This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@ericniebler : this is not yet ready, but I wanted to push it up, so that you have a chance to see it earlier.

I'm planning to change the schedulers (static_thread_pool, libdispatch, parallel_scheduler) to be customized in terms of bulk_unchunked.

I'm also expecting some issues on the CUDA sides; I'm not sure how can I move forward on that side. Any help is greatly appreciated.

/ok to test

The GCC failure should be fixed. I'm not sure how to approach the Windows failure :(

/ok to test

@lucteo you can disregard the GPU CI failures. they are infra-related. the msvc failures are real though.

@ericniebler : there are some issues on the GPU side as well caused by my patch. I will temporarily pause this for 2 weeks, as I need to finish my ACCU presentation for next week.

/ok to test

@lucteo i have addressed the GPU build problem. the msvc build failure remains.

/ok to test

/ok to test

/ok to test

/ok to test 67740ab

@ericniebler : can you please retrigger? After a long and painful process I manage to isolate the Windows failure to an MSVC bug (see https://compiler-explorer.com/z/Eneab1zYT and https://developercommunity.visualstudio.com/t/noexcept-expression-in-lambda-template-n/10718680)

/ok to test 8e9547c

/ok to test a6dfb60

The error is strange. First, it works perfectly fine on my Mac; I was expecting the CI Mac build to be similar to my builds, but apparently it isn't.
Second, it seems that only the first declaration of operator() from the base class is used in the derived class, while the second is not. I find this strange (especially that it happens on 3 compilers)

Fixed by moving the deprecation in the base class. The deprecation method should also apply to bulk_chunked w/o policy, but there shouldn't be any code affected by this, so, we should be fine.

/ok to test a78089f

finally merged. thanks, @lucteo.