Optimize png_do_expand_palette for ARM

ARM-specific optimization processes 8 or 4 pixels at once.
Improves performance by around 10-22% on a recent ARM Chromebook.

1 files changed, 23 insertions, 0 deletions

diff --git a/pngpriv.h b/pngpriv.h
index a23febb81..973c3eac1 100644
--- a/pngpriv.h
+++ b/pngpriv.h
@@ -2117,6 +2117,29 @@ PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_sse2,
 PNG_INTERNAL_FUNCTION(png_uint_32, png_check_keyword, (png_structrp png_ptr,
    png_const_charp key, png_bytep new_key), PNG_EMPTY);
 
+#if PNG_ARM_NEON_IMPLEMENTATION == 1
+PNG_INTERNAL_FUNCTION(void,
+                      png_riffle_palette_rgba,
+                      (png_structrp, png_row_infop),
+                      PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(int,
+                      png_do_expand_palette_neon_rgba,
+                      (png_structrp,
+                       png_row_infop,
+                       png_const_bytep,
+                       const png_bytepp,
+                       const png_bytepp),
+                      PNG_EMPTY);
+PNG_INTERNAL_FUNCTION(int,
+                      png_do_expand_palette_neon_rgb,
+                      (png_structrp,
+                       png_row_infop,
+                       png_const_bytep,
+                       const png_bytepp,
+                       const png_bytepp),
+                      PNG_EMPTY);
+#endif
+
 /* Maintainer: Put new private prototypes here ^ */
 
 #include "pngdebug.h"