/* * Copyright (C) 2011 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ #include <cutils/bitops.h> /* for popcount() */ #include <audio_utils/primitives.h> #include "private/private.h" void ditherAndClamp(int32_t *out, const int32_t *sums, size_t pairs) { for (; pairs > 0; --pairs) { const int32_t l = clamp16(*sums++ >> 12); const int32_t r = clamp16(*sums++ >> 12); *out++ = (r << 16) | (l & 0xFFFF); } } void memcpy_to_i16_from_q4_27(int16_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst++ = clamp16(*src++ >> 12); } } void memcpy_to_i16_from_u8(int16_t *dst, const uint8_t *src, size_t count) { dst += count; src += count; for (; count > 0; --count) { *--dst = (int16_t)(*--src - 0x80) << 8; } } void memcpy_to_u8_from_i16(uint8_t *dst, const int16_t *src, size_t count) { for (; count > 0; --count) { *dst++ = (*src++ >> 8) + 0x80; } } void memcpy_to_u8_from_float(uint8_t *dst, const float *src, size_t count) { for (; count > 0; --count) { *dst++ = clamp8_from_float(*src++); } } void memcpy_to_i16_from_i32(int16_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst++ = *src++ >> 16; } } void memcpy_to_i16_from_float(int16_t *dst, const float *src, size_t count) { for (; count > 0; --count) { *dst++ = clamp16_from_float(*src++); } } void memcpy_to_float_from_q4_27(float *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst++ = float_from_q4_27(*src++); } } void memcpy_to_float_from_i16(float *dst, const int16_t *src, size_t count) { dst += count; src += count; for (; count > 0; --count) { *--dst = float_from_i16(*--src); } } void memcpy_to_float_from_u8(float *dst, const uint8_t *src, size_t count) { dst += count; src += count; for (; count > 0; --count) { *--dst = float_from_u8(*--src); } } void memcpy_to_float_from_p24(float *dst, const uint8_t *src, size_t count) { dst += count; src += count * 3; for (; count > 0; --count) { src -= 3; *--dst = float_from_p24(src); } } void memcpy_to_i16_from_p24(int16_t *dst, const uint8_t *src, size_t count) { for (; count > 0; --count) { #if HAVE_BIG_ENDIAN *dst++ = src[1] | (src[0] << 8); #else *dst++ = src[1] | (src[2] << 8); #endif src += 3; } } void memcpy_to_i32_from_p24(int32_t *dst, const uint8_t *src, size_t count) { dst += count; src += count * 3; for (; count > 0; --count) { src -= 3; #if HAVE_BIG_ENDIAN *--dst = (src[2] << 8) | (src[1] << 16) | (src[0] << 24); #else *--dst = (src[0] << 8) | (src[1] << 16) | (src[2] << 24); #endif } } void memcpy_to_p24_from_i16(uint8_t *dst, const int16_t *src, size_t count) { dst += count * 3; src += count; for (; count > 0; --count) { dst -= 3; const int16_t sample = *--src; #if HAVE_BIG_ENDIAN dst[0] = sample >> 8; dst[1] = sample; dst[2] = 0; #else dst[0] = 0; dst[1] = sample; dst[2] = sample >> 8; #endif } } void memcpy_to_p24_from_float(uint8_t *dst, const float *src, size_t count) { for (; count > 0; --count) { int32_t ival = clamp24_from_float(*src++); #if HAVE_BIG_ENDIAN *dst++ = ival >> 16; *dst++ = ival >> 8; *dst++ = ival; #else *dst++ = ival; *dst++ = ival >> 8; *dst++ = ival >> 16; #endif } } void memcpy_to_p24_from_q8_23(uint8_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { int32_t ival = clamp24_from_q8_23(*src++); #if HAVE_BIG_ENDIAN *dst++ = ival >> 16; *dst++ = ival >> 8; *dst++ = ival; #else *dst++ = ival; *dst++ = ival >> 8; *dst++ = ival >> 16; #endif } } void memcpy_to_p24_from_i32(uint8_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { int32_t ival = *src++ >> 8; #if HAVE_BIG_ENDIAN *dst++ = ival >> 16; *dst++ = ival >> 8; *dst++ = ival; #else *dst++ = ival; *dst++ = ival >> 8; *dst++ = ival >> 16; #endif } } void memcpy_to_q8_23_from_i16(int32_t *dst, const int16_t *src, size_t count) { dst += count; src += count; for (; count > 0; --count) { *--dst = (int32_t)*--src << 8; } } void memcpy_to_q8_23_from_float_with_clamp(int32_t *dst, const float *src, size_t count) { for (; count > 0; --count) { *dst++ = clamp24_from_float(*src++); } } void memcpy_to_q8_23_from_p24(int32_t *dst, const uint8_t *src, size_t count) { dst += count; src += count * 3; for (; count > 0; --count) { src -= 3; #if HAVE_BIG_ENDIAN *--dst = (int8_t)src[0] << 16 | src[1] << 8 | src[2]; #else *--dst = (int8_t)src[2] << 16 | src[1] << 8 | src[0]; #endif } } void memcpy_to_q4_27_from_float(int32_t *dst, const float *src, size_t count) { for (; count > 0; --count) { *dst++ = clampq4_27_from_float(*src++); } } void memcpy_to_i16_from_q8_23(int16_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst++ = clamp16(*src++ >> 8); } } void memcpy_to_float_from_q8_23(float *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst++ = float_from_q8_23(*src++); } } void memcpy_to_i32_from_i16(int32_t *dst, const int16_t *src, size_t count) { dst += count; src += count; for (; count > 0; --count) { *--dst = (int32_t)*--src << 16; } } void memcpy_to_i32_from_float(int32_t *dst, const float *src, size_t count) { for (; count > 0; --count) { *dst++ = clamp32_from_float(*src++); } } void memcpy_to_float_from_i32(float *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst++ = float_from_i32(*src++); } } void memcpy_to_float_from_float_with_clamping(float *dst, const float *src, size_t count, float absMax) { // Note: using NEON intrinsics (vminq_f32, vld1q_f32...) did NOT accelerate // the function when benchmarked. The compiler already vectorize using FMINNM f32x4 & similar. // Note: clamping induce a ~20% overhead compared to memcpy for count in [64, 512] // See primitives_benchmark for (; count > 0; --count) { const float sample = *src++; *dst++ = fmax(-absMax, fmin(absMax, sample)); } } void downmix_to_mono_i16_from_stereo_i16(int16_t *dst, const int16_t *src, size_t count) { for (; count > 0; --count) { *dst++ = (int16_t)(((int32_t)src[0] + (int32_t)src[1]) >> 1); src += 2; } } void upmix_to_stereo_i16_from_mono_i16(int16_t *dst, const int16_t *src, size_t count) { dst += count * 2; src += count; for (; count > 0; --count) { const int32_t temp = *--src; dst -= 2; dst[0] = temp; dst[1] = temp; } } void downmix_to_mono_float_from_stereo_float(float *dst, const float *src, size_t frames) { for (; frames > 0; --frames) { *dst++ = (src[0] + src[1]) * 0.5; src += 2; } } void upmix_to_stereo_float_from_mono_float(float *dst, const float *src, size_t frames) { dst += frames * 2; src += frames; for (; frames > 0; --frames) { const float temp = *--src; dst -= 2; dst[0] = temp; dst[1] = temp; } } size_t nonZeroMono32(const int32_t *samples, size_t count) { size_t nonZero = 0; for (; count > 0; --count) { nonZero += *samples++ != 0; } return nonZero; } size_t nonZeroMono16(const int16_t *samples, size_t count) { size_t nonZero = 0; for (; count > 0; --count) { nonZero += *samples++ != 0; } return nonZero; } size_t nonZeroStereo32(const int32_t *frames, size_t count) { size_t nonZero = 0; for (; count > 0; --count) { nonZero += frames[0] != 0 || frames[1] != 0; frames += 2; } return nonZero; } size_t nonZeroStereo16(const int16_t *frames, size_t count) { size_t nonZero = 0; for (; count > 0; --count) { nonZero += frames[0] != 0 || frames[1] != 0; frames += 2; } return nonZero; } /* * C macro to do channel mask copying independent of dst/src sample type. * Don't pass in any expressions for the macro arguments here. */ #define copy_frame_by_mask(dst, dmask, src, smask, count, zero) \ { \ uint32_t bit, ormask; \ for (; (count) > 0; --(count)) { \ ormask = (dmask) | (smask); \ while (ormask) { \ bit = ormask & -ormask; /* get lowest bit */ \ ormask ^= bit; /* remove lowest bit */ \ if ((dmask) & bit) { \ *(dst)++ = (smask) & bit ? *(src)++ : (zero); \ } else { /* source channel only */ \ ++(src); \ } \ } \ } \ } void memcpy_by_channel_mask(void *dst, uint32_t dst_mask, const void *src, uint32_t src_mask, size_t sample_size, size_t count) { #if 0 /* alternate way of handling memcpy_by_channel_mask by using the idxary */ int8_t idxary[32]; uint32_t src_channels = popcount(src_mask); uint32_t dst_channels = memcpy_by_index_array_initialization(idxary, 32, dst_mask, src_mask); memcpy_by_idxary(dst, dst_channels, src, src_channels, idxary, sample_size, count); #else if (dst_mask == src_mask) { memcpy(dst, src, sample_size * popcount(dst_mask) * count); return; } switch (sample_size) { case 1: { uint8_t *udst = (uint8_t*)dst; const uint8_t *usrc = (const uint8_t*)src; copy_frame_by_mask(udst, dst_mask, usrc, src_mask, count, 0); } break; case 2: { uint16_t *udst = (uint16_t*)dst; const uint16_t *usrc = (const uint16_t*)src; copy_frame_by_mask(udst, dst_mask, usrc, src_mask, count, 0); } break; case 3: { /* could be slow. use a struct to represent 3 bytes of data. */ uint8x3_t *udst = (uint8x3_t*)dst; const uint8x3_t *usrc = (const uint8x3_t*)src; static const uint8x3_t zero; /* tricky - we use this to zero out a sample */ copy_frame_by_mask(udst, dst_mask, usrc, src_mask, count, zero); } break; case 4: { uint32_t *udst = (uint32_t*)dst; const uint32_t *usrc = (const uint32_t*)src; copy_frame_by_mask(udst, dst_mask, usrc, src_mask, count, 0); } break; default: abort(); /* illegal value */ break; } #endif } /* * C macro to do copying by index array, to rearrange samples * within a frame. This is independent of src/dst sample type. * Don't pass in any expressions for the macro arguments here. */ #define copy_frame_by_idx(dst, dst_channels, src, src_channels, idxary, count, zero) \ { \ unsigned i; \ int index; \ for (; (count) > 0; --(count)) { \ for (i = 0; i < (dst_channels); ++i) { \ index = (idxary)[i]; \ *(dst)++ = index < 0 ? (zero) : (src)[index]; \ } \ (src) += (src_channels); \ } \ } void memcpy_by_index_array(void *dst, uint32_t dst_channels, const void *src, uint32_t src_channels, const int8_t *idxary, size_t sample_size, size_t count) { switch (sample_size) { case 1: { uint8_t *udst = (uint8_t*)dst; const uint8_t *usrc = (const uint8_t*)src; copy_frame_by_idx(udst, dst_channels, usrc, src_channels, idxary, count, 0); } break; case 2: { uint16_t *udst = (uint16_t*)dst; const uint16_t *usrc = (const uint16_t*)src; copy_frame_by_idx(udst, dst_channels, usrc, src_channels, idxary, count, 0); } break; case 3: { /* could be slow. use a struct to represent 3 bytes of data. */ uint8x3_t *udst = (uint8x3_t*)dst; const uint8x3_t *usrc = (const uint8x3_t*)src; static const uint8x3_t zero; copy_frame_by_idx(udst, dst_channels, usrc, src_channels, idxary, count, zero); } break; case 4: { uint32_t *udst = (uint32_t*)dst; const uint32_t *usrc = (const uint32_t*)src; copy_frame_by_idx(udst, dst_channels, usrc, src_channels, idxary, count, 0); } break; default: abort(); /* illegal value */ break; } } size_t memcpy_by_index_array_initialization(int8_t *idxary, size_t idxcount, uint32_t dst_mask, uint32_t src_mask) { size_t n = 0; int srcidx = 0; uint32_t bit, ormask = src_mask | dst_mask; while (ormask && n < idxcount) { bit = ormask & -ormask; /* get lowest bit */ ormask ^= bit; /* remove lowest bit */ if (src_mask & dst_mask & bit) { /* matching channel */ idxary[n++] = srcidx++; } else if (src_mask & bit) { /* source channel only */ ++srcidx; } else { /* destination channel only */ idxary[n++] = -1; } } return n + popcount(ormask & dst_mask); } size_t memcpy_by_index_array_initialization_src_index(int8_t *idxary, size_t idxcount, uint32_t dst_mask, uint32_t src_mask) { size_t dst_count = popcount(dst_mask); if (idxcount == 0) { return dst_count; } if (dst_count > idxcount) { dst_count = idxcount; } size_t src_idx, dst_idx; for (src_idx = 0, dst_idx = 0; dst_idx < dst_count; ++dst_idx) { if (src_mask & 1) { idxary[dst_idx] = src_idx++; } else { idxary[dst_idx] = -1; } src_mask >>= 1; } return dst_idx; } size_t memcpy_by_index_array_initialization_dst_index(int8_t *idxary, size_t idxcount, uint32_t dst_mask, uint32_t src_mask) { size_t src_idx, dst_idx; size_t dst_count = __builtin_popcount(dst_mask); size_t src_count = __builtin_popcount(src_mask); if (idxcount == 0) { return dst_count; } if (dst_count > idxcount) { dst_count = idxcount; } for (src_idx = 0, dst_idx = 0; dst_idx < dst_count; ++src_idx) { if (dst_mask & 1) { idxary[dst_idx++] = src_idx < src_count ? (signed)src_idx : -1; } dst_mask >>= 1; } return dst_idx; } void accumulate_i16(int16_t *dst, const int16_t *src, size_t count) { while (count--) { *dst = clamp16((int32_t)*dst + *src++); ++dst; } } void accumulate_u8(uint8_t *dst, const uint8_t *src, size_t count) { int32_t sum; for (; count > 0; --count) { // 8-bit samples are centered around 0x80. sum = *dst + *src++ - 0x80; // Clamp to [0, 0xff]. *dst++ = (sum & 0x100) ? (~sum >> 9) : sum; } } void accumulate_p24(uint8_t *dst, const uint8_t *src, size_t count) { for (; count > 0; --count) { // Unpack. int32_t dst_q8_23 = 0; int32_t src_q8_23 = 0; memcpy_to_q8_23_from_p24(&dst_q8_23, dst, 1); memcpy_to_q8_23_from_p24(&src_q8_23, src, 1); // Accumulate and overwrite. dst_q8_23 += src_q8_23; memcpy_to_p24_from_q8_23(dst, &dst_q8_23, 1); // Move on to next sample. dst += 3; src += 3; } } void accumulate_q8_23(int32_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst = clamp24_from_q8_23(*dst + *src++); ++dst; } } void accumulate_i32(int32_t *dst, const int32_t *src, size_t count) { for (; count > 0; --count) { *dst = clamp32((int64_t)*dst + *src++); ++dst; } } void accumulate_float(float *dst, const float *src, size_t count) { for (; count > 0; --count) { *dst++ += *src++; } }