Something

 void Render(sF32 *outbuf, sInt samples)
  {
    const sF32 step=sF32(PAULARATE)/sF32(OUTRATE);
    const sF32 pan=1;
    const sF32 vm0=MasterVolume*sFSqrt(pan);
    const sF32 vm1=MasterVolume*sFSqrt(1-pan);

    for (sInt s=0; s<samples; s++)
    {
      sInt ReadEnd=ReadPos+FIR_WIDTH+1;
      if (WritePos<ReadPos) ReadEnd-=RBSIZE;
      if (ReadEnd>WritePos) Calc();
      sF32 outl0=0, outl1=0;
      sF32 outr0=0, outr1=0;

      // this needs optimization. SSE would come to mind.
      sInt offs=(ReadPos-FIR_WIDTH-1)&(RBSIZE-1);
      sF32 vl=RingBuf[offs];
      sF32 vr=RingBuf[offs+RBSIZE];
      for (sInt i=1; i<2*FIR_WIDTH-1; i++)
      {
        sF32 w=FIRMem[i];
        outl0+=vl*w;
        outr0+=vr*w;
        offs=(offs+1)&(RBSIZE-1);
        vl=RingBuf[offs];
        vr=RingBuf[offs+RBSIZE];
        outl1+=vl*w;
        outr1+=vr*w;
      }
      sF32 outl=sLerp(outl0,outl1,ReadFrac);
      sF32 outr=sLerp(outr0,outr1,ReadFrac);
      *outbuf++=vm0*outl+vm1*outr;
      *outbuf++=vm1*outl+vm0*outr;
	
      ReadFrac+=step;
      sInt rfi=sInt(ReadFrac);
      ReadPos=(ReadPos+rfi)&(RBSIZE-1);
      ReadFrac-=rfi;
    }

  }