简体繁体 English

使用NVIDIA GPU在Windows上对OpenCL应用程序进行性能分析

[英]Profiling OpenCL application on Windows with NVIDIA GPU

原文 2016-08-10 11:40:16 6 1 profiling/ opencl/ nvidia/ vtune/ nvvp

can you help me? 你能帮助我吗？ I'm developing OpenCL application on windows 7 x64. 我正在Windows 7 x64上开发OpenCL应用程序。 Hardware is Intel Core i5, NVIDIA GTX 770. OpenCL uses NVIDIA for acceleration. 硬件是Intel Core i5，NVIDIA GTX770。OpenCL使用NVIDIA进行加速。

If I'm trying to use Intel VTune Amplifier XE 2015 my application hangs on the end of profiling and doesnt return any report. 如果我尝试使用Intel VTune Amplifier XE 2015，我的应用程序将在分析结束时挂起，并且不返回任何报告。 Documentation recommends to disable all "Cn(ACPI Cn) report to OS" BIOS options. 文档建议禁用所有“向OS发送的Cn（ACPI Cn）报告” BIOS选项。 But it's absolutely unclear how to change my application code to avoid this hanging. 但是，绝对不清楚如何更改我的应用程序代码以避免这种挂起。

Also I've tried to utilize NVVP using this manual https://github.com/UoB-HPC/UoB-HPC.github.io/blob/master/_posts/2015-05-27-nvvp-import-opencl.md . 我也尝试使用本手册https://github.com/UoB-HPC/UoB-HPC.github.io/blob/master/_posts/2015-05-27-nvvp-import-opencl.md来利用NVVP 。 Unfortunately profiler generates an error: Unable to profile application. 不幸的是，探查器会生成错误：无法探查应用程序。 "Another CUDA tool (profiler, debugger, memcheck) is currently running which prevents the application from being profiled". “当前正在运行另一个CUDA工具（探查器，调试器，memcheck），它阻止了应用程序的探查”。 This error also appears when only NVVP is running. 仅运行NVVP时也会出现此错误。

1 个解决方案

First of all I suggest you to upgrade to the most recent VTune version, VTune Amplifier XE 2015 is too old. 首先，我建议您升级到最新的VTune版本，VTune Amplifier XE 2015太旧了。

With your version I can also recommend you to try these options: 对于您的版本，我还建议您尝试以下选项：

use another analysis type - eg Advanced Hotspots 使用其他分析类型-例如“高级热点”
do not enable OpenCL tracing in the analysis type - it is considered to be used with Intel Integrated Graphics only. 不要在分析类型中启用OpenCL跟踪-它被认为仅与Intel Integrated Graphics一起使用。