Re: [Gegl-developer] speeding up GEGL operations in GIMP
Hi,
On 10:06, Thu 15 May 08, Sven Neumann wrote:
Changing this in GIMP is not feasible. What needs to be done is to
implement shortcuts for the conversions that GIMP actually uses.
Eventually he gimp-8bit extension should provide shortcuts for both
gamma-corrected and linear 8bit - double conversions.
attached is a patch that adds RGBA float-RGB u8 conversion
using code already present in gggl-lies extension based on
lookup table. I can try to move the code to the gimp-8bit
extension, although I don't think there is a point to doing
so, since this would only introduce duplicity and whould
increase memory footprint of BABL. However, for this
conversion to be selected, BABL_ERROR must be set at
least to 0.0005.
The lookup tables in gegl-fixups are rather large, btw. As pointed out
in the paper, a much smaller lookup table would be sufficient to cover
the range from 0.0 to 1.0. I guess that adding a clever range check and
reducing the size of the lookup tables would yield a performance
improvement due to better cache coherency.
The lookup tables for float-u8 conversions are rather quite
large and it is indeed possible to add a range check to size
them down. However this range check would have to be
present in the inner loop of a conversion, which would
rather slow things down. Further, it wouldn't IMHO improve cache
coherency much, since the pixels already are in the 0-1 range
most of the time meaning only approx. 2.5% of the lookup
table items are frequently touched.
Regards,
Jan
Index: extensions/gggl-lies.c
===
--- extensions/gggl-lies.c (revision 311)
+++ extensions/gggl-lies.c (working copy)
@@ -62,13 +62,17 @@ static float table_16_F[1 16
static unsigned char table_F_8[1 16];
static unsigned short table_F_16[1 16];
-
static int table_inited = 0;
static void
table_init (void)
{
int i;
+ union
+ {
+float f;
+unsigned short s[2];
+ } u;
if (table_inited)
return;
@@ -84,64 +88,48 @@ table_init (void)
table_16_F[i] = (i * 1.0) / 65535.0;
}
/* fill tables for conversion from float to integer */
- {
-union
+
+ u.s[0] = 0;
+ for (i = 0; i 1 16; i++)
{
- float f;
- unsigned short s[2];
-} u;
-u.f = 0.0;
-
-u.s[0] = 0.0;
-
-for (i = 0; i 1 16; i++)
- {
-unsigned char c;
-unsigned short s;
+ unsigned char c;
+ unsigned short s;
-u.s[1] = i;
+ u.s[1] = i;
-if (u.f = 0.0)
- {
-c = 0;
-s = 0;
- }
-else if (u.f = 1.0)
- {
-c = 255;
-s = 65535;
- }
-else
- {
-c = rint (u.f * 255.0);
-s = rint (u.f * 65535.0);
- }
+ if (u.f = 0.0)
+{
+ c = 0;
+ s = 0;
+}
+ else if (u.f = 1.0)
+{
+ c = 255;
+ s = 65535;
+}
+ else
+{
+ c = rint (u.f * 255.0);
+ s = rint (u.f * 65535.0);
+}
/*fprintf (stderr, %2.3f=%03i %05i , f, c, (*hi));
/ if (! ((*hi)%9))
/ fprintf (stderr, \n); */
-table_F_8[u.s[1]] = c;
-table_F_16[u.s[1]] = s;
- }
- }
- /* fix tables to ensure 1:1 conversions back and forth */
- if (0)
-{ /*FIXME: probably not the right way to do
it,.. must sit down and scribble on paper */
- int i;
- for (i = 0; i 256; i++)
-{
- float f = table_8_F[i];
- unsigned short *hi = ((unsigned short *) (void *) f);
- unsigned short *lo = ((unsigned short *) (void *) f);
- *lo = 0;
- table_F_8[(*hi)] = i;
-}
+ table_F_8[u.s[1]] = c;
+ table_F_16[u.s[1]] = s;
}
+ /* fix tables to ensure 1:1 conversions back and forth */
+for (i = 0; i 256; i++)
+ {
+u.f = table_8_F[i];
+table_F_8[u.s[1]] = i;
+ }
}
/* function to find the index in table for a float */
-static unsigned int
+static INLINE unsigned int
gggl_float_to_index16 (float f)
{
union
@@ -904,6 +892,33 @@ conv_rgbaF_rgbA16 (unsigned char *src, u
return samples;
}
+#ifdef USE_TABLES
+
+static INLINE long
+conv_rgbaF_rgb8 (unsigned char *src, unsigned char *dst, long samples)
+{
+ long n = samples;
+ register float f;
+ float *fsrc = (float *) src;
+
+ while (n--)
+{
+ f = (*(float *) fsrc++);
+ *(unsigned char *) dst++ = table_F_8[gggl_float_to_index16 (f)];
+
+ f = (*(float *) fsrc++);
+ *(unsigned char *) dst++ = table_F_8[gggl_float_to_index16 (f)];
+
+ f = (*(float *) fsrc++);
+ *(unsigned char *) dst++ = table_F_8[gggl_float_to_index16 (f)];
+
+ fsrc++;
+}
+ return samples;
+}
+
+#else
+
static INLINE long
conv_rgbaF_rgb8 (unsigned char *src, unsigned
Re: [Gegl-developer] speeding up GEGL operations in GIMP
Hi Sven, I can give it a try. If I understand it correctly, gimp-8bit.c only implements u8-float conversions, which seem to be picked up by BABL correctly, so the problem lies in the float-u8 conversions that are computed by ReferenceFish. Is that correct? Regards, Jan On 09:14, Wed 14 May 08, Sven Neumann wrote: Hi, currently the GEGL operations in GIMP are very slow. I have done some basic profiling yesterday and it appears that the main problem is the conversion from floating point to 8bit. Here is where the most time is being spent. So it doesn't really make much sense to optimize the operations in GIMP or the point filters in GEGL. We need to look at the babl conversions first. Here's an interesting paper that outlines how to do conversion from linear light floating point image data to 8-bit sRGB using a relatively small lookup table: http://mysite.verizon.net/spitzak/conversion/sketches_0265.pdf That is exactly the conversion that the tile sink executes. It would help us a lot if someone could implement this. The file extensions/gimp-8bit.c would probably be the right place to put this code. I am afraid I am not going to find time for this. Is anyone else interested in working on this? Sven ___ Gegl-developer mailing list Gegl-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer
Re: [Gegl-developer] speeding up GEGL operations in GIMP
Hi, On 11:19, Wed 14 May 08, Richard Kralovic wrote: I just had a short look into the babl code; it looks like there is a float-u8 conversion implemented using the 16bit lookup table in extensions/gegl-fixups.c. In fact, it looks like the same algorithm as described in the pdf paper. I do not have much time to look at it now, I just noticed that the function gggl_float_to_index16 (line 130) is not inlined, but I guess it should be. Could you try to check if that helps? The gegl-fixups.c code deals with gamma corrected RGB u8 conversions, so the inlining would affect RGBA float - R'G'B' u8 conversion. However, after loading a jpeg file into gimp-2.5 and the using a color tool, conversions from RGB u8-RGBA float and RGBA float - RGB u8 are requested, that is non-gamma corrected RGB u8. Is there a simple way how to make gimp request conversions from/to R'G'B'u8 space? On 09:14, Wed 14 May 08, Sven Neumann wrote: Hi, currently the GEGL operations in GIMP are very slow. I have done some basic profiling yesterday and it appears that the main problem is the conversion from floating point to 8bit. Here is where the most time is being spent. So it doesn't really make much sense to optimize the operations in GIMP or the point filters in GEGL. We need to look at the babl conversions first. As you correctly pointed out on IRC, there is currently no speedup code to conform to 0.0001 default BABL error setting, so the RGBA float - RGB u8 is handled by ReferenceFish and that is the reason the code is so slow. After setting the BABL_ERROR env. variable to 0.1, the RGBA float-RGB float-RGB u8 is selected, making it a bit faster. There is already RGBA float-RGB u8 shortcut in the babl codebase in the gggl-lies extension, however this one seems to work slower and introducing higher error than the RGBA float-RGB float-RGB u8 path. The RGB u8-RGBA float conversion is correctly handled by the gimp-8bit extension, performing approx. twice as fast as the RGBA float-RGB float-RGB u8 path. Regards Jan ___ Gegl-developer mailing list Gegl-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer
Re: [Gegl-developer] Color temperature correction GeglOperation
Hi, On 18:36, Sat 26 Apr 08, Kevin Cozens wrote: Jan Heller wrote: On 11:09, Sat 26 Apr 08, Øyvind Kolås wrote: It would also be nice to replace the planckian locus lookup table with a function that approximates it. I played with MATLAB for a while and came up with rational functions of degree 5 that approximate the Planckian locus dataset reasonably well. Attached is a modified version of the operation using these approximations. I haven't looked at the code yet, but I would have thought execution would be faster using the lookup table than having to calculate values using a formula especially given that the formula only approximates the values of the table. In the previous version of the code the values outside the lookup table got interpolated anyway. The approximating functions are really quite good and pretty fast. Since this computation only takes place only at the beginning of the operation and not per pixel basis, the speed impairment is really negligible. Regards, Jan ___ Gegl-developer mailing list Gegl-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer
Re: [Gegl-developer] Color temperature correction GeglOperation
Hi,
On 11:09, Sat 26 Apr 08, Øyvind Kolås wrote:
On Fri, Apr 25, 2008 at 6:52 PM, Jan Heller [EMAIL PROTECTED] wrote:
I wrote it to better familiarize myself with
Gegl and I am posting it here in hope it will be useful for
others.
I think it is a good operation to have, so I have commited a slightly
modified version to svn.
Nice, thanks!
GEGL and babl deals with the out of gamut handling themselves at a
later stage, during processing we preserve headroom and footroom. This
allows us to change the contrast of the image to bring details back in
later. The implemented gamut handling also seemed to introduce banding
that the conversions in babl does not.
Good to know.
It would also be nice to replace the planckian locus lookup table with
a function that approximates it.
I played with MATLAB for a while and came up with rational
functions of degree 5 that approximate the Planckian locus
dataset reasonably well. Attached is a modified version of
the operation using these approximations.
Regards,
Jan
/* This file is an image processing operation for GEGL
*
* GEGL is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 3 of the License, or (at your option) any later version.
*
* GEGL 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with GEGL; if not, see http://www.gnu.org/licenses/.
*
* Copyright 2008 Jan Heller jan.heller (at) matfyz.cz
*/
#ifdef GEGL_CHANT_PROPERTIES
gegl_chant_double (original_temp, Original temperature, 1000, 12000, 6500, Estimated temperature of the light source in K the image was taken with.)
gegl_chant_double (intended_temp, Intended temperature, 1000, 12000, 6500, Corrected estimation of the temperature of the light source in K.)
#else
#define GEGL_CHANT_TYPE_POINT_FILTER
#define GEGL_CHANT_C_FILE color-temperature.c
#include gegl-chant.h
#define LOWEST_TEMPERATURE 1000
#define HIGHEST_TEMPERATURE 12000
gfloat rgb_r55[][12];
static void
convert_k_to_rgb (gfloat temperature,
gfloat *rgb)
{
gfloat nomin, denom;
intchannel, deg;
if (temperature LOWEST_TEMPERATURE)
temperature = LOWEST_TEMPERATURE;
if (temperature HIGHEST_TEMPERATURE)
temperature = HIGHEST_TEMPERATURE;
/* Evaluation of an approximation of the Planckian locus in linear RGB space
* by rational functions of degree 5 using Horner's scheme
* f(x) = (p1*x^5 + p2*x^4 + p3*x^3 + p4*x^2 + p5*x + p6) /
*(x^5 + q1*x^4 + q2*x^3 + q3*x^2 + q4*x + q5)
*/
for (channel = 0; channel 3; channel++)
{
nomin = rgb_r55[channel][0];
for (deg = 1; deg 6; deg++)
nomin = nomin * temperature + rgb_r55[channel][deg];
denom = rgb_r55[channel][6];
for (deg = 1; deg 6; deg++)
denom = denom * temperature + rgb_r55[channel][6 + deg];
rgb[channel] = nomin / denom;
}
}
static void prepare (GeglOperation *operation)
{
Babl *format = babl_format (RGBA float);
gegl_operation_set_format (operation, input, format);
gegl_operation_set_format (operation, output, format);
}
/* GeglOperationPointFilter gives us a linear buffer to operate on
* in our requested pixel format
*/
static gboolean
process (GeglOperation *op,
void *in_buf,
void *out_buf,
glong n_pixels)
{
GeglChantO *o = GEGL_CHANT_PROPERTIES (op);
gfloat *in_pixel;
gfloat *out_pixel;
gfloat original_temp, original_temp_rgb[3];
gfloat intended_temp, intended_temp_rgb[3];
gfloat coefs[3];
glong i;
in_pixel = in_buf;
out_pixel = out_buf;
original_temp = o-original_temp;
intended_temp = o-intended_temp;
convert_k_to_rgb (original_temp, original_temp_rgb);
convert_k_to_rgb (intended_temp, intended_temp_rgb);
coefs[0] = original_temp_rgb[0] / intended_temp_rgb[0];
coefs[1] = original_temp_rgb[1] / intended_temp_rgb[1];
coefs[2] = original_temp_rgb[2] / intended_temp_rgb[2];
for (i = 0; i n_pixels; i++)
{
out_pixel[0] = in_pixel[0] * coefs[0];
out_pixel[1] = in_pixel[1] * coefs[1];
out_pixel[2] = in_pixel[2] * coefs[2];
in_pixel += 4;
out_pixel += 4;
}
return TRUE;
}
static void
gegl_chant_class_init (GeglChantClass *klass)
{
GeglOperationClass*operation_class;
GeglOperationPointFilterClass *point_filter_class;
operation_class= GEGL_OPERATION_CLASS (klass);
point_filter_class = GEGL_OPERATION_POINT_FILTER_CLASS (klass);
point_filter_class-process = process;
operation_class-prepare = prepare
[Gegl-developer] Color temperature correction GeglOperation
Hi,
attached is a GeglOperation performing color temperature
correction. This is of course analogical to the
whitebalance operation that is already in the Gegl source
tree, though it uses photographer friendly input values. It
has two parameters, estimated temperature of the light
source in K the image was taken with and Corrected
estimation of the temperature of the light source in K. The
computation is a rather simple linear interpolation of a
data table. I wrote it to better familiarize myself with
Gegl and I am posting it here in hope it will be useful for
others.
Regards,
Jan
/* This file is an image processing operation for GEGL
*
* GEGL is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 3 of the License, or (at your option) any later version.
*
* GEGL 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with GEGL; if not, see http://www.gnu.org/licenses/.
*
* Copyright 2008 Jan Heller [EMAIL PROTECTED]
*/
#ifdef GEGL_CHANT_PROPERTIES
gegl_chant_double (original_temp, Original temperature, 1000, 12000, 6500, Estimated temperature of the light source in K the image was taken with.)
gegl_chant_double (intended_temp, Intended temperature, 1000, 12000, 6500, Corrected estimation of the temperatureof the light source in K.)
#else
#define GEGL_CHANT_TYPE_POINT_FILTER
#define GEGL_CHANT_C_FILE color-temperature.c
#include gegl-chant.h
#define LOWEST_TEMPERATURE 1000
#define HIGHEST_TEMPERATURE 12000
#define TEMPERATURE_STEP 20
#define LERP(x, x0, y0, x1, y1) (y0 + (x - x0) * (gfloat) (y1 - y0) / (gfloat) (x1 - x0))
gfloat planckian_locus[][3];
static void
convert_k_to_rgb (gfloat temperature,
gfloat *rgb)
{
intilow, ihigh;
intdelta_temperature;
gfloat resid_temperature;
if (temperature LOWEST_TEMPERATURE)
temperature = LOWEST_TEMPERATURE;
if (temperature HIGHEST_TEMPERATURE)
temperature = HIGHEST_TEMPERATURE;
delta_temperature = temperature - LOWEST_TEMPERATURE;
ilow = delta_temperature / TEMPERATURE_STEP;
ihigh = (delta_temperature % TEMPERATURE_STEP) ? ilow + 1 : ilow;
resid_temperature = temperature - (ilow * TEMPERATURE_STEP + LOWEST_TEMPERATURE);
rgb[0] = LERP(resid_temperature,
0,
planckian_locus[ilow][0],
TEMPERATURE_STEP,
planckian_locus[ihigh][0]);
rgb[1] = LERP(resid_temperature,
0,
planckian_locus[ilow][1],
TEMPERATURE_STEP,
planckian_locus[ihigh][1]);
rgb[2] = LERP(resid_temperature,
0,
planckian_locus[ilow][2],
TEMPERATURE_STEP,
planckian_locus[ihigh][2]);
}
static void prepare (GeglOperation *operation)
{
Babl *format = babl_format (RGBA float);
gegl_operation_set_format (operation, input, format);
gegl_operation_set_format (operation, output, format);
}
/* GeglOperationPointFilter gives us a linear buffer to operate on
* in our requested pixel format
*/
static gboolean
process (GeglOperation *op,
void *in_buf,
void *out_buf,
glong n_pixels)
{
GeglChantO *o = GEGL_CHANT_PROPERTIES (op);
gfloat *in_pixel;
gfloat *out_pixel;
gfloat original_temp, original_temp_rgb[3];
gfloat intended_temp, intended_temp_rgb[3];
gfloat coefs[3];
glong i;
gfloat min, max;
in_pixel = in_buf;
out_pixel = out_buf;
original_temp = o-original_temp;
intended_temp = o-intended_temp;
convert_k_to_rgb (original_temp, original_temp_rgb);
convert_k_to_rgb (intended_temp, intended_temp_rgb);
coefs[0] = original_temp_rgb[0] / intended_temp_rgb[0];
coefs[1] = original_temp_rgb[1] / intended_temp_rgb[1];
coefs[2] = original_temp_rgb[2] / intended_temp_rgb[2];
for (i = 0; i n_pixels; i++)
{
out_pixel[0] = in_pixel[0] * coefs[0];
out_pixel[1] = in_pixel[1] * coefs[1];
out_pixel[2] = in_pixel[2] * coefs[2];
/* handle out-of-gamut colors */
min = -MIN(MIN(out_pixel[0], out_pixel[1]), MIN(out_pixel[2], 0));
out_pixel[0] += min;
out_pixel[1] += min;
out_pixel[2] += min;
max = MAX(MAX(out_pixel[0], out_pixel[1]), MAX(out_pixel[2], 1)) - 1;
out_pixel[0] -= max;
out_pixel[1] -= max;
out_pixel[2] -= max;
in_pixel += 4;
out_pixel += 4;
}
return TRUE;
}
static void
gegl_chant_class_init (GeglChantClass *klass)
{
GeglOperationClass*operation_class
[Gegl-developer] BABL path vs. reference fish
Hi,
attached is a path that introduces logic to disallow
creation of a fish path that is actually slower than
appropriate reference fish.
Suggested Changelog entry:
* babl/babl-fish-path.c (get_conversion_path), (babl_fish_path),
(test_create), (init_path_instrumentation),
(destroy_path_instrumentation), (get_path_instrumentation): Improved
fish path instrumentation during the search for a new one to optimize
for speed. Added logic to disallow creation of a fish path that is
actually slower than appropriate reference fish.
* babl/babl-util.[ch] (babl_process_cost): New function for unified
timing formula for babl processings.
* babl/babl-conversion.c (babl_conversion_error): Make use of the
new babl_process_cost function.
* extensions/gimp-8bit.c: Added a function declaration to prevent
compiler warning.
Regards,
Jan
Index: extensions/gimp-8bit.c
===
--- extensions/gimp-8bit.c (revision 304)
+++ extensions/gimp-8bit.c (working copy)
@@ -296,6 +296,7 @@ conv_g8_gamma_2_2_rgbaF_linear (unsigned
return samples;
}
+int init (void);
int
init (void)
Index: babl/babl-fish-path.c
===
--- babl/babl-fish-path.c (revision 304)
+++ babl/babl-fish-path.c (working copy)
@@ -23,7 +23,17 @@
#define BABL_MAX_COST_VALUE 200
static double
-get_conversion_path_error (BablList *path);
+init_path_instrumentation (Babl *fmt_source,
+ Babl *fmt_destination);
+
+static void
+destroy_path_instrumentation (void);
+
+static void
+get_path_instrumentation (BablList *path,
+ double *path_cost,
+ double *ref_cost,
+ double *path_error);
static long
process_conversion_path (BablList *path,
@@ -118,28 +128,32 @@ get_conversion_path (Babl *current_forma
{
/* We have found a candidate path, let's
* see about it's properties */
- double temp_cost = 0.0;
- double temp_error = 1.0;
+ double path_cost = 0.0;
+ double ref_cost = 0.0;
+ double path_error = 1.0;
inti;
for (i = 0; i babl_list_size (current_path); i++)
{
- temp_error *= (1.0 + babl_conversion_error ((BablConversion *)
current_path-items[i]));
- temp_cost += babl_conversion_cost ((BablConversion *)
current_path-items[i]);
+ path_error *= (1.0 + babl_conversion_error ((BablConversion *)
current_path-items[i]));
}
- if (temp_cost fish_path-fish_path.cost
- temp_error - 1.0 = legal_error () /* check this before the
next;
- which does a more
accurate
- measurement of the
error */
- (temp_error = get_conversion_path_error (current_path)) =
legal_error ()
- )
+ if (path_error - 1.0 = legal_error ()) /* check this before the next;
+ which does a more accurate
+ measurement of the error */
{
- /* We have found the best path so far,
- * let's copy it into our new fish */
- fish_path-fish_path.cost = temp_cost;
- fish_path-fish.error = temp_error;
- babl_list_copy (current_path, fish_path-fish_path.conversion_list);
+ get_path_instrumentation (current_path, path_cost, ref_cost,
path_error);
+
+ if ((path_cost ref_cost) /* do not use paths that took longer
to compute than reference */
+ (path_cost fish_path-fish_path.cost)
+ (path_error = legal_error ()))
+{
+ /* We have found the best path so far,
+ * let's copy it into our new fish */
+ fish_path-fish_path.cost = path_cost;
+ fish_path-fish.error = path_error;
+ babl_list_copy (current_path,
fish_path-fish_path.conversion_list);
+}
}
}
else
@@ -163,11 +177,10 @@ get_conversion_path (Babl *current_forma
if (!next_format-format.visited)
{
/* next_format is not in the current path, we can pay a
visit */
- babl_list_insert_last (current_path, next_conversion);
- get_conversion_path (next_format, current_length + 1,
max_length);
- babl_list_remove_last (current_path);
- }
-
+ babl_list_insert_last (current_path, next_conversion);
+ get_conversion_path (next_format, current_length + 1,
max_length);
+ babl_list_remove_last (current_path);
+}
}
/* Remove the current
[Gegl-developer] BABL fishing patch
Hi,
attached is a patch that improves logic of go_fishing code
in babl-fish.c. The current code searches list of all
fishes while looking for suitable preexistent
BABL_FISH_PATH instance. The new code only searches
relevant part of database's hash table. Further, the
current code searches for fish path every time such a babl
fish is requested, even though the same fish path has been
requested before and not found. The new code creates dummy
BABL_FISH instance with appropriate source/destination
formats and inserts it into the fish database to indicate
that fish path has been searched for and not found. See
comments in the patch for details.
Regards,
Jan
Index: babl/babl-internal.h
===
--- babl/babl-internal.h(revision 302)
+++ babl/babl-internal.h(working copy)
@@ -88,6 +88,8 @@ long babl_fish_path_process
void *source,
void *destination,
longn);
+int babl_fish_get_id (const Babl *source,
+ const Babl *destination);
double babl_format_loss (Babl *babl);
Babl * babl_image_from_linear (char *buffer,
Index: babl/babl-fish-path.c
===
--- babl/babl-fish-path.c (revision 302)
+++ babl/babl-fish-path.c (working copy)
@@ -193,9 +193,10 @@ Babl *
babl_fish_path (const Babl *source,
const Babl *destination)
{
- Babl *babl = NULL;
- char *name = create_name (source, destination, 1);
+ Babl *babl = NULL;
+ char *name;
+ name = create_name (source, destination, 1);
babl = babl_db_exist_by_name (babl_fish_db (), name);
if (babl)
{
@@ -205,17 +206,11 @@ babl_fish_path (const Babl *source,
return babl;
}
- babl_assert (BABL_IS_BABL (source));
- babl_assert (BABL_IS_BABL (destination));
-
- babl_assert (source-class_type == BABL_FORMAT);
- babl_assert (destination-class_type == BABL_FORMAT);
-
babl = babl_calloc (1, sizeof (BablFishPath) +
strlen (name) + 1);
babl-class_type= BABL_FISH_PATH;
- babl-instance.id = 0;
+ babl-instance.id = babl_fish_get_id (source, destination);
babl-instance.name = ((char *) babl) + sizeof (BablFishPath);
strcpy (babl-instance.name, name);
babl-fish.source = source;
Index: babl/babl-fish-simple.c
===
--- babl/babl-fish-simple.c (revision 302)
+++ babl/babl-fish-simple.c (working copy)
@@ -33,7 +33,6 @@ babl_fish_simple (BablConversion *conver
babl_assert (BABL_IS_BABL (conversion));
name = create_name (conversion);
-
babl = babl_db_exist_by_name (babl_fish_db (), name);
if (babl)
{
@@ -46,7 +45,7 @@ babl_fish_simple (BablConversion *conver
babl = babl_malloc (sizeof (BablFishSimple) +
strlen (name) + 1);
babl-class_type= BABL_FISH_SIMPLE;
- babl-instance.id = 0;
+ babl-instance.id = babl_fish_get_id (conversion-source,
conversion-destination);
babl-instance.name = ((char *) babl) + sizeof (BablFishSimple);
strcpy (babl-instance.name, name);
babl-fish.source = conversion-source;
Index: babl/babl-fish-reference.c
===
--- babl/babl-fish-reference.c (revision 302)
+++ babl/babl-fish-reference.c (working copy)
@@ -69,7 +69,7 @@ babl_fish_reference (const Babl *source,
babl = babl_malloc (sizeof (BablFishReference) +
strlen (name) + 1);
babl-class_type= BABL_FISH_REFERENCE;
- babl-instance.id = 0;
+ babl-instance.id = babl_fish_get_id (source, destination);
babl-instance.name = ((char *) babl) + sizeof (BablFishReference);
strcpy (babl-instance.name, name);
babl-fish.source = source;
Index: babl/babl-fish.c
===
--- babl/babl-fish.c(revision 302)
+++ babl/babl-fish.c(working copy)
@@ -20,6 +20,87 @@
#include string.h
#include stdarg.h
+typedef struct _BablFindFish BablFindFish;
+
+typedef struct _BablFindFish
+{
+ Babl *fish_path;
+ Babl *fish_ref;
+ Babl *fish_fish;
+ intfishes;
+ const Babl *source;
+ const Babl *destination;
+} _BablFishFish;
+
+
+static int
+match_conversion (Babl *conversion,
+ void *inout);
+
+static int
+find_fish_path (Babl *item,
+void *data);
+
+static int
+find_memcpy_fish (Babl *item,
+ void *data);
+
+static int
+each_babl_fish_destroy (Babl *babl,
+void *data);
+
+
+/*
[Gegl-developer] BABL fish path patch
Hi,
attached is a patch that ports BablFishPath class to the
new list API and the list API is a bit expanded.
Further, the algorithm for generating the shortest
conversion path is reformulated to be more readable and
comprehensible and thoroughly commented. The algorithm for
processing the conversion paths is reformulated and
commented. The patch also contains minor readability
cleanups end speedups.
Regards,
Jan
Index: babl/babl-list.c
===
--- babl/babl-list.c(revision 300)
+++ babl/babl-list.c(working copy)
@@ -63,29 +63,78 @@ int
babl_list_size (BablList *list)
{
babl_assert (list);
+
return list-count;
}
-void
-babl_list_insert (BablList *list,
- Babl *item)
+inline void
+babl_list_insert_last (BablList *list,
+ Babl *item)
{
babl_assert(list);
babl_assert(BABL_IS_BABL(item));
if (list-size list-count + 1)
{
-Babl **new_items;
+ Babl **new_items;
-new_items = babl_realloc (list-items, (list-size * 2) * sizeof
(BablInstance *));
-babl_assert (new_items);
-list-items = new_items;
-memset (list-items + list-size, 0, list-size * sizeof (BablInstance
*));
-list-size *= 2;
+ new_items = babl_realloc (list-items, (list-size * 2) * sizeof
(BablInstance *));
+ babl_assert (new_items);
+ list-items = new_items;
+ memset (list-items + list-size, 0, list-size * sizeof (BablInstance
*));
+ list-size *= 2;
}
list-items[list-count++] = item;
}
+inline void
+babl_list_remove_last (BablList *list)
+{
+ babl_assert (list);
+ babl_assert (list-count 0);
+
+ list-count--;
+}
+
+inline Babl *
+babl_list_get_first (BablList *list)
+{
+ babl_assert (list);
+ babl_assert (list-count 0);
+
+ return (list-items[0]);
+}
+
+inline Babl *
+babl_list_get_last (BablList *list)
+{
+ babl_assert (list);
+ babl_assert (list-count 0);
+
+ return (list-items[list-count - 1]);
+}
+
+inline void
+babl_list_copy (BablList *from,
+BablList *to)
+{
+ babl_assert (from);
+ babl_assert (to);
+
+ if (to-size from-count)
+{
+ Babl **new_items;
+
+ new_items = babl_realloc (to-items, from-count * sizeof (BablInstance
*));
+ babl_assert (new_items);
+ to-items = new_items;
+ to-size = from-count;
+}
+
+memcpy (to-items, from-items, from-count * sizeof (BablInstance *));
+to-count = from-count;
+}
+
void
babl_list_each (BablList *list,
BablEachFunction each_fun,
Index: babl/babl-list.h
===
--- babl/babl-list.h(revision 300)
+++ babl/babl-list.h(working copy)
@@ -47,9 +47,22 @@ babl_list_destroy (BablList *list);
int
babl_list_size (BablList *list);
-void
-babl_list_insert (BablList *list,
- Babl *item);
+inline void
+babl_list_insert_last (BablList *list,
+ Babl *item);
+
+inline void
+babl_list_remove_last (BablList *list);
+
+inline Babl *
+babl_list_get_first (BablList *list);
+
+inline Babl *
+babl_list_get_last (BablList *list);
+
+inline void
+babl_list_copy (BablList *from,
+BablList *to);
void
babl_list_each (BablList *list,
Index: babl/babl-fish-path.c
===
--- babl/babl-fish-path.c (revision 300)
+++ babl/babl-fish-path.c (working copy)
@@ -19,14 +19,35 @@
#include math.h
#include babl-internal.h
+#define BABL_LEGAL_ERROR0.01
+#define BABL_MAX_COST_VALUE 200
+
static double
-chain_error (const Babl *fmt_source,
- const Babl *fmt_destination,
- BablConversion **chain,
- int conversions);
+get_conversion_path_error (BablList *path);
+
+static long
+process_conversion_path (BablList *path,
+ void *source_buffer,
+ void *destination_buffer,
+ long n);
+
+static void
+get_conversion_path (Babl *current_format,
+ int current_length,
+ int max_length);
+
+static double *
+test_create (void);
+
+static char *
+create_name (const Babl *source,
+ const Babl *destination,
+ int is_reference);
+
+static double legal_error (void);
+
+static int max_path_length (void);
-//#define BABL_LEGAL_ERROR 0.01
-//#define BABL_LEGAL_ERROR 0.01
static double legal_error (void)
{
@@ -40,7 +61,7 @@ static double legal_error (void)
if (env)
error = atof (env);
else
-error = 0.01;
+error = BABL_LEGAL_ERROR;
return error;
}
@@ -64,207 +85,99 @@ static int max_path_length (void)
return max_length;
}
-typedef struct BablChainContext
-{
- const Babl *from;
- const Babl *to;
- double *best_cost;
- double *best_loss;
[Gegl-developer] BABL list API patch
Hi,
attached is a patch that introduces minimal changes needed
for removal of the old list routines. It ports several
lists to the new API and deletes the unused old functions.
There are still several plain C arrays in the code used as lists,
however, these do not depend on the old routines and are
left untouched by the patch. I can try to identify and port these as
well. This would certainly improve things from the
readability point of view, although it wouldn't IMO do much
about the speed of the code.
To Sven: Thanks for including me into the AUTHORS file :)
Regards,
Jan
Index: babl/babl-model.c
===
--- babl/babl-model.c (revision 299)
+++ babl/babl-model.c (working copy)
@@ -27,7 +27,8 @@ static int
each_babl_model_destroy (Babl *babl,
void *data)
{
- babl_free (babl-model.from);
+ if (babl-model.from_list)
+babl_list_destroy (babl-model.from_list);
babl_free (babl);
return 0; /* continue iterating */
}
@@ -72,7 +73,7 @@ model_new (const char *name,
strcpy (babl-instance.name, name);
memcpy (babl-model.component, component, sizeof (BablComponent *) *
components);
- babl-model.from = NULL;
+ babl-model.from_list = NULL;
return babl;
}
Index: babl/babl-sampling.c
===
--- babl/babl-sampling.c(revision 299)
+++ babl/babl-sampling.c(working copy)
@@ -46,7 +46,8 @@ static int
each_babl_sampling_destroy (Babl *babl,
void *data)
{
- babl_free (babl-sampling.from);
+ if (babl-sampling.from_list)
+babl_list_destroy (babl-sampling.from_list);
return 0; /* continue iterating */
}
Index: babl/babl-list.c
===
--- babl/babl-list.c(revision 299)
+++ babl/babl-list.c(working copy)
@@ -16,11 +16,8 @@
* http://www.gnu.org/licenses/.
*/
-/* Implementation of list data structure. This is a bit superior
- * to the list implementation in babl-util.c.
+/* Implementation of list data structure.
* Copyright (C) 2008, Jan Heller
- *
- * TODO: migrate babl to BablList
*/
#include babl-internal.h
@@ -30,11 +27,19 @@
BablList *
babl_list_init (void)
{
+ return babl_list_init_with_size (BABL_LIST_INITIAL_SIZE);
+}
+
+BablList *
+babl_list_init_with_size (int initial_size)
+{
BablList *list = babl_calloc (sizeof (BablList), 1);
babl_assert (list);
- list-size = BABL_LIST_INITIAL_SIZE;
+ if (initial_size == 0)
+initial_size = 1;
+ list-size = initial_size;
list-count = 0;
list-items = NULL;
if (list-size)
@@ -81,13 +86,10 @@ babl_list_insert (BablList *list,
list-items[list-count++] = item;
}
-/* TODO: Rename babl_list_each_temp to babl_list_each after the list migration
- */
-
void
-babl_list_each_temp (BablList *list,
- BablEachFunction each_fun,
- void *user_data)
+babl_list_each (BablList *list,
+BablEachFunction each_fun,
+void *user_data)
{
int i;
Index: babl/babl-list.h
===
--- babl/babl-list.h(revision 299)
+++ babl/babl-list.h(working copy)
@@ -20,12 +20,13 @@
#define _BABL_LIST_H
#ifndef _BABL_CLASSES_H
+/* babl-classes.h contains forward declaration
+ * typedef struct _BablList BablList;
+ */
#error babl-list.h is only to be included after babl-classes.h
#endif
-typedef struct _BablList BablList;
-
typedef struct _BablList
{
int count;
@@ -37,6 +38,9 @@ typedef struct _BablList
BablList *
babl_list_init (void);
+BablList *
+babl_list_init_with_size (int initial_size);
+
void
babl_list_destroy (BablList *list);
@@ -48,9 +52,9 @@ babl_list_insert (BablList *list,
Babl *item);
void
-babl_list_each_temp (BablList *list,
- BablEachFunction each_fun,
- void *user_data);
+babl_list_each (BablList *list,
+BablEachFunction each_fun,
+void *user_data);
#endif
Index: babl/babl-internal.h
===
--- babl/babl-internal.h(revision 299)
+++ babl/babl-internal.h(working copy)
@@ -25,6 +25,7 @@
#define BABL_MAX_COMPONENTS 32
#define BABL_HARD_MAX_PATH_LENGTH 8
+#define BABL_CONVERSIONS 5
#include stdlib.h
#include stdio.h
Index: babl/babl-fish-path.c
===
--- babl/babl-fish-path.c (revision 299)
+++ babl/babl-fish-path.c (working copy)
@@ -141,10 +141,10 @@ get_conversion_chain (const Babl *f
temp_chain[temp_conversions] = NULL;
babl_assert (from);
babl_assert (from-class_type == BABL_FORMAT
Re: [Gegl-developer] Hi, need any help?
On 17:55, Mon 10 Mar 08, Martin Nordholts wrote: That looks very interesting. Do you think you could provide some benchmarking data of the performance improvements in variuos situations? It would be useful to have Hi, I used babl/tests/babl_fish_path_dhtml as a simple benchmark. Here are running times of babl_fish_path_dhtml w/o and w/ the patch applied: [EMAIL PROTECTED] ~/projects/gegl/babl/babl/tests $ time ./babl_fish_path_dhtml babl_fish_path_0.0.20.html real0m10.463s user0m9.441s sys 0m0.040s [EMAIL PROTECTED] ~/projects/gegl/babl-patch/babl/tests $ time ./babl_fish_path_dhtml babl_fish_path_patch.html real0m3.844s user0m3.416s sys 0m0.048s Here are the resulting HTML files: http://www.ms.mff.cuni.cz/~hellj1am/WWW/babl_fish_path_0.0.20.html http://www.ms.mff.cuni.cz/~hellj1am/WWW/babl_fish_path_patch.html Such a dramatic performance gain is not to be expected from any practical usage of the babl library, since the database code is not as heavily used. A simple 512x512 16-bit/color RGBA PNG image load/save test using gegl gives following results for runs w/o and w/ the patch. I took the respective best results from several tries: W/O real0m6.367s user0m3.476s sys 0m0.168s W real0m2.930s user0m2.400s sys 0m0.096s I hope these results give a better idea about the performance of the changes. Regards, Jan ___ Gegl-developer mailing list Gegl-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer
Re: [Gegl-developer] Hi, need any help?
On 21:09, Mon 10 Mar 08, Sven Neumann wrote: I would prefer if this functionality could be added in new files babl-list.[ch] and babl-hash-table.[ch]. Also the hash-table functions should be prefixed with babl_hash_table_ instead of just babl_hash_ as we should keep the babl_hash prefix for hash functions. Then I don't Such changes can be easily and happily arranged in case the patch would make it to the babl. understand the purpose of babl_list_each_NEW(). This is just temporary, isn't it? Yes, that is correct. There is an another implementation of the list data structure already present in the babl with a colliding function name. The next step after the patch would be migrating the code to the new implementation of the list data structure, which would not necessarily improve speed of the code, however, this would IMHO improve readability and maintainability a bit. Regards, Jan ___ Gegl-developer mailing list Gegl-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer
[Gegl-developer] Hi, need any help?
Hi all, I am new to this list and I would like to help with the GEGL development. I am a programmer and a computer graphics enthusiast. I started to look into the GEGL/BABL code base and have been able to make some improvements in babl database code that speed things up a bit, without loss of readability. Should you guys be interested I can post the patch to this list and I can try to look into things further. If you have other ideas as to where any help might be needed I'll be glad to help. Regards, Jan ___ Gegl-developer mailing list Gegl-developer@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer
Re: [Gegl-developer] Hi, need any help?
On 19:07, Sun 09 Mar 08, Martin Nordholts wrote:
Hi Jan
We're absolutely interested in your contributions! Eagerly awaiting the
patch :)
Best regards,
Martin Nordholts
Glad to hear it :) So here is the patch. Essentially, it
implements coalesced hashing as BablHashTable structure and
uses it as the babl database data structure. The code is not
long and is quite self-explanatory I believe, although I
wouldn't mind adding a few comments if needed.
Regards,
Jan
Index: babl/babl-util.c
===
--- babl/babl-util.c(revision 290)
+++ babl/babl-util.c(working copy)
@@ -22,6 +22,287 @@
#include math.h
#include babl-memory.h
#include babl-internal.h
+#include babl-util.h
+
+#define UTIL_INITIAL_LIST_SIZE 0x7F
+#define UTIL_INITIAL_HT_MASK 0x7F
+
+/* static functions declarations */
+static inline int
+hash_by_name (BablHashTable *htab,
+ const char*str);
+
+static inline int
+hash_by_id (BablHashTable *htab,
+int id);
+
+static inline int
+hash_insert (BablHashTable *htab,
+ Babl *item);
+
+static void
+hash_rehash (BablHashTable *htab);
+
+/* BablList functions */
+
+BablList *
+babl_list_init (void)
+{
+ BablList *list = babl_calloc (sizeof (BablList), 1);
+
+ list-size = UTIL_INITIAL_LIST_SIZE;
+ list-count = 0;
+ list-items = NULL;
+ if (list-size)
+{
+ list-items = babl_calloc (sizeof (BablInstance *), list-size);
+}
+
+ return list;
+}
+
+void
+babl_list_destroy (BablList *list)
+{
+babl_assert(list);
+
+babl_free (list-items);
+babl_free (list);
+}
+
+void
+babl_list_insert (BablList *list,
+ Babl *item)
+{
+ babl_assert(list);
+ babl_assert(BABL_IS_BABL(item));
+
+ if (list-size list-count + 1)
+{
+Babl **new_items;
+
+new_items = babl_realloc (list-items, (list-size * 2) * sizeof
(BablInstance *));
+babl_assert (new_items);
+list-items = new_items;
+memset (list-items + list-size, 0, list-size * sizeof (BablInstance
*));
+list-size *= 2;
+}
+list-items[list-count++] = item;
+}
+
+void
+babl_list_each_NEW (BablList *list,
+BablEachFunction each_fun,
+void *user_data)
+{
+ int i;
+
+ babl_assert(list);
+ babl_assert(each_fun);
+
+ for (i = 0; i list-count; i++)
+{
+ if (list-items[i])
+{
+ if (each_fun ((Babl *) list-items[i], user_data))
+break;
+}
+}
+}
+
+/* BablHashTable functions */
+
+
+inline int
+babl_hash_by_str (BablHashTable *htab,
+ const char*str)
+{
+ int hash = 0;
+
+ while (*str)
+ {
+hash += *str++;
+hash += (hash 10);
+hash ^= (hash 6);
+ }
+ hash += (hash 3);
+ hash ^= (hash 11);
+ hash += (hash 15);
+
+ return (hash htab-mask);
+}
+
+inline int
+babl_hash_by_int (BablHashTable *htab,
+ int id)
+{
+ int hash = 0;
+ int i;
+
+ for (i = 0; i sizeof (int); i++)
+ {
+hash += id 0xFF;
+hash += (hash 10);
+hash ^= (hash 6);
+id = 8;
+ }
+ hash += (hash 3);
+ hash ^= (hash 11);
+ hash += (hash 15);
+
+ return (hash htab-mask);
+}
+
+static inline int
+hash_insert (BablHashTable *htab,
+ Babl *item)
+{
+ int hash = htab-hash_func(htab, item);
+
+ if (htab-data_table[hash] == NULL)
+{
+ /* create new chain */
+ htab-data_table[hash] = item;
+}
+ else
+{
+ int it, oit, cursor = 0;
+
+ while ((cursor (htab-mask + 1)) (htab-data_table[cursor] != NULL))
+++cursor;
+
+ htab-data_table[cursor] = item;
+
+ for (oit = hash, it = htab-chain_table[oit]; it != -1; oit = it, it =
htab-chain_table[oit])
+;
+ htab-chain_table[oit] = cursor;
+}
+
+ htab-count++;
+ return 0;
+}
+
+static void
+hash_rehash (BablHashTable *htab)
+{
+ Babl *item;
+ int i;
+ BablHashTable *nhtab = babl_calloc (sizeof (BablHashTable), 1);
+
+ nhtab-data_table = NULL;
+ nhtab-chain_table = NULL;
+ nhtab-mask = (htab-mask 1) + 1;
+ nhtab-count = 0;
+ nhtab-hash_func = htab-hash_func;
+ nhtab-find_func = htab-find_func;
+ if (nhtab-mask)
+{
+ nhtab-data_table = babl_calloc (sizeof (BablInstance *),
babl_hash_size(nhtab));
+ nhtab-chain_table = babl_malloc (sizeof (int *) *
babl_hash_size(nhtab));
+ memset (nhtab-chain_table, -1, sizeof (int) * babl_hash_size(nhtab));
+}
+
+ for (i = 0; i babl_hash_size (htab); i++)
+{
+ item = htab-data_table[i];
+ babl_hash_insert (nhtab, item);
+}
+
+ htab-mask = nhtab-mask;
+ babl_free (htab-data_table);
+ babl_free (htab-chain_table);
+ htab-data_table = nhtab-data_table;
+ htab-chain_table = nhtab-chain_table;
+ babl_free (nhtab);
+}
+
+inline int
+babl_hash_size (BablHashTable *htab)
+{
+return htab-mask + 1;
+}
+
+BablHashTable *
