aom_dsp/get_prob: make clip_prob branchless
+ inline the function directly as there was only one consumer
(get_prob())
this is an attempt to reduce the amount of branches to workaround an amd
bug. this change is mildly faster or neutral across x86-64, arm.
http://support.amd.com/TechDocs/44739_12h_Rev_Gd.pdf
665 Integer Divide Instruction May Cause Unpredictable Behavior
cherry-picked from libvpx:
7481edb33 vpx_dsp/get_prob: make clip_prob branchless
Change-Id: I433059c61ce43ec5058cc16ca590d186bfa8aab5
diff --git a/aom_dsp/prob.h b/aom_dsp/prob.h
index 94571dc..e88a570 100644
--- a/aom_dsp/prob.h
+++ b/aom_dsp/prob.h
@@ -53,13 +53,14 @@
typedef const aom_tree_index aom_tree[];
-static INLINE aom_prob clip_prob(int p) {
- return (p > 255) ? 255 : (p < 1) ? 1 : p;
-}
-
static INLINE aom_prob get_prob(unsigned int num, unsigned int den) {
if (den == 0) return 128u;
- return clip_prob((int)(((int64_t)num * 256 + (den >> 1)) / den));
+ {
+ const int p = (int)(((int64_t)num * 256 + (den >> 1)) / den);
+ // (p > 255) ? 255 : (p < 1) ? 1 : p;
+ const int clipped_prob = p | ((255 - p) >> 23) | (p == 0);
+ return (aom_prob)clipped_prob;
+ }
}
static INLINE aom_prob get_binary_prob(unsigned int n0, unsigned int n1) {
