diff options
author | dcommander <dcommander@632fc199-4ca6-4c93-a231-07263d6284db> | 2015-01-16 07:17:42 +0000 |
---|---|---|
committer | dcommander <dcommander@632fc199-4ca6-4c93-a231-07263d6284db> | 2015-01-16 07:17:42 +0000 |
commit | 041f19a8a6beb2cdb0037c697649f21d7a8f04be (patch) | |
tree | 2c6dc6a572ff171cae7053743b87e237a4780711 | |
parent | 47d3f427dbaf471ef2873862026f2562d18ec169 (diff) |
Some clarifications (actually MIPS doesn't implement the float DCT/IDCT using SIMD instructions)
git-svn-id: svn://svn.code.sf.net/p/libjpeg-turbo/code/trunk@1518 632fc199-4ca6-4c93-a231-07263d6284db
-rw-r--r-- | Makefile.am | 17 | ||||
-rwxr-xr-x | README-turbo.txt | 4 |
2 files changed, 10 insertions, 11 deletions
diff --git a/Makefile.am b/Makefile.am index 309e6d6..956f30d 100644 --- a/Makefile.am +++ b/Makefile.am @@ -440,17 +440,18 @@ endif # The output of the floating point tests is not validated by default, because # the output differs depending on the type of floating point math used, and -# this is only deterministic if there is full SIMD coverage of all of the -# floating point algorithms in libjpeg-turbo. Pass one of the following on the -# make command line to validate the floating point tests against one of the -# expected results: +# this is only deterministic if the DCT/IDCT are implemented using SIMD +# instructions on a particular platform. Pass one of the following on the make +# command line to validate the floating point tests against one of the expected # results: # # FLOATTEST=sse validate against the expected results from the libjpeg-turbo # SSE SIMD extensions -# FLOATTEST=32bit validate against the expected results from the MIPS DSPr2 -# SIMD extensions or 32-bit FPUs or GCC when -mfpmath=sse is -# used (which is the default on x86-64 systems) -# FLOATTEST=64bit validate against the exepected results from 64-bit FPUs +# FLOATTEST=32bit validate against the expected results from the C code +# when running on a 32-bit FPU (or when SSE is being used for +# floating point math, which is generally the default with +# x86-64 compilers) +# FLOATTEST=64bit validate against the exepected results from the C code +# when running on a 64-bit FPU # CC: RGB->YCC SAMP: fullsize/int FDCT: float ENT: prog huff ./cjpeg -sample 3x2 -dct float -prog -outfile testout_3x2_float_prog.jpg $(srcdir)/testimages/testorig.ppm diff --git a/README-turbo.txt b/README-turbo.txt index 1fcaedc..32b1d10 100755 --- a/README-turbo.txt +++ b/README-turbo.txt @@ -313,9 +313,7 @@ following reasons: range of the quality scale is typically more like a 1.0 dB difference.) -- If the floating point algorithms in libjpeg-turbo are not implemented using SIMD instructions on a particular platform, then the accuracy of the - floating point DCT/IDCT can depend on the compiler settings. For instance, - different results will be obtained when using -mfpmath=387 or -mfpmath=sse - with GCC on x86 systems. + floating point DCT/IDCT can depend on the compiler settings. While libjpeg-turbo does emulate the libjpeg v8 API/ABI, under the hood, it is still using the same algorithms as libjpeg v6b, so there are several specific |