generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

vpx_dsp/get_prob: make clip_prob branchless

Browse Source 
+ 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

BUG=chromium:639712

Suggested-by: Pascal Massimino <pascal.massimino@gmail.com>
Change-Id: Ia91823aded79aab469dd68095d44300e8df04ed2
This commit is contained in:
James Zern 2016-09-27 19:43:03 -07:00
parent 240726ac85
commit 7481edb33f
1 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
vpx_dsp/prob.h
View File

@ -43,13 +43,14 @@ typedef int8_t vpx_tree_index;
typedef const vpx_tree_index vpx_tree[];
static INLINE vpx_prob clip_prob(int p) {
return (p > 255) ? 255 : (p < 1) ? 1 : p;
}
static INLINE vpx_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 (vpx_prob)clipped_prob;
}
}
static INLINE vpx_prob get_binary_prob(unsigned int n0, unsigned int n1) {