diff --git a/audio/sinc.c b/audio/sinc.c index 4041e8caed..e4ab461755 100644 --- a/audio/sinc.c +++ b/audio/sinc.c @@ -123,6 +123,8 @@ rarch_resampler_t *resampler_new(void) RARCH_LOG("Sinc resampler [AVX]\n"); #elif defined(__SSE__) RARCH_LOG("Sinc resampler [SSE]\n"); +#elif defined(HAVE_NEON) + RARCH_LOG("Sinc resampler [NEON]\n"); #else RARCH_LOG("Sinc resampler [C]\n"); #endif @@ -208,6 +210,18 @@ static void process_sinc(rarch_resampler_t *resamp, float *out_buffer) // movehl { X, R, X, L } == { X, R, X, R } _mm_store_ss(out_buffer + 1, _mm_movehl_ps(sum, sum)); } +#elif defined(HAVE_NEON) +void process_sinc_neon_asm(float *out, const float *left, const float *right, const float *coeff); +static void process_sinc(rarch_resampler_t *resamp, float *out_buffer) +{ + const float *buffer_l = resamp->buffer_l + resamp->ptr; + const float *buffer_r = resamp->buffer_r + resamp->ptr; + + unsigned phase = resamp->time >> SUBPHASE_BITS; + const float *phase_table = resamp->phase_table[phase]; + + process_sinc_neon_asm(out_buffer, buffer_l, buffer_r, phase_table); +} #else // Plain ol' C99 static void process_sinc(rarch_resampler_t *resamp, float *out_buffer) { diff --git a/audio/sinc_neon.S b/audio/sinc_neon.S new file mode 100644 index 0000000000..f6c78ce122 --- /dev/null +++ b/audio/sinc_neon.S @@ -0,0 +1,52 @@ +/* RetroArch - A frontend for libretro. + * Copyright (C) 2010-2012 - Hans-Kristian Arntzen + * + * RetroArch is free software: you can redistribute it and/or modify it under the terms + * of the GNU General Public License as published by the Free Software Found- + * ation, either version 3 of the License, or (at your option) any later version. + * + * RetroArch is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; + * without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR + * PURPOSE. See the GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along with RetroArch. + * If not, see . + */ + +.arm +.align 4 +.global process_sinc_neon_asm +# void process_sinc_neon(float *out, const float *left, const float *right, const float *coeff) +# Hardcoded to 16 taps. +process_sinc_neon_asm: + # Left + vld1.f32 {q0-q1}, [r1]! + vld1.f32 {q2-q3}, [r1]! + # Right + vld1.f32 {q4-q5}, [r2]! + vld1.f32 {q6-q7}, [r2]! + # Coeff + vld1.f32 {q8-q9}, [r3, :1024]! + vld1.f32 {q10-q11}, [r3, :1024]! + + # Left + vmul.f32 q0, q0, q8 + vmul.f32 q1, q1, q9 + vmla.f32 q0, q2, q10 + vmla.f32 q1, q3, q11 + + # Right + vmul.f32 q4, q4, q8 + vmul.f32 q5, q5, q9 + vmla.f32 q4, q6, q10 + vmla.f32 q5, q7, q11 + + # Add everything together + vadd.f32 q0, q0, q1 + vadd.f32 q4, q4, q5 + vadd.f32 d0, d0, d1 + vadd.f32 d8, d8, d9 + vpadd.f32 d0, d0, d8 + vst1.f32 d0, [r0] + + bx lr