• CUDA 10.0 adds support for the Turing architecture (compute_75 and sm_75).

• CUDA 10.0 adds support for new programming constructs called CUDA Graphs, a new asynchronous task-graph programming model that enables more efficient launch and execution. See the API documentation for more information.

• Warp matrix functions now support additional matrix shapes 32x8x16 and 8x32x16. Warp matrix functions also include the ability (experimental in CUDA 10.0) to perform sub-byte operations (4-bit unsigned, 4-bit signed and 1-bit) using the Tensor Cores.

• Added support for CUDA-Vulkan and CUDA-DX12 interoperability APIs.

• Added support for a new instruction nanosleep that suspends a thread for a specified duration.

• Added 6.3 version of the Parallel Thread Execution instruction set architecture (ISA). For more details on new (sm_75 target, wmma, nanosleep, FP16 atomics) and deprecated instructions, see this section in the PTX documentation.

• Starting with CUDA 10.0, the CUDA runtime is compatible with specific older NVIDIA drivers. A new package called “cuda-compat-<version>” is included in the toolkit installer packages. For more information on compatibility, see the section in the Best Practices Guide.

• The following new operating systems are supported by CUDA. See the System Requirements section in the NVIDIA CUDA Installation Guide for Linux for a full list of supported operating systems.

  • Ubuntu 18.04 LTS*

  • Ubuntu 14.04 LTS

  • SUSE SLES 15

  • OpenSUSE Leap 15

• Added support for peer-to-peer (P2P) with CUDA on Windows (WDDM 2.0+ only).

• Added a new CUDA sample to demonstrate multi-device cooperative group APIs.

• CUDA samples are now also available on GitHub: https://github.com/NVIDIA/cuda-samples.

• Added APIs to retrieve the LUID of CUDA devices (cuDeviceGetLuid).

• Added cudaLimitMaxL2FetchGranularity in the device management APIs (cudaDeviceGetLimit) to set the maximum fetch granularity of L2 (in Bytes).

• The cudaDeviceProp struct now includes the device UUID.

• Added support for synchronization across multiple devices with Cooperative Groups (cuLaunchCooperativeKernelMultiDevice) on Windows in TCC mode.
• Added the ability to lock clocks in nvidia-smi and NVML (nvmlDeviceSetGpuLockedClocks and nvmlDeviceResetGpuLockedClocks APIs). The following commands can be used in nvidia-smi:

  • $ nvidia-smi -lgc/--lock-gpu-clock <minGpuClock, maxGpuClock>

  • $ nvidia-smi -rgc/--reset-gpu-clock

This release of the toolkit includes optimized libraries for Turing architecture and performance-tuned across single and multi-GPU environments. Also this release introduces a new library, nvJPEG, for GPU accelerated hybrid JPEG decoding.

• Fixed a bug in conjugateGradientMultiDeviceCG sample where it would fail to run on some Volta GPUs.

• Fixed an issue with the GPU memory resource consumption of the MPS server on Volta V100.

• Fixed an issue that caused a 6% performance difference in FAHBench (DP_average test) with OpenCL.

• Fixed a memory allocation issue in the CUDA runtime (cudart) library that was reported as an error by AddressSanitizer.

• Fixed an issue when NVML would not show all active NVLinks (via nvidia-smi nvlink -s) after a reboot of the system.

• Fixed an issue with the CUDA sample reduction (under 6_Advanced/reduction), where the sample would fail when reducing less than 16384 elements.

• Fixed a bug in nvprof where --print-nvlink-topology would crash on POWER 8 systems.

• Fixed an issue with nvprof where system memory related metrics would fail to be collected on Pascal GPUs.

• Fixed an issue where nvprof would result in a crash when no kernel is specified in an nvtx range.

• Fixed an issue where the --profile-child-processes option with nvprof would result in no profile data being collected.

• Fixed a NullPointerException in some OpenMP codes when running the Visual Profiler.

• Fixed an issue with Nsight Eclipse Edition that was preventing auto-complete from working with CUDA.

• Fixed an issue in nvvp where the CPU profiling button should be disabled if tracing is enabled. This issue would result in a failure to generate timelines.

• Fixed an issue in nvvp to show the correct topology and link information for GPUs in a POWER 9 system.

• Fixed an issue where profiling of CUDA applications would result in invalid profiling results on Azure GPU instances.

• Fixed an issue in ptxas where the vmax instruction in specific cases (vmax2.s32.s32.s32) would result in incorrect results.

• Fixed an issue in nvcc where an error would be reported in some cases in relaxed_constexpr mode for references to host variables in device code.

• Fixed an issue in the visual profiler where the CPU profiling does not default correctly for PGI.

• Fixed an issue where cuda-gdb would assert when running some OpenACC codes with break_on_launch set.

• Fixed an issue where nvprof was incorrectly calculating the percentage overhead for NVLink (see nvlink_overhead_* metrics).

• Fixed an issue which prevented tracing and profiling (event and metric collection) for multidevice cooperative kernels, that is, kernels launched by using the API functions cudaLaunchCooperativeKernelMultiDevice or cuLaunchCooperativeKernelMultiDevice.

• Fixed an issue in nvcc where an internal compiler error is encountered when generating debug info for code containing a pointer to data member or member function.

• Fixed an issue where nvcc would return an error for a template that is defined inside an unnamed namespace.
• Fixed an issue in the CUDA compiler where using __shfl_xor_sync() in some cases would cause kernels to hang on Volta GPUs.

• Fixed the following performance issues with cuBLAS:
  • SGEMM (FP32) on V100/Ubuntu 16.04,
  • INT8 GEMM on P4 for small input m and n sizes, and
  • Batched CGEMM performance on Tesla V100.
• Fixed performance issue in cuRAND for Philox one-at-a-time random number generation.
• Improved cuBLAS performance and heuristics selection for large K matrices.
• Added METIS reordering in the cuSOLVER sample for sparse linear solver.
• Fixed an issue with cuSOLVER QR factorization (geqrf) that sometimes returned R factors containing a non numeric value.
• Fixed overflow and underflow issues in cuSOLVER householder reflection and givens rotations.
• Fixed an issue with cuBLAS SGEMM algorithm 102 that occasionally returned non-deterministic results.
• Fixed an issue in cuSOLVER cusolverDnDsyevj that returned incorrectly sorted eigenvalues.
• Improved dense SVD performance in cuSOLVER for tall skinny sizes by adding QR when m>n.
• Improved performance of dense QR with domino scheme in cuSOLVER.
• Fixed an issue in NPP nppiCountInRange routine that returned NPP_RANGE_ERROR when called from multiple threads.
• Fixed missing check for uplo parameter in cuSOLVER routines cusolverDn<X>potrf and cusolverDn<X>potrf_bufferSize that resulted in incorrect return code.
• Fixed an issue with cuBLAS batched GEMM where it selected wrong kernels for bandwidth bound GEMMs.

• Fixed an issue in Thrust where merge_by_key uses the wrong element type in the default comparator.

• CUDA graphs with host nodes do not work under MPS with pre-Volta GPUs. Attempting to launch a graph in this configuration will result in cuLaunchGraph() potentially executing a portion of the graph and returning an error.
• On Red Hat Enterprise Linux and Fedora based systems, installation of the CUDA toolkit without installing the NVIDIA OpenGL libraries is not supported. See https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#advanced-setup for this option. This issue will be fixed in the next release of CUDA.

• Using the CUDA Graphs stream capture APIs with CUDA libraries (e.g. cuBLAS) may result in errors. This issue will be fixed in a patch update to CUDA 10.0.

• The nvJPEG library is currently only available on Linux x86-64 and is not supported on Linux POWER, Windows or Mac platforms. Support for these additional platforms may be added in a future release of CUDA.

• On ppc64le (Ubuntu 18.04 or RHEL 7.5) systems, the persistence daemon may not be started automatically after CUDA 10.0 is installed. The nvidia-persistenced.service (typically found under /lib/systemd/system/) file should be changed to as follows:

  [Unit] 
  Description=NVIDIA Persistence Daemon 
  After=syslog.target 
   
  [Service] 
  Type=forking
  PIDFile=/var/run/nvidia-persistenced/nvidia-persistenced.pid
  Restart=always
  ExecStart=/usr/sbin/nvidia-persistenced --verbose
  ExecStopPost=/bin/rm -rf /var/run/nvidia-persistenced/*
  TimeoutSec=300 
   
  [Install] 
  WantedBy=multi-user.target

  Check the status of the persistence daemon with the following command:

  $ systemctl status nvidia-persistenced

  If it is not active, run the following command:

  $ sudo systemctl enable nvidia-persistenced

• cuda-memcheck may report uninitialized reads in CUB’s radix sort in the CUDA sample smokeParticles.
• Some graphic CUDA samples may terminate with an error when switching to the desktop using Windows + D key combination.

• For known issues in cuda-memcheck, see the Known Issues section in the cuda-memcheck documentation.

• Fixed a crash in nvprune when dealing with ELF objects with SHT_NOBITS sections.

CUDA Tegra Driver

• Support for Drive AGX Pegasus added on QNX.

• Support for CUDA-Vulkan interoperability added on Nvidia Jetson and Android platforms.

• Support for Ubuntu 18.04 LTS added.

• CUDA kernel launch latency improved on QNX.

• Starting with CUDA 10.0, support for 14.04 LTS has been dropped.

CUDA Compiler

• Support for GCC 6.4 added.
• Support for Clang 5.0 & libc++ for Android added.

CUDA Samples

• A new Tegra sample for Unified Memory is added. This sample demonstrates significant speedup while using attached hints.

CUDA Tegra Driver

• Support for 32-bit applications will no longer be available.
• Support for cudaHostRegister on QNX is not available, due to lack of compatible QNX kernels.
• The cudaDeviceGetAttribute methods returns false for the attribute cudaDevAttrHostNativeAtomicSupported for T194, because system wide atomics are currently not supported (for Nvidia Drive and Jetson platforms).
• Memory clock rate of dGPU is not reported correctly by cudaGetDeviceProperties/cudaDeviceGetAttribute.

CUDA-GDB

• Linux : The /dev/nvhost-dbg-gpu and /dev/nvhost-prof-gpu nodes default to 0660 permissions for security reasons. It causes the following Dev Tools to be broken out of the box:
  • Tegra Graphics Debugger,
  • PerfWorks,
  • nvprof,
  • cuda-memcheck,
  • cuda-gdb.
  Note: Before using any of the Dev Tools listed above, please log into the board and run this command:

  sudo chmod a+rw /dev/nvhost-dbg-gpu /dev/nvhost-prof-gpu.

• Linux: The set cuda memcheck on command in cuda-gdb does not have any effect.
• QNX: ntoaarch64-gdb and cuda-qnx-gdb may hang when executing the runcommand.

NVPROF

• PC sampling is not supported.
• The Volta dGPU (GV100) is not supported.
• This release does not support HWPM context switching. That means that counters that are collected through the HWPM counter provider are available at the device level only at this time. This will be fixed in a future release.

General CUDA

• During initialization, CUDA driver reserves a large chunk of CPU virtual address for its internal memory management. On QNX, this CPU virtual address reservation might take a considerable amount of time on systems with large physical memory. Because of this behavior, CUDA initialization might take more time on QNX Xavier as compared to earlier releases.
• CUDA allocators cannot make a single allocation greater than 4 GB on Tegra SoC memory. This applies to all allocations on Tegra iGPU and zero copy memory allocations on Tegra dGPU. Application needs to make multiple allocations and aggregate them to create a large allocation.