forked from OSchip/llvm-project
2 Commits
Author | SHA1 | Message | Date |
---|---|---|---|
Chandler Carruth | 693eedb138 |
[PM/Unswitch] Teach SimpleLoopUnswitch to do non-trivial unswitching,
making it no longer even remotely simple. The pass will now be more of a "full loop unswitching" pass rather than anything substantively simpler than any other approach. I plan to rename it accordingly once the dust settles. The key ideas of the new loop unswitcher are carried over for non-trivial unswitching: 1) Fully unswitch a branch or switch instruction from inside of a loop to outside of it. 2) Update the CFG and IR. This avoids needing to "remember" the unswitched branches as well as avoiding excessively cloning and reliance on complex parts of simplify-cfg to cleanup the cfg. 3) Update the analyses (where we can) rather than just blowing them away or relying on something else updating them. Sadly, #3 is somewhat compromised here as the dominator tree updates were too complex for me to want to reason about. I will need to make another attempt to do this now that we have a nice dynamic update API for dominators. However, we do adhere to #3 w.r.t. LoopInfo. This approach also adds an important principls specific to non-trivial unswitching: not *all* of the loop will be duplicated when unswitching. This fact allows us to compute the cost in terms of how much *duplicate* code is inserted rather than just on raw size. Unswitching conditions which essentialy partition loops will work regardless of the total loop size. Some remaining issues that I will be addressing in subsequent commits: - Handling unstructured control flow. - Unswitching 'switch' cases instead of just branches. - Moving to the dynamic update API for dominators. Some high-level, interesting limitationsV that folks might want to push on as follow-ups but that I don't have any immediate plans around: - We could be much more clever about not cloning things that will be deleted. In fact, we should be able to delete *nothing* and do a minimal number of clones. - There are many more interesting selection criteria for which branch to unswitch that we might want to look at. One that I'm interested in particularly are a set of conditions which all exit the loop and which can be merged into a single unswitched test of them. Differential revision: https://reviews.llvm.org/D34200 llvm-svn: 318549 |
|
Chandler Carruth | 1353f9a48b |
[PM/LoopUnswitch] Introduce a new, simpler loop unswitch pass.
Currently, this pass only focuses on *trivial* loop unswitching. At that reduced problem it remains significantly better than the current loop unswitch: - Old pass is worse than cubic complexity. New pass is (I think) linear. - New pass is much simpler in its design by focusing on full unswitching. (See below for details on this). - New pass doesn't carry state for thresholds between pass iterations. - New pass doesn't carry state for correctness (both miscompile and infloop) between pass iterations. - New pass produces substantially better code after unswitching. - New pass can handle more trivial unswitch cases. - New pass doesn't recompute the dominator tree for the entire function and instead incrementally updates it. I've ported all of the trivial unswitching test cases from the old pass to the new one to make sure that major functionality isn't lost in the process. For several of the test cases I've worked to improve the precision and rigor of the CHECKs, but for many I've just updated them to handle the new IR produced. My initial motivation was the fact that the old pass carried state in very unreliable ways between pass iterations, and these mechansims were incompatible with the new pass manager. However, I discovered many more improvements to make along the way. This pass makes two very significant assumptions that enable most of these improvements: 1) Focus on *full* unswitching -- that is, completely removing whatever control flow construct is being unswitched from the loop. In the case of trivial unswitching, this means removing the trivial (exiting) edge. In non-trivial unswitching, this means removing the branch or switch itself. This is in opposition to *partial* unswitching where some part of the unswitched control flow remains in the loop. Partial unswitching only really applies to switches and to folded branches. These are very similar to full unrolling and partial unrolling. The full form is an effective canonicalization, the partial form needs a complex cost model, cannot be iterated, isn't canonicalizing, and should be a separate pass that runs very late (much like unrolling). 2) Leverage LLVM's Loop machinery to the fullest. The original unswitch dates from a time when a great deal of LLVM's loop infrastructure was missing, ineffective, and/or unreliable. As a consequence, a lot of complexity was added which we no longer need. With these two overarching principles, I think we can build a fast and effective unswitcher that fits in well in the new PM and in the canonicalization pipeline. Some of the remaining functionality around partial unswitching may not be relevant today (not many test cases or benchmarks I can find) but if they are I'd like to add support for them as a separate layer that runs very late in the pipeline. Purely to make reviewing and introducing this code more manageable, I've split this into first a trivial-unswitch-only pass and in the next patch I'll add support for full non-trivial unswitching against a *fixed* threshold, exactly like full unrolling. I even plan to re-use the unrolling thresholds, as these are incredibly similar cost tradeoffs: we're cloning a loop body in order to end up with simplified control flow. We should only do that when the total growth is reasonably small. One of the biggest changes with this pass compared to the previous one is that previously, each individual trivial exiting edge from a switch was unswitched separately as a branch. Now, we unswitch the entire switch at once, with cases going to the various destinations. This lets us unswitch multiple exiting edges in a single operation and also avoids numerous extremely bad behaviors, where we would introduce 1000s of branches to test for thousands of possible values, all of which would take the exact same exit path bypassing the loop. Now we will use a switch with 1000s of cases that can be efficiently lowered into a jumptable. This avoids relying on somehow forming a switch out of the branches or getting horrible code if that fails for any reason. Another significant change is that this pass actively updates the CFG based on unswitching. For trivial unswitching, this is actually very easy because of the definition of loop simplified form. Doing this makes the code coming out of loop unswitch dramatically more friendly. We still should run loop-simplifycfg (at the least) after this to clean up, but it will have to do a lot less work. Finally, this pass makes much fewer attempts to simplify instructions based on the unswitch. Something like loop-instsimplify, instcombine, or GVN can be used to do increasingly powerful simplifications based on the now dominating predicate. The old simplifications are things that something like loop-instsimplify should get today or a very, very basic loop-instcombine could get. Keeping that logic separate is a big simplifying technique. Most of the code in this pass that isn't in the old one has to do with achieving specific goals: - Updating the dominator tree as we go - Unswitching all cases in a switch in a single step. I think it is still shorter than just the trivial unswitching code in the old pass despite having this functionality. Differential Revision: https://reviews.llvm.org/D32409 llvm-svn: 301576 |