<p><br>
On Nov 7, 2013 11:24 PM, "Steve Borho" <<a href="mailto:steve@borho.org">steve@borho.org</a>> wrote:<br>
><br>
><br>
><br>
><br>
> On Thu, Nov 7, 2013 at 5:17 AM, Aarthi Thirumalai <<a href="mailto:aarthi@multicorewareinc.com">aarthi@multicorewareinc.com</a>> wrote:<br>
>><br>
>> # HG changeset patch<br>
>> # User Aarthi Thirumalai<br>
>> # Date 1383823017 -19800<br>
>> # Thu Nov 07 16:46:57 2013 +0530<br>
>> # Node ID 0823d243b4e2f9eee6323a9750eeb811bccbad77<br>
>> # Parent 85002898f5b4308547af6ce464bbdff5f360fa13<br>
>> aq: bug fix.extend border of TComPic::m_origPicYuv to a multiple of 16<br>
>><br>
>> diff -r 85002898f5b4 -r 0823d243b4e2 source/Lib/TLibCommon/TComPicYuv.cpp<br>
>> --- a/source/Lib/TLibCommon/TComPicYuv.cpp Thu Nov 07 14:31:05 2013 +0530<br>
>> +++ b/source/Lib/TLibCommon/TComPicYuv.cpp Thu Nov 07 16:46:57 2013 +0530<br>
>> @@ -345,6 +345,10 @@<br>
>> /* width and height - without padsize */<br>
>> int width = m_picWidth - padx;<br>
>> int height = m_picHeight - pady;<br>
>> + uint8_t rem = m_picWidth % 16;<br>
>> + padx += rem ? 16 - rem : 0;<br>
>> + rem = m_picHeight % 16;<br>
>> + pady += rem ? 16 - rem : 0;<br>
><br>
><br>
> if the row width was 17, padx would already be 3 in order to pad out to the nearest multiple of 4. You would be doing: padx += 15, which would result in a pad of 18, which is more than what you bargained for. I believe the += should just be =. Also, Gopu just added a new else {} clause this is missed here, so I've hoisted this operation higher in the file.<br>
> <br>
>>when row width is 17, padx is already 3 and m_picWidth would be 20 at this stage. So padx += 12 not 15..<br>
>><br>
>> // Manually copy pixels to up-size them<br>
>> for (int r = 0; r < height; r++)<br>
>> @@ -475,10 +479,12 @@<br>
>> uint8_t *v = (uint8_t*)pic.planes[2];<br>
>><br>
>> /* width and height - without padsize */<br>
>> - int width = (m_picWidth * (pic.bitDepth > 8 ? 2 : 1)) - padx;<br>
>> - int height = m_picHeight - pady;<br>
>> -<br>
>> - // copy pixels by row into encoder's buffer<br>
>> + int width = (m_picWidth * (pic.bitDepth > 8 ? 2 : 1)) - padx;<br>
>> + int height = m_picHeight - pady;<br>
>> + uint8_t rem = m_picWidth % 16;<br>
>> + padx += rem ? 16 - rem : 0;<br>
>> + rem = m_picHeight % 16;<br>
>> + pady += rem ? 16 - rem : 0;<br>
><br>
><br>
> still white-space alignment problems here, but this code has been removed<br>
> <br>
>><br>
>> for (int r = 0; r < height; r++)<br>
>> {<br>
>> memcpy(Y, y, width);<br>
>> diff -r 85002898f5b4 -r 0823d243b4e2 source/common/lowres.cpp<br>
>> --- a/source/common/lowres.cpp Thu Nov 07 14:31:05 2013 +0530<br>
>> +++ b/source/common/lowres.cpp Thu Nov 07 16:46:57 2013 +0530<br>
>> @@ -150,31 +150,6 @@<br>
>> intraMbs[i] = 0;<br>
>> }<br>
>><br>
>> - int y, extWidth = (orig->getWidth() + X265_LOWRES_CU_SIZE - 1);<br>
>> - int srcStride = orig->getStride();<br>
>> - int srcHeight = orig->getHeight();<br>
>> - int srcWidth = orig->getWidth();<br>
>> - Pel *src;<br>
>> - src = orig->getLumaAddr();<br>
>> -<br>
>> - /* extending right margin */<br>
>> - if (2 * width > srcWidth)<br>
>> - {<br>
>> - for (y = 0; y < srcHeight; y++)<br>
>> - {<br>
>> - ::memset(src + srcWidth, src[srcWidth - 1], sizeof(Pel) * (X265_LOWRES_CU_SIZE - 1));<br>
>> - src += srcStride;<br>
>> - }<br>
>> - }<br>
>> -<br>
>> - /* extending bottom margin */<br>
>> - src = orig->getLumaAddr() + (srcHeight - 1) * srcStride;<br>
>> -<br>
>> - for (y = 1; y <= 2 * lines - srcHeight; y++)<br>
>> - {<br>
>> - ::memcpy(src + y * srcStride, src, sizeof(Pel) * (extWidth));<br>
>> - }<br>
>> -<br>
>> /* downscale and generate 4 HPEL planes for lookahead */<br>
>> primitives.frame_init_lowres_core(orig->getLumaAddr(),<br>
>> lowresPlane[0], lowresPlane[1], lowresPlane[2], lowresPlane[3],<br>
>> diff -r 85002898f5b4 -r 0823d243b4e2 source/encoder/frameencoder.cpp<br>
>> --- a/source/encoder/frameencoder.cpp Thu Nov 07 14:31:05 2013 +0530<br>
>> +++ b/source/encoder/frameencoder.cpp Thu Nov 07 16:46:57 2013 +0530<br>
>> @@ -1149,9 +1149,8 @@<br>
>> {<br>
>> /* Derive qpOffet for each CU by averaging offsets for all 16x16 blocks in the cu. */<br>
>> double qp_offset = 0;<br>
>> - int blockSize = g_maxCUWidth >> 2;<br>
>> - int maxBlockCols = (pic->getPicYuvOrg()->getWidth() + (blockSize - 1)) / blockSize;<br>
>> - int maxBlockRows = (pic->getPicYuvOrg()->getHeight() + (blockSize - 1)) / blockSize;<br>
>> + int maxBlockCols = (pic->getPicYuvOrg()->getWidth() + (16 - 1)) / 16;<br>
>> + int maxBlockRows = (pic->getPicYuvOrg()->getHeight() + (16 - 1)) / 16;<br>
><br>
><br>
> much better<br>
> <br>
>><br>
>> int block_y = (cuAddr / pic->getPicSym()->getFrameWidthInCU()) * 4;<br>
>> int block_x = (cuAddr * 4) - block_y * pic->getPicSym()->getFrameWidthInCU();<br>
>> int cnt = 0;<br>
>> _______________________________________________<br>
>> x265-devel mailing list<br>
>> <a href="mailto:x265-devel@videolan.org">x265-devel@videolan.org</a><br>
>> <a href="https://mailman.videolan.org/listinfo/x265-devel">https://mailman.videolan.org/listinfo/x265-devel</a><br>
><br>
><br>
><br>
><br>
> -- <br>
> Steve Borho<br>
><br>
> _______________________________________________<br>
> x265-devel mailing list<br>
> <a href="mailto:x265-devel@videolan.org">x265-devel@videolan.org</a><br>
> <a href="https://mailman.videolan.org/listinfo/x265-devel">https://mailman.videolan.org/listinfo/x265-devel</a><br>
><br>
</p>