CodeXL
Original author(s) | Advanced Micro Devices |
---|---|
Stable release |
2.2
/ August 8, 2016 |
Operating system | Linux, Microsoft Windows |
Type | GPU debugging, GPU & CPU profiling, Static Kernel Analysis |
License | MIT License |
Website |
github |
CodeXL (formerly AMD CodeXL) is an open source software development tool suite that includes a GPU debugger, a GPU profiler, a CPU profiler, Graphics frame analyzer and a static shader/kernel analyzer.
CodeXL has been mainly developed by AMD. With version 2.0 CodeXL was made part of GPUOpen and is free and open-source software subject to the requirements of the MIT License. no longer branded as an AMD product.[1]
Features
GPU debugger
CodeXL’s GPU debugger allows engineers to debug OpenGL and OpenCL API calls and runtime objects, and debug OpenCL kernels: set breakpoints, step through source code in real-time, view all variables across different GPU cores during kernel execution, identify logic and memory errors, reduce memory transaction overhead, visualize OpenCL/OpenGL buffers and images and OpenGL textures as pictures or as spreadsheet data, and in this way to improve general software quality and optimize its performance.
GPU profiler
- CodeXL’s GPU profiler collects and visualizes hardware performance counters data, application trace, kernel occupancy, and offers hotspot analysis for AMD GPUs and APUs. The profiler gathers data from the OpenCL runtime, and from the GPU/APU itself during the execution of the kernels, and can be used to discover performance bottlenecks and optimize kernel execution.
CPU profiler
- CodeXL’s CPU profiling suite can be used to identify, investigate and improve the performance of applications, drivers and system software on AMD CPUs. CodeXL’s CPU profiler uses a statistical sampling based approach with various profiling techniques and measures: Time-Based Profiling (TBP), Event-Based Profiling (EBP), Instruction-Based Sampling (IBS) and CPU hardware performance monitors. The CodeXL CPU profiler replaces AMD CodeAnalyst
Static Shader/Kernel Analyzer
- CodeXL's static kernel analyzer allows engineers to compile, analyze and disassemble the code of DirectX, OpenGL and Vulkan shaders and OpenCL kernels. The tool provides performance estimation for each shader/kernel on different kinds of AMD chips, without actually executing the code. This way, the tool assists in fine-tuning to achieve best performance on AMD GPUs.
HSA profiler
- With version 2.0 the CodeXL HSA Profiler and HSAIL Kernel Debugger support the Boltzmann initiative driver, on GCN 1.2 hardware (i.e. Radeon R9 Fury, Fury X and Fury Nano, and "Carrizo" APUs.[1]
Graphics Frame Analyzer
The Frame Analyzer can collect and display a frame timeline for applications that use Microsoft DirectX 12 or Vulkan. The analyzer's view lists each API call that was made on the CPU side and its corresponding command that executed on the GPU side in a inter-linked and unified timeline view, as well as aggregated statistics for user-selected specific time fragment – cumulative time for each type of API, number of calls, 20 longest calls and more.
Supported platforms
- CodeXL officially supports both Linux and Microsoft Windows operating systems. On Windows, CodeXL is available both as a standalone application and as a fully integrated Microsoft Visual Studio extension.
CodeXL was successfully used to debug Bullet.[2]
Versions and availability
- AMD CodeXL 1.0, released on Dec-4th, 2012.[3]
- AMD CodeXL 1.1, released on Feb-27th, 2013.[4]
- AMD CodeXL 1.2, released on Jul-9th, 2013.[5]
- AMD CodeXL 1.3, released on Nov-11th, 2013.[6]
- AMD CodeXL 1.4, released on Apr-11th, 2014.[7]
- AMD CodeXL 1.5, released on Sep-22nd, 2014.[8]
- AMD CodeXL 1.6, released on Dec-16th, 2014.[9]
- AMD CodeXL 1.7, released on Apr-28th. 2015.[10]
- AMD CodeXL 1.8, released on Aug-20th. 2015.[11]
- CodeXL 2.0, released on Apr-19, 2016.[1]
- CodeXL 2.1, released on May-31, 2016.[12]
- CodeXL 2.2, released on Aug-08, 2016.[13]
The latest version of CodeXL is freely available for download on the CodeXL GitHub releases page: https://github.com/GPUOpen-Tools/CodeXL/releases
GPU PerfStudio
GPU PerfStudio is AMD's performance and debugging tool for graphics applications. It was initially developed to support Direct3D and OpenGL on Microsoft Windows only and was ported to Linux during 2013 and is available for Linux since the end of Q1 2014. The suite of tools is considered useful when developing games for Steam Linux and especially useful when optimizing games for AMD GPUs. GPU PerfStudio has an integrated Frame Profiles, Frame Debugger and API Trace with CPU timing information.
GPU PerfStudio supports Direct3D 10, Direct3D 10.1, Direct3D 11 and OpenGL 4.2. It is described as a lightweight, no installer, no change to your game, drag and drop suite of GPU tools. It can be run from a USB drive, there is no need for Visual Studio integration, it runs with game executables, it does not require a special driver or a new compilation.
GPU PerfStudio 3.6 supports Direct3D 12 on Windows 10 and Vulkan on Windows and Linux.[14]
GPU PerfStudio is available for Linux and Microsoft Windows.
At the AMD Developer Summit (APU) in November 2013 Gordon Selley presented GPU PerfStudio 2.[15]
At the SteamDevDays in February 2014, Tony Hosier and Gordon Selley presented GPU PerfStudio 2 in a 43 minutes video: http:
Frame Debugger
Frame Profiles
Shader Debugger
API Trace
GPUPerfAPI
GPUPerfAPI is AMD’s library for accessing GPU performance counters on AMD Radeon graphics cards and APUs. It is used by GPU PerfStudio and CodeXL and is also available to third party developers who wish to incorporate it within their own applications. GPUPerfAPI supports DirectX11, OpenGL, and OpenCL applications. GPUPerfAPI is available for Linux and Microsoft Windows.
See also
- Old website: developer
.amd .com /tools-and-sdks /opencl-zone /opencl-tools-sdks /codexl / - List of performance analysis tools
- CodeAnalyst
- Heterogeneous System Architecture (HSA)
References
- 1 2 3 AMD GPUOpen (2016-04-19). "CodeXL 2.0 made open-source".
- ↑ "Bullet 3 OpenCL Rigid Body Simulation". 2013-11-21.
- ↑ http://developer.amd.com/community/blog/2012/12/04/continue-your-heterogeneous-computing-hc-development-with-amd-codexl-1-0-final-release-now-available/ AMD Developer Central CodeXL 1.0 release blog post
- ↑ http://developer.amd.com/community/blog/2013/02/27/amd-codexl-version-1-1-released-to-developer-amd-com/ AMD Developer Central CodeXL 1.1 release blog post
- ↑ http://developer.amd.com/community/blog/2013/07/09/latest-versions-of-amd-codexl-bolt-and-amd-app-sdk-now-available/ AMD Developer Central CodeXL 1.2 release blog post
- ↑ http://developer.amd.com/community/blog/2013/11/08/codexl-1-3-released/ AMD Developer Central CodeXL 1.3 release blog post
- ↑ http://developer.amd.com/community/blog/2014/04/15/codexl-1-4-released/ AMD Developer Central CodeXL 1.4 release blog post
- ↑ http://developer.amd.com/community/blog/2014/09/22/whats-new-codexl-1-5/ AMD Developer Central "What's New in CodeXL 1.5?" blog post
- ↑ http://developer.amd.com/community/blog/2014/12/16/whats-new-amd-codexl-1-6/ AMD Developer Central "What's New in CodeXL 1.6?" blog post
- ↑ http://developer.amd.com/community/blog/2015/04/28/amd-codexl-1-7/ AMD Developer Central "AMD CodeXL 1.7 is here!" blog post
- ↑ http://developer.amd.com/community/blog/2015/08/20/amd-codexl-1-8-released/ AMD Developer Central "CodeXL 1.8 Release Loaded With Features" blog post
- ↑ "CodeXL 2.1 is out and Searing hot with Vulkan - GPUOpen". 2016-05-31. Retrieved 2016-08-15.
- ↑ "Blazing CodeXL 2.2 is here! - GPUOpen". 2016-08-08. Retrieved 2016-08-15.
- ↑ "GPUPerfStudio Vulkan Tools". AMD.
- ↑ "Optimizing Game Development using AMD GPU PerfStudio 2". 2013-11-21.
External links
- "CodeXL | AMD". developer.amd.com. Retrieved 2014-01-24.
- "AMD Announces CodeXL At AFDS | SemiAccurate". semiaccurate.com. Retrieved 2014-01-24.
- "AMD Code XL Public Beta Available for Download | Dr Dobb's". drdobbs.com. Retrieved 2014-01-24.
- Potsdam University’s Software Profiling Course slides about using AMD CodeXL
- CodeXL optimizes encryption using OpenCL: http://www.diva-portal.org/smash/get/diva2:796645/FULLTEXT01.pdf
- CodeXL on GitHub