iOS Tools, Profiling, and Debugging

客官°小女子只卖身不卖艺 2022-05-15 15:56 608阅读 0赞

https://developer.apple.com/documentation/metal/tools_profiling_and_debugging

Tools, Profiling, and Debugging

Build a Metal function library at the command line, or use Xcode to debug issues, and improve performance.

Framework

  • Metal

Topics

Files and Libraries

See information and steps to create Metal shaders and shader libraries.

About the Metal Shading Language Filename Extension

Use the .metal filename extension to gain access to Metal’s build, profile, and debug tools.

Building a Library with Metal’s Command-Line Tools

Use command-line tools to run the Metal compiler toolchain.

Frame Capture

View a summary of the Metal commands you submit in a single frame and drill down on specific areas for development and debugging purposes.

Metal GPU Capture

Use Xcode tools or Metal API to capture a runtime snapshot of your app’s Metal commands and state.

Labeling Metal Objects and Commands

Assign meaningful labels to your Metal objects and commands so you can easily identify them in the call list of a captured frame.

Seeing a Frame’s Render Passes with the Dependency Viewer

View your render passes as a flow chart and inspect individual resource dependencies to understand which commands wait on others to complete.

Debugging

Use a suite of tools to debug and develop Metal shaders.

Inspecting Vertices with the Geometry Viewer

Find problems with geometry by navigating a free-fly camera outside of your camera’s frustum and checking vertex values.

Resolving Shader Issues with the Shader Debugger

Step through shader execution with the ability to inspect variable values and update shader code in place.

Profiling and Metrics

Use informational panels to suggest the best use of your optimization effort and the biggest performance wins.

GPU Activity Monitors

Use Xcode or macOS tools to view a high-level summary of the GPU activity of your app or a Mac.

Optimizing Performance with the Shader Profiler

View the elapsed execution time of individual statements in your shader to understand where it spends the most time.

Viewing Performance Metrics with GPU Counters

Ensure that properties related to an encoder’s rendering are within the desired range.

Viewing Pipeline Statistics of a Draw

See relative percentages of where a given draw call spent its time across the GPU architecture.

发表评论

表情:
评论列表 (有 0 条评论,608人围观)

还没有评论,来说两句吧...

相关阅读

    相关 NVIDIA Profiling Tools

    前言:目前在用的监控GPU的工具有很多,比较常用的有nvprof、nvvp和Nsight。对这三种工具的使用不是很熟练,尚在学习当中。下面针对前两种自己的使用情况进行记录,如有