Cuda内核准时（jit）编译可能？ [英] Cuda kernel just-in-time (jit) compilation possible?

问题描述

CUDA是否支持cuda内核的JIT编译？

Does CUDA support JIT compilation of a cuda kernel?

我知道OpenCL提供此功能。

I know that OpenCL offers this feature.

我有一些变量在运行时没有改变（即只依赖于输入文件），因此我想在内核compiletime（即在运行时）使用宏定义这些值。

I have some variables which are not changed during runtime (i.e. only depend on the input file), therefore I would like to define these values with a macro at kernel compiletime(i.e at runtime).

如果我在compiletime手动定义这些值，我的注册使用从53下降到46，并大大提高性能。

If I define these values manually at compiletime my register usage drops from 53 to 46 and greatly improves performance.

谢谢。 b $ b

推荐答案

它提供了cuda 7.0的nvrtc库。通过此库，您可以在运行时编译您的cuda代码。

It became available with nvrtc library of cuda 7.0. By this library you can compile your cuda codes during runtime.

http://devblogs.nvidia.com/parallelforall/cuda-7-release-candidate-feature-overview/

Bu你能获得什么样的优势？在我看来，我找不到动态编译的巨大优势。

Bu what kind of advantages you can gain? In my view, i couldn't find so much dramatic advantages of dynamic compilation.

这篇关于Cuda内核准时（jit）编译可能？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！