summaryrefslogtreecommitdiffstats
path: root/src
Commit message (Collapse)AuthorAgeFilesLines
...
* Patterns doesn't depend on planJukka Ojanen2014-10-292-184/+174
|
* MSVC has problems with "complex" typeJukka Ojanen2014-10-291-15/+14
|
* Cleaning and reorganizingJukka Ojanen2014-10-298-779/+1015
|
* Add macro definitions for various function/variable attributes; aligned, ↵Jukka Ojanen2014-10-291-0/+77
| | | | | | inlining.. Cleaning of test application
* YASM requires ".code 64" in assemblyJukka Ojanen2014-10-291-378/+376
|
* do not mprotect if dynamic code generation is disabledJérôme Carretero2014-10-141-0/+2
|
* automatic trailing space removalJérôme Carretero2014-10-063-63/+63
| | | | ... because I use a real editor
* real: fix alignment issue in 1d execution (bug #30)Jérôme Carretero2014-10-061-80/+80
| | | | | Because of the size of M/2+1, you can't expect the data to be aligned at 128 bits.
* real_nd: improve transpose size by unrolling of block of h=8, w=1Jérôme Carretero2014-10-061-7/+40
| | | | That works well on ARM and should also work better elsewhere.
* Merge pull request #20 from dennisss/android-neon-fixesAnthony Blake2014-01-054-93/+182
|\ | | | | Android neon fixes
| * Check to make sure that the buffers were aligned properlyDennis Shtatnov2013-12-301-0/+12
| |
| * Add simple logging and test whether N is a power of 2Dennis Shtatnov2013-12-242-0/+13
| |
| * Fix compiling static code with NEONDennis Shtatnov2013-12-242-93/+157
| |
* | Adding in Vim modelines to all .c and .h files.Robert Massaioli2013-12-0527-0/+27
|/
* Fixed #9 properly.=2013-11-201-1/+11
|
* Fixed 9.=2013-11-201-3/+12
|
* Added header files to SOURCES, and build_*.sh, so make dist works properlyAnthony Blake2013-09-252-3/+12
|
* More fomatting for the codegen fileMatt Hunter2013-07-311-32/+31
|
* Add formatting to patterns.cMatt Hunter2013-07-311-9/+10
|
* Add formatting for ffts.cMatt Hunter2013-07-311-113/+113
|
* Add formatting to codegen.cMatt Hunter2013-07-311-32/+36
|
* Commented the plan_t structure with Anthony BlakeMatt Hunter2013-07-251-5/+78
|
* Fixed transpose -- thanks to Sowon Kim <mogyak@gmail.com>Anthony Blake2013-07-161-1/+1
|
* Fixed #5=2013-06-021-5/+5
|
* Added a check for NULL pointer.=2013-06-011-1/+1
|
* Fixed a memory leak for real multi-dimensional transforms.=2013-06-011-1/+24
|
* JNI java/android support, by Michael Zucchi <notzed@gmail.com>Anthony Blake2013-05-061-1/+17
|
* Update codegen_arm.hmatszpk2013-04-241-3/+3
| | | Fix codegen for size>20 (ARM version).
* All 1D complex VFP transforms work (there is still some NEON code in the ND ↵Anthony Blake2013-04-243-287/+306
| | | | and real transforms)
* Included new files I forgot to commit earlier -- thanks Michael CreeAnthony Blake2013-04-225-0/+608
|
* Renamed sse_float.h -> macros-sse.hAnthony Blake2013-04-223-2/+2
|
* Fixed up the smaller VFP transforms. Inverse VFP and real/nd VFP still not ↵Anthony Blake2013-04-229-308/+168
| | | | working yet.
* Disabled some NEON code still in the VFP buildsAnthony Blake2013-04-104-29/+45
|
* VFP supportAnthony Blake2013-04-035-39/+48
|
* Added VFP support for ARMAnthony Blake2013-04-037-37/+666
|
* Fixed cacheflush bug on ARM/Linux -- thanks to Eric Humbert from Invoxia for ↵Anthony Blake2013-03-201-0/+4
| | | | reporting this bug
* Changes to ND codeAnthony Blake2013-01-251-36/+69
|
* codegen.c: Fixed improperly placed postfix increment which was causing problemsAnthony Blake2013-01-111-10/+10
| | | | | when compiling for INTEGRITY. Many thanks to Serge Plagnol for characterizing and reporting this fault.
* Included config.h in ffts.hAnthony Blake2013-01-081-0/+2
|
* Fixed incorrect memory allocationAnthony Blake2012-12-151-1/+1
|
* Fixed memory leak on p->ws_isAnthony Blake2012-12-141-0/+2
|
* Temp hack to fix aligned stuff in ffts.hAnthony Blake2012-12-141-1/+6
|
* miscAnthony Blake2012-12-131-1/+37
|
* Slight speed improvement for ARM NDAnthony Blake2012-12-121-2/+2
|
* Fixed ND segfaultAnthony Blake2012-12-121-4/+4
|
* Added static modeAnthony Blake2012-12-0310-17/+2128
|
* 2D real-valued inverse works for N1 != N2Anthony Blake2012-11-161-11/+10
|
* Removed unneeded real transposeAnthony Blake2012-11-161-10/+0
|
* Real-valued inverse 2D transforms workingAnthony Blake2012-11-161-12/+58
|
* Fixed inverse real-valued transformAnthony Blake2012-11-151-7/+22
|
OpenPOWER on IntegriCloud