similar to: [PATCH 3] for bitwriter.c

--- src/libFLAC/bitwriter.c | 31 +++++++++++++++---------------- 1 file changed, 15 insertions(+), 16 deletions(-) diff --git a/src/libFLAC/bitwriter.c b/src/libFLAC/bitwriter.c index 651440d..7da4b15 100644 --- a/src/libFLAC/bitwriter.c +++ b/src/libFLAC/bitwriter.c @@ -43,12 +43,11 @@ /* Things should be fastest when this matches the machine word size */ /* WATCHOUT: if you change this

--- src/libFLAC/bitreader.c | 445 +++++++++++------------------------------------ 1 files changed, 105 insertions(+), 340 deletions(-) diff --git a/src/libFLAC/bitreader.c b/src/libFLAC/bitreader.c index ae515a0..7ae086d 100644 --- a/src/libFLAC/bitreader.c +++ b/src/libFLAC/bitreader.c @@ -755,379 +755,144 @@ FLAC__bool FLAC__bitreader_read_rice_signed(FLAC__BitReader *br, int *val, unsig }

On Fri, Mar 14, 2008 at 07:36:31PM +0100, Miroslav Lichvar wrote: > attached are patches that improve decoding speed a bit. The first > patch improves the bit scan macro used for decoding unary values, the > second one adds a GCC inline assembly for bswap and the third patch > replaces the read_rice_block function. The third patch has a bug causing reading past input buffer, attaching

Hi, attached are patches that improve decoding speed a bit. The first patch improves the bit scan macro used for decoding unary values, the second one adds a GCC inline assembly for bswap and the third patch replaces the read_rice_block function. In my testing it turned out to be even faster than the _ia32_bswap function. If the code produced by MSVC is faster as well, I'd suggest to remove

1) The definition of FLAC__BitWriter was updated with the current text from /libFLAC/bitwriter.c (the text in comments was changed). And the definition of TOTAL_BITS was made closer to the code from /libFLAC/bitwriter.c. 2) The values for 'words' and 'bits' values now calculated, not just some magic constants. 3) Added FLAC__U64L() for 64-bit constants. It seems that it

There are preprocessor definitions in bitreader.c and bitwriter.c: /* Things should be fastest when this matches the machine word size */ /* WATCHOUT: if you change this you must also change the following #defines down to SWAP_BE_WORD_TO_HOST below to match */ /* WATCHOUT: there are a few places where the code will not work unless uint32_t is >= 32 bits wide */ #define

Erik de Castro Lopo wrote: > Well FLAC__int32 is just a 32 bit integer and on all the platforms/ > architecures/compilers that FLAC supports FLAC__int32 and int are > the same. > > Personally I think the FLAC__xxxx stuff should go, to be replaced with > C standard int32_t, uint32_t, int16_t etc, at least for internal code. > I am slowly doing that when I touch code. > >

Hi All, I work on a backend for a target similar to Mips, where large immediates are loaded into registers with 2 instructions, 1 to load the MSBits and 1 to load the LSBits. I've noticed a recurring pattern where, despite low register pressure, these constants will be rematerialized in every iteration of a loop, rather than being hoisted. Here's an example using the

The commit <http://git.xiph.org/?p=flac.git;a=commitdiff;h=e3c8095336249182b6a8871747df3cc99640103a> fixed UB in bitwriter.c. But there are 2 unused functions - FLAC__bitwriter_rice_bits() and FLAC__bitwriter_write_rice_signed() that also have this UB. IMHO it makes sense to fix their code as well. Maybe someone will use them in future. -------------- next part -------------- A non-text

Yes machine-licm can and should hoist constant materialization instructions out of the loop. If it's not doing that, it's probably because the target is not modeling the instruction correctly. I would walk through MachineLICM::IsLoopInvariantInst() in the debugger to figure it out. You can also try compiling the same bitcode for a target like ARM or X86 as a comparison. Evan On Mar 7,

Eric, I finally got around to your patches after Miroslav's. the first one (the memcpy/memset replacement) I had problems with, one because the buffers can overlap so I had to use memmove (is this usually assembly in libc too?) and also the endpoints looked wrong, for my full patch see below. speedup for me was around 3% the second patch got another 2%. a question though, why do you have: +

Thanks for the tip! I looked into it and it looks like the problem as of SVN HEAD is that the lui and ori instructions in Mips are considered cheap (1-cycle def-use latency) by MachineLICM::IsCheapInstruction(), but are not trivially materializable because their register operands are not always available. This makes MachineLICM::IsProfitableToHoist() return false, preventing the hoist even

thanks for the patch. also, if you have miroslav's patch again a more updated version of bitbuffer.c that would be great. I have been meaning to get around to applying it for a long time. btw how are you playing it on the ipod? not sure how to help out with lpc_restore_signal(). there are x86 and PPC versions of the routines in CVS that might be good references. it is basically a

--- libcelt/bands.c | 43 ++++++++++++++++++++++++------------------- libcelt/bands.h | 6 ++---- libcelt/celt.c | 6 ++---- 3 files changed, 28 insertions(+), 27 deletions(-) diff --git a/libcelt/bands.c b/libcelt/bands.c index d0c3498..315cce8 100644 --- a/libcelt/bands.c +++ b/libcelt/bands.c @@ -247,23 +247,6 @@ void compute_pitch_gain(const CELTMode *m, const celt_norm_t *X, const

Pulled from my Arch archive, this following patch seems to have made quite a difference in getting my ARM7TDMI chip to play FLAC (compression levels 0-2) on my ipod. I don't have benchmarks with hard numbers, but playing with skips vs playing without skips is a fairly noticeable difference. memcpy and memset on uClibc are optimized in asm for the ARM7TDMI in uClibc. Other hardware/libc

--- src/libFLAC/include/private/bitmath.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/src/libFLAC/include/private/bitmath.h b/src/libFLAC/include/private/bitmath.h index 61b0e03..d32b1a7 100644 --- a/src/libFLAC/include/private/bitmath.h +++ b/src/libFLAC/include/private/bitmath.h @@ -42,7 +42,7 @@ #endif /* Will never be emitted for MSVC, GCC, Intel compilers */

Hello everyone At EuroLLVM I presented some testing work we have been doing on improving correctness of the MC Layer for ARM. There seemed to be interest from the community in seeing the results of this test suite. Background ----------- We are using a test suite, called MC Hammer, that compares MC with an ARM in-house implementation of the same functionality. The test space for this suite is

This lpc_restore_order was partially inspired by Miroslav's affd, though my (not very great) ARM asm version resembled this, as well. The other two reduce CPU array indexing overhead in loops a little. Additionally, a request for help: My not very optimized lpc_restore_signal is at the below URL, I couldn't get the ldm* instructions to work as advertised, even though I've talked

This patch adds support for other compilers and systems including MSVC, Intel C compiler etc.. --- src/libFLAC/bitmath.c | 48 ------------- src/libFLAC/bitreader.c | 54 ++------------- src/libFLAC/include/private/bitmath.h | 120 ++++++++++++++++++++++++++++++--- 3 files changed, 116 insertions(+), 106 deletions(-) diff --git a/src/libFLAC/bitmath.c

metaflac --import-vc-from ... used on flac file without vorbis comment destroys the file. This patch fix it. -- Miroslav Lichvar -------------- next part -------------- --- src/metaflac/main.c.orig 2002-09-28 14:45:49.000000000 +0200 +++ src/metaflac/main.c 2002-10-06 23:36:26.000000000 +0200 @@ -1754,20 +1754,24 @@ found_vc_block = true; } while(!found_vc_block &&