claude_and_no_claude.h

#define USE_AVX2_AMBIS
#ifdef USE_AVX2_AMBIS
            // Process 8 channels at a time using AVX
            int chan = 0;
            for (; chan <= num_outputchans - 8; chan += 8)
            {
                // Load 8 ambisonics coefficients for each source
                __m256 coeffs0 = _mm256_load_ps(&ambcoeffs[chan]);      // coeffs for outsample0
                __m256 coeffs1 = _mm256_load_ps(&ambcoeffs[chan + 64]); // coeffs for outsample1

                // Broadcast the scalar audio samples across all 8 lanes
                __m256 sample0 = _mm256_set1_ps(outsample0);
                __m256 sample1 = _mm256_set1_ps(outsample1);

                // Multiply-accumulate: sample * coefficients
                __m256 result =
                    _mm256_fmadd_ps(sample0, coeffs0,                 // outsample0 * coeffs0
                                    _mm256_mul_ps(sample1, coeffs1)); // + outsample1 * coeffs1

                // Store results into the output buffer
                _mm256_store_ps(&outputs[i * 64 + chan], result);
            }

            // Scalar fallback for any remaining channels (if num_outputchans isn't a multiple of 8)
            for (; chan < num_outputchans; ++chan)
            {
                outputs[i * 64 + chan] =
                    outsample0 * ambcoeffs[chan] + outsample1 * ambcoeffs[chan + 64];
            }
#else
            // Teemun alkuperäinen koodi
            for (int chan = 0; chan < num_outputchans; ++chan)
            {
                outputs[i * 64 + chan] = 0.0f;
                outputs[i * 64 + chan] += outsample0 * ambcoeffs[chan];
                outputs[i * 64 + chan] += outsample1 * ambcoeffs[chan + 64];
            }
#endif