On 12/01/2011 02:42 AM, Michel Dänzer wrote:
On Mit, 2011-11-30 at 20:36 -0700, Brian Paul wrote:
We weren't doing the necessary byte swap.
I just stumbled over this as well.
---
src/mesa/main/format_unpack.c | 9 +++++----
1 files changed, 5 insertions(+), 4 deletions(-)
diff --git a/src/mesa/main/format_unpack.c b/src/mesa/main/format_unpack.c
index 4b4ee6b..fc0db34 100644
--- a/src/mesa/main/format_unpack.c
+++ b/src/mesa/main/format_unpack.c
@@ -275,10 +275,11 @@ unpack_ARGB1555_REV(const void *src, GLfloat dst[][4],
GLuint n)
const GLushort *s = ((const GLushort *) src);
GLuint i;
for (i = 0; i< n; i++) {
- dst[i][RCOMP] = UBYTE_TO_FLOAT( ((s[i]>> 7)& 0xf8) | ((s[i]>> 12)&
0x7) );
- dst[i][GCOMP] = UBYTE_TO_FLOAT( ((s[i]>> 2)& 0xf8) | ((s[i]>> 7)&
0x7) );
- dst[i][BCOMP] = UBYTE_TO_FLOAT( ((s[i]<< 3)& 0xf8) | ((s[i]>> 2)&
0x7) );
- dst[i][ACOMP] = UBYTE_TO_FLOAT( ((s[i]>> 15)& 0x01) * 255 );
+ GLushort tmp = (s[i]<< 8) | (s[i]>> 8); /* byteswap */
A BSWAP16() macro might be nice for this, but I wouldn't let that hold
up this fix.
+ dst[i][RCOMP] = UBYTE_TO_FLOAT( ((tmp>> 7)& 0xf8) | ((tmp>> 12)&
0x7) );
+ dst[i][GCOMP] = UBYTE_TO_FLOAT( ((tmp>> 2)& 0xf8) | ((tmp>> 7)&
0x7) );
+ dst[i][BCOMP] = UBYTE_TO_FLOAT( ((tmp<< 3)& 0xf8) | ((tmp>> 2)&
0x7) );
Don't these lines need to be changed to be the same as in
unpack_ARGB1555()?
It should be equivalent (lookup table vs. arithmetic). It would nice
if someone could write a little benchmark to see which is really faster.
I'll commit this later if there's no other concerns.
-Brian
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
