summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
...
* Add more tests for type conversions and roundingErik Schnetter2013-03-212-4/+53
|
* Correct rounding and conversion functionsErik Schnetter2013-03-215-64/+69
|
* Remove typeinfo work-around that is not necessary any moreErik Schnetter2013-03-211-13/+3
|
* Disable GNU extensionsErik Schnetter2013-03-212-8/+8
|
* Simplify remainder() and fmod()Erik Schnetter2013-03-211-4/+8
|
* Don't use typeinfoErik Schnetter2013-03-214-18/+22
|
* Merged in visit0r/vecmathlib (pull request #3) Erik Schnetter2013-03-215-7/+23
|\ | | | | Fixes to make it build with Clang++ 3.2 using libstdc++ from Debian 6.0.
| * Didn't mean to commit that :)Pekka Jääskeläinen2013-03-211-4/+0
| |
| * Fixes to make it build with Clang++ 3.2 using libstdc++ from Debian 6.0.Pekka Jääskeläinen2013-03-215-7/+27
|/
* Update list of pocl functions (remove temporarily disabled "clamp")Erik Schnetter2013-03-201-1/+1
|
* Output configuration settingsErik Schnetter2013-03-201-0/+21
|
* Add _cl_ prefix to some functionsErik Schnetter2013-03-191-20/+54
|
* Disable clamp, since vecmathlib only provides the floating-point versionErik Schnetter2013-03-191-2/+3
|
* Remove __vml_ prefix for pocl wrappersErik Schnetter2013-03-151-17/+18
|
* Remove explicit "inline" declarationsErik Schnetter2013-03-158-8/+8
|
* Improve fdimErik Schnetter2013-03-151-1/+1
|
* Introduce "builtin" vectors (still disabled)Erik Schnetter2013-03-155-6/+1178
|
* Provide vector types with 8 and 16 bit precisionErik Schnetter2013-02-196-31/+1373
| | | | Use non-functional dummy float types fp8 and fp16.
* Beautify codeErik Schnetter2013-02-192-2/+2
|
* Improve lsr() implementationErik Schnetter2013-02-191-12/+2
|
* Correct definitions of lsr, >>, and << for vector shift argumentsErik Schnetter2013-02-194-99/+75
|
* Define all() and any() inside classErik Schnetter2013-02-194-96/+72
|
* Increase output precision when tests failErik Schnetter2013-02-191-5/+7
|
* Output absolute error when a test failsErik Schnetter2013-02-191-0/+2
|
* Rename makefile variable for clarityErik Schnetter2013-02-191-3/+3
|
* Use double2 instead of float1 as exampleErik Schnetter2013-02-191-2/+2
|
* Instantiate operator[] and set_elt()Erik Schnetter2013-02-191-0/+31
|
* Turn instantiations into a libraryErik Schnetter2013-02-192-6/+2
|
* Delegate implementing operator[] and set_elt to the compilerErik Schnetter2013-02-194-149/+17
|
* Add rint(), correct round()Erik Schnetter2013-02-1914-27/+93
|
* Implement pocl functions that take pointer argumentsErik Schnetter2013-02-183-158/+199
|
* Generate pocl-compatible OpenCL math function bindingsErik Schnetter2013-02-1713-358/+922
|
* Create libvecmathlib.a containing implementations of most math functionsErik Schnetter2013-02-162-0/+359
|
* Add cbrt, hypot, trunc; rename scalbn to ldexpErik Schnetter2013-02-1617-95/+237
|
* Replace assert(0) by __builtin_unreachable()Erik Schnetter2013-02-154-11/+11
|
* Change order of testsErik Schnetter2013-02-151-4/+5
|
* Replace all assert() by VML_ASSERT()Erik Schnetter2013-02-1514-22/+8
|
* Examine log() machine codeErik Schnetter2013-02-151-0/+3
|
* Optimize log() moreErik Schnetter2013-02-151-6/+9
|
* Optimize log()Erik Schnetter2013-02-153-183/+179
|
* Update READMEErik Schnetter2013-02-151-1/+2
|
* Look at exp() codeErik Schnetter2013-02-151-0/+3
|
* Speed up exp()Erik Schnetter2013-02-151-3/+21
|
* Use optimised coefficients for exp()Erik Schnetter2013-02-141-18/+31
|
* Update sin() coefficientsErik Schnetter2013-02-141-18/+16
|
* Update scripts and coefficientsErik Schnetter2013-02-142-1088/+1091
|
* Add Mathematica script to generate optimised polynomial coefficientsErik Schnetter2013-02-142-0/+1535
| | | | Also add script output with these coefficients.
* Add "instantiations.cc" to look at generated machine codeErik Schnetter2013-02-143-0/+37
|
* Clean up vector classesErik Schnetter2013-02-146-132/+1561
| | | | | | realvec<T,N>: machine-specific, optimised implementations realpseudovec<T,N>: scalarised implementation based on libm (for benchmark comparisons) realtestvec<T,N>: scalarised implementation without libm, using only vecmathlib's algorithms (for correctness testing)
* Correct sum() implementationErik Schnetter2013-02-141-1/+1
|
OpenPOWER on IntegriCloud