[cfe-dev] Comparison of 2 schemes to implement OpenMP 5.0 declare mapper codegen

[Alexey Bataev via cfe-dev]

Hi Hal,

Best regards,
Alexey Bataev

28 июня 2019 г., в 23:46, Finkel, Hal J. via cfe-dev <cfe-dev at lists.llvm.org<mailto:cfe-dev at lists.llvm.org>> написал(а):


Hi, Alexey, Lingda,

I haven't been following this closely, so a few questions/comments:

 1. Recursive mappers are not supported in OpenMP 5, but do we expect that to change in the future?

Good question. Do not know, actually, but I think both of those schemes can be adapted to support recursive mappers.

 2. Our experience so far suggests that the most important optimization in this space is to limit the number of distinct host-to-device transfers (or data copies) on systems where data needs to be copied. In these schemes, where does that coalescing occur?

In both schemes we transfer the data only ones, after we gather all the required data mapping info and after that we transfer it to the device at once. The only difference in these schemes is the number of runtime functions calls required to fill this mapping data.


 3. So long as the mappers aren't recursive, I agree with Alexey that the total number of to-be-mapped components should be efficient to calculate. The counting function should simplify to a trivial expression in nearly all cases. The only case where it might not is where the type contains an array section with dynamic bounds, and the element type also has a mapper with an array section with dynamic bounds. In this case (similar to the unsupported recursive cases, which as an aside, we should probably support it as an extension) we could need to walk the data structure twice to precalculate the number of total components to map. However, this case is certainly detectable by static analysis of the declared mappers, and so I think that we can get the best of both worlds: we could use Alexey's proposed scheme except in cases where we truly need to walk the data-structure twice, in which case we could use Lingda's combined walk/push_back scheme. Is there any reason why that wouldn't work?

I think it is better to use only one scheme. I rather doubt that we can implement some kind of analysis in the frontend. Later, when the real codegen is moved to the backend, we can try to implement 2 schemes. But not today. We need to choose one and I just want to hear all pros and cons for both (actually, there are 3 schemes already) schemes to choose the most flexible, reliable and fast one.

Thanks again,

Hal

On 6/28/19 9:00 AM, Alexey Bataev wrote:

Hi Lingda, thanks for your comments.
We can allocate the buffer either by allocating it on the stack or calling OpenMP allocate function.
With this solution, we allocate memory only once (no need to resize buffer after push_backs) and we do not need to call the runtime function to put map data to the buffer, compiler generated code can do it.
But anyway, I agree, it would be good to hear some other opinions.
--------------
Best regards,
Alexey Bataev


...

--
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20190629/9f951082/attachment.html>