4 ms·
GPUs already beat MICs in core count, performance, price per watt, and initial cost. Why worry about writing specialized vectorized code? Just by using OpenCL
by gravypod 10y ago
GPUs already beat MICs in core count, performance, price per watt, and initial cost.
Why worry about writing specialized vectorized code? Just by using OpenCL your code will work like that. That's what OpenCL does. You give it chunks of data, you give it a kernel to run, then you run it on your data.
An extremely old AMD cards you can easily find 1280 "core" systems that can push GFLOPS of data through them.
OpenCL is the tool you should use for this. Not a hodge podge of messy fortran code that's been ""vectorized"".
Don't do the compiler's job.
- federchen13 10y agoI think we are talking about different things. PGAS is intended to develop more sophisticated parallel LOGIC codes more easily, which is a requirement to handle upcomming MIC hardware. The logic codes itself will play a major role to make efficient use of such hardware and to develop new algorithms for upcomming hardware. I am not against OpenCL or anything else, but consider these as low-level parallelism. The MIC is just starting by now and is very distinct from GPUs. At this time, we are only preparing for the near future.