minhngoc25a
/
freetype2
mirror of git://git.savannah.gnu.org/freetype/freetype2.git
2
1
Fork 2
Code Issues 4 Releases Wiki Activity

[truetype] Scale F_dot_P down. 

The dot product between freeVector and projVector or cosine of
the angle between these FT_F2Dot14 unit vectors used to be scaled up
by 4 and routinely occupied 32 bits in an FT_Long field F_dot_P.
This patch scales the value down by 2^14 instead, which simplifies
its use throughout the bytecode interpreter.

This does not lead to the loss of precision because the lower bits
are unreliable anyway. Consider two unit vectors (1,0) and (.6,.8)
for which the true value of F_dot_P is .6 * 0x40000000 = 0x26666666.
These vectors are stored as (0x4000,0) and (0x2666,0x3333) after
rounding and F_dot_P is assigned 0x26660000. The lower bits were
already lost while rounding the unit vector components.

Besides code simplification, this change can lead to better
performance when FT_MulDiv with the scaled-down F_dot_P is less
likely to use the costly 64-bit path. We are not changing the type
of F_dot_P to FT_F2Dot14 at this point.

* src/truetype/ttinterp.c (Compute_Funcs): Scale F_dot_P down by 14
bits and modify its use accordingly.
(Direct_Move, Direct_Move_Orig, Compute_Point_Displacement): Modify
the use of F_dot_P field.
* src/truetype/ttobjs.c (tt_size_run_fpgm): Change arbitrary
assignment of F_dot_P to its theoretical maximum in case we decide
to scale back its type later.
This commit is contained in:
Alexei Podtelezhnikov 2012-12-10 06:59:29 -05:00
parent 13e87e0426
commit 65d6572105
3 changed files with 45 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
ChangeLog
View File

@ -1,3 +1,33 @@
2012-12-10 Alexei Podtelezhnikov <apodtele@gmail.com>
[truetype] Scale F_dot_P down.
The dot product between freeVector and projVector or cosine of
the angle between these FT_F2Dot14 unit vectors used to be scaled up
by 4 and routinely occupied 32 bits in an FT_Long field F_dot_P.
This patch scales the value down by 2^14 instead, which simplifies
its use throughout the bytecode interpreter.
This does not lead to the loss of precision because the lower bits
are unreliable anyway. Consider two unit vectors (1,0) and (.6,.8)
for which the true value of F_dot_P is .6 * 0x40000000 = 0x26666666.
These vectors are stored as (0x4000,0) and (0x2666,0x3333) after
rounding and F_dot_P is assigned 0x26660000. The lower bits were
already lost while rounding the unit vector components.
Besides code simplification, this change can lead to better
performance when FT_MulDiv with the scaled-down F_dot_P is less
likely to use the costly 64-bit path. We are not changing the type
of F_dot_P to FT_F2Dot14 at this point.
* src/truetype/ttinterp.c (Compute_Funcs): Scale F_dot_P down by 14
bits and modify its use accordingly.
(Direct_Move, Direct_Move_Orig, Compute_Point_Displacement): Modify
the use of F_dot_P field.
* src/truetype/ttobjs.c (tt_size_run_fpgm): Change arbitrary
assignment of F_dot_P to its theoretical maximum in case we decide
to scale back its type later.
2012-12-09 Johnson Y. Yan <yinsen_yan@foxitsoftware.com>
[type1] Another fix for 2012-09-17 commit.

39
src/truetype/ttinterp.c
View File

@ -1861,9 +1861,7 @@
if ( !CUR.ignore_x_mode ||
( CUR.sph_tweak_flags & SPH_TWEAK_ALLOW_X_DMOVE ) )
#endif /* TT_CONFIG_OPTION_SUBPIXEL_HINTING */
zone->cur[point].x += FT_MulDiv( distance,
v * 0x10000L,
CUR.F_dot_P );
zone->cur[point].x += FT_MulDiv( distance, v, CUR.F_dot_P );
zone->tags[point] |= FT_CURVE_TAG_TOUCH_X;
}
@ -1872,9 +1870,7 @@
if ( v != 0 )
{
zone->cur[point].y += FT_MulDiv( distance,
v * 0x10000L,
CUR.F_dot_P );
zone->cur[point].y += FT_MulDiv( distance, v, CUR.F_dot_P );
zone->tags[point] |= FT_CURVE_TAG_TOUCH_Y;
}
@ -1913,16 +1909,12 @@
v = CUR.GS.freeVector.x;
if ( v != 0 )
zone->org[point].x += FT_MulDiv( distance,
v * 0x10000L,
CUR.F_dot_P );
zone->org[point].x += FT_MulDiv( distance, v, CUR.F_dot_P );
v = CUR.GS.freeVector.y;
if ( v != 0 )
zone->org[point].y += FT_MulDiv( distance,
v * 0x10000L,
CUR.F_dot_P );
zone->org[point].y += FT_MulDiv( distance, v, CUR.F_dot_P );
}
@ -2646,12 +2638,13 @@
#endif /* TT_CONFIG_OPTION_UNPATENTED_HINTING */
if ( CUR.GS.freeVector.x == 0x4000 )
CUR.F_dot_P = CUR.GS.projVector.x * 0x10000L;
CUR.F_dot_P = CUR.GS.projVector.x;
else if ( CUR.GS.freeVector.y == 0x4000 )
CUR.F_dot_P = CUR.GS.projVector.y * 0x10000L;
CUR.F_dot_P = CUR.GS.projVector.y;
else
CUR.F_dot_P = (FT_Long)CUR.GS.projVector.x * CUR.GS.freeVector.x * 4 +
(FT_Long)CUR.GS.projVector.y * CUR.GS.freeVector.y * 4;
CUR.F_dot_P = ( (FT_Long)CUR.GS.projVector.x * CUR.GS.freeVector.x +
(FT_Long)CUR.GS.projVector.y * CUR.GS.freeVector.y ) >>
14;
if ( CUR.GS.projVector.x == 0x4000 )
CUR.func_project = (TT_Project_Func)Project_x;
@ -2670,7 +2663,7 @@
CUR.func_move = (TT_Move_Func)Direct_Move;
CUR.func_move_orig = (TT_Move_Func)Direct_Move_Orig;
if ( CUR.F_dot_P == 0x40000000L )
if ( CUR.F_dot_P == 0x4000L )
{
if ( CUR.GS.freeVector.x == 0x4000 )
{
@ -2687,8 +2680,8 @@
/* at small sizes, F_dot_P can become too small, resulting */
/* in overflows and `spikes' in a number of glyphs like `w'. */
if ( FT_ABS( CUR.F_dot_P ) < 0x4000000L )
CUR.F_dot_P = 0x40000000L;
if ( FT_ABS( CUR.F_dot_P ) < 0x400L )
CUR.F_dot_P = 0x4000L;
/* Disable cached aspect ratio */
CUR.tt_metrics.ratio = 0;
@ -5827,12 +5820,8 @@
else
#endif
{
*x = FT_MulDiv( d,
(FT_Long)CUR.GS.freeVector.x * 0x10000L,
CUR.F_dot_P );
*y = FT_MulDiv( d,
(FT_Long)CUR.GS.freeVector.y * 0x10000L,
CUR.F_dot_P );
*x = FT_MulDiv( d, (FT_Long)CUR.GS.freeVector.x, CUR.F_dot_P );
*y = FT_MulDiv( d, (FT_Long)CUR.GS.freeVector.y, CUR.F_dot_P );
}
return SUCCESS;

2
src/truetype/ttobjs.c
View File

@ -764,7 +764,7 @@
exec->threshold = 0;
exec->instruction_trap = FALSE;
exec->F_dot_P = 0x10000L;
exec->F_dot_P = 0x4000L;
exec->pedantic_hinting = pedantic;