Replace thrust::swap by cuda::std::swap #2985

bernhardmgruber · 2024-11-28T16:31:16Z

bernhardmgruber · 2024-11-28T16:33:22Z

thrust/thrust/detail/contiguous_storage.inl

+  // FIXME(bgruber): swap_allocators already swaps m_allocator, so we are swapping twice here !!
  swap_allocators(integral_constant<bool, allocator_traits<Alloc>::propagate_on_container_swap::value>(),
                  x.m_allocator);
-
-  thrust::swap(m_allocator, x.m_allocator);
+  // FIXME(bgruber): this should use ADL-two-step swap, but this creates an ambiguity with std::swap until
+  // https://github.com/NVIDIA/cccl/issues/2984 is resolved.
+  // swap(m_allocator, x.m_allocator);
+  ::cuda::std::swap(m_allocator, x.m_allocator);


So this looks like a bug to me. swap_allocators was introduced in 403effb, but did not replace swapping the allocator below. @griwes, I assume the ::cuda::std::swap(m_allocator, x.m_allocator); is not needed/wrong here?

bernhardmgruber · 2024-11-28T16:38:06Z

thrust/thrust/detail/contiguous_storage.inl

+  // FIXME(bgruber): it is really concerning, that swapping an allocator can throw. swap() should be noexcept in
+  // general.
  NV_IF_TARGET(NV_IS_DEVICE,
               (
                 // allocators must be equal when swapping containers with allocators that propagate on swap
                 assert(!is_allocator_not_equal(other));),
               (if (is_allocator_not_equal(other)) { throw allocator_mismatch_on_swap(); }));


I have no ideas about the depths of allocators, but throwing an exception when swapping allocators strikes me as odd.

The reference for std::vector<T>::swap says:

[...] if get_allocator() != other.get_allocator(), the behavior is undefined

I would rather prefer to have a hard fault here, so we can mark all swap functions of thrust vectors as noexcept.

I mean swapping is move construction + move assignment so it can always potentially throw.

I would say it should not throw from our side, but especially allocators are potentially throwing a ton.

As an example think about a block allocator that reserves a huge slab of memory on construction and hands out slices of it. That could easily fail with an OOM

I mean swapping is move construction + move assignment so it can always potentially throw.

For a generic implementation, yes. In the long term, relocation could be used to avoid the moved from state, which could simplify things.

As an example think about a block allocator that reserves a huge slab of memory on construction and hands out slices of it. That could easily fail with an OOM

How can swapping two block allocators using memory slabs fail with an OOM? I would still just swap pointers and some bookkeeping?

github-actions · 2024-12-02T09:29:59Z

🟨 CI finished in 1h 45m: Pass: 96%/224 | Total: 6d 12h | Avg: 41m 58s | Max: 1h 16m | Hits: 40%/10436

🟨 thrust: Pass: 92%/111 | Total: 2d 16h | Avg: 34m 42s | Max: 1h 16m | Hits: 29%/7408

🔍 cpu: amd64 🔍
  🔍 amd64              Pass:  92%/103 | Total:  2d 11h | Avg: 34m 50s | Max:  1h 16m | Hits:  29%/7408  
  🟩 arm64              Pass: 100%/8   | Total:  4h 25m | Avg: 33m 10s | Max: 38m 20s
🔍 ctk: 12.6 🔍
  🟩 11.1               Pass: 100%/15  | Total:  8h 16m | Avg: 33m 05s | Max:  1h 06m | Hits:  29%/1852  
  🟩 11.8               Pass: 100%/3   | Total:  1h 57m | Avg: 39m 02s | Max: 44m 53s
  🟩 12.5               Pass: 100%/4   | Total:  4h 29m | Avg:  1h 07m | Max:  1h 16m
  🔍 12.6               Pass:  91%/89  | Total:  2d 01h | Avg: 33m 22s | Max:  1h 12m | Hits:  29%/5556  
🔍 cudacxx: nvcc12.6 🔍
  🟩 ClangCUDA18        Pass: 100%/4   | Total:  1h 57m | Avg: 29m 26s | Max: 31m 09s
  🟩 nvcc11.1           Pass: 100%/15  | Total:  8h 16m | Avg: 33m 05s | Max:  1h 06m | Hits:  29%/1852  
  🟩 nvcc11.8           Pass: 100%/3   | Total:  1h 57m | Avg: 39m 02s | Max: 44m 53s
  🟩 nvcc12.5           Pass: 100%/4   | Total:  4h 29m | Avg:  1h 07m | Max:  1h 16m
  🔍 nvcc12.6           Pass:  90%/85  | Total:  1d 23h | Avg: 33m 33s | Max:  1h 12m | Hits:  29%/5556  
🔍 cudacxx_family: nvcc 🔍
  🟩 ClangCUDA          Pass: 100%/4   | Total:  1h 57m | Avg: 29m 26s | Max: 31m 09s
  🔍 nvcc               Pass:  92%/107 | Total:  2d 14h | Avg: 34m 54s | Max:  1h 16m | Hits:  29%/7408  
🟨 cxx
  🟩 Clang9             Pass: 100%/6   | Total:  3h 14m | Avg: 32m 27s | Max: 37m 47s
  🟩 Clang10            Pass: 100%/3   | Total:  1h 47m | Avg: 35m 53s | Max: 37m 34s
  🟩 Clang11            Pass: 100%/4   | Total:  2h 12m | Avg: 33m 03s | Max: 35m 18s
  🟩 Clang12            Pass: 100%/4   | Total:  2h 16m | Avg: 34m 00s | Max: 36m 18s
  🟩 Clang13            Pass: 100%/4   | Total:  2h 20m | Avg: 35m 10s | Max: 37m 11s
  🟩 Clang14            Pass: 100%/4   | Total:  2h 20m | Avg: 35m 05s | Max: 38m 30s
  🟩 Clang15            Pass: 100%/4   | Total:  2h 16m | Avg: 34m 11s | Max: 36m 54s
  🟩 Clang16            Pass: 100%/4   | Total:  2h 17m | Avg: 34m 29s | Max: 37m 22s
  🟩 Clang17            Pass: 100%/4   | Total:  2h 24m | Avg: 36m 02s | Max: 39m 24s
  🟨 Clang18            Pass:  81%/11  | Total:  5h 00m | Avg: 27m 21s | Max: 37m 06s
  🟩 GCC6               Pass: 100%/2   | Total:  1h 00m | Avg: 30m 26s | Max: 33m 12s
  🟩 GCC7               Pass: 100%/6   | Total:  3h 09m | Avg: 31m 36s | Max: 36m 58s
  🟩 GCC8               Pass: 100%/6   | Total:  3h 16m | Avg: 32m 40s | Max: 38m 42s
  🟩 GCC9               Pass: 100%/6   | Total:  3h 18m | Avg: 33m 06s | Max: 37m 24s
  🟩 GCC10              Pass: 100%/4   | Total:  2h 18m | Avg: 34m 44s | Max: 37m 31s
  🟩 GCC11              Pass: 100%/7   | Total:  4h 21m | Avg: 37m 22s | Max: 44m 53s
  🟩 GCC12              Pass: 100%/4   | Total:  2h 27m | Avg: 36m 53s | Max: 42m 24s
  🟨 GCC13              Pass:  68%/16  | Total:  6h 20m | Avg: 23m 45s | Max: 38m 20s
  🟩 Intel2023.2.0      Pass: 100%/3   | Total:  2h 21m | Avg: 47m 19s | Max: 53m 14s
  🟩 MSVC14.16          Pass: 100%/1   | Total:  1h 06m | Avg:  1h 06m | Max:  1h 06m | Hits:  29%/1852  
  🟩 MSVC14.29          Pass: 100%/2   | Total:  2h 18m | Avg:  1h 09m | Max:  1h 12m | Hits:  29%/3704  
  🟨 MSVC14.39          Pass:  50%/2   | Total:  1h 31m | Avg: 45m 47s | Max:  1h 06m | Hits:  29%/1852  
  🟩 NVHPC24.7          Pass: 100%/4   | Total:  4h 29m | Avg:  1h 07m | Max:  1h 16m
🟨 cxx_family
  🟨 Clang              Pass:  95%/48  | Total:  1d 02h | Avg: 32m 44s | Max: 39m 24s
  🟨 GCC                Pass:  90%/51  | Total:  1d 02h | Avg: 30m 51s | Max: 44m 53s
  🟩 Intel              Pass: 100%/3   | Total:  2h 21m | Avg: 47m 19s | Max: 53m 14s
  🟨 MSVC               Pass:  80%/5   | Total:  4h 57m | Avg: 59m 27s | Max:  1h 12m | Hits:  29%/7408  
  🟩 NVHPC              Pass: 100%/4   | Total:  4h 29m | Avg:  1h 07m | Max:  1h 16m
🟨 jobs
  🟩 Build              Pass: 100%/103 | Total:  2d 14h | Avg: 36m 24s | Max:  1h 16m | Hits:  29%/7408  
  🟥 TestCPU            Pass:   0%/4   | Total: 47m 08s | Avg: 11m 47s | Max: 24m 38s
  🟥 TestGPU            Pass:   0%/4   | Total: 55m 57s | Avg: 13m 59s | Max: 18m 27s
🟨 std
  🟨 11                 Pass:  93%/30  | Total: 14h 25m | Avg: 28m 50s | Max: 59m 36s
  🟩 14                 Pass: 100%/29  | Total: 18h 12m | Avg: 37m 41s | Max:  1h 06m | Hits:  29%/3704  
  🟩 17                 Pass: 100%/27  | Total: 17h 44m | Avg: 39m 24s | Max:  1h 12m | Hits:  29%/1852  
  🟨 20                 Pass:  78%/23  | Total: 13h 07m | Avg: 34m 13s | Max:  1h 16m | Hits:  29%/1852  
🟨 cmake_options
  🟨 -DTHRUST_DISPATCH_TYPE=Force32bit Pass:  50%/2   | Total: 43m 35s | Avg: 21m 47s | Max: 31m 57s
🟨 gpu
  🟨 v100               Pass:  92%/111 | Total:  2d 16h | Avg: 34m 42s | Max:  1h 16m | Hits:  29%/7408  
🟩 sm
  🟩 60;70;80;90        Pass: 100%/3   | Total:  1h 57m | Avg: 39m 02s | Max: 44m 53s
  🟩 90a                Pass: 100%/4   | Total:  1h 28m | Avg: 22m 13s | Max: 25m 42s

🟩 cub: Pass: 100%/110 | Total: 3d 20h | Avg: 50m 14s | Max: 1h 13m | Hits: 66%/3028

🟩 cpu
  🟩 amd64              Pass: 100%/102 | Total:  3d 12h | Avg: 49m 54s | Max:  1h 13m | Hits:  66%/3028  
  🟩 arm64              Pass: 100%/8   | Total:  7h 15m | Avg: 54m 26s | Max: 57m 07s
🟩 ctk
  🟩 11.1               Pass: 100%/15  | Total: 11h 40m | Avg: 46m 41s | Max:  1h 01m | Hits:  66%/757   
  🟩 11.8               Pass: 100%/3   | Total:  3h 34m | Avg:  1h 11m | Max:  1h 13m
  🟩 12.5               Pass: 100%/4   | Total:  4h 05m | Avg:  1h 01m | Max:  1h 03m
  🟩 12.6               Pass: 100%/88  | Total:  3d 00h | Avg: 49m 36s | Max:  1h 04m | Hits:  66%/2271  
🟩 cudacxx
  🟩 ClangCUDA18        Pass: 100%/4   | Total:  3h 48m | Avg: 57m 03s | Max: 58m 29s
  🟩 nvcc11.1           Pass: 100%/15  | Total: 11h 40m | Avg: 46m 41s | Max:  1h 01m | Hits:  66%/757   
  🟩 nvcc11.8           Pass: 100%/3   | Total:  3h 34m | Avg:  1h 11m | Max:  1h 13m
  🟩 nvcc12.5           Pass: 100%/4   | Total:  4h 05m | Avg:  1h 01m | Max:  1h 03m
  🟩 nvcc12.6           Pass: 100%/84  | Total:  2d 20h | Avg: 49m 14s | Max:  1h 04m | Hits:  66%/2271  
🟩 cudacxx_family
  🟩 ClangCUDA          Pass: 100%/4   | Total:  3h 48m | Avg: 57m 03s | Max: 58m 29s
  🟩 nvcc               Pass: 100%/106 | Total:  3d 16h | Avg: 49m 58s | Max:  1h 13m | Hits:  66%/3028  
🟩 cxx
  🟩 Clang9             Pass: 100%/6   | Total:  4h 49m | Avg: 48m 13s | Max: 51m 36s
  🟩 Clang10            Pass: 100%/3   | Total:  2h 42m | Avg: 54m 04s | Max: 56m 42s
  🟩 Clang11            Pass: 100%/4   | Total:  3h 31m | Avg: 52m 46s | Max: 54m 17s
  🟩 Clang12            Pass: 100%/4   | Total:  3h 30m | Avg: 52m 39s | Max: 57m 19s
  🟩 Clang13            Pass: 100%/4   | Total:  3h 32m | Avg: 53m 10s | Max: 57m 39s
  🟩 Clang14            Pass: 100%/4   | Total:  3h 35m | Avg: 53m 51s | Max: 57m 11s
  🟩 Clang15            Pass: 100%/4   | Total:  3h 21m | Avg: 50m 16s | Max: 51m 44s
  🟩 Clang16            Pass: 100%/4   | Total:  3h 34m | Avg: 53m 38s | Max: 56m 20s
  🟩 Clang17            Pass: 100%/4   | Total:  3h 34m | Avg: 53m 41s | Max: 57m 07s
  🟩 Clang18            Pass: 100%/11  | Total:  8h 59m | Avg: 49m 02s | Max: 58m 29s
  🟩 GCC6               Pass: 100%/2   | Total:  1h 26m | Avg: 43m 08s | Max: 43m 12s
  🟩 GCC7               Pass: 100%/6   | Total:  4h 50m | Avg: 48m 25s | Max: 51m 34s
  🟩 GCC8               Pass: 100%/6   | Total:  4h 56m | Avg: 49m 28s | Max: 56m 13s
  🟩 GCC9               Pass: 100%/6   | Total:  5h 12m | Avg: 52m 07s | Max:  1h 01m
  🟩 GCC10              Pass: 100%/4   | Total:  3h 40m | Avg: 55m 05s | Max: 57m 21s
  🟩 GCC11              Pass: 100%/7   | Total:  7h 03m | Avg:  1h 00m | Max:  1h 13m
  🟩 GCC12              Pass: 100%/4   | Total:  3h 40m | Avg: 55m 14s | Max: 57m 59s
  🟩 GCC13              Pass: 100%/16  | Total:  9h 08m | Avg: 34m 16s | Max: 57m 07s
  🟩 Intel2023.2.0      Pass: 100%/3   | Total:  2h 55m | Avg: 58m 29s | Max:  1h 00m
  🟩 MSVC14.16          Pass: 100%/1   | Total: 52m 09s | Avg: 52m 09s | Max: 52m 09s | Hits:  66%/757   
  🟩 MSVC14.29          Pass: 100%/2   | Total:  1h 57m | Avg: 58m 54s | Max: 59m 44s | Hits:  66%/1514  
  🟩 MSVC14.39          Pass: 100%/1   | Total:  1h 04m | Avg:  1h 04m | Max:  1h 04m | Hits:  66%/757   
  🟩 NVHPC24.7          Pass: 100%/4   | Total:  4h 05m | Avg:  1h 01m | Max:  1h 03m
🟩 cxx_family
  🟩 Clang              Pass: 100%/48  | Total:  1d 17h | Avg: 51m 29s | Max: 58m 29s
  🟩 GCC                Pass: 100%/51  | Total:  1d 15h | Avg: 47m 02s | Max:  1h 13m
  🟩 Intel              Pass: 100%/3   | Total:  2h 55m | Avg: 58m 29s | Max:  1h 00m
  🟩 MSVC               Pass: 100%/4   | Total:  3h 54m | Avg: 58m 32s | Max:  1h 04m | Hits:  66%/3028  
  🟩 NVHPC              Pass: 100%/4   | Total:  4h 05m | Avg:  1h 01m | Max:  1h 03m
🟩 gpu
  🟩 v100               Pass: 100%/110 | Total:  3d 20h | Avg: 50m 14s | Max:  1h 13m | Hits:  66%/3028  
🟩 jobs
  🟩 Build              Pass: 100%/102 | Total:  3d 17h | Avg: 52m 32s | Max:  1h 13m | Hits:  66%/3028  
  🟩 DeviceLaunch       Pass: 100%/1   | Total: 18m 08s | Avg: 18m 08s | Max: 18m 08s
  🟩 GraphCapture       Pass: 100%/1   | Total: 15m 03s | Avg: 15m 03s | Max: 15m 03s
  🟩 HostLaunch         Pass: 100%/3   | Total:  1h 04m | Avg: 21m 21s | Max: 25m 33s
  🟩 TestGPU            Pass: 100%/3   | Total:  1h 09m | Avg: 23m 08s | Max: 26m 08s
🟩 sm
  🟩 60;70;80;90        Pass: 100%/3   | Total:  3h 34m | Avg:  1h 11m | Max:  1h 13m
  🟩 90a                Pass: 100%/4   | Total:  1h 32m | Avg: 23m 08s | Max: 24m 31s
🟩 std
  🟩 11                 Pass: 100%/30  | Total:  1d 00h | Avg: 48m 52s | Max:  1h 10m
  🟩 14                 Pass: 100%/29  | Total:  1d 01h | Avg: 53m 07s | Max:  1h 11m | Hits:  66%/1514  
  🟩 17                 Pass: 100%/27  | Total: 23h 54m | Avg: 53m 08s | Max:  1h 13m | Hits:  66%/757   
  🟩 20                 Pass: 100%/24  | Total: 18h 04m | Avg: 45m 10s | Max:  1h 04m | Hits:  66%/757

🟩 cccl_c_parallel: Pass: 100%/2 | Total: 9m 23s | Avg: 4m 41s | Max: 7m 05s

🟩 cpu
  🟩 amd64              Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 ctk
  🟩 12.6               Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 cudacxx
  🟩 nvcc12.6           Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 cudacxx_family
  🟩 nvcc               Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 cxx
  🟩 GCC13              Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 cxx_family
  🟩 GCC                Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 gpu
  🟩 v100               Pass: 100%/2   | Total:  9m 23s | Avg:  4m 41s | Max:  7m 05s
🟩 jobs
  🟩 Build              Pass: 100%/1   | Total:  2m 18s | Avg:  2m 18s | Max:  2m 18s
  🟩 Test               Pass: 100%/1   | Total:  7m 05s | Avg:  7m 05s | Max:  7m 05s

🟩 python: Pass: 100%/1 | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s

🟩 cpu
  🟩 amd64              Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 ctk
  🟩 12.6               Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 cudacxx
  🟩 nvcc12.6           Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 cudacxx_family
  🟩 nvcc               Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 cxx
  🟩 GCC13              Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 cxx_family
  🟩 GCC                Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 gpu
  🟩 v100               Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s
🟩 jobs
  🟩 Test               Pass: 100%/1   | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s

👃 Inspect Changes

Modifications in project?

	Project
	CCCL Infrastructure
	libcu++
	CUB
+/-	Thrust
	CUDA Experimental
	python
	CCCL C Parallel Library
	Catch2Helper

Modifications in project or dependencies?

	Project
	CCCL Infrastructure
	libcu++
+/-	CUB
+/-	Thrust
	CUDA Experimental
+/-	python
+/-	CCCL C Parallel Library
+/-	Catch2Helper

🏃‍ Runner counts (total jobs: 224)

#	Runner
185	`linux-amd64-cpu16`
16	`linux-arm64-cpu16`
14	`linux-amd64-gpu-v100-latest-1`
9	`windows-amd64-cpu16`

bernhardmgruber · 2024-12-02T18:08:01Z

Great, the changes so far break this:

  thrust::device_reference<T> ref1 = ...;
  thrust::device_reference<T> ref2 = ...;
thrust::swap(ref1, ref2);

It works with unqualified swap. However, the tests also fail for std::swap and cuda::std::swap. This needs more investigation.

bernhardmgruber · 2024-12-03T11:03:30Z

I discussed this with @miscco today and we agreed that the proper way to swap anything in CCCL is:

using ::cuda::std::swap;
swap(a, b);

This must be made to work for any types of a and b and do the right thing.

Furthermore, thrust::swap should be deprecated+removed, since swapping two values is a primitive operation that belongs to the standard library.

A proper solution requires us to fix: #2984

Fixes: NVIDIA#2984

Fixes: NVIDIA#2948

bernhardmgruber requested review from a team as code owners November 28, 2024 16:31

bernhardmgruber requested review from miscco and gonidelis November 28, 2024 16:31

bernhardmgruber commented Nov 28, 2024

View reviewed changes

bernhardmgruber force-pushed the fix_swap branch 2 times, most recently from 7a1abe9 to 81287ee Compare November 28, 2024 17:15

bernhardmgruber mentioned this pull request Nov 28, 2024

[BUG]: Suboptimal swap performance on universal vectors #2948

Open

1 task

bernhardmgruber force-pushed the fix_swap branch 2 times, most recently from e9cb9ac to 6b6050b Compare December 2, 2024 07:42

bernhardmgruber force-pushed the fix_swap branch from 6b6050b to db57e74 Compare December 3, 2024 13:50

bernhardmgruber requested review from a team as code owners December 3, 2024 13:50

bernhardmgruber requested a review from elstehle December 3, 2024 13:50

bernhardmgruber force-pushed the fix_swap branch 3 times, most recently from 52268d2 to f4566a9 Compare December 3, 2024 15:30

bernhardmgruber added 5 commits December 3, 2024 17:28

Fix swapping types from namespace std

6b82f20

Fixes: NVIDIA#2984

Also test std::pair

79b7b40

handle nvrtc

cd26b9a

Fix swap.pass.cpp for MSVC and NVRTC

125b329

Fix ambiguity

e8e10be

bernhardmgruber force-pushed the fix_swap branch from f4566a9 to 9cb163f Compare December 3, 2024 16:30

bernhardmgruber added 2 commits December 3, 2024 18:58

Fix nvrtc

da3bbe4

Replace thrust::swap by cuda::std::swap

751c208

Fixes: NVIDIA#2948

bernhardmgruber force-pushed the fix_swap branch from 9cb163f to 751c208 Compare December 3, 2024 17:59

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Replace thrust::swap by cuda::std::swap #2985

Replace thrust::swap by cuda::std::swap #2985

bernhardmgruber commented Nov 28, 2024

bernhardmgruber Nov 28, 2024 •

edited

Loading

bernhardmgruber Nov 28, 2024

miscco Nov 29, 2024

bernhardmgruber Nov 29, 2024

github-actions bot commented Dec 2, 2024

🟨 thrust: Pass: 92%/111 | Total: 2d 16h | Avg: 34m 42s | Max: 1h 16m | Hits: 29%/7408

🟩 cub: Pass: 100%/110 | Total: 3d 20h | Avg: 50m 14s | Max: 1h 13m | Hits: 66%/3028

🟩 cccl_c_parallel: Pass: 100%/2 | Total: 9m 23s | Avg: 4m 41s | Max: 7m 05s

🟩 python: Pass: 100%/1 | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s

👃 Inspect Changes

Modifications in project?

Modifications in project or dependencies?

🏃‍ Runner counts (total jobs: 224)

bernhardmgruber commented Dec 2, 2024

bernhardmgruber commented Dec 3, 2024

Replace thrust::swap by cuda::std::swap #2985

Are you sure you want to change the base?

Replace thrust::swap by cuda::std::swap #2985

Conversation

bernhardmgruber commented Nov 28, 2024

bernhardmgruber Nov 28, 2024 • edited Loading

Choose a reason for hiding this comment

bernhardmgruber Nov 28, 2024

Choose a reason for hiding this comment

miscco Nov 29, 2024

Choose a reason for hiding this comment

bernhardmgruber Nov 29, 2024

Choose a reason for hiding this comment

github-actions bot commented Dec 2, 2024

🟨 thrust: Pass: 92%/111 | Total: 2d 16h | Avg: 34m 42s | Max: 1h 16m | Hits: 29%/7408

🟩 cub: Pass: 100%/110 | Total: 3d 20h | Avg: 50m 14s | Max: 1h 13m | Hits: 66%/3028

🟩 cccl_c_parallel: Pass: 100%/2 | Total: 9m 23s | Avg: 4m 41s | Max: 7m 05s

🟩 python: Pass: 100%/1 | Total: 14m 13s | Avg: 14m 13s | Max: 14m 13s

👃 Inspect Changes

Modifications in project?

Modifications in project or dependencies?

🏃‍ Runner counts (total jobs: 224)

bernhardmgruber commented Dec 2, 2024

bernhardmgruber commented Dec 3, 2024

bernhardmgruber Nov 28, 2024 •

edited

Loading