commit 930140e919a10a732738fae7cfc1ec8210cf5079
author Alexei Podtelezhnikov <apodtele@gmail.com> Thu Aug 26 16:49:59 2021 +0000
committer Alexei Podtelezhnikov <apodtele@gmail.com> Sun Aug 29 21:55:19 2021 +0000
tree 1c083c94b429e7c64afb968741e44e7ac0fc21f9
parent d1c20005042a31fba7e86603b0f65fcbf21fe51c

[smooth] Reduce shift in multiply-shift optimization.

Smaller shifts that keep the division operands of FT_UDIVPREP within
32 bits result in slightly faster divisions, which is noticeable in
the overall performance.  The loss of precision is tolerable until the
divisors (the components dx and dy) approach 32 - PIXEL_BITS. With
PIXEL_BITS = 8, this corresponds to 65,000 pixels or the bitmap size
that we refuse to render anyway.

Using `ftbench -p -s60 -t5 -bc timesi.ttf`,

Before: 8.52 us/op
After:  8.32 us/op

src/smooth/ftgrays.c

diff --git a/src/smooth/ftgrays.c b/src/smooth/ftgrays.c
index 576dbb3..2c0c1ba 100644
--- a/src/smooth/ftgrays.c
+++ b/src/smooth/ftgrays.c
@@ -386,12 +386,11 @@
   /* divisors to provide sufficient accuracy of the multiply-shift.    */
   /* It should not exceed (64 - PIXEL_BITS) to prevent overflowing and */
   /* leave enough room for 64-bit unsigned multiplication however.     */
-#define FT_UDIVPREP( c, b )                                                 \
-  FT_Int64  b ## _r = c ? (FT_Int64)( ~(FT_UInt64)0 >> PIXEL_BITS ) / ( b ) \
+#define FT_UDIVPREP( c, b )                            \
+  FT_Int64  b ## _r = c ? (FT_Int64)0xFFFFFFFF / ( b ) \
                     : 0
-#define FT_UDIV( a, b )                                         \
-  (TCoord)( ( (FT_UInt64)( a ) * (FT_UInt64)( b ## _r ) ) >>    \
-            ( sizeof( FT_UInt64 ) * FT_CHAR_BIT - PIXEL_BITS ) )
+#define FT_UDIV( a, b )                                           \
+  (TCoord)( ( (FT_UInt64)( a ) * (FT_UInt64)( b ## _r ) ) >> 32 )
 
 
   /* Scale area and apply fill rule to calculate the coverage byte. */