slight tune of neon asm - based on stuff from nash. not worth a

changelog entry though. not a major enough improvement as i measure it.

git-svn-id: svn+ssh://svn.enlightenment.org/var/svn/e/trunk/evas@56614 7cbeb6ba-43b4-40fd-8cce-4c39aea84d33

diff --git a/src/lib/engines/common/evas_op_blend/op_blend_pixel_neon.c b/src/lib/engines/common/evas_op_blend/op_blend_pixel_neon.c
index cba9c66..1cb50b6 100644 (file)
--- a/src/lib/engines/common/evas_op_blend/op_blend_pixel_neon.c
+++ b/src/lib/engines/common/evas_op_blend/op_blend_pixel_neon.c
@@ -113,7 +113,8 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
                //** Oct loop
                "vldm   %[s]!,  {d0,d1,d2,d3)           \n\t"
                "vldm   %[d],   {d4,d5,d6,d7}           \n\t"
-
+                "pld    [%[s], #64]                     \n\t"
+                
 
                // Copy s.a into q2 (>> 24) & subtract from 255
                "vmvn.u8        q4,     q0              \n\t"