[x265] fix lossless

[Satoshi Nakagawa]

Negative shift or bts cause unexpected behavior.

 

My Core i5 4300U, (Win7 32bit, VS2013) fails following test.

 

--- a/source/test/pixelharness.cpp     Fri Aug 22 15:53:34 2014 -0500

+++ b/source/test/pixelharness.cpp   Mon Aug 25 11:30:37 2014 +0900

@@ -562,7 +562,8 @@

     intptr_t stride = STRIDE;

     for (int i = 0; i < ITERS; i++)

     {

-        int shift = (rand() % 7 + 1);

+//        int shift = (rand() % 7 + 1);

+        int shift = 0;

         int index = i % TEST_CASES;

         checked(opt, opt_dest, int_test_buff[index] + j, stride, shift,
(int)STRIDE);

 

 

 

 

From: x265-devel [mailto:x265-devel-bounces at videolan.org] On Behalf Of chen
Sent: Monday, August 25, 2014 11:20 AM
To: Development for x265
Subject: Re: [x265] fix lossless

 

He just modify shift=0 path, shif-left and shift-righ is equal in that time

At 2014-08-25 10:06:23,"Steve Borho" <steve at borho.org> wrote:
>On 08/23, Satoshi Nakagawa wrote:
>> # HG changeset patch
>> # User Satoshi Nakagawa <nakagawa424 at oki.com>
>> # Date 1408803114 -32400
>> #      Sat Aug 23 23:11:54 2014 +0900
>> # Node ID 218b9ddfe148c55afccc2c5a073858692e09f5c6
>> # Parent  6e6756f94b27c3ef30f6159f1880112a7ff978e3
>> fix lossless
> 
>do you have a repro case that this fixes?  I tried Main and Main10
>encodes with --lossless --hash 1 and they were both fine prior to this
>patch.
> 
>> diff -r 6e6756f94b27 -r 218b9ddfe148 source/common/quant.cpp
>> --- a/source/common/quant.cpp   Fri Aug 22 15:53:34 2014 -0500
>> +++ b/source/common/quant.cpp   Sat Aug 23 23:11:54 2014 +0900
>> @@ -399,7 +399,7 @@
>>  {
>>      if (transQuantBypass)
>>      {
>> -        primitives.cvt32to16_shr(residual, coeff, stride, 0, 1 <<
log2TrSize);
>> +        primitives.cvt32to16_shl[log2TrSize - 2](residual, coeff,
stride, 0);
>>          return;
>>      }
>>  
>> @@ -430,7 +430,7 @@
>>  #if X265_DEPTH <= 10
>>          primitives.cvt32to16_shr(residual, m_resiDctCoeff, stride,
shift, trSize);
>>  #else
>> -        if (shift >= 0)
>> +        if (shift > 0)
>>              primitives.cvt32to16_shr(residual, m_resiDctCoeff, stride,
shift, trSize);
>>          else
>>              primitives.cvt32to16_shl[log2TrSize - 2](residual,
m_resiDctCoeff, stride, -shift);
>> _______________________________________________
>> x265-devel mailing list
>> x265-devel at videolan.org
>> https://mailman.videolan.org/listinfo/x265-devel
> 
>-- 
>Steve Borho
>_______________________________________________
>x265-devel mailing list
>x265-devel at videolan.org
>https://mailman.videolan.org/listinfo/x265-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20140825/48c1ae92/attachment-0001.html>