search for: math_approx

...have troubles with posting to mailing list. Sorry for top posting, I could not reply and forward at once. Btw, your "read-only access" may happen if you're posting from different mailing address then one from which you subscribed. re: conflicting names To avoid conflict functions in math_approx.h should be wrapped by OVERRIDE_FUNC_NAME defines, then it will be easy to conditionally turn default implementations on and off. This is done so for all asm-optimized functions in other places of Speex. ---------- Forwarded message ---------- From: Oleg Homenko <brox42@gmail.com> Date: Dec...

Oleg Homenko a ?crit : > Hi, > I started to optimize math_approx fixed point functions to ARM CPU. And > found out that approximations are not Taylor series, That's right. for example: > spx_exp2(x) = D0 + D1*x + D2*x^2 +D3*x^3, > where D0 = 1, D1 = log(2), D2 = (log(2))^2 / 2, D3 = (log(2))^3 / 6 for > Taylor serie, > but Speex source code...

...s). Some of the things I had to do to be able to split are: - Remove the vorbis-psy tuning because I no longer have an fft in the codec - Remove the Speex-specific jitter buffer (but kept the generic one) because it was the only part of libspeexvoip requiring the codec. - The content of misc.c and math_approx.c was moved to the header files as inline functions (misc.c and math_approx.c no longer exist). Any comments, suggestions? Does that break anything I hadn't thought about? Any reason it shouldn't be done? ... Cheers, Jean-Marc -------------- next part -------------- A non-text attachmen...

...7 3:48 PM: > Hi Oleg, > > Turns out the problem was only triggered with DTX on. I just committed a > fix in svn/git. Let me know if it works for you now. > > Thanks again for reporting the bug. > > Jean-Marc > Thank you, problem is gone now! Oleg PS. I just finished math_approx.asm (speed optimized version for ARM). The code can be compiled using armasm utility that comes with VS2005. Math functions work up to 2...2.5 times faster now. Are you interested in adding it to the project? Also, have you made some profiling on the project? I mean do you know any bottle necks...

...speex_bits_unpack_signed speex_bits_peek_unsigned speex_bits_peek speex_bits_nbytes speex_bits_insert_terminator In cd_search.c: noise_codebook_quant noise_codebook_unquant In filters.c: compute_rms syn_percep_zero16 qmf_decomp qmf_synth In ltp.c: forced_pitch_quant forced_pitch_unquant In math_approx.c: spx_ilog2 _spx_cos_pi_2 spx_cos_norm spx_exp2 spx_atan In nb_celp.c: nb_encoder_destroy nb_decoder_destroy In quant_lsp.c: lsp_quant_nb lsp_unquant_nb lsp_quant_high lsp_unquant_high In speex.c: speex_encoder_destroy speex_decoder_destroy speex_encode speex_decode nb_mode_query wb_mode_que...

The section that establishes the MIME type is stating audio/x-speex. Per the MIME Types and File Extensions it should be audio/ogg. If the change will be made, there's no need to put it as audio/x-ogg as the IETF values MIME types already in use. -Ivo

...s). Some of the things I had to do to be able to split are: - Remove the vorbis-psy tuning because I no longer have an fft in the codec - Remove the Speex-specific jitter buffer (but kept the generic one) because it was the only part of libspeexvoip requiring the codec. - The content of misc.c and math_approx.c was moved to the header files as inline functions (misc.c and math_approx.c no longer exist). Any comments, suggestions? Does that break anything I hadn't thought about? Any reason it shouldn't be done? ... Cheers, Jean-Marc

...with it => In function speex_echo_state_init in mdf.c I am getting error: "undefined reference to exp". I don't see a fixed point implementation of "exp". Can you point me to the correct files. I am using the following files from speex 1.2 beta1 => mdf.c, math_approx.c, misc.c, fft_wrap.c, kiss_fft.c, config.h, speex_types.h, speex_echo.h, arch.h . -Deepa -Marc Valin <jean-marc.valin@usherbrooke.ca> wrote: Deepa Bhatt a ?crit : > Analog Devices sponsored a fixed point version for AEC which was > posted early this year, but it seems f...

...e.lo gain_table_lbr.lo cb_search.lo filters.lo bits.lo modes.lo speex.lo vq.lo high_lsp_tables.lo vbr.lo hexc_table.lo exc_5_256_table.lo exc_5_64_table.lo exc_8_128_table.lo exc_10_32_table.lo exc_10_16_table.lo exc_20_32_table.lo hexc_10_32_table.lo misc.lo speex_header.lo speex_callbacks.lo math_approx.lo stereo.lo preprocess.lo smallft.lo lbr_48k_tables.lo jitter.lo mdf.lo -lm arm-linux-gcc -shared .libs/nb_celp.o .libs/sb_celp.o .libs/lpc.o .libs/ltp.o .libs/lsp.o .libs/quant_lsp.o .libs/lsp_tables_nb.o .libs/gain_table.o .libs/gain_table_lbr.o .libs/cb_search.o .libs/filters.o .libs/bits...

Hi Jean-Marc, >I think it's worth trying the float version on your CPU. However, I >wouldn't be surprised if the fixed-point was still faster than float. ok, I'll try both configurations. >BTW, you can use --enable-arm5-asm instead of --enable-arm4-asm. While >both are currently almost the same, eventually it would be possible to >get better performance using the

hi, this patch should fix the symbian build. /alfred -------------- next part -------------- A non-text attachment was scrubbed... Name: speex-symbian2.patch Type: text/x-patch Size: 2669 bytes Desc: not available Url : http://lists.xiph.org/pipermail/speex-dev/attachments/20061115/5ac4915e/speex-symbian2.bin

...exc_10_16_table.c exc_10_32_table.c exc_20_32_table.c fftwrap.c kiss_fft.c kiss_fftr.c filterbank.c filters.c gain_table.c >> SOURCE gain_table_lbr.c hexc_10_32_table.c hexc_table.c high_lsp_tables.c jitter.c >> SOURCE lbr_48k_tables.c lpc.c lsp.c lsp_tables_nb.c ltp.c math_approx.c mdf.c misc.c >> SOURCE modes.c nb_celp.c preprocess.c quant_lsp.c sb_celp.c smallft.c >> -SOURCE speex.c speex_callbacks.c speex_header.c stereo.c vbr.c vq.c >> +SOURCE speex.c speex_callbacks.c speex_header.c stereo.c vbr.c vq.c window.c >> USERINCL...

...e_codebook_quant > noise_codebook_unquant No need for that (assuming you only want 8 kbps). > In filters.c: > compute_rms > syn_percep_zero16 > qmf_decomp > qmf_synth No need for that. > In ltp.c: > forced_pitch_quant > forced_pitch_unquant No need for that. > In math_approx.c: > spx_ilog2 > _spx_cos_pi_2 > spx_cos_norm > spx_exp2 > spx_atan No need for that. > In nb_celp.c: > nb_encoder_destroy > nb_decoder_destroy No need for that if you never plan on freeing the memory > In quant_lsp.c: > lsp_quant_nb > lsp_unquant_nb > lsp_qu...

...> speex_bits_insert_terminator > > In cd_search.c: > noise_codebook_quant > noise_codebook_unquant > > In filters.c: > compute_rms > syn_percep_zero16 > qmf_decomp > qmf_synth > > In ltp.c: > forced_pitch_quant > forced_pitch_unquant > > In math_approx.c: > spx_ilog2 > _spx_cos_pi_2 > spx_cos_norm > spx_exp2 > spx_atan > > In nb_celp.c: > nb_encoder_destroy > nb_decoder_destroy > > In quant_lsp.c: > lsp_quant_nb > lsp_unquant_nb > lsp_quant_high > lsp_unquant_high > > In speex.c: > speex...

...rom speex_ltp.c ? bits.obj cb_search.obj exc_10_16_table.obj exc_10_32_table.obj exc_20_32_table.obj exc_5_256_table.obj exc_5_64_table.obj exc_8_128_table.obj filters.obj gain_table.obj gain_table_lbr.obj hexc_10_32_table.obj hexc_table.obj high_lsp_tables.obj jitter.obj lsp.obj lsp_tables_nb.obj math_approx.obj mdf.obj misc.obj modes.obj nb_celp.obj preprocess.obj quant_lsp.obj sb_celp.obj smallft.obj speex.obj speex_callbacks.obj speex_header.obj stereo.obj vbr.obj Thanks - -- - ---------------------------------------------------------------------- *Dott. Matteo Orefice* Via San Lorenzo 59 Bussol...

Analog Devices sponsored a fixed point version for AEC which was posted early this year, but it seems from the source code that the fft routines are still in floating point. Is the port still not complete or am I missing something? Has anyone out there ported speex AEC on RISC architecture? Please let me know. Thanks in advance, -Deepa -------------- next part -------------- An HTML

...CE exc_10_16_table.c exc_10_32_table.c exc_20_32_table.c fftwrap.c kiss_fft.c kiss_fftr.c filterbank.c filters.c gain_table.c > SOURCE gain_table_lbr.c hexc_10_32_table.c hexc_table.c high_lsp_tables.c jitter.c > SOURCE lbr_48k_tables.c lpc.c lsp.c lsp_tables_nb.c ltp.c math_approx.c mdf.c misc.c > SOURCE modes.c nb_celp.c preprocess.c quant_lsp.c sb_celp.c smallft.c > -SOURCE speex.c speex_callbacks.c speex_header.c stereo.c vbr.c vq.c > +SOURCE speex.c speex_callbacks.c speex_header.c stereo.c vbr.c vq.c window.c > USERINCLUDE . ..\inclu...

Hi Jean-Marc >> I tested it on Windows Mobile 5.0 PocketPC platform with several >> different applications (duplex voice capturing/encoding/decoding/playback). >> If I use any recent libspeex version >=13186, application exits in 2-3 >> seconds after I stop to talk :(. >> > by exit, you mean segfault (or whatever the Windows equivalent is), a > division

Jean-Marc, > That reminds me that I've just moved lots of stuff around with the > allocation functions. Can you check that the TI stuff still works with > that? In the end, it'll probably make things easier for you now. For > example, if you link statically, wideband should be automatically left > out if unused. Also, all the libc stuff is now in os_support.h, which is >

...c -SOURCE exc_10_16_table.c exc_10_32_table.c exc_20_32_table.c fftwrap.c kiss_fft.c kiss_fftr.c filterbank.c filters.c gain_table.c -SOURCE gain_table_lbr.c hexc_10_32_table.c hexc_table.c high_lsp_tables.c jitter.c -SOURCE lbr_48k_tables.c lpc.c lsp.c lsp_tables_nb.c ltp.c math_approx.c mdf.c misc.c -SOURCE modes.c nb_celp.c preprocess.c quant_lsp.c sb_celp.c smallft.c +SOURCE exc_10_16_table.c exc_10_32_table.c exc_20_32_table.c filters.c gain_table.c +SOURCE gain_table_lbr.c hexc_10_32_table.c hexc_table.c high_lsp_tables.c +SOURCE lpc.c lsp.c l...