Skip to content

Commit 0ef5c50

Browse files
brycelelbachbrycelelbach
committed
Update changelog for the 1.10.0 release.
1 parent 0e2cb97 commit 0ef5c50

File tree

3 files changed

+155
-6
lines changed

3 files changed

+155
-6
lines changed

‎CHANGELOG.md‎

Lines changed: 153 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,148 @@
1+
# Thrust 1.10.0 (NVIDIA HPC SDK 20.9)
2+
3+
## Summary
4+
5+
Thrust 1.10.0 is the major release accompanying the NVIDIA HPC SDK 20.9 release.
6+
It drops support for C++03, GCC < 5, Clang < 6, and MSVC < 2017.
7+
It also overhauls CMake support.
8+
Finally, we now have a Code of Conduct for contributors:
9+
https://github.com/thrust/thrust/blob/main/CODE_OF_CONDUCT.md
10+
11+
## Breaking Changes
12+
13+
- C++03 is no longer supported.
14+
- GCC < 5, Clang < 6, and MSVC < 2017 are no longer supported.
15+
- C++11 is deprecated.
16+
Using this dialect will generate a compile-time warning.
17+
These warnings can be suppressed by defining
18+
`THRUST_IGNORE_DEPRECATED_CPP_DIALECT` or `THRUST_IGNORE_DEPRECATED_CPP_11`.
19+
Suppression is only a short term solution.
20+
We will be dropping support for C++11 in the near future.
21+
- Asynchronous algorithms now require C++14.
22+
- CMake < 3.15 is no longer supported.
23+
- The default branch on GitHub is now called `main`.
24+
- Allocator and vector classes have been replaced with alias templates.
25+
26+
## New Features
27+
28+
- Contributor documentation: https://github.com/thrust/thrust/blob/main/CONTRIBUTING.md
29+
- thrust/thrust#1159: CMake multi-config support, which allows multiple
30+
combinations of host and device systems to be built and tested at once.
31+
More details can be found here: https://github.com/thrust/thrust/blob/main/CONTRIBUTING.md#multi-config-cmake-options
32+
- CMake refactoring:
33+
- Added install targets to CMake builds.
34+
- Added support for CUB tests and examples.
35+
- Thrust can be added to another CMake project by calling `add_subdirectory`
36+
with the Thrust source root (see thrust/thrust#976).
37+
An example can be found here:
38+
https://github.com/thrust/thrust/blob/main/examples/cmake/add_subdir/CMakeLists.txt
39+
- CMake < 3.15 is no longer supported.
40+
- Dialects are now configured through target properties.
41+
A new `THRUST_CPP_DIALECT` option has been added for single config mode.
42+
Logic that modified `CMAKE_CXX_STANDARD` and `CMAKE_CUDA_STANDARD` has been
43+
eliminated.
44+
- Testing related CMake code has been moved to `testing/CMakeLists.txt`
45+
- Example related CMake code has been moved to `examples/CMakeLists.txt`
46+
- Header testing related CMake code has been moved to `cmake/ThrustHeaderTesting.cmake`
47+
- CUDA configuration CMake code has been moved to to `cmake/ThrustCUDAConfig.cmake`.
48+
- Now we explicitly `include(cmake/*.cmake)` files rather than searching
49+
`CMAKE_MODULE_PATH` - we only want to use the ones in the repo.
50+
- `thrust::transform_input_output_iterator`, a variant of transform iterator
51+
adapter that works as both an input iterator and an output iterator.
52+
The given input function is applied after reading from the wrapped iterator
53+
while the output function is applied before writing to the wrapped iterator.
54+
Thanks to Trevor Smith for this contribution.
55+
56+
## Other Enhancements
57+
58+
- Support for all combinations of host and device systems.
59+
- C++17 support.
60+
- thrust/thrust#1221: Allocator and vector classes have been replaced with
61+
alias templates.
62+
Thanks to Michael Francis for this contribution.
63+
- thrust/thrust#1186: Use placeholder expressions to simplify the definitions
64+
of a number of algorithms.
65+
Thanks to Michael Francis for this contribution.
66+
- thrust/thrust#1170: More conforming semantics for scan algorithms:
67+
- Follow P0571's guidance regarding intermediate types.
68+
- https://wg21.link/P0571
69+
- The accumulator's type is now:
70+
- The type of the user-supplied initial value (if provided), or
71+
- The input iterator's value type if no initial value.
72+
- Follow C++ standard guidance for default binary operator type.
73+
- https://eel.is/c++draft/exclusive.scan#1
74+
- Thrust binary/unary functors now specialize a default void template
75+
parameter.
76+
Types are deduced and forwarded transparently.
77+
- Updated the scan's default binary operator to the new `thrust::plus<>`
78+
specialization.
79+
- The `thrust::intermediate_type_from_function_and_iterators` helper is no
80+
longer needed and has been removed.
81+
- thrust/thrust#1255: Always use `cudaStreamSynchronize` instead of
82+
`cudaDeviceSynchronize` if the execution policy has a stream attached to it.
83+
Thanks to Rong Ou for this contribution.
84+
- thrust/thrust#1201: Tests for correct handling of legacy and per-thread
85+
default streams.
86+
Thanks to Rong Ou for this contribution.
87+
88+
## Bug Fixes
89+
90+
- thrust/thrust#1260: Fix `thrust::transform_inclusive_scan` with heterogeneous
91+
types.
92+
Thanks to Rong Ou for this contribution.
93+
- thrust/thrust#1258, NVC++ FS #28463: Ensure the CUDA radix sort backend
94+
synchronizes before returning; otherwise, copies from temporary storage will
95+
race with destruction of said temporary storage.
96+
- thrust/thrust#1264: Evaluate `CUDA_CUB_RET_IF_FAIL` macro argument only once.
97+
Thanks to Jason Lowe for this contribution.
98+
- thrust/thrust#1262: Add missing `<stdexcept>` header.
99+
- thrust/thrust#1250: Restore some `THRUST_DECLTYPE_RETURNS` macros in async
100+
test implementations.
101+
- thrust/thrust#1249: Use `std::iota` in `CUDATestDriver::target_devices`.
102+
Thanks to Michael Francis for this contribution.
103+
- thrust/thrust#1244: Check for macro collisions with system headers during
104+
header testing.
105+
- thrust/thrust#1224: Remove unnecessary SFINAE contexts from asynchronous
106+
algorithms.
107+
- thrust/thrust#1190: Make `out_of_memory_recovery` test trigger faster.
108+
- thrust/thrust#1187: Elminate superfluous iterators specific to the CUDA
109+
backend.
110+
- thrust/thrust#1181: Various fixes for GoUDA.
111+
Thanks to Andrei Tchouprakov for this contribution.
112+
- thrust/thrust#1178, thrust/thrust#1229: Use transparent functionals in
113+
placeholder expressions, fixing issues with `thrust::device_reference` and
114+
placeholder expressions and `thrust::find` with asymmetric equality
115+
operators.
116+
- thrust/thrust#1153: Switch to placement new instead of assignment to
117+
construct items in uninitialized memory.
118+
Thanks to Hugh Winkler for this contribution.
119+
- thrust/thrust#1050: Fix compilation of asynchronous algorithms when RDC is
120+
enabled.
121+
- thrust/thrust#1042: Correct return type of
122+
`thrust::detail::predicate_to_integral` from `bool` to `IntegralType`.
123+
Thanks to Andreas Hehn for this contribution.
124+
- thrust/thrust#1009: Avoid returning uninitialized allocators.
125+
Thanks to Zhihao Yuan for this contribution.
126+
- thrust/thrust#990: Add missing `<thrust/system/cuda/memory.h>` include to
127+
`<thrust/system/cuda/detail/malloc_and_free.h>`.
128+
Thanks to Robert Maynard for this contribution.
129+
- thrust/thrust#966: Fix spurious MSVC conversion with loss of data warning in
130+
sort algorithms.
131+
Thanks to Zhihao Yuan for this contribution.
132+
- Add more metadata to mock specializations for testing iterator in
133+
`testing/copy.cu`.
134+
- Add missing include to shuffle unit test.
135+
- Specialize `thrust::wrapped_function` for `void` return types because MSVC is
136+
not a fan of the pattern `return static_cast<void>(expr);`.
137+
- Replace deprecated `tbb/tbb_thread.h` with `<thread>`.
138+
- Fix overcounting of initial value in TBB scans.
139+
- Use `thrust::advance` instead of `+=` for generic iterators.
140+
- Wrap the OMP flags in `-Xcompiler` for NVCC
141+
- Extend `ASSERT_STATIC_ASSERT` skip for the OMP backend.
142+
- Add missing header caught by `tbb.cuda` configs.
143+
- Fix "unsafe API" warnings in examples on MSVC: `s/fopen/fstream/`
144+
- Various C++17 fixes.
145+
1146
# Thrust 1.9.10-1 (NVIDIA HPC SDK 20.7, CUDA Toolkit 11.1)
2147

3148
## Summary
@@ -1076,11 +1221,14 @@ Support for TBB allows Thrust programs to integrate more naturally into
10761221
- `set_operations`
10771222

10781223
## Other Enhancements
1079-
- thrust::for_each now returns the end of the input range similar to most other algorithms
1080-
- thrust::pair and thrust::tuple have swap functionality
1081-
- All CUDA algorithms now support large data types
1082-
- Iterators may be dereferenced in user __device__ or __global__ functions
1083-
- The safe use of different backend systems is now possible within a single binary
1224+
1225+
- `thrust::for_each` now returns the end of the input range similar to most
1226+
other algorithms.
1227+
- `thrust::pair` and `thrust::tuple` have swap functionality.
1228+
- All CUDA algorithms now support large data types.
1229+
- Iterators may be dereferenced in user `__device__` or `__global__` functions.
1230+
- The safe use of different backend systems is now possible within a single
1231+
binary
10841232

10851233
## Bug Fixes
10861234

‎README.md‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -81,6 +81,7 @@ See the [changelog](CHANGELOG.md) for details about specific releases.
8181

8282
| Thrust Release | Included In |
8383
| ----------------- | --------------------------------------- |
84+
| 1.10.0 | NVIDIA HPC SDK 20.9 |
8485
| 1.9.10-1 | NVIDIA HPC SDK 20.7 & CUDA Toolkit 11.1 |
8586
| 1.9.10 | NVIDIA HPC SDK 20.5 |
8687
| 1.9.9 | CUDA Toolkit 11.0 |

‎dependencies/cub‎

Submodule cub updated from 8e09201 to a3ee304

0 commit comments

Comments
 (0)