Commits · 10248d040f423c340994be52505a9e9abf244c21 · Andreas Klöckner / pycuda · GitLab

Oct 22, 2008
- Disable 32-right shift if on 32-bit system. Thanks, Visual C. · 10248d04
  Andreas Klöckner authored Oct 22, 2008
  
  10248d04
Oct 14, 2008
- Default the device parameter of DeviceData. · 5aaebf70
  Andreas Kloeckner authored Oct 14, 2008
  
  5aaebf70
- Return function from prepare() for chainability. · 725c71e3
  Andreas Kloeckner authored Oct 14, 2008
  
  725c71e3
Oct 08, 2008
- Fix memset documentation. · 9343199c
  Andreas Kloeckner authored Oct 08, 2008
  
  9343199c
- Fix error reporting for memset*. (reported by Mark on the PyCuda ML) · 71599ced
  Andreas Kloeckner authored Oct 08, 2008
  
  71599ced
Oct 07, 2008
- Make CUDA compiler errors discernible. · 8e81a606
  Andreas Kloeckner authored Oct 07, 2008
  
  8e81a606
Oct 06, 2008
- Support proper flop counting in mul_add/axpbyz. · 288dccc8
  Andreas Kloeckner authored Oct 06, 2008
  
  288dccc8
- Add timings for vector math. · e6baa2b7
  Andreas Kloeckner authored Oct 06, 2008
  
  e6baa2b7
- Add enough timing to be able to time RK4. · 61ab4a76
  Andreas Kloeckner authored Oct 06, 2008
  
  61ab4a76
- Remove cache-dependency of compiler cache on pycuda.driver module. · d4d191a0
  Andreas Kloeckner authored Oct 06, 2008
  
  d4d191a0
- Fix pycuda.driver to 'fake' per-function lmem/smem/registers in SourceModule. · 17c8878f
  Andreas Kloeckner authored Oct 06, 2008
  
  17c8878f
Sep 30, 2008
- Merge branch 'master' of haamster:src/pycuda · 18029264
  Andreas Kloeckner authored Sep 29, 2008
  
  Conflicts: src/python/driver.py
  18029264
Sep 29, 2008
- Fix malformed assert, reported by ̈́Émilien Tlapale. · ff58a51d
  Andreas Klöckner authored Sep 28, 2008
  
  ff58a51d
- Return timing callable from prepared_timed_call. · 849cbaa2
  Andreas Kloeckner authored Sep 28, 2008
  
  849cbaa2
Sep 28, 2008
- Raise exceptions for invalid arguments to pycuda.tools.OccupancyRecord. · 1703c321
  Andreas Kloeckner authored Sep 28, 2008
  
  1703c321
- Give PyCuda its own exception classes for better error handling. · 0dff27e5
  Andreas Kloeckner authored Sep 28, 2008
  
  0dff27e5
Sep 27, 2008
- Have Function.prepared_timed_call return time "lazily". · c3e9db8c
  Andreas Kloeckner authored Sep 27, 2008
  
  c3e9db8c
Sep 24, 2008
- Doc addition about texrefs and life support. · a8b00b95
  Andreas Kloeckner authored Sep 24, 2008
  
  a8b00b95
Sep 23, 2008
- Add allow_user_edit(). · 2536a219
  Andreas Kloeckner authored Sep 23, 2008
  
  2536a219
- Move dtype_to_array_format to driver module. · f37e3a77
  Andreas Kloeckner authored Sep 23, 2008
  
  f37e3a77
Sep 21, 2008
- Add a numpy.dtype-to-array_format conversion function. · 8af3cded
  Andreas Kloeckner authored Sep 21, 2008
  
  8af3cded
- Add documentation for TexRef.{get,set}_format(). · 3c593236
  Andreas Kloeckner authored Sep 21, 2008
  
  3c593236
Sep 20, 2008
- Small doc fix. · dec5e0fe
  Andreas Kloeckner authored Sep 20, 2008
  
  dec5e0fe
- Default to presently-available compute capability for compilation. · ab9d576a
  Andreas Kloeckner authored Sep 20, 2008
  
  ab9d576a
- Add new test for arithmetic behavior of NaNs. · be93449d
  Andreas Kloeckner authored Sep 20, 2008
  
  be93449d
Sep 17, 2008
- Fix cumath and curandom. Tests pass again. · b44274b6
  Andreas Kloeckner authored Sep 17, 2008
  
  b44274b6
- Fix vector_expression execution: Block shape must be set. · 215abb25
  Andreas Klöckner authored Sep 17, 2008
  
  215abb25
- DebugMemoryPool in a usable state. · 9da8b60e
  Andreas Klöckner authored Sep 17, 2008
  
  9da8b60e
Sep 16, 2008
- Back out ref-showing debug mempool. · 20f799b9
  Andreas Klöckner authored Sep 16, 2008
  
  20f799b9
- Failed attempt at ref-showing debugging memory pool. · 742c078a
  Andreas Klöckner authored Sep 16, 2008
  
  (This failed mainly because numpy's object arrays don't participate in cyclic gc.)
  742c078a
Sep 08, 2008
- Remove old and empirical vector splay. · ac10e66c
  Andreas Kloeckner authored Sep 08, 2008
  
  ac10e66c
- More tweaking of vector-op splay. · f54ac1e3
  Andreas Klöckner authored Sep 08, 2008
  
  f54ac1e3
- More experiments with scalar vector op grid sizes. · 6d694f21
  Andreas Klöckner authored Sep 08, 2008
  
  6d694f21
- Cache nvcc output. · 900afb46
  Andreas Klöckner authored Sep 08, 2008
  
  900afb46
Sep 07, 2008
- Figure out vector splaying by (cached) trial-and-error. · 2352949e
  Andreas Kloeckner authored Sep 07, 2008
  
  2352949e
- Make pycuda.driver.Device hashable and comparable. · 5dfc1eb2
  Andreas Kloeckner authored Sep 07, 2008
  
  5dfc1eb2
- Synchronize before kernel call to get meaningful timing data. · 063597a7
  Andreas Kloeckner authored Sep 07, 2008
  
  063597a7
- Make sure vector expression kernels have a meaningful name for profiling. · 76d1fd61
  Andreas Kloeckner authored Sep 07, 2008
  
  76d1fd61
- Make better autoinit mechanism. · dd797da1
  Andreas Kloeckner authored Sep 07, 2008
  
  dd797da1
Sep 04, 2008
- Convert remaining kernels to prepared execution. Improve GPUArray perf. · c787e765
  Andreas Kloeckner authored Sep 03, 2008
  
  c787e765