By skimming through the code, this seems to have the same data-flow model as Intel's TBB, but extended to distributed memory configurations using MPI as backend.
One can still use TBB inside the functions to make use of it alongside getting MPI parallelism from easyLambda.
Moreover, easyLambda has data flows, map, reduce which in a way improves the way programs are written and composed. There are no classes to be declared for anything. It has many syntactic sugars, for column selection, control over parallelism, i/o etc. I personally like the style of coding in
it for various problems, keeping parallelism aside.
Since it supports distributed computation does it have any abstraction for dealing with failures, for example, network problems etc. Support for checkpointing or some other abstraction that deals with these kinds of errors would be very useful. In fact absolutely essential for anything of non-trivial scale.
BTW not complaining, far, far from it, just a useful feature addition you may consider in case you have the bandwidth to invest.
Yeah, you are right, fault tolerance is a needed feature. Although, I'm not pretty sure if that needs to be added to easyLambda itself. Since, MPI 3.0 is expected to support fault tolerance by default. But instead of waiting for MPI 3.0 fault tolerance support, I'm thinking of running it on top of HPX or some other system that provides good runtime support on top of MPI. Also, since most of parallelism scheme is confined to MPIBridge which is just a unit in the flow, these different implementations can coexist.
It's very common when chaining lots of operations, each of which transforms the result of the previous one (functional style). I've seen tons of Scala code do it. Writing it all in one line would be quite the unreadable mess.
It's an argument that communities and organizations give disproportionate weight to trivial issues (like coding style) instead of discussing content. I agree that there are more interesting things to discuss about easyLambda than the placement of newlines.
It lets you pass around a FooObject/FooContext and offers you a lot of the powers of a DSL while still being boring, stock BarLanguage. I find it to be very helpful for making declarative-style APIs.
So this is a bunch of constructs for creating function-pointers and the like which then would be sent to mpi or other parallel libraries to do the actual interfacing with the parallel mechanism.
Any idea how this would work with (random stuff I'm thinking about) Cuda or Tensorflow?
Thanks for your interest. There are indeed no function pointers, no void pointers, no preprocessor hacks, no wild type-casting involved anywhere in the library. This is completely compile time type-safe. Still, the kind of things you can do would be a far cry for all the hacks with those shoddy pointers. There are a lot of features like column manipulation, passing vector of a values for multiple rows and just a value for a single row etc. that make composability and ease of use way better. All of these features are implemented using template meta programming and traits available in modern C++. The language has become quite impressive from what we used to think of. You must check it out.
A map/reduce/rise take a template parameter which can be of any class type and library has nothing to do what a function object or function does in the function body. The function just need to take in the arguments that the prior unit passes it and return something. I don't think there will be any problem if one uses any other library to perform computation inside a function, even if some library does not work out of the box, I'm pretty positive it can be done with little tinkering. I hope this makes things clear.
I'm probably not as up on the most modern c++ approaches as I should be but I'm quite interested. I have downloaded the code and I'm looking through it.
Basically, it uses template metaprogramming to go from the higher-level map/reduce/rise semantics to lower level for-loops and calls.
Beyond the particulars, the question/challenge I'd have would be - "how would you use this dynamically generate map/reduce/rise calls?"
20 comments
[ 2.8 ms ] story [ 64.7 ms ] threadMoreover, easyLambda has data flows, map, reduce which in a way improves the way programs are written and composed. There are no classes to be declared for anything. It has many syntactic sugars, for column selection, control over parallelism, i/o etc. I personally like the style of coding in it for various problems, keeping parallelism aside.
BTW not complaining, far, far from it, just a useful feature addition you may consider in case you have the bandwidth to invest.
ezl::rise(readFile<string> (argv[1]).rowSeparator('s').colSeparator(""))
Really, dang, you can remove posts like mine but you do nothing to address the real problem here.
Any idea how this would work with (random stuff I'm thinking about) Cuda or Tensorflow?
A map/reduce/rise take a template parameter which can be of any class type and library has nothing to do what a function object or function does in the function body. The function just need to take in the arguments that the prior unit passes it and return something. I don't think there will be any problem if one uses any other library to perform computation inside a function, even if some library does not work out of the box, I'm pretty positive it can be done with little tinkering. I hope this makes things clear.
I'm probably not as up on the most modern c++ approaches as I should be but I'm quite interested. I have downloaded the code and I'm looking through it.
Basically, it uses template metaprogramming to go from the higher-level map/reduce/rise semantics to lower level for-loops and calls.
Beyond the particulars, the question/challenge I'd have would be - "how would you use this dynamically generate map/reduce/rise calls?"