search for: descrimin

...st at a coarse > level). Yes, if transformations made earlier in the pipeline make different > decisions, then that will invalidate later fine-grained data (at least > potentially). I don't see how any of this makes this infeasible. We just > need a way for the profiling counts, per descriminator, to remain > available, and for the transformations themselves to know which > discriminators (loop ids, or whatever) to consider. > > -Hal > > --paulr > > > > > > > -- > Hal Finkel > Lead, Compiler Technology and Programming Languages > Leadersh...

...e duplication factor. > If more profile info can be used for better unroll/vectorize > decision making, we can definitely add it. I'm still missing something here. In your proposed system, does the tool collecting the profiling information *interpret* the duplication factor encoded in the descriminators at all? If it does, it seems like this should be specific to vectorization. For unrolling, you still have the same number of instructions, and so you just need to make the different instructions from the different loop iterations carry different discriminator values (so they they'll sum to...

As illustrated in the above example, it is not like "vectorization has a distinct bit". All different optimizations make clones of code which will be labeled by UIDs represented by N (e.g. 8) bits. In this way, the space will be capped by the number of clones all optimizations have made, instead of # of optimizations that has applied. And it will be capped at 2^N-1. The cons of using uid

...n display that information usefully to the user, and so that our optimizations can make useful choices (i.e. don't bother vectorizing a loop when the scalar loop is run a lot but the vectorized version almost never runs). In short, I think that differentiating these different regions using the descriminator seems like a natural choice, but "And we do not need to introduce new building blocks to debug info" might not save us all that much in the long run. To keep information on what regions correspond to what optimizations, we may need to do that. That's not a bad thing, and I'd r...

...ed for better unroll/vectorize > > > decision making, we can definitely add it. > > > > > I'm still missing something here. In your proposed system, does the > > tool collecting the profiling information *interpret* the > > duplication factor encoded in the descriminators at all? > > Yes it does. Okay. I'm still missing something. Can you please write out an example using your notation below for a vectorized loop? > > If it does, it seems like this should be specific to vectorization. > > For unrolling, you still have the same number of...

...l). Yes, if transformations made earlier in the pipeline make different >>> decisions, then that will invalidate later fine-grained data (at least >>> potentially). I don't see how any of this makes this infeasible. We just >>> need a way for the profiling counts, per descriminator, to remain >>> available, and for the transformations themselves to know which >>> discriminators (loop ids, or whatever) to consider. >>> >>> -Hal >>> >>> --paulr >>> >>> >>> >>> >>> >>&gt...

Am 28.08.2011 20:02, schrieb geovanisouza92 at gmail.com: > Hi, Jo! > > I'm trying create a new programming language, and I want that it have > Unicode support (support for read and manipulate rightly the source-code and > string literals). > > But, in addition, my programming language supports "string interpolation" > string, and in these interpolations, tiny

...n that will invalidate later > > > > > fine-grained data (at least potentially). I don't see how any > > > > > of > > > > > this makes this infeasible. We just need a way for the > > > > > profiling > > > > > counts, per descriminator, to remain available, and for the > > > > > transformations themselves to know which discriminators (loop > > > > > ids, > > > > > or whatever) to consider. > > > > > > > > > > > > > > > -Hal > &gt...

...r, how can we tell whether the vectorized version we might now create will be executed (based on having profiling data from a run where the compiler might have previously made a similar choice)? > > In short, I think that differentiating these different regions > > using > > the descriminator seems like a natural choice, but "And we do not > > need to introduce new building blocks to debug info" might not save > > us all that much in the long run. To keep information on what > > regions correspond to what optimizations, we may need to do that. > > T...

...for better unroll/vectorize decision making, we >>> can definitely add it. >>> >>> I'm still missing something here. In your proposed system, does the tool >>> collecting the profiling information *interpret* the duplication factor >>> encoded in the descriminators at all? >>> >> >> Yes it does. >> >> Okay. I'm still missing something. Can you please write out an example >> using your notation below for a vectorized loop? >> > > Sure. Original code: > > for (int i = 0; i < N; i++) > a[i...

...ed from the duplication factor. If more profile info can be used for better unroll/vectorize decision making, we can definitely add it. I'm still missing something here. In your proposed system, does the tool collecting the profiling information *interpret* the duplication factor encoded in the descriminators at all? Yes it does. Okay. I'm still missing something. Can you please write out an example using your notation below for a vectorized loop? Sure. Original code: for (int i = 0; i < N; i++) a[i] = b[i]; Transformed code: for (int i = 0; i < N / 4; i++) vectorized_assign_4(&a...

I wish to mount a Win95 share on my linux machine. I can use smbclient to get to the shared disk, using this format: smbclient //win95.machine/share <password> The win95.machine is listed in my /etc/hosts file. I've tried every permutation of the smbmount syntax that I've been able to find in the documentation, dejanews, or this mailing list. Perhaps I've missed some. I

...uid: 5 > So basically, each clone of the loop will have a uid, which you can > use to identify which version of the loop the instruction was coming > from. > > > > In short, I think that differentiating these different regions > > > > using > > > > the descriminator seems like a natural choice, but "And we do > > > > not > > > > need to introduce new building blocks to debug info" might not > > > > save > > > > us all that much in the long run. To keep information on what > > > > regions...

...have some money saved for retirement, but I would actually like that to be part of my retirement plan is to not really retire fully. Anyone else feel that way ? Are there alot of older programmers ? Does the job market for older programmers look promising or not ? Are there similar or different age descrimination issues down the road ? Ever wonder about that ? --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Ruby on Rails: Talk" group. To post to this group, send email to rubyonrails-talk-/JYPxA39Uh5TLH3MbocFFw...

Motivation: Many optimizations duplicate code. E.g. loop unroller duplicates the loop body, GVN duplicates computation, etc. The duplicated code will share the same debug info with the original code. For SamplePGO, the debug info is used to present the profile. Code duplication will affect profile accuracy. Taking loop unrolling for example: #1 foo(); #2 for (i = 0; i < N; i++) { #3 bar();

The following series of patches fill in most of the gaps in the OCaml bindings to libxl, to make them useful for clients such as xapi/xenopsd (from XCP). There are a number of bugfixes to the existing bindings as well. I have an experimental version of xenopsd that successfully uses the new bindings. An earlier version of the first half of the series was submitted to the last by Ian Campbell on