IMPORTANT – GPUPERFAPI HAS MOVED TO GPUOPEN!
The GPUPerfAPI page has moved! Starting with release 2.20 GPUPerfAPI is now open-source and a part of the GPUOpen.com initiative.
Please go to GPUOpen.com to access GPUPerfAPI news, binaries and source code.
GPUPerfAPI is AMD’s library for accessing GPU performance counters on AMD Radeon graphics cards and APUs. It is used by GPU PerfStudio and AMD CodeXL and is also available to third party developers who wish to incorporate it within their own applications. GPUPerfAPI supports DirectX11, OpenGL, OpenGL ES, and OpenCL™ applications.
Release Notes – Version 2.17 (8/12/15)
- Add support for additional GPUs and APUs.
- Add OpenGL ES support for both Windows and Linux.
- DirectX 11/OpenGL: Add CSFlatVMInsts and CSFlatLDSInsts counters to measure flat instructions used for compute shaders on 2nd Generation GCN hardware or newer.
- DirectX11/OpenGL: Fix CSLDSInsts, CSVWriteInsts, CSVFetchInsts counter on 2nd Generation GCN hardware or newer.
- DirectX 11: Fixed a crash that could occur on Multi-GPU systems.
- OpenGL: rework the GPUTime implementation so that GPA no longer needs to stop/start any existing queries that the application may be using.
- OpenCL™: Add FlatVMInsts and FlatLDSInsts counters to measure flat instructions used for OpenCL kernels on 2nd Generation GCN hardware or newer.
- OpenCL™: Fix LDSInsts, VWriteInsts, VFetchInsts counter on 2nd Generation GCN hardware or newer.
- OpenCL™: Fix MemUnitBusy counter on 2nd Generation GCN hardware or newer.
- Fix a potential crash in GPA_SelectContext.
- Fix a bug in GPA_DisableCounter that could result in the counter not actually getting disabled.