First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
#!/bin/sh
|
|
|
|
|
|
|
|
|
|
# The ClutterFixed type and macros now use floats, but we are keeping the
|
|
|
|
|
# CoglFixed type + macros using fixed point so now we convert all uses of
|
|
|
|
|
# the Cogl fixed point macros within Clutter proper to use the ClutterFixed
|
|
|
|
|
# macros instead.
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -maxdepth 1 -iname '*.c' -exec sed -i \
|
2009-01-16 17:52:26 +00:00
|
|
|
-e 's/COGL_FIXED_MUL/CLUTTER_FIXED_MUL/g' \
|
|
|
|
|
-e 's/COGL_FIXED_DIV/CLUTTER_FIXED_DIV/g' \
|
|
|
|
|
-e 's/COGL_FIXED_FAST_MUL/CLUTTER_FIXED_MUL/g' \
|
|
|
|
|
-e 's/COGL_FIXED_FAST_DIV/CLUTTER_FIXED_DIV/g' \
|
|
|
|
|
-e 's/COGL_FIXED_FROM_FLOAT/CLUTTER_FLOAT_TO_FIXED/g' \
|
|
|
|
|
-e 's/COGL_FIXED_TO_FLOAT/CLUTTER_FIXED_TO_FLOAT/g' \
|
|
|
|
|
-e 's/COGL_FIXED_TO_DOUBLE/CLUTTER_FIXED_TO_DOUBLE/g' \
|
|
|
|
|
-e 's/COGL_FIXED_PI/CFX_PI/g' \
|
|
|
|
|
{} \;
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
|
|
|
|
|
# All remaining uses of the Cogl fixed point API now get expanded out to simply
|
|
|
|
|
# use float calculations... (we will restore the cogl-fixed code itself later)
|
|
|
|
|
|
2009-01-16 17:52:26 +00:00
|
|
|
# XXX: The following three assume that no nested function - with
|
|
|
|
|
# multiple arguments - is ever found as the RHS argument to
|
|
|
|
|
# COGL_FIXED_MUL. This is because we simply replace the last ',' with
|
|
|
|
|
# the * operator. If you want to double check that's still true:
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
# $ grep -r --include=*.c COGL_FIXED_MUL *|less
|
2009-01-16 17:52:26 +00:00
|
|
|
#
|
|
|
|
|
# XXX: (Note in the third regexp there were examples of COGL_FIXED_MUL
|
|
|
|
|
# being used as the RHS argument, but since we have already replaced
|
|
|
|
|
# instances of COGL_FIXED_MUL, that works out ok.
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' -exec sed -i -r \
|
2009-01-16 17:52:26 +00:00
|
|
|
-e 's/COGL_FIXED_MUL (.*),/\1 */g' \
|
|
|
|
|
-e 's|COGL_FIXED_FAST_DIV (.*),|\1 /|g' \
|
|
|
|
|
-e 's|COGL_FIXED_DIV (.*),|\1 /|g' \
|
|
|
|
|
{} \;
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
|
|
|
|
|
# A fix due to the assumptions used above
|
|
|
|
|
sed -i 's/#define DET2X(a,b,c,d).*/#define DET2X(a,b,c,d) ((a * d) - (b * c))/g' ./clutter/clutter-actor.c
|
|
|
|
|
|
2009-01-08 22:56:17 +00:00
|
|
|
find ./clutter/cogl/gles -iname '*.[ch]' -exec sed -i 's/GLfixed/GLfloat/g' {} \;
|
|
|
|
|
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
#we get some redundant brackets like this, but C's automatic type promotion
|
|
|
|
|
#works out fine for most cases...
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' -exec sed -r -i \
|
2009-01-16 17:52:26 +00:00
|
|
|
-e 's/COGL_FIXED_TO_INT//g' \
|
|
|
|
|
-e 's/COGL_FIXED_FROM_INT /(float)/g' \
|
|
|
|
|
-e 's/COGL_FIXED_FROM_INT/(float)/g' \
|
|
|
|
|
-e 's/COGL_FIXED_TO_FLOAT//g' \
|
|
|
|
|
-e 's/COGL_FIXED_FROM_FLOAT//g' \
|
|
|
|
|
-e 's/COGL_FIXED_TO_DOUBLE /(double)/g' \
|
|
|
|
|
\
|
|
|
|
|
-e 's/COGL_FIXED_FLOOR/floorf/g' \
|
|
|
|
|
-e 's/COGL_FIXED_CEIL/ceilf/g' \
|
|
|
|
|
-e 's/COGL_FIXED_360/360.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_240/240.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_255/255.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_180/180.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_120/120.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_60/60.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_1/1.0/g' \
|
|
|
|
|
-e 's/COGL_FIXED_0_5/0.5/g' \
|
|
|
|
|
-e 's/COGL_FIXED_PI/G_PI/g' \
|
|
|
|
|
\
|
|
|
|
|
-e 's/COGL_ANGLE_FROM_DEG \((.*)\),/\1,/g' \
|
|
|
|
|
{} \; \
|
|
|
|
|
\
|
|
|
|
|
-exec perl -p -i \
|
|
|
|
|
-e "s|cogl_angle_cos \((.*?)\)|cosf (\1 * (G_PI/180.0))|;" \
|
|
|
|
|
-e "s|cogl_angle_sin \((.*?)\)|sinf (\1 * (G_PI/180.0))|;" \
|
|
|
|
|
-e "s|cogl_angle_tan \((.*?)\)|tanf (\1 * (G_PI/180.0))|;" \
|
|
|
|
|
{} \;
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
|
|
|
|
|
#XXX: NB: cogl_fixed_div must be done before mul since there is a case were they
|
|
|
|
|
#are nested which would otherwise break the assumption used here that the last
|
|
|
|
|
#coma of the line can simply be replaced with the corresponding operator
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' -exec sed -i -r \
|
2009-01-16 17:52:26 +00:00
|
|
|
-e 's|cogl_fixed_div (.*),|\1 /|g' \
|
|
|
|
|
-e 's|cogl_fixed_mul (.*),|\1 *|g' \
|
|
|
|
|
-e 's/cogl_fixed_pow2/pow2f/g' \
|
|
|
|
|
-e 's/cogl_fixed_pow/powf/g' \
|
|
|
|
|
-e 's/cogl_fixed_log2/log2f/g' \
|
|
|
|
|
-e 's/cogl_fixed_sqrt/sqrtf/g' \
|
|
|
|
|
-e 's/cogl_fixed_cos/cosf/g' \
|
|
|
|
|
-e 's/cogl_fixed_sin/sinf/g' \
|
|
|
|
|
-e 's/cogl_fixed_atan2/atan2f/g' \
|
|
|
|
|
-e 's/cogl_fixed_atan/atanf/g' \
|
|
|
|
|
-e 's/cogl_fixed_tan/tanf/g' \
|
|
|
|
|
{} \;
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
|
|
|
|
|
#TODO: fixup gles/cogl.c set_clip_plane
|
|
|
|
|
|
|
|
|
|
cat clutter/cogl/common/cogl-primitives.c| \
|
|
|
|
|
grep -v '#define CFX_MUL2'| \
|
|
|
|
|
grep -v '#undef CFX_MUL2'| \
|
|
|
|
|
grep -v '#define CFX_MUL3'| \
|
|
|
|
|
grep -v '#undef CFX_MUL3'| \
|
|
|
|
|
grep -v '#define CFX_SQ'| \
|
|
|
|
|
grep -v '#undef CFX_SQ'| \
|
|
|
|
|
sed -r 's/CFX_MUL2 \((.{7})\)/(\1 * 2)/g' | \
|
|
|
|
|
sed -r 's/CFX_MUL3 \((.{7})\)/(\1 * 3)/g' | \
|
|
|
|
|
sed -r 's/CFX_SQ \((.{7})\)/(\1 * \1)/g' \
|
|
|
|
|
>./tmp
|
|
|
|
|
mv ./tmp clutter/cogl/common/cogl-primitives.c
|
|
|
|
|
|
|
|
|
|
#this has too many false positives...
|
|
|
|
|
#find ./clutter -iname '*.[ch]' -exec sed -i 's|>> 1|/ 2|g' {} \;
|
|
|
|
|
#find ./clutter -iname '*.[ch]' -exec sed -i 's|<< 1|* 2|g' {} \;
|
|
|
|
|
|
2009-01-16 17:52:26 +00:00
|
|
|
sed -i -e 's|>> 1|/ 2|g' -e 's|<< 1|* 2|g' \
|
|
|
|
|
./clutter/cogl/common/cogl-primitives.c
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
#find ./clutter -iname '*.[ch]' -exec sed -i 's|<< 1|* 2|g' {} \;
|
|
|
|
|
|
|
|
|
|
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' \
|
|
|
|
|
-exec sed -i 's/CoglFixed/float/g' {} \;
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
#XXX: This might need changing later...
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' \
|
|
|
|
|
-exec sed -i 's/CoglFixedVec2/CoglVec2/g' {} \;
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
sed -i 's/CoglFixed/float/g' ./clutter/cogl/cogl.h.in
|
|
|
|
|
|
|
|
|
|
# maintain the existing CoglFixed code as utility code for applications:
|
|
|
|
|
sed -i 's/float:/CoglFixed:/g' clutter/cogl/cogl-types.h
|
|
|
|
|
sed -i 's/gint32 float/gint32 CoglFixed/g' clutter/cogl/cogl-types.h
|
|
|
|
|
git-checkout clutter/cogl/cogl-fixed.h clutter/cogl/common/cogl-fixed.c
|
|
|
|
|
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' -exec sed -i 's/CoglAngle/float/g' {} \;
|
2009-01-07 15:33:24 +00:00
|
|
|
|
|
|
|
|
# maintain the existing CoglAngle code as utility code for applications:
|
|
|
|
|
sed -i 's/float:/CoglAngle:/g' clutter/cogl/cogl-types.h
|
|
|
|
|
sed -i 's/gint32 float/gint32 CoglAngle/g' clutter/cogl/cogl-types.h
|
|
|
|
|
git-checkout clutter/cogl/cogl-fixed.h clutter/cogl/common/cogl-fixed.c
|
|
|
|
|
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter ./tests -iname '*.[ch]' ! -iname 'clutter-fixed.h' \
|
|
|
|
|
-exec sed -i 's/ClutterAngle/float/g' {} \;
|
2009-01-07 15:33:24 +00:00
|
|
|
|
2009-01-16 12:16:28 +00:00
|
|
|
# use the floating point names for GL ES functions instead of the
|
|
|
|
|
# fixed. These get #define'd to the float versions in one of the
|
|
|
|
|
# patches anyway but the names should be fixed up to avoid confusion
|
|
|
|
|
find ./clutter/cogl -iname '*.[ch]' -exec perl -p -i -e \
|
|
|
|
|
's/\b(cogl_wrap_(?:glMultMatrix|glFrustum|glScale|glTranslate
|
|
|
|
|
|glRotate|glOrtho|glTexEnv|glClipPlane|glFog|glColor4))x(v?)\b/$1f$2/gx' {} \;
|
|
|
|
|
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
echo "Cogl API to remove/replace with float versions:"
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter/ ./tests -iname '*.c' -exec grep '^cogl_[a-zA-Z_]*x ' {} \; | cut -d' ' -f1|grep -v 'box$'|grep -v 'matrix$'
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
echo "Clutter API to remove/replace with float versions:"
|
2009-01-16 18:29:29 +00:00
|
|
|
find ./clutter/ ./tests -iname '*.c' -exec grep '^clutter_[a-zA-Z_]*x ' {} \; | cut -d' ' -f1|grep -v 'box$'|grep -v 'matrix$'|grep -v '_x$'
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
|
|
|
|
|
#
|
|
|
|
|
# Now the last mile is dealt with manually with a bunch of patches...
|
|
|
|
|
#
|
|
|
|
|
|
|
|
|
|
git-commit -a -m "[By fixed-to-float.sh] Fixed to Float automatic changes" --no-verify
|
|
|
|
|
|
|
|
|
|
patch -p1<fixed-to-float-patches/gl-cogl-texture.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-actor.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-alpha.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-alpha.h.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-behaviour-ellipse.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-bezier.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-path.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/cogl-fixed.h.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/cogl-fixed.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/test-cogl-tex-tile.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-texture.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-fixed.c.0.patch
|
2009-01-13 11:38:55 +00:00
|
|
|
patch -p1<fixed-to-float-patches/gl-cogl.c.0.patch
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
patch -p1<fixed-to-float-patches/cogl-pango-render.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/cogl-primitives.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/gl-cogl-primitives.c.0.patch
|
2009-01-08 22:56:17 +00:00
|
|
|
patch -p1<fixed-to-float-patches/gles-cogl.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/gles-cogl-gles2-wrapper.h.0.patch
|
2009-01-15 15:24:05 +00:00
|
|
|
patch -p1<fixed-to-float-patches/gles-cogl-gles2-wrapper.c.0.patch
|
2009-01-08 22:56:17 +00:00
|
|
|
patch -p1<fixed-to-float-patches/gles-cogl-primitives.c.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/gles-cogl-texture.c.0.patch
|
2009-01-12 17:07:55 +00:00
|
|
|
patch -p1<fixed-to-float-patches/cogl.h.in.0.patch
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
|
2009-01-13 11:38:55 +00:00
|
|
|
# Finally remove any cogl_blahx Cogl interfaces that used to take CoglFixed
|
|
|
|
|
# params. The corresponding interfaces that take integer params are also
|
|
|
|
|
# patched to take floats instead:
|
|
|
|
|
patch -p1<fixed-to-float-patches/remove_cogl_apis_taking_fixed_params.0.patch
|
|
|
|
|
|
First cut at a fixed point to floating point conversion script + patches
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
2008-12-19 21:55:35 +00:00
|
|
|
#XXX: COGL_PANGO_UNIT_TO_FIXED
|
|
|
|
|
|
|
|
|
|
git-commit -a -m "[By fixed-to-float.sh] Fixed to Float patches" --no-verify
|
|
|
|
|
|
|
|
|
|
# The fixes in these files are entirely handcoded, so to avoid clashes with the
|
|
|
|
|
# automatic stuff above the patches below are based against the pristine
|
|
|
|
|
# versions, and we don't want to commit any of the automatic changes here.
|
|
|
|
|
git-checkout HEAD~2 clutter/clutter-fixed.h
|
|
|
|
|
git-checkout HEAD~2 clutter/clutter-units.h
|
|
|
|
|
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-fixed.h.0.patch
|
|
|
|
|
patch -p1<fixed-to-float-patches/clutter-units.h.0.patch
|
|
|
|
|
|
|
|
|
|
git-commit -a -m "[By fixed-to-float.sh] clutter-fixed.h and clutter-units.h changes" --no-verify
|
|
|
|
|
|