Actually, even better would be to remove the double and floats being
used here and replace them with fixed point math so EMMS is not
required.

On Tue, Jun 10, 2014 at 9:26 PM, chen <[email protected]> wrote:
> The best way is find the asm code that need EMMS.
>
> At 2014-06-11 09:20:06,"Satoshi Nakagawa" <[email protected]> wrote:
>
>># HG changeset patch
>># User Satoshi Nakagawa <[email protected]>
>># Date 1402449456 -32400
>>#      Wed Jun 11 10:17:36 2014 +0900
>># Node ID 40ad5bf953cd6b80d97aba803f321ba655a388f7
>># Parent  d0bacf50eb951fe5f91e419072399b3dae8926d9
>>fix: emms issue
>>
>>diff -r d0bacf50eb95 -r 40ad5bf953cd source/encoder/compress.cpp
>>--- a/source/encoder/compress.cpp Tue Jun 10 18:37:26 2014 -0500
>>+++ b/source/encoder/compress.cpp Wed Jun 11 10:17:36 2014 +0900
>>@@ -378,6 +378,8 @@
>> #if TOPSKIP
>>     if (depth == 0)
>>     {
>>+        x265_emms();
>>+
>>         TComDataCU* colocated0 = slice->getNumRefIdx(REF_PIC_LIST_0) > 0 ?
>> slice->getRefPic(REF_PIC_LIST_0, 0)->getCU(outTempCU->getAddr()) : NULL;
>>         TComDataCU* colocated1 = slice->getNumRefIdx(REF_PIC_LIST_1) > 0 ?
>> slice->getRefPic(REF_PIC_LIST_1, 0)->getCU(outTempCU->getAddr()) : NULL;
>>         char currentQP = outTempCU->getQP(0);
>>@@ -648,6 +650,8 @@
>>         if (outBestCU != 0)
>> #endif
>>         {
>>+            x265_emms();
>>+
>>             uint64_t totalCostNeigh = 0, totalCostCU = 0, totalCountNeigh
>> = 0, totalCountCU = 0;
>>             double avgCost = 0;
>>             TComDataCU* above = outTempCU->getCUAbove();
>>@@ -836,6 +840,8 @@
>>             xCopyYuv2Pic(pic, outBestCU->getAddr(),
>> outBestCU->getZorderIdxInCU(), depth);
>>     }
>>
>>+    x265_emms();
>>+
>>     if (!bInsidePicture) return;
>>
>>     /* Assert if Best prediction mode is NONE
>>_______________________________________________
>>x265-devel mailing list
>>[email protected]
>>https://mailman.videolan.org/listinfo/x265-devel
>
> _______________________________________________
> x265-devel mailing list
> [email protected]
> https://mailman.videolan.org/listinfo/x265-devel
>



-- 
Steve Borho
_______________________________________________
x265-devel mailing list
[email protected]
https://mailman.videolan.org/listinfo/x265-devel

Reply via email to