Commit message (Collapse) | Author | Age | Files | Lines | ||
---|---|---|---|---|---|---|
... | ||||||
* | Support Altivec with Clang | Erik Schnetter | 2013-09-03 | 1 | -3/+34 | |
| | ||||||
* | Altivec, VSX: Make build with IBM compiler | Erik Schnetter | 2013-09-02 | 2 | -121/+90 | |
| | ||||||
* | Remove invalid comment | Erik Schnetter | 2013-09-02 | 1 | -4/+0 | |
| | ||||||
* | Protect __builtin_clz and __builtin_popcount with #ifdefs | Erik Schnetter | 2013-09-02 | 1 | -0/+8 | |
| | ||||||
* | Allow (again) larger errors for rounding functions | Erik Schnetter | 2013-09-02 | 1 | -25/+26 | |
| | ||||||
* | Provide clean set of libm elemental functions in namespace vml_std | Erik Schnetter | 2013-09-02 | 15 | -243/+308 | |
| | | | | Different C++ standard versions and different compilers provide different sets of elemental functions in std. Thus, we create a clean set in a new namespace vml_std. | |||||
* | Correct auto-detecting VSX | Erik Schnetter | 2013-09-02 | 1 | -1/+1 | |
| | ||||||
* | Implement new integer functions for NEON | Erik Schnetter | 2013-09-02 | 2 | -13/+65 | |
| | ||||||
* | Correct new integer functions for Altivec/VSX | Erik Schnetter | 2013-09-02 | 2 | -13/+69 | |
| | ||||||
* | SSE: Rename function argument | Erik Schnetter | 2013-09-02 | 2 | -6/+7 | |
| | ||||||
* | QPX: Support additional integer | Erik Schnetter | 2013-09-02 | 1 | -8/+61 | |
| | ||||||
* | Correct SSE frexp() | Erik Schnetter | 2013-09-02 | 1 | -4/+5 | |
| | ||||||
* | Reduce accuracy requirement for fma | Erik Schnetter | 2013-09-02 | 1 | -1/+1 | |
| | ||||||
* | Add build instructions for Compute (at Perimeter) | Erik Schnetter | 2013-09-01 | 1 | -0/+3 | |
| | ||||||
* | Correct error in debug instantiations | Erik Schnetter | 2013-09-01 | 1 | -1/+1 | |
| | ||||||
* | Implement more integer functions: abs bitifthen clz isignbit max min ↵ | Erik Schnetter | 2013-09-01 | 22 | -110/+896 | |
| | | | | | | popcount rotate Rename integer signbit to isignbit to avoid name conflicts. | |||||
* | NEON: Describe idea for prod() improvement | Erik Schnetter | 2013-08-26 | 1 | -0/+1 | |
| | ||||||
* | Describe how to build on Stampede (MIC) and Raspberry Pi (ARM) | Erik Schnetter | 2013-08-26 | 1 | -0/+9 | |
| | ||||||
* | Provide a string describing the configuration options | Erik Schnetter | 2013-08-26 | 2 | -38/+74 | |
| | ||||||
* | Make frexp take a pointer argument instead of a reference | Erik Schnetter | 2013-08-21 | 20 | -31/+34 | |
| | ||||||
* | Add float example to build script | Erik Schnetter | 2013-08-20 | 1 | -0/+1 | |
| | ||||||
* | Use better timers for benchmarking loops | Erik Schnetter | 2013-08-20 | 1 | -4/+9 | |
| | ||||||
* | Benchmark more math functions | Erik Schnetter | 2013-08-20 | 1 | -55/+151 | |
| | ||||||
* | Correct detecting ARM NEON intrinsics | Erik Schnetter | 2013-08-20 | 1 | -2/+2 | |
| | ||||||
* | Add single-precision example | Erik Schnetter | 2013-08-20 | 2 | -0/+41 | |
| | ||||||
* | Provide trivial fallback "best vector" if no "best vector" is available | Erik Schnetter | 2013-08-20 | 1 | -14/+22 | |
| | ||||||
* | Neon: Implement unaligned load/store operations | Erik Schnetter | 2013-08-20 | 2 | -10/+16 | |
| | ||||||
* | Explicitly align allocated memory | Erik Schnetter | 2013-08-19 | 1 | -1/+6 | |
| | ||||||
* | Add potential optimization for SSE/AVX all/any functions | Erik Schnetter | 2013-08-19 | 5 | -5/+5 | |
| | ||||||
* | SSE: Correct float4 all/any | Erik Schnetter | 2013-08-19 | 1 | -2/+2 | |
| | ||||||
* | Begin to add MIC support | Erik Schnetter | 2013-08-19 | 3 | -1/+663 | |
| | ||||||
* | QPX: Undo using const& to pass QPX vectors | Erik Schnetter | 2013-08-18 | 1 | -66/+66 | |
| | | | | I hear that the QPX API does allow passing structs containing a single vector in a vector register. | |||||
* | Replace tabs by spaces | Erik Schnetter | 2013-08-17 | 3 | -31/+31 | |
| | ||||||
* | QPX: Correct various errors | Erik Schnetter | 2013-08-17 | 1 | -12/+31 | |
| | ||||||
* | QPX: Correct some type errors | Erik Schnetter | 2013-08-17 | 1 | -3/+6 | |
| | ||||||
* | QPX: Pass vector arguments as const&, not by value | Erik Schnetter | 2013-08-17 | 1 | -83/+95 | |
| | | | | This should improve performance, since passing by value involves scalarizing the vector on QPX. | |||||
* | QPX: Correct type errors in any/all | Erik Schnetter | 2013-08-17 | 1 | -6/+6 | |
| | ||||||
* | QPX: Remove debug output | Erik Schnetter | 2013-08-17 | 1 | -2/+0 | |
| | ||||||
* | QPX: Correct copysign | Erik Schnetter | 2013-08-17 | 1 | -1/+1 | |
| | ||||||
* | QPX: Optimize all/any | Erik Schnetter | 2013-08-17 | 1 | -3/+11 | |
| | ||||||
* | QPX: Remove debug output | Erik Schnetter | 2013-08-17 | 1 | -2/+0 | |
| | ||||||
* | Remove pocl subdirectory. This is now instead part of pocl. | Erik Schnetter | 2013-08-15 | 10 | -1089/+0 | |
| | ||||||
* | Update build instructions | Erik Schnetter | 2013-08-15 | 2 | -15/+4 | |
| | ||||||
* | Remove GPL and LGPL licences. This makes Vecmathlib a pure MIT licence project. | Erik Schnetter | 2013-08-15 | 2 | -839/+0 | |
| | ||||||
* | Add MIT licence | Erik Schnetter | 2013-08-15 | 1 | -0/+19 | |
| | ||||||
* | Clean up build mechanism | Erik Schnetter | 2013-08-15 | 4 | -60/+34 | |
| | ||||||
* | Simplify example, add comments | Erik Schnetter | 2013-07-28 | 2 | -19/+57 | |
| | ||||||
* | Merge branch 'master' of bitbucket.org:eschnett/vecmathlib | Erik Schnetter | 2013-07-28 | 1 | -4/+4 | |
|\ | ||||||
| * | Do not build pocl subdirectory by default | Erik Schnetter | 2013-07-27 | 1 | -4/+4 | |
| | | ||||||
* | | Do not use std_ prefix for fma, fmax, or fmin | Erik Schnetter | 2013-07-28 | 1 | -7/+0 | |
|/ |