Today

• 3 hours ago

  issue 53 (Parallelize OpenMP scan functions) changed by wnbell   -  

  Status: Started

  Labels: Priority-Low Milestone-Release1.x −Priority-Medium −Milestone-Release1.2

  Status: Started

  Labels: Priority-Low Milestone-Release1.x −Priority-Medium −Milestone-Release1.2
• 3 hours ago

  issue 57 (improve cuda::stable_merge_sort() performance with large siz...) changed by wnbell   -   Relates to issue #19 Resolved by r657 r658 and r659

  Status: Fixed

  Labels: Milestone-Release1.1 −Milestone-Release1.x

  Relates to issue #19 Resolved by r657 r658 and r659

  Status: Fixed

  Labels: Milestone-Release1.1 −Milestone-Release1.x
• 5 hours ago

  issue 64 (Check correctness of OMP backend without -fopenmp or /openmp) reported by wnbell   -   OpenMP implementations should execute correctly even if the user omits the necessary compiler flags. For example, parallel code blocks should not assume a specific number of threads have been created using the num_threads() option to #pragma omp parallel. Instead, the code should query omp_get_num_threads() within the block.

  OpenMP implementations should execute correctly even if the user omits the necessary compiler flags. For example, parallel code blocks should not assume a specific number of threads have been created using the num_threads() option to #pragma omp parallel. Instead, the code should query omp_get_num_threads() within the block.
• 22 hours ago

  r687 (WAR extern __shared__ array problem with extern_shared_ptr c...) committed by jaredhoberock   -   WAR extern __shared__ array problem with extern_shared_ptr class template.

  WAR extern __shared__ array problem with extern_shared_ptr class template.
• 23 hours ago

  issue 63 (TestPairSegmentedScan crashes on XP 32b with MSVC 2008) reported by jaredhoberock   -   What steps will reproduce the problem? tester TestPairSegmentedScan

  What steps will reproduce the problem? tester TestPairSegmentedScan

Yesterday

• 34 hours ago

  r686 (g++ needs to know the device backend too; send it along ) committed by jaredhoberock   -   g++ needs to know the device backend too; send it along

  g++ needs to know the device backend too; send it along
• 45 hours ago

  issue 62 (Provide better documentation for device_reference<T> / T int...) reported by jaredhoberock   -   Examples include: 1. member access 2. printf on a device_reference

  Examples include: 1. member access 2. printf on a device_reference

Last 7 days

• Dec 15, 2009

  r685 (Implement are_spaces_interoperable Catch interoperable space...) committed by jaredhoberock   -   Implement are_spaces_interoperable Catch interoperable space case inside device_dereference when assigning or converting and do the fast thing

  Implement are_spaces_interoperable Catch interoperable space case inside device_dereference when assigning or converting and do the fast thing
• Dec 14, 2009

  r684 (fix bug in segmented_scan ) committed by jaredhoberock   -   fix bug in segmented_scan

  fix bug in segmented_scan
• Dec 14, 2009

  issue 61 (vector_base's method should be __host__ only for now) reported by jaredhoberock   -   Declaring vector_base's methods __host__ __device__ causes badness with -- device-emulation turned on. Declare them as __host__ only for now.

  Declaring vector_base's methods __host__ __device__ causes badness with -- device-emulation turned on. Declare them as __host__ only for now.
• Dec 14, 2009

  r683 (clarified transform_reduce() semantics ) committed by wnbell   -   clarified transform_reduce() semantics

  clarified transform_reduce() semantics
• Dec 13, 2009

  r682 (fixed invalid iterator reference in generic::exclusive_segme...) committed by wnbell   -   fixed invalid iterator reference in generic::exclusive_segmented_scan

  fixed invalid iterator reference in generic::exclusive_segmented_scan
• Dec 13, 2009

  r681 (Use the index-less version of dereference in the omp impleme...) committed by jaredhoberock   -   Use the index-less version of dereference in the omp implementation of copy.

  Use the index-less version of dereference in the omp implementation of copy.
• Dec 13, 2009

  r680 (Most versions of dereference should take const references to...) committed by jaredhoberock   -   Most versions of dereference should take const references to their iterator argument.

  Most versions of dereference should take const references to their iterator argument.
• Dec 13, 2009

  r679 (Correct typo in omp reduce implementation. ) committed by jaredhoberock   -   Correct typo in omp reduce implementation.

  Correct typo in omp reduce implementation.
• Dec 12, 2009

  issue 53 (Parallelize OpenMP scan functions) commented on by wnbell   -   The above code can be compiled with: $ g++ -O3 -fopenmp -o scan scan.cpp -lgomp

  The above code can be compiled with: $ g++ -O3 -fopenmp -o scan scan.cpp -lgomp
• Dec 12, 2009

  issue 53 (Parallelize OpenMP scan functions) commented on by wnbell   -   The attached file implements inclusive_scan() with OpenMP. Unfortunately it is no faster than a serial scan (i.e. std::partial_sum) on a either the dual-core or quad-core Core2 systems tested.

  The attached file implements inclusive_scan() with OpenMP. Unfortunately it is no faster than a serial scan (i.e. std::partial_sum) on a either the dual-core or quad-core Core2 systems tested.

Last 30 days

• Dec 11, 2009

  issue 60 (The SVN version of thrust should issue a warning) commented on by wnbell   -   With such a macro we could support a -DDISABLE_THRUST_WARNING (or equivalent, if there's a standard way). If a refinement of THRUST_WARNING is desired we could define THRUST_DEPRECATION_WARNING.

  With such a macro we could support a -DDISABLE_THRUST_WARNING (or equivalent, if there's a standard way). If a refinement of THRUST_WARNING is desired we could define THRUST_DEPRECATION_WARNING.
• Dec 11, 2009

  issue 60 (The SVN version of thrust should issue a warning) commented on by wnbell   -   It would be nice to have a THRUST_WARNING(message) macro that handled the msvc/gcc switch (used in sorting below): http://code.google.com/p/thrust/source/browse/trunk/thrust/sorting/radix_sort.h#22

  It would be nice to have a THRUST_WARNING(message) macro that handled the msvc/gcc switch (used in sorting below): http://code.google.com/p/thrust/source/browse/trunk/thrust/sorting/radix_sort.h#22
• Dec 11, 2009

  issue 60 (The SVN version of thrust should issue a warning) reported by jaredhoberock   -   Warn that the svn version of thrust is not tested, and mention how the warning can be disabled.

  Warn that the svn version of thrust is not tested, and mention how the warning can be disabled.
• Dec 10, 2009

  r678 (Convert CUDA scan kernel to use the index-less version of de...) committed by jaredhoberock   -   Convert CUDA scan kernel to use the index-less version of dereference.

  Convert CUDA scan kernel to use the index-less version of dereference.
• Dec 10, 2009

  r677 (moved cuda/detail/block/* to cuda/block/ ) committed by wnbell   -   moved cuda/detail/block/* to cuda/block/

  moved cuda/detail/block/* to cuda/block/
• Dec 10, 2009

  r676 (fixed several bugs found by trivial_tests ) committed by wnbell   -   fixed several bugs found by trivial_tests

  fixed several bugs found by trivial_tests
• Dec 10, 2009

  issue 8 (Thrust should be strict about checking whether it is safe to...) Labels changed by jaredhoberock   -   We should fix this in 1.2.

  Labels: Milestone-Release1.2 −Milestone-Release1.x

  We should fix this in 1.2.

  Labels: Milestone-Release1.2 −Milestone-Release1.x
• Dec 10, 2009

  FrequentlyAskedQuestions (Frequently Asked Questions) Wiki page edited by wnbell   -   Revision r675 Edited wiki page through web user interface.

  Revision r675 Edited wiki page through web user interface.
• Dec 10, 2009

  Roadmap (Tentative schedule of Thrust development) Wiki page commented on by Damien.Massart   -   Do you have any plan to use openCL instead of Cuda ? Thx

  Do you have any plan to use openCL instead of Cuda ? Thx
• Dec 08, 2009

  r674 (Use index-less dereference in omp & cuda reduce implementati...) committed by jaredhoberock   -   Use index-less dereference in omp & cuda reduce implementations.

  Use index-less dereference in omp & cuda reduce implementations.
• Dec 08, 2009

  r673 (Use index-less dereference in omp & cuda versions of for_eac...) committed by jaredhoberock   -   Use index-less dereference in omp & cuda versions of for_each

  Use index-less dereference in omp & cuda versions of for_each
• Dec 08, 2009

  r672 (Fix segmentation fault in vector_cpp_subset.cpp: * unguard C...) committed by jaredhoberock   -   Fix segmentation fault in vector_cpp_subset.cpp: * unguard CUDA's free dispatch * make malloc & free functions templated on a dummy argument to prevent their instantiation if they're not used

  Fix segmentation fault in vector_cpp_subset.cpp: * unguard CUDA's free dispatch * make malloc & free functions templated on a dummy argument to prevent their instantiation if they're not used
• Dec 08, 2009

  r671 (added missing file: host/unique.h ) committed by wnbell   -   added missing file: host/unique.h

  added missing file: host/unique.h
• Dec 08, 2009

  r670 (device::generic::segmented_scan() and device::generic::uniqu...) committed by wnbell   -   device::generic::segmented_scan() and device::generic::unique() now invoke device::scan()

  device::generic::segmented_scan() and device::generic::unique() now invoke device::scan()
• Dec 08, 2009

  issue 59 (implement unique_by_key, unique_copy, and unique_copy_by_key) Status changed by wnbell   -   r669 provides a preliminary implementation of all three functions no functions have been documented yet

  Status: Started

  r669 provides a preliminary implementation of all three functions no functions have been documented yet

  Status: Started
• Dec 08, 2009

  r669 (added unique_by_key and unique_copy_by_key partially address...) committed by wnbell   -   added unique_by_key and unique_copy_by_key partially addresses issue #59

  added unique_by_key and unique_copy_by_key partially addresses issue #59
• Dec 08, 2009

  r668 (Use index-less version of dereference in merge sort's implem...) committed by jaredhoberock   -   Use index-less version of dereference in merge sort's implementation.

  Use index-less version of dereference in merge sort's implementation.
• Dec 07, 2009

  issue 41 (Test Thrust with --device-emulation) changed by wnbell   -   We should check for __DEVICE_EMULATION__ in thrust/detail/config.h and generate a compile time #error if necessary.

  Owner: wnbell

  Labels: Milestone-Release1.2 −Milestone-Release1.x

  We should check for __DEVICE_EMULATION__ in thrust/detail/config.h and generate a compile time #error if necessary.

  Owner: wnbell

  Labels: Milestone-Release1.2 −Milestone-Release1.x
• Dec 07, 2009

  r667 (added project1st and project2nd functionals ) committed by wnbell   -   added project1st and project2nd functionals

  added project1st and project2nd functionals
• Dec 07, 2009

  r666 (added unique_copy() partially resolves issue #59 unique_copy...) committed by wnbell   -   added unique_copy() partially resolves issue #59 unique_copy still requires documentation

  added unique_copy() partially resolves issue #59 unique_copy still requires documentation
• Dec 05, 2009

  issue 59 (implement unique_by_key, unique_copy, and unique_copy_by_key) reported by wnbell   -   template <typename InputIterator, typename OutputIterator, typename BinaryPredicate> OutputIterator unique_copy(InputIterator first, InputIterator last, OutputIterator result, BinaryPredicate binary_pred) template <typename ForwardIterator1, typename ForwardIterator2, typename BinaryPredicate, typename BinaryFunction> ForwardIterator1 unique_by_key(ForwardIterator1 keys_first, ForwardIterator1 keys_last, ForwardIterator2 values_first, BinaryPredicate binary_pred, BinaryFunction binary_op) template <typename InputIterator1, typename InputIterator2, typename OutputIterator1, typename OutputIterator2, typename BinaryPredicate, typename BinaryFunction> OutputIterator unique_copy_by_key(InputIterator1 keys_first, InputIterator1 keys_last, InputIterator2 values_first, OutputIterator1 keys_output, OutputIterator2 values_output, BinaryPredicate binary_pred, BinaryFunction binary_op) BinaryPredicate defaults to equal_to<T> BinaryFunction defaults to project1st<T> BinaryFunction reduces adjacent values with the same key. Must be associative. Could require commutativity like reduce() does.

  template <typename InputIterator, typename OutputIterator, typename BinaryPredicate> OutputIterator unique_copy(InputIterator first, InputIterator last, OutputIterator result, BinaryPredicate binary_pred) template <typename ForwardIterator1, typename ForwardIterator2, typename BinaryPredicate, typename BinaryFunction> ForwardIterator1 unique_by_key(ForwardIterator1 keys_first, ForwardIterator1 keys_last, ForwardIterator2 values_first, BinaryPredicate binary_pred, BinaryFunction binary_op) template <typename InputIterator1, typename InputIterator2, typename OutputIterator1, typename OutputIterator2, typename BinaryPredicate, typename BinaryFunction> OutputIterator unique_copy_by_key(InputIterator1 keys_first, InputIterator1 keys_last, InputIterator2 values_first, OutputIterator1 keys_output, OutputIterator2 values_output, BinaryPredicate binary_pred, BinaryFunction binary_op) BinaryPredicate defaults to equal_to<T> BinaryFunction defaults to project1st<T> BinaryFunction reduces adjacent values with the same key. Must be associative. Could require commutativity like reduce() does.
• Dec 05, 2009

  Roadmap (Tentative schedule of Thrust development) Wiki page edited by wnbell   -   Revision r665 moved unique_copy, unique_by_key, and unique_copy_by_key to Milestone 1.2

  Revision r665 moved unique_copy, unique_by_key, and unique_copy_by_key to Milestone 1.2
• Dec 05, 2009

  r664 (Explicity implement tuple_transform & tuple_meta_transform w...) committed by jaredhoberock   -   Explicity implement tuple_transform & tuple_meta_transform without cons.

  Explicity implement tuple_transform & tuple_meta_transform without cons.
• Dec 04, 2009

  r663 (added padded_grid_reduction example removed redundant output...) committed by wnbell   -   added padded_grid_reduction example removed redundant output from histogram example

  added padded_grid_reduction example removed redundant output from histogram example
• Dec 03, 2009

  r662 (Change the name of the function that determines the result o...) committed by jaredhoberock   -   Change the name of the function that determines the result of dereference from iterator_device_reference to dereference_result.

  Change the name of the function that determines the result of dereference from iterator_device_reference to dereference_result.
• Dec 03, 2009

  r661 (added a histogram example in response to a question on thrus...) committed by wnbell   -   added a histogram example in response to a question on thrust-users [1] [1] http://groups.google.com/group/thrust-users/browse_thread/thread/aea782788cf5686a

  added a histogram example in response to a question on thrust-users [1] [1] http://groups.google.com/group/thrust-users/browse_thread/thread/aea782788cf5686a
• Dec 03, 2009

  Tutorial (Tutorial for new Thrust developers) Wiki page edited by wnbell   -   Revision r660 Edited wiki page through web user interface.

  Revision r660 Edited wiki page through web user interface.
• Dec 01, 2009

  issue 58 (refine raw_device_buffer usage in device backends) reported by wnbell   -   Our present usage of raw_device_buffer in the device backends will fail if the non-default backend is used explicitly. Either we should always use raw_buffer<T, cuda_device_space_tag> or define a raw_cuda_buffer and raw_omp_buffer in each of the device backends.

  Our present usage of raw_device_buffer in the device backends will fail if the non-default backend is used explicitly. Either we should always use raw_buffer<T, cuda_device_space_tag> or define a raw_cuda_buffer and raw_omp_buffer in each of the device backends.
• Dec 01, 2009

  issue 19 (make algorithms work for arbitrarily large data types) commented on by wnbell   -   r659 allows cuda::stable_sort_by_key() to handle arbitrarily large values

  r659 allows cuda::stable_sort_by_key() to handle arbitrarily large values
• Dec 01, 2009

  r659 (stable_sort_by_key() now uses indirection when sorting large...) committed by wnbell   -   stable_sort_by_key() now uses indirection when sorting large values partially addresses issue #19

  stable_sort_by_key() now uses indirection when sorting large values partially addresses issue #19
• Dec 01, 2009

  issue 19 (make algorithms work for arbitrarily large data types) commented on by wnbell   -   r658 allows cuda::stable_sort_by_key() to handle arbitrarily large keys

  r658 allows cuda::stable_sort_by_key() to handle arbitrarily large keys
• Dec 01, 2009

  r658 (stable_sort_by_key() now uses indirection when sorting large...) committed by wnbell   -   stable_sort_by_key() now uses indirection when sorting large keys partially addresses issue #19

  stable_sort_by_key() now uses indirection when sorting large keys partially addresses issue #19
• Dec 01, 2009

  issue 19 (make algorithms work for arbitrarily large data types) commented on by wnbell   -   r657 allows cuda::stable_sort() to handle arbitrarily large keys

  r657 allows cuda::stable_sort() to handle arbitrarily large keys