search for: superword

...e to be > supported. Therefore my aim is to start with the most minimal > implementation possible, to explore the difficulties encountered in > the specific context of LLVM and to build a foundation from which future > work can progress. There's two types of autovectorization, SLP (superword level parallelism) and ILP (instruction level parallelism). You can do ILP which is loop-ignorant autovectorization of straight-line code which turns out to be the code that runs inside the loop. > So, initially, I aim at supporting only the simplest loops such as: > > int a[256],...

...RFC thread on [bumping the bitcode version in LLVM 6.0](http://lists.llvm.org/pipermail/llvm-dev/2018-February/121114.html), due to the changes in fast math flags in bitcode. * Florian Hahn [proposes](http://lists.llvm.org/pipermail/llvm-dev/2018-February/121000.html) making LoopVectorize aware of superword-level parallelism (SLP) operations. * David Li has written a short [summary](http://lists.llvm.org/pipermail/llvm-dev/2018-February/120964.html) of developments in profile-guided optimisation with LLVM over the past 18 months. * Michael Haidl started a discussion on [upstreaming PACXX](http://lis...

.... Therefore my aim is to start with the most minimal >> implementation possible, to explore the difficulties encountered in >> the specific context of LLVM and to build a foundation from which future >> work can progress. > > There's two types of autovectorization, SLP (superword level > parallelism) and ILP (instruction level parallelism). You can do ILP > which is loop-ignorant autovectorization of straight-line code which > turns out to be the code that runs inside the loop. Pardon me. I've been disabused by Owen Anderson for writing the above. SLP is l...

Hi all, I'd like to have a first stab at a loop-based auto-vectorization pass as part of 2009's Google Summer of Code program. As far as I can tell from searching the mailing list archives, no work on such an auto-vectorizer seems to be currently in progress. Whereas auto-vectorization is a well-researched topic, complexity seems to quickly explode once more advanced loop constructs are