From: Siarhei Siamashka <[email protected]>
After fast path cache introduction, the overhead of having this fallback is
insignificant. On the other hand, some of the ARM assembly optimizations (for
example nearest neighbor scaling) do not need NEON.
---
pixman/pixman-arm-neon.c | 8 ++++++--
1 files changed, 6 insertions(+), 2 deletions(-)
diff --git a/pixman/pixman-arm-neon.c b/pixman/pixman-arm-neon.c
index cdaa29a..9c6a241 100644
--- a/pixman/pixman-arm-neon.c
+++ b/pixman/pixman-arm-neon.c
@@ -347,9 +347,13 @@ BIND_COMBINE_U (out_reverse)
pixman_implementation_t *
_pixman_implementation_create_arm_neon (void)
{
- pixman_implementation_t *general = _pixman_implementation_create_fast_path
();
+#ifdef USE_ARM_SIMD
+ pixman_implementation_t *fallback = _pixman_implementation_create_arm_simd
();
+#else
+ pixman_implementation_t *fallback =
_pixman_implementation_create_fast_path ();
+#endif
pixman_implementation_t *imp =
- _pixman_implementation_create (general, arm_neon_fast_paths);
+ _pixman_implementation_create (fallback, arm_neon_fast_paths);
imp->combine_32[PIXMAN_OP_OVER] = neon_combine_over_u;
imp->combine_32[PIXMAN_OP_ADD] = neon_combine_add_u;
--
1.7.2.2
_______________________________________________
Pixman mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/pixman
