summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* Add initial AUTHORS fileHEADmasterTimothy Pearson2015-12-211-0/+2
|
* Add VSX double support for POWER8Timothy Pearson2015-12-212-1/+583
|
* Initial fixes for POWER8 and GCCTimothy Pearson2015-12-215-8/+12
|
* Report AVX2 capabilities in capability stringErik Schnetter2015-11-141-1/+6
|
* Improve optimization barriersErik Schnetter2015-10-303-29/+29
|
* Correct syntax error in AVX2 codeErik Schnetter2015-10-291-6/+6
|
* Update build instructionsErik Schnetter2015-10-291-2/+2
|
* Correct type name __m256iErik Schnetter2015-10-291-14/+14
|
* Use C++03 standard for source code formattingErik Schnetter2015-10-163-38/+104
|
* Reformat source code with clang-formatErik Schnetter2015-10-1644-17970/+15431
| | | | | | | | | Note: If you have an existing checkout with modifications, then you should: (1) save your current state on a branch (2) run clang-format on this branch (3) compare this branch to a fresh checkout of the clang-formatted master Since both your branch and master have been formatted with clang-format, this will lead to a very small diff, avoiding spurious changes due to formatting differences.
* Modern versions of Clang support __builtin_assume_alignedErik Schnetter2015-10-161-6/+0
|
* Begin to implement AVX2 support for 16-bit typesErik Schnetter2015-10-161-0/+29
|
* Add Meson build scriptErik Schnetter2015-10-161-0/+15
|
* Improve fallback static_assert implementationErik Schnetter2015-10-161-1/+5
|
* Correct AVX2 to_bool implementationErik Schnetter2015-10-162-2/+2
| | | | Thanks to Guillaume de Roujoux for finding this and providing a patch.
* Use AVX2 intrinsics if availableErik Schnetter2015-09-035-27/+356
|
* Rename test to selftestErik Schnetter2015-09-032-1/+1
|
* Add missing typecastErik Schnetter2015-02-281-1/+1
|
* Correct SSE2 min/maxErik Schnetter2015-01-211-2/+2
|
* Describe exponent offset of float8 and float16Erik Schnetter2015-01-211-2/+2
|
* Optimize AVX isignbitErik Schnetter2014-10-101-6/+1
|
* Add missing #endifErik Schnetter2014-06-171-0/+1
|
* Remove debug outputErik Schnetter2014-06-071-4/+0
|
* Do not accidentally disable builtin vectors' barrier functionErik Schnetter2014-06-071-1/+1
|
* Provide integer comparison operafors for AVX fp8 and fp16 vectorsErik Schnetter2014-06-073-0/+25
|
* Add constructors from double for fp8 and fp16 typesErik Schnetter2014-06-072-0/+23
|
* Correct calculating inverse of a bit pattern (counteract integer promotion)Erik Schnetter2014-06-071-2/+4
|
* Add constants for infinity and nan to floating point propertiesErik Schnetter2014-06-071-0/+6
|
* Correct return types of some floating point propertiesErik Schnetter2014-06-071-6/+6
|
* Benchmark Clang's builtin vectors as wellErik Schnetter2014-03-241-6/+6
|
* Improve log performanceErik Schnetter2013-11-281-43/+30
|
* Add to-do itemsErik Schnetter2013-11-282-0/+5
|
* IndentErik Schnetter2013-11-281-2/+8
|
* Update log coefficientsErik Schnetter2013-11-282-387/+115
|
* Don’t test builtin vectors; they are too slowErik Schnetter2013-11-281-2/+30
|
* Speed up benchmarksErik Schnetter2013-11-281-1/+1
|
* Update build instructions for out-of-source-tree buildingErik Schnetter2013-11-281-5/+2
|
* Don’t define << or >> for VML_NO_IOSTREAMErik Schnetter2013-11-263-0/+6
|
* vec_builtin is only for ClangErik Schnetter2013-11-261-45/+6
|
* Correct syntax errors in NEON vectorsErik Schnetter2013-11-262-2/+2
|
* Improve auto-detecting of math functions provided by STLErik Schnetter2013-11-261-33/+40
|
* Use __restrict__ instead of restrictErik Schnetter2013-11-261-2/+4
|
* Some ideas to improve exp on QPXErik Schnetter2013-11-251-2/+28
|
* Rename “automatic” vector types from float/double to float32/float64 for ↵Erik Schnetter2013-11-254-28/+28
| | | | clarity
* Run benchmarks of all functions, not just someErik Schnetter2013-11-251-90/+184
|
* Begin to support builtin vectors on non-clang compilersErik Schnetter2013-11-251-7/+48
|
* Only enable builtin vectors when clang is usedErik Schnetter2013-11-252-1/+21
|
* Correct clz implementationErik Schnetter2013-11-251-12/+20
|
* If VML_NO_IOSTREAM is set, then don’t provide any iostream operationsErik Schnetter2013-11-255-5/+20
| | | | This reduces the size of the library.
* Provide lrintErik Schnetter2013-11-259-1/+77
|
OpenPOWER on IntegriCloud