Summary
- Introduce constructor for multi-GPU support. (details)
- Fix typo (details)
- Add test (details)
- Fix typo. (details)
- Explicitly check for valid device id (details)
- Set the device id in cuda_kernel_arch (details)
- Check for default device (details)
- Check that device associated with stream matches requested device (details)
- Remove extra constructor (details)
- Address reviewer comments (details)
- m_cudaDev isn't static anymore (details)
- Set the device id explicitly for CUDA API calls in impl_initialize (details)
- Fixup test math functions ulp should double -> int (details)
- Drop DualView converting copy assignment operator (details)
- Don't use rocm-docker for clang-format (details)
- Diable HIP CI (details)
- Do not negate the dependent true traits helper (details)
- Add missing gfx940 (details)
- Add Impl::always_false type-dendent false trait (details)
- Per review prefer always_false<Arg>::value to is_void_v<Arg> (details)
- Improve "no copy mechanism" exception message (details)
- Add a unit test for new deep_copy exception msg (details)
- Add missing include sstream (details)
- src->source, dst->destination (details)
- Workaround for ROCm 6.0 failing to compile with AVX2 SIMD support (details)
- SYCL: Force inlining of Kokkos::printf (#6650) (details)
- Improve handling of printf in OMPT on Intel GPUs (details)
- OpenMP: Use `omp_get_nested` for older gcc versions (#6685) (details)
- Drop unnecessary guarding for a tool library being loaded in ProfilingSection (details)
- Drop unnecessary header include in Kokkos_Profiling_ProfileSection.hpp (details)
Summary
- Introduce constructor for multi-GPU support. (details)
- Fix typo (details)
- Add test (details)
- Fix typo. (details)
- Explicitly check for valid device id (details)
- Set the device id in cuda_kernel_arch (details)
- Check for default device (details)
- Check that device associated with stream matches requested device (details)
- Remove extra constructor (details)
- Address reviewer comments (details)
- m_cudaDev isn't static anymore (details)
- Set the device id explicitly for CUDA API calls in impl_initialize (details)
- Fixup test math functions ulp should double -> int (details)
- Drop DualView converting copy assignment operator (details)
- Don't use rocm-docker for clang-format (details)
- Diable HIP CI (details)
- Do not negate the dependent true traits helper (details)
- Add missing gfx940 (details)
- Add Impl::always_false type-dendent false trait (details)
- Per review prefer always_false<Arg>::value to is_void_v<Arg> (details)
- Improve "no copy mechanism" exception message (details)
- Add a unit test for new deep_copy exception msg (details)
- Add missing include sstream (details)
- src->source, dst->destination (details)
- Workaround for ROCm 6.0 failing to compile with AVX2 SIMD support (details)
- SYCL: Force inlining of Kokkos::printf (#6650) (details)
- Improve handling of printf in OMPT on Intel GPUs (details)
- OpenMP: Use `omp_get_nested` for older gcc versions (#6685) (details)
- Drop unnecessary guarding for a tool library being loaded in ProfilingSection (details)
- Drop unnecessary header include in Kokkos_Profiling_ProfileSection.hpp (details)
Summary
- Introduce constructor for multi-GPU support. (details)
- Fix typo (details)
- Add test (details)
- Fix typo. (details)
- Explicitly check for valid device id (details)
- Set the device id in cuda_kernel_arch (details)
- Check for default device (details)
- Check that device associated with stream matches requested device (details)
- Remove extra constructor (details)
- Address reviewer comments (details)
- m_cudaDev isn't static anymore (details)
- Set the device id explicitly for CUDA API calls in impl_initialize (details)
- Fixup test math functions ulp should double -> int (details)
- Drop DualView converting copy assignment operator (details)
- Don't use rocm-docker for clang-format (details)
- Diable HIP CI (details)
- Do not negate the dependent true traits helper (details)
- Add missing gfx940 (details)
- Add Impl::always_false type-dendent false trait (details)
- Per review prefer always_false<Arg>::value to is_void_v<Arg> (details)
- Improve "no copy mechanism" exception message (details)
- Add a unit test for new deep_copy exception msg (details)
- Add missing include sstream (details)
- src->source, dst->destination (details)
- Workaround for ROCm 6.0 failing to compile with AVX2 SIMD support (details)
- SYCL: Force inlining of Kokkos::printf (#6650) (details)
- Improve handling of printf in OMPT on Intel GPUs (details)
- OpenMP: Use `omp_get_nested` for older gcc versions (#6685) (details)
- Drop unnecessary guarding for a tool library being loaded in ProfilingSection (details)
- Drop unnecessary header include in Kokkos_Profiling_ProfileSection.hpp (details)
Summary
- Introduce constructor for multi-GPU support. (details)
- Fix typo (details)
- Add test (details)
- Fix typo. (details)
- Explicitly check for valid device id (details)
- Set the device id in cuda_kernel_arch (details)
- Check for default device (details)
- Check that device associated with stream matches requested device (details)
- Remove extra constructor (details)
- Address reviewer comments (details)
- m_cudaDev isn't static anymore (details)
- Set the device id explicitly for CUDA API calls in impl_initialize (details)
- Fixup test math functions ulp should double -> int (details)
- Drop DualView converting copy assignment operator (details)
- Don't use rocm-docker for clang-format (details)
- Diable HIP CI (details)
- Do not negate the dependent true traits helper (details)
- Add missing gfx940 (details)
- Add Impl::always_false type-dendent false trait (details)
- Per review prefer always_false<Arg>::value to is_void_v<Arg> (details)
- Improve "no copy mechanism" exception message (details)
- Add a unit test for new deep_copy exception msg (details)
- Add missing include sstream (details)
- src->source, dst->destination (details)
- Workaround for ROCm 6.0 failing to compile with AVX2 SIMD support (details)
- SYCL: Force inlining of Kokkos::printf (#6650) (details)
- Improve handling of printf in OMPT on Intel GPUs (details)
- OpenMP: Use `omp_get_nested` for older gcc versions (#6685) (details)
- Drop unnecessary guarding for a tool library being loaded in ProfilingSection (details)
- Drop unnecessary header include in Kokkos_Profiling_ProfileSection.hpp (details)
Summary
- Introduce constructor for multi-GPU support. (details)
- Fix typo (details)
- Add test (details)
- Fix typo. (details)
- Explicitly check for valid device id (details)
- Set the device id in cuda_kernel_arch (details)
- Check for default device (details)
- Check that device associated with stream matches requested device (details)
- Remove extra constructor (details)
- Address reviewer comments (details)
- m_cudaDev isn't static anymore (details)
- Set the device id explicitly for CUDA API calls in impl_initialize (details)
- Fixup test math functions ulp should double -> int (details)
- Drop DualView converting copy assignment operator (details)
- Don't use rocm-docker for clang-format (details)
- Diable HIP CI (details)
- Do not negate the dependent true traits helper (details)
- Add missing gfx940 (details)
- Add Impl::always_false type-dendent false trait (details)
- Per review prefer always_false<Arg>::value to is_void_v<Arg> (details)
- Improve "no copy mechanism" exception message (details)
- Add a unit test for new deep_copy exception msg (details)
- Add missing include sstream (details)
- src->source, dst->destination (details)
- Workaround for ROCm 6.0 failing to compile with AVX2 SIMD support (details)
- SYCL: Force inlining of Kokkos::printf (#6650) (details)
- Improve handling of printf in OMPT on Intel GPUs (details)
- OpenMP: Use `omp_get_nested` for older gcc versions (#6685) (details)
- Drop unnecessary guarding for a tool library being loaded in ProfilingSection (details)
- Drop unnecessary header include in Kokkos_Profiling_ProfileSection.hpp (details)
Summary
- Introduce constructor for multi-GPU support. (details)
- Fix typo (details)
- Add test (details)
- Fix typo. (details)
- Explicitly check for valid device id (details)
- Set the device id in cuda_kernel_arch (details)
- Check for default device (details)
- Check that device associated with stream matches requested device (details)
- Remove extra constructor (details)
- Address reviewer comments (details)
- m_cudaDev isn't static anymore (details)
- Set the device id explicitly for CUDA API calls in impl_initialize (details)
- Fixup test math functions ulp should double -> int (details)
- Drop DualView converting copy assignment operator (details)
- Don't use rocm-docker for clang-format (details)
- Diable HIP CI (details)
- Do not negate the dependent true traits helper (details)
- Add missing gfx940 (details)
- Add Impl::always_false type-dendent false trait (details)
- Per review prefer always_false<Arg>::value to is_void_v<Arg> (details)
- Improve "no copy mechanism" exception message (details)
- Add a unit test for new deep_copy exception msg (details)
- Add missing include sstream (details)
- src->source, dst->destination (details)
- Workaround for ROCm 6.0 failing to compile with AVX2 SIMD support (details)
- SYCL: Force inlining of Kokkos::printf (#6650) (details)
- Improve handling of printf in OMPT on Intel GPUs (details)
- OpenMP: Use `omp_get_nested` for older gcc versions (#6685) (details)
- Drop unnecessary guarding for a tool library being loaded in ProfilingSection (details)
- Drop unnecessary header include in Kokkos_Profiling_ProfileSection.hpp (details)