search for: flac__lpc_restore_signal_wid

As I wrote earlier, GCC generates slow ia32 code for FLAC__lpc_compute_residual_from_qlp_coefficients_wide() and FLAC__lpc_restore_signal_wide(). So 24-bit encoding/decoding is slower for GCC compile than for MSVS or ICC compile. I took FLAC__lpc_compute_residual_from_qlp_coefficients_asm_ia32 and FLAC__lpc_restore_signal_asm_ia32 asm functions and wrote their _wide versions. -------------- next part -------------- A non-text attachment...

Christian Weisgerber <naddy@mips.inka.de> wrote: > #0 0x0000000040d18810 in FLAC__lpc_compute_residual_from_qlp_coefficients_wide > (data=0x49e4c014, data_len=110, qlp_coeff=0x7f7ffffece70, order=1, > lp_quantization=14, residual=0x4fced000) at lpc.c:745 > 745 residual[i] = > data[i] - (FLAC__int32)((qlp_coeff[0] *

Hi all, I'm trying to compile the flac-1.1.1 tarball on a Linux PPC system (a G3 iBook running Debian Testing). Configure is fine, but make bombs out almost immediately with: lpc_asm.s: Assembler messages: lpc_asm.s:1: Error: junk at end of line, first unrecognized character is `l' lpc_asm.s:2: Error: junk at end of line, first unrecognized character is `C' lpc_asm.s:4:

...signed lag, FLAC__real autoc[]); #endif #endif +#endif /* * FLAC__lpc_compute_lp_coefficients() @@ -146,17 +147,16 @@ void FLAC__lpc_restore_signal(const FLAC__int32 residual[], unsigned data_len, const FLAC__int32 qlp_coeff[], unsigned order, int lp_quantization, FLAC__int32 data[]); void FLAC__lpc_restore_signal_wide(const FLAC__int32 residual[], unsigned data_len, const FLAC__int32 qlp_coeff[], unsigned order, int lp_quantization, FLAC__int32 data[]); #ifndef FLAC__NO_ASM -#ifdef FLAC__CPU_IA32 -#ifdef FLAC__HAS_NASM +# ifdef FLAC__CPU_IA32 +# ifdef FLAC__HAS_NASM void FLAC__lpc_restore_signal_asm_ia32(...

...stream_decoder.c:296: error: `FLAC__lpc_restore_signal' undeclared > (first use in this function) > stream_decoder.c:296: error: (Each undeclared identifier is reported > only once > stream_decoder.c:296: error: for each function it appears in.) > stream_decoder.c:297: error: `FLAC__lpc_restore_signal_wide' > undeclared > (first use in this function) > make[4]: *** [stream_decoder.lo] Error 1 > make[4]: Leaving directory > `/var/tmp/portage/flac-1.1.1/work/flac-1.1.1/src/libFLAC' > > Could you please at least make sure that disabling altivec > optimization > wo...

...es. It turns out that 32-bit encoder made by GCC is ~1.7x times slower than 32-bit encoder made by MSVS. It seems that GCC creates inefficient code for 32bit * 32bit -> 64bit multiplication for 32-bit architecture. This problem affects FLAC__lpc_compute_residual_from_qlp_coefficients_wide() and FLAC__lpc_restore_signal_wide() functions. Is there any way to fix this (apart from rewriting these 2 functions in 32-bit asm) ?

Hello! As I understand, there is versions of FLAC for Linux x64. Such question, whether is native x64-86 versions for OS Windows? Best regards, Valentin.

...+ +#if 0 +void FLAC__lpc_restore_signal_orig(const FLAC__int32 residual[], unsigned data_len, const FLAC__int32 qlp_coeff[], unsigned order, int lp_quantization, FLAC__int32 data[]) +{ #ifdef FLAC__OVERFLOW_DETECT FLAC__int64 sumo; #endif @@ -339,6 +542,7 @@ } */ } +#endif /* 0 */ void FLAC__lpc_restore_signal_wide(const FLAC__int32 residual[], unsigned data_len, const FLAC__int32 qlp_coeff[], unsigned order, int lp_quantization, FLAC__int32 data[]) { --- orig/src/libFLAC/bitbuffer.c +++ mod/src/libFLAC/bitbuffer.c @@ -1466,6 +1469,7 @@ { unsigned i, bits_ = bits; FLAC__uint32 v = 0; + FLAC__blurb *bb...

...stream_decoder_init': stream_decoder.c:296: error: `FLAC__lpc_restore_signal' undeclared (first use in this function) stream_decoder.c:296: error: (Each undeclared identifier is reported only once stream_decoder.c:296: error: for each function it appears in.) stream_decoder.c:297: error: `FLAC__lpc_restore_signal_wide' undeclared (first use in this function) make[4]: *** [stream_decoder.lo] Error 1 make[4]: Leaving directory `/var/tmp/portage/flac-1.1.1/work/flac-1.1.1/src/libFLAC' Could you please at least make sure that disabling altivec optimization won't break the build? That's a bit hig...

I think I've gotten FLAC__lpc_restore_signal() about as good as I'm going to get it. Here's what I have: -a new file, lpc_asm.s, which has the assembly routines -changes to cpu.h, cpu.c, and stream_decoder.c to enable them -changes to configure.in to support the new cpu stuff -a preliminary Makefile.am -maybe something else I'm forgetting Now automake complains that configure.in

...tents[FLAC__MAX_CHANNELS]; unsigned output_capacity, output_channels; FLAC__uint32 last_frame_number; *************** *** 281,286 **** --- 282,288 ---- decoder->private_->local_lpc_restore_signal = FLAC__lpc_restore_signal; decoder->private_->local_lpc_restore_signal_64bit = FLAC__lpc_restore_signal_wide; decoder->private_->local_lpc_restore_signal_16bit = FLAC__lpc_restore_signal; + decoder->private_->local_lpc_restore_signal_16bit_order8 = FLAC__lpc_restore_signal; /* now override with asm where appropriate */ #ifndef FLAC__NO_ASM if(decoder->private_->cpuinfo.use_a...