Nsight Compute 2021.1 - New Features

Remote source resolution can now use the IP address, in addition to the hostname, to find the necessary SSH target.

Added support for the existing command line options for kernel filtering while importing data from an existing report file using --import.
Option -k is not considered as deprecated option --kernel-regex anymore.

Fixed failure to profile kernels from applications that use the CUDA graphics interop APIs to share semaphores.
Fixed wavefront metric in the L1TEX table for writes to shared memory on GA10x chips.
Fixed an issue resulting in incomplete data collection for the interactive profile activity after switching from single-pass mode to collecting multiple passes in the same session.
Fixed values shown in the mimimap of the Source page when all functions are collapsed.
Fixed an issue causing names set by the NVTX naming APIs of one application to be applied to all subsequent sessions of the same instance of NVIDIA Nsight Compute.
Fixed behavior of horizontal scroll bars when clicking in the source views on the Source page.
Fixed appearance of multi-line entries in column chooser on the Source page.
Fixed enablement state of the reset button on the Connection dialog.
Fixed potential crash of NVIDIA Nsight Compute when windows size becomes small while being on the Source page.
Fixed potential crash of NVIDIA Nsight Compute when relative paths for section/rules files could not be found.
Fixed potential crash of NVIDIA Nsight Compute after removing baselines.

Added support for the CUDA toolkit 11.3, including
- Nsight Compute Resource Tracking for new CUDA 11.3's features:
Added support for the OptiX 7 API.
GpuArch enumeration values used for filtering in section files were renamed from architecture names to compute capabilities.
NVTX states can now be accessed via the NvRules API.
Added a rule for the Occupancy section.

Added support for new CUDA asynchronous allocator attributes in the Memory Pools resources view.
Added a topology chart and link properties table in the NVLink section.
The selected metric column is scrolled into view on the Source page when a new metric is selected.
Users can choose the Source heatmap color scale in the Options dialog.

Added file-based application replay as the new default application replay mode. File-based replay uses a temporary file for keeping replay data, instead of allocating them in memory. This keeps the required memory footprint close to constant, independent of the number of profiled kernels. Users can switch between buffer modes using the --app-replay-buffer option.
CLI output now shows NVTX color and message information.
--kernel-regex and --kernel-regex-base options are deprecated and replaced by --kernel-name and --kernel-name-base, respectively.
All options which support regex need to provide regex: as a prefix before an argument to match per the regex, e.g <option> <regex:expression>

Fixed that baselines were not updated properly on the Comments page.
Fixed that NVTX ranges named using their payloads can be used in NVTX filtering expressions.
Fixed crashes in MacOSX hosts when terminating the target application.
The NVLINK(nvl*) metrics are now added back.

For a complete overview of all NVIDIA Nsight™ Compute features and access to resources, please visit the main Nsight™ Compute page.

NVIDIA® Nsight™ Compute 2021.1 is available for download under the NVIDIA Registered Developer Program.

Videos