) ) {
thr->error = TRUE;
pool->error = TRUE;
}
VIPS_GATE_STOP( "vips_thread_work_unit: work" );
}
/* What runs as a thread ... loop, waiting to be told to do stuff.

2
tools/Makefile.am
View File

@ -26,6 +26,7 @@ bin_SCRIPTS = \
batch_image_convert \
batch_rubber_sheet \
batch_crop \
vipsprofile \
vips-7.37
EXTRA_DIST = \
@ -42,4 +43,5 @@ install-exec-hook:
chmod ugo+x ${DESTDIR}${bindir}/batch_image_convert
chmod ugo+x ${DESTDIR}${bindir}/batch_rubber_sheet
chmod ugo+x ${DESTDIR}${bindir}/batch_crop
chmod ugo+x ${DESTDIR}${bindir}/vipsprofile

41
tools/vipsprofile.py → tools/vipsprofile
View File

@ -135,6 +135,23 @@ for thread in threads:
print 'last time =', last_time
# calculate some simple stats
print 'name\t\t\talive\twait\twait%'
for thread in threads:
start = last_time
stop = 0
wait = 0
for event in thread.events:
if event.start < start:
start = event.start
if event.stop > stop:
stop = event.stop
if event.wait:
wait += event.stop - event.start
alive = stop - start
print thread.thread_name, alive, wait, 100 * wait / alive
# do two gates overlap?
def is_overlap(events, gate_name1, gate_name2):
for event1 in events:
@ -184,17 +201,25 @@ PIXELS_PER_SECOND = 1000
PIXELS_PER_GATE = 20
LEFT_BORDER = 320
BAR_HEIGHT = 5
WIDTH = int(LEFT_BORDER + last_time * PIXELS_PER_SECOND)
WIDTH = int(LEFT_BORDER + last_time * PIXELS_PER_SECOND) + 50
HEIGHT = int((total_y + 1) * PIXELS_PER_GATE)
surface = cairo.ImageSurface(cairo.FORMAT_ARGB32, WIDTH + 50, HEIGHT)
output_filename = "vips-profile.svg"
print 'writing to', output_filename
surface = cairo.SVGSurface(output_filename, WIDTH, HEIGHT)
ctx = cairo.Context(surface)
ctx.select_font_face('Sans')
ctx.set_font_size(15)
ctx.rectangle(0, 0, WIDTH, HEIGHT)
ctx.set_source_rgba(0.0, 0.0, 0.3, 1.0)
ctx.fill()
def draw_event(ctx, event):
left = event.start * PIXELS_PER_SECOND + LEFT_BORDER
top = event.total_y * PIXELS_PER_GATE
top = event.total_y * PIXELS_PER_GATE + BAR_HEIGHT / 2
width = (event.stop - event.start) * PIXELS_PER_SECOND - 1
height = BAR_HEIGHT
@ -216,19 +241,19 @@ def draw_event(ctx, event):
ctx.move_to(left + width / 2 - twidth / 2, top + theight)
ctx.set_source_rgb(1.00, 0.83, 0.00)
ctx.show_text(event.gate_name)
#ctx.stroke()
for thread in threads:
ctx.rectangle(0, thread.total_y * PIXELS_PER_GATE, WIDTH, 1)
ctx.set_source_rgb(1.00, 1.00, 1.00)
ctx.fill()
xbearing, ybearing, twidth, theight, xadvance, yadvance = \
ctx.text_extents(thread.thread_name)
ctx.move_to(0, theight + thread.total_y * PIXELS_PER_GATE)
ctx.set_source_rgb(1.00, 1.00, 1.00)
ctx.show_text(thread.thread_name)
#ctx.stroke()
for event in thread.events:
draw_event(ctx, event)
output_filename = "example.png"
print 'writing to', output_filename
surface.write_to_png(output_filename)
surface.finish()