[Opus] Update to v1.2.1

third_party/opus/src/celt/mdct.c

 /* Copyright (c) 2007-2008 CSIRO
    Copyright (c) 2007-2008 Xiph.Org Foundation
    Written by Jean-Marc Valin */
 /*
    Redistribution and use in source and binary forms, with or without
    modification, are permitted provided that the following conditions
    are met:
 
    - Redistributions of source code must retain the above copyright
    notice, this list of conditions and the following disclaimer.
@@ skipping 252 matching lines @@
       const kiss_fft_scalar * OPUS_RESTRICT xp1 = in;
       const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+stride*(N2-1);
       kiss_fft_scalar * OPUS_RESTRICT yp = out+(overlap>>1);
       const kiss_twiddle_scalar * OPUS_RESTRICT t = &trig[0];
       const opus_int16 * OPUS_RESTRICT bitrev = l->kfft[shift]->bitrev;
       for(i=0;i<N4;i++)
       {
          int rev;
          kiss_fft_scalar yr, yi;
          rev = *bitrev++;
-         yr = S_MUL(*xp2, t[i]) + S_MUL(*xp1, t[N4+i]);
-         yi = S_MUL(*xp1, t[i]) - S_MUL(*xp2, t[N4+i]);
+         yr = ADD32_ovflw(S_MUL(*xp2, t[i]), S_MUL(*xp1, t[N4+i]));
+         yi = SUB32_ovflw(S_MUL(*xp1, t[i]), S_MUL(*xp2, t[N4+i]));
          /* We swap real and imag because we use an FFT instead of an IFFT. */
          yp[2*rev+1] = yr;
          yp[2*rev] = yi;
          /* Storing the pre-rotation directly in the bitrev order. */
          xp1+=2*stride;
          xp2-=2*stride;
       }
    }
 
    opus_fft_impl(l->kfft[shift], (kiss_fft_cpx*)(out+(overlap>>1)));
 
    /* Post-rotate and de-shuffle from both ends of the buffer at once to make
       it in-place. */
    {
       kiss_fft_scalar * yp0 = out+(overlap>>1);
       kiss_fft_scalar * yp1 = out+(overlap>>1)+N2-2;
       const kiss_twiddle_scalar *t = &trig[0];
       /* Loop to (N4+1)>>1 to handle odd N4. When N4 is odd, the
          middle pair will be computed twice. */
       for(i=0;i<(N4+1)>>1;i++)
       {
          kiss_fft_scalar re, im, yr, yi;
          kiss_twiddle_scalar t0, t1;
          /* We swap real and imag because we're using an FFT instead of an IFFT. */
          re = yp0[1];
          im = yp0[0];
          t0 = t[i];
          t1 = t[N4+i];
          /* We'd scale up by 2 here, but instead it's done when mixing the windows */
-         yr = S_MUL(re,t0) + S_MUL(im,t1);
-         yi = S_MUL(re,t1) - S_MUL(im,t0);
+         yr = ADD32_ovflw(S_MUL(re,t0), S_MUL(im,t1));
+         yi = SUB32_ovflw(S_MUL(re,t1), S_MUL(im,t0));
          /* We swap real and imag because we're using an FFT instead of an IFFT. */
          re = yp1[1];
          im = yp1[0];
          yp0[0] = yr;
          yp1[1] = yi;
 
          t0 = t[(N4-i-1)];
          t1 = t[(N2-i-1)];
          /* We'd scale up by 2 here, but instead it's done when mixing the windows */
-         yr = S_MUL(re,t0) + S_MUL(im,t1);
-         yi = S_MUL(re,t1) - S_MUL(im,t0);
+         yr = ADD32_ovflw(S_MUL(re,t0), S_MUL(im,t1));
+         yi = SUB32_ovflw(S_MUL(re,t1), S_MUL(im,t0));
          yp1[0] = yr;
          yp0[1] = yi;
          yp0 += 2;
          yp1 -= 2;
       }
    }
 
    /* Mirror on both sides for TDAC */
    {
       kiss_fft_scalar * OPUS_RESTRICT xp1 = out+overlap-1;
       kiss_fft_scalar * OPUS_RESTRICT yp1 = out;
       const opus_val16 * OPUS_RESTRICT wp1 = window;
       const opus_val16 * OPUS_RESTRICT wp2 = window+overlap-1;
 
       for(i = 0; i < overlap/2; i++)
       {
          kiss_fft_scalar x1, x2;
          x1 = *xp1;
          x2 = *yp1;
-         *yp1++ = MULT16_32_Q15(*wp2, x2) - MULT16_32_Q15(*wp1, x1);
-         *xp1-- = MULT16_32_Q15(*wp1, x2) + MULT16_32_Q15(*wp2, x1);
+         *yp1++ = SUB32_ovflw(MULT16_32_Q15(*wp2, x2), MULT16_32_Q15(*wp1, x1));
+         *xp1-- = ADD32_ovflw(MULT16_32_Q15(*wp1, x2), MULT16_32_Q15(*wp2, x1));
          wp1++;
          wp2--;
       }
    }
 }
 #endif /* OVERRIDE_clt_mdct_backward */