# HG changeset patch
# User Min Chen <[email protected]>
# Date 1443652339 18000
# Node ID 3a346ffd4f2c21d3385e3e722c513ae42c454a57
# Parent fd5d57606bb724c7ece1073ee2d31b46b5bffba5
reduce shift operators in rdoQuant() since trSizeCG is constant
---
source/common/quant.cpp | 2 +-
source/common/quant.h | 8 ++++----
2 files changed, 5 insertions(+), 5 deletions(-)
diff -r fd5d57606bb7 -r 3a346ffd4f2c source/common/quant.cpp
--- a/source/common/quant.cpp Wed Sep 30 17:32:16 2015 -0500
+++ b/source/common/quant.cpp Wed Sep 30 17:32:19 2015 -0500
@@ -739,7 +739,7 @@
uint32_t ctxSet = (cgScanPos && bIsLuma) ? 2 : 0;
const uint32_t cgBlkPos = codeParams.scanCG[cgScanPos];
const uint32_t cgPosY = cgBlkPos >> log2TrSizeCG;
- const uint32_t cgPosX = cgBlkPos - (cgPosY << log2TrSizeCG);
+ const uint32_t cgPosX = cgBlkPos & ((1 << log2TrSizeCG) - 1);
const uint64_t cgBlkPosMask = ((uint64_t)1 << cgBlkPos);
const int patternSigCtx = calcPatternSigCtx(sigCoeffGroupFlag64,
cgPosX, cgPosY, cgBlkPos, cgStride);
const int ctxSigOffset = codeParams.firstSignificanceMapContext +
(cgScanPos && bIsLuma ? 3 : 0);
diff -r fd5d57606bb7 -r 3a346ffd4f2c source/common/quant.h
--- a/source/common/quant.h Wed Sep 30 17:32:16 2015 -0500
+++ b/source/common/quant.h Wed Sep 30 17:32:19 2015 -0500
@@ -126,8 +126,8 @@
const uint32_t sigPos = (uint32_t)(sigCoeffGroupFlag64 >> (cgBlkPos +
1)); // just need lowest 7-bits valid
// TODO: instruction BT is faster, but _bittest64 still generate
instruction 'BT m, r' in VS2012
- const uint32_t sigRight = ((uint32_t)(cgPosX - (trSizeCG - 1)) >> 31)
& sigPos;
- const uint32_t sigLower = ((uint32_t)(cgPosY - (trSizeCG - 1)) >> 31)
& (sigPos >> (trSizeCG - 1));
+ const uint32_t sigRight = (cgPosX != (trSizeCG - 1)) & sigPos;
+ const uint32_t sigLower = (cgPosY != (trSizeCG - 1)) & (sigPos >>
(trSizeCG - 1));
return sigRight + sigLower * 2;
}
@@ -137,8 +137,8 @@
X265_CHECK(cgBlkPos < 64, "cgBlkPos is too large\n");
// NOTE: unsafe shift operator, see NOTE in calcPatternSigCtx
const uint32_t sigPos = (uint32_t)(cgGroupMask >> (cgBlkPos + 1)); //
just need lowest 8-bits valid
- const uint32_t sigRight = ((uint32_t)(cgPosX - (trSizeCG - 1)) >> 31)
& sigPos;
- const uint32_t sigLower = ((uint32_t)(cgPosY - (trSizeCG - 1)) >> 31)
& (sigPos >> (trSizeCG - 1));
+ const uint32_t sigRight = (cgPosX != (trSizeCG - 1)) & sigPos;
+ const uint32_t sigLower = (cgPosY != (trSizeCG - 1)) & (sigPos >>
(trSizeCG - 1));
return (sigRight | sigLower);
}
_______________________________________________
x265-devel mailing list
[email protected]
https://mailman.videolan.org/listinfo/x265-devel
