| /* |
| * Copyright (c) 2011 The WebRTC project authors. All Rights Reserved. |
| * |
| * Use of this source code is governed by a BSD-style license |
| * that can be found in the LICENSE file in the root of the source |
| * tree. An additional intellectual property rights grant can be found |
| * in the file PATENTS. All contributing project authors may |
| * be found in the AUTHORS file in the root of the source tree. |
| */ |
| |
| #include "settings.h" |
| #include "arith_routines.h" |
| |
| |
| /* |
| * code symbols into arithmetic bytestream |
| */ |
| void WebRtcIsac_EncHistMulti(Bitstr *streamdata, /* in-/output struct containing bitstream */ |
| const int *data, /* input: data vector */ |
| const WebRtc_UWord16 **cdf, /* input: array of cdf arrays */ |
| const int N) /* input: data vector length */ |
| { |
| WebRtc_UWord32 W_lower, W_upper; |
| WebRtc_UWord32 W_upper_LSB, W_upper_MSB; |
| WebRtc_UWord8 *stream_ptr; |
| WebRtc_UWord8 *stream_ptr_carry; |
| WebRtc_UWord32 cdf_lo, cdf_hi; |
| int k; |
| |
| |
| /* point to beginning of stream buffer */ |
| stream_ptr = streamdata->stream + streamdata->stream_index; |
| W_upper = streamdata->W_upper; |
| |
| for (k=N; k>0; k--) |
| { |
| /* fetch cdf_lower and cdf_upper from cdf tables */ |
| cdf_lo = (WebRtc_UWord32) *(*cdf + *data); |
| cdf_hi = (WebRtc_UWord32) *(*cdf++ + *data++ + 1); |
| |
| /* update interval */ |
| W_upper_LSB = W_upper & 0x0000FFFF; |
| W_upper_MSB = W_upper >> 16; |
| W_lower = W_upper_MSB * cdf_lo; |
| W_lower += (W_upper_LSB * cdf_lo) >> 16; |
| W_upper = W_upper_MSB * cdf_hi; |
| W_upper += (W_upper_LSB * cdf_hi) >> 16; |
| |
| /* shift interval such that it begins at zero */ |
| W_upper -= ++W_lower; |
| |
| /* add integer to bitstream */ |
| streamdata->streamval += W_lower; |
| |
| /* handle carry */ |
| if (streamdata->streamval < W_lower) |
| { |
| /* propagate carry */ |
| stream_ptr_carry = stream_ptr; |
| while (!(++(*--stream_ptr_carry))); |
| } |
| |
| /* renormalize interval, store most significant byte of streamval and update streamval */ |
| while ( !(W_upper & 0xFF000000) ) /* W_upper < 2^24 */ |
| { |
| W_upper <<= 8; |
| *stream_ptr++ = (WebRtc_UWord8) (streamdata->streamval >> 24); |
| streamdata->streamval <<= 8; |
| } |
| } |
| |
| /* calculate new stream_index */ |
| streamdata->stream_index = (int)(stream_ptr - streamdata->stream); |
| streamdata->W_upper = W_upper; |
| |
| return; |
| } |
| |
| |
| |
| /* |
| * function to decode more symbols from the arithmetic bytestream, using method of bisection |
| * cdf tables should be of size 2^k-1 (which corresponds to an alphabet size of 2^k-2) |
| */ |
| int WebRtcIsac_DecHistBisectMulti(int *data, /* output: data vector */ |
| Bitstr *streamdata, /* in-/output struct containing bitstream */ |
| const WebRtc_UWord16 **cdf, /* input: array of cdf arrays */ |
| const WebRtc_UWord16 *cdf_size, /* input: array of cdf table sizes+1 (power of two: 2^k) */ |
| const int N) /* input: data vector length */ |
| { |
| WebRtc_UWord32 W_lower, W_upper; |
| WebRtc_UWord32 W_tmp; |
| WebRtc_UWord32 W_upper_LSB, W_upper_MSB; |
| WebRtc_UWord32 streamval; |
| const WebRtc_UWord8 *stream_ptr; |
| const WebRtc_UWord16 *cdf_ptr; |
| int size_tmp; |
| int k; |
| |
| W_lower = 0; //to remove warning -DH |
| stream_ptr = streamdata->stream + streamdata->stream_index; |
| W_upper = streamdata->W_upper; |
| if (W_upper == 0) |
| /* Should not be possible in normal operation */ |
| return -2; |
| |
| if (streamdata->stream_index == 0) /* first time decoder is called for this stream */ |
| { |
| /* read first word from bytestream */ |
| streamval = *stream_ptr << 24; |
| streamval |= *++stream_ptr << 16; |
| streamval |= *++stream_ptr << 8; |
| streamval |= *++stream_ptr; |
| } else { |
| streamval = streamdata->streamval; |
| } |
| |
| for (k=N; k>0; k--) |
| { |
| /* find the integer *data for which streamval lies in [W_lower+1, W_upper] */ |
| W_upper_LSB = W_upper & 0x0000FFFF; |
| W_upper_MSB = W_upper >> 16; |
| |
| /* start halfway the cdf range */ |
| size_tmp = *cdf_size++ >> 1; |
| cdf_ptr = *cdf + (size_tmp - 1); |
| |
| /* method of bisection */ |
| for ( ;; ) |
| { |
| W_tmp = W_upper_MSB * *cdf_ptr; |
| W_tmp += (W_upper_LSB * *cdf_ptr) >> 16; |
| size_tmp >>= 1; |
| if (size_tmp == 0) break; |
| if (streamval > W_tmp) |
| { |
| W_lower = W_tmp; |
| cdf_ptr += size_tmp; |
| } else { |
| W_upper = W_tmp; |
| cdf_ptr -= size_tmp; |
| } |
| } |
| if (streamval > W_tmp) |
| { |
| W_lower = W_tmp; |
| *data++ = (int)(cdf_ptr - *cdf++); |
| } else { |
| W_upper = W_tmp; |
| *data++ = (int)(cdf_ptr - *cdf++ - 1); |
| } |
| |
| /* shift interval to start at zero */ |
| W_upper -= ++W_lower; |
| |
| /* add integer to bitstream */ |
| streamval -= W_lower; |
| |
| /* renormalize interval and update streamval */ |
| while ( !(W_upper & 0xFF000000) ) /* W_upper < 2^24 */ |
| { |
| /* read next byte from stream */ |
| streamval = (streamval << 8) | *++stream_ptr; |
| W_upper <<= 8; |
| } |
| |
| if (W_upper == 0) |
| /* Should not be possible in normal operation */ |
| return -2; |
| |
| |
| } |
| |
| streamdata->stream_index = (int)(stream_ptr - streamdata->stream); |
| streamdata->W_upper = W_upper; |
| streamdata->streamval = streamval; |
| |
| |
| /* find number of bytes in original stream (determined by current interval width) */ |
| if ( W_upper > 0x01FFFFFF ) |
| return streamdata->stream_index - 2; |
| else |
| return streamdata->stream_index - 1; |
| } |
| |
| |
| |
| /* |
| * function to decode more symbols from the arithmetic bytestream, taking single step up or |
| * down at a time |
| * cdf tables can be of arbitrary size, but large tables may take a lot of iterations |
| */ |
| int WebRtcIsac_DecHistOneStepMulti(int *data, /* output: data vector */ |
| Bitstr *streamdata, /* in-/output struct containing bitstream */ |
| const WebRtc_UWord16 **cdf, /* input: array of cdf arrays */ |
| const WebRtc_UWord16 *init_index, /* input: vector of initial cdf table search entries */ |
| const int N) /* input: data vector length */ |
| { |
| WebRtc_UWord32 W_lower, W_upper; |
| WebRtc_UWord32 W_tmp; |
| WebRtc_UWord32 W_upper_LSB, W_upper_MSB; |
| WebRtc_UWord32 streamval; |
| const WebRtc_UWord8 *stream_ptr; |
| const WebRtc_UWord16 *cdf_ptr; |
| int k; |
| |
| |
| stream_ptr = streamdata->stream + streamdata->stream_index; |
| W_upper = streamdata->W_upper; |
| if (W_upper == 0) |
| /* Should not be possible in normal operation */ |
| return -2; |
| |
| if (streamdata->stream_index == 0) /* first time decoder is called for this stream */ |
| { |
| /* read first word from bytestream */ |
| streamval = *stream_ptr << 24; |
| streamval |= *++stream_ptr << 16; |
| streamval |= *++stream_ptr << 8; |
| streamval |= *++stream_ptr; |
| } else { |
| streamval = streamdata->streamval; |
| } |
| |
| |
| for (k=N; k>0; k--) |
| { |
| /* find the integer *data for which streamval lies in [W_lower+1, W_upper] */ |
| W_upper_LSB = W_upper & 0x0000FFFF; |
| W_upper_MSB = W_upper >> 16; |
| |
| /* start at the specified table entry */ |
| cdf_ptr = *cdf + (*init_index++); |
| W_tmp = W_upper_MSB * *cdf_ptr; |
| W_tmp += (W_upper_LSB * *cdf_ptr) >> 16; |
| if (streamval > W_tmp) |
| { |
| for ( ;; ) |
| { |
| W_lower = W_tmp; |
| if (cdf_ptr[0]==65535) |
| /* range check */ |
| return -3; |
| W_tmp = W_upper_MSB * *++cdf_ptr; |
| W_tmp += (W_upper_LSB * *cdf_ptr) >> 16; |
| if (streamval <= W_tmp) break; |
| } |
| W_upper = W_tmp; |
| *data++ = (int)(cdf_ptr - *cdf++ - 1); |
| } else { |
| for ( ;; ) |
| { |
| W_upper = W_tmp; |
| --cdf_ptr; |
| if (cdf_ptr<*cdf) { |
| /* range check */ |
| return -3; |
| } |
| W_tmp = W_upper_MSB * *cdf_ptr; |
| W_tmp += (W_upper_LSB * *cdf_ptr) >> 16; |
| if (streamval > W_tmp) break; |
| } |
| W_lower = W_tmp; |
| *data++ = (int)(cdf_ptr - *cdf++); |
| } |
| |
| /* shift interval to start at zero */ |
| W_upper -= ++W_lower; |
| /* add integer to bitstream */ |
| streamval -= W_lower; |
| |
| /* renormalize interval and update streamval */ |
| while ( !(W_upper & 0xFF000000) ) /* W_upper < 2^24 */ |
| { |
| /* read next byte from stream */ |
| streamval = (streamval << 8) | *++stream_ptr; |
| W_upper <<= 8; |
| } |
| } |
| |
| streamdata->stream_index = (int)(stream_ptr - streamdata->stream); |
| streamdata->W_upper = W_upper; |
| streamdata->streamval = streamval; |
| |
| |
| /* find number of bytes in original stream (determined by current interval width) */ |
| if ( W_upper > 0x01FFFFFF ) |
| return streamdata->stream_index - 2; |
| else |
| return streamdata->stream_index - 1; |
| } |