[FFmpeg-cvslog] vp8/armv6: mc: avoid boolean expression in calculation

[Janne Grunau]

ffmpeg | branch: master | Janne Grunau <janne-libav at jannau.net> | Sat Jul  9 15:30:34 2016 +0200| [5f74bd31a9bd1ac7655103b11743c12d38e0419f] | committer: Janne Grunau

vp8/armv6: mc: avoid boolean expression in calculation

GNU as evaluates true as '-1' while Apple's variant and llvm's internal
assembler evaluate it as '1'. The best way to avoid this madness is to
eliminate boolean expressions instead of trying to fix it with
preprocessor directives. Use a direct formula to calculate the
required temporary space on the stack in
ff_put_vp8_{epel,bilin}{4,8,16}_h[246]v[246]_armv6().

Fixes a checkasm segfault in vp8dsp.mc when using llvm's internal
assembler for a non-Apple target.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=5f74bd31a9bd1ac7655103b11743c12d38e0419f
---

 libavcodec/arm/vp8dsp_armv6.S | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

diff --git a/libavcodec/arm/vp8dsp_armv6.S b/libavcodec/arm/vp8dsp_armv6.S
index 4e83fe1..565361e 100644
--- a/libavcodec/arm/vp8dsp_armv6.S
+++ b/libavcodec/arm/vp8dsp_armv6.S
@@ -1226,13 +1226,8 @@ vp8_mc_1                bilin,  8, v
 vp8_mc_1                bilin,  4, h
 vp8_mc_1                bilin,  4, v
 
-/* True relational expressions have the value -1 in the GNU assembler,
-   +1 in Apple's. */
-#ifdef __APPLE__
-#   define TMPSIZE \size * (8 + 8*(\size > 4) + \ytaps - 1)
-#else
-#   define TMPSIZE \size * (8 - 8*(\size > 4) + \ytaps - 1)
-#endif
+@ 4 and 8 pixel wide mc blocks might have height of 8 or 16 lines
+#define TMPSIZE \size * (16 / ((16 / \size + 1) / 2) + \ytaps - 1)
 
 .macro  vp8_mc_hv       name, size, h, v, ytaps
 function ff_put_vp8_\name\size\()_\h\v\()_armv6, export=1