Refactor CServer::MixEncodeTransmitData for performance

[ann0see]

To me it seems as if CServer::MixEncodeTransmitData() has a lot of branches, is messy and likely worth optimizing. For example , we could likely also use a pattern like

                        vecfIntermProcBuf[k] += vecsData[i] * fGainL;
                        vecfIntermProcBuf[k + 1] += vecsData[i] * fGainR;

instead of

                        if ( ( i & 1 ) == 0 )
                        {
                            // if even : left channel
                            vecfIntermProcBuf[i] += vecsData[i] * fGainL;
                        }
                        else
                        {
                            // if odd  : right channel
                            vecfIntermProcBuf[i] += vecsData[i] * fGainR;
                        }

jamulus/src/server.cpp

Line 914 in 3827902

void CServer::MixEncodeTransmitData ( const int iChanCnt, const int iNumClients )

As this seems to be the probably most important function in Jamulus, this warrants a close investigation

[ann0see]

CC @pljones @softins

[ann0see]

            // if channel gain is 1, avoid multiplication for speed optimization
            if ( fGain == 1.0f )
            {

Questionable if this really gives speedup...

[ann0see]

                        vecfIntermProcBuf[i] += ( static_cast<float> ( vecsData[k] ) + vecsData[k + 1] ) / 2;

the /2 can definitely be made * 0.5 even if the compiler does this itself

[ann0see]

                    const int iOffset = iB * SYSTEM_FRAME_SIZE_SAMPLES * vecNumAudioChannels[iChanCnt];

Probably just iOffset+=iB is enough with respective refactoring.

[pljones]

Anything else needed to close this that won't make it into 4.0.0? If not, could you change the milestone and close. Ta.

[ann0see]

I've checked the assembly and there's basically no vectorization for the code. Also I still believe that the function is insanely messy and should be refactored