ChangeLog 2.2 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364
  1. StarPU 0.2.901 aka 0.3-rc1 (svn revision 1236)
  2. ==============================================
  3. The asynchronous heterogeneous multi-accelerator release
  4. * Many API changes and code cleanups
  5. - Implement starpu_get_worker_id
  6. - Implement starpu_get_worker_name
  7. - Implement starpu_get_worker_type
  8. - Implement starpu_get_worker_count
  9. - Implement starpu_display_codelet_stats
  10. - Implement starpu_prefetch_data_on_node
  11. - Expose the starpu_data_set_wb_mask function
  12. * Support nvidia (heterogeneous) multi-GPU
  13. * Add the data request mechanism
  14. - All data transfers use data requests now
  15. - Implement asynchronous data transfers
  16. - Implement prefetch mechanism
  17. - Chain data requests to support GPU->RAM->GPU transfers
  18. * Make it possible to bypass the scheduler and to assign a task to a specific
  19. worker
  20. * Support restartable tasks to reinstanciate dependencies task graphs
  21. * Improve performance prediction
  22. - Model data transfer overhead
  23. - One model is created for each accelerator
  24. * Support for CUDA's driver API is deprecated
  25. * The WORKERS_GPUID and WORKERS_CPUID env. variables make it possible to
  26. specify where to bind the workers
  27. * Use the hwloc library to detect the actual number of cores
  28. StarPU 0.2.0 (svn revision 1013)
  29. ==============================================
  30. The Stabilizing-the-Basics release
  31. * Various API cleanups
  32. * Mac OS X is supported now
  33. * Add dynamic code loading facilities onto Cell's SPUs
  34. * Improve performance analysis/feedback tools
  35. * Application can interact with StarPU tasks
  36. - The application may access/modify data managed by the DSM
  37. - The application may wait for the termination of a (set of) task(s)
  38. * An initial documentation is added
  39. * More examples are supplied
  40. StarPU 0.1.0 (svn revision 794)
  41. ==============================================
  42. First release.
  43. Status:
  44. * Only supports Linux platforms yet
  45. * Supported architectures
  46. - multicore CPUs
  47. - NVIDIA GPUs (with CUDA 2.x)
  48. - experimental Cell/BE support
  49. Changes:
  50. * Scheduling facilities
  51. - run-time selection of the scheduling policy
  52. - basic auto-tuning facilities
  53. * Software-based DSM
  54. - transparent data coherency management
  55. - High-level expressive interface