[PATCH] propagate IR-level fast-math-flags to DAG nodes, disabled by default

[escha]

We’re getting significant CSE regressions on our backend, caused by the fact that none of the DAG folds propagate fast-math flags. This results in nodes not being CSE’d, since we only CSE nodes if they have identical flags (and thus worse codegen).

I was able to get rid of many of them by adding flag propagation to FNEG folding, but there’s way more than just that.

—escha

> On Jul 29, 2015, at 11:51 AM, Sanjay Patel <spatel at rotateright.com> wrote:
> 
> For reference, fixes for PR24141 are in:
> http://reviews.llvm.org/rL243293 <https://urldefense.proofpoint.com/v2/url?u=http-3A__reviews.llvm.org_rL243293&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=EH36SVVrOANvPpWUfmtjHvcZene16JZYtYcOsOWHmmQ&s=oJkUxyP-U28dL9zPjcyCjESFgqf9fEa1WxScYHC1BLE&e=>
> http://reviews.llvm.org/rL243498 <https://urldefense.proofpoint.com/v2/url?u=http-3A__reviews.llvm.org_rL243498&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=EH36SVVrOANvPpWUfmtjHvcZene16JZYtYcOsOWHmmQ&s=9UShb3DxctMQors5uP4ErkfZ2ETsaqYXEeRLX8qKYt0&e=>
> http://reviews.llvm.org/rL243500 <https://urldefense.proofpoint.com/v2/url?u=http-3A__reviews.llvm.org_rL243500&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=EH36SVVrOANvPpWUfmtjHvcZene16JZYtYcOsOWHmmQ&s=S5dEJW-NYcq9yLglX74iN2esNypm3qfP6wCNidxysx8&e=>...assuming those don't cause new breakage, it's finally, really time to flip the FMF flag...and wait for more breakage.
> 
> On Mon, Jul 20, 2015 at 11:34 AM, Sanjay Patel <spatel at rotateright.com <mailto:spatel at rotateright.com>> wrote:
> 
> 
> On Fri, Jul 17, 2015 at 4:42 PM, Nick Lewycky <nlewycky at google.com <mailto:nlewycky at google.com>> wrote:
> On 9 July 2015 at 15:27, Sanjay Patel <spatel at rotateright.com <mailto:spatel at rotateright.com>> wrote:
> Hi Nick,
> 
> The reciprocal fix is in at r241826. Please let me know when you find another failure. Thanks!
> 
> Thank you! I tested it and found exactly nothing wrong. Hopefully I didn't goof the testing, but I think it's time to flip the flag to true, then wait a few days before removing the flag entirely.
> 
> Wow...I'm surprised, but I'll hope for the best. Thanks for the testing!
> 
> 
> Note however, that we still have problems with PR24141 which is related to unsafe fp math, but doesn't change with the -enable-fmf-dag flag.
> 
> Proposed fix:
> http://reviews.llvm.org/D11345 <https://urldefense.proofpoint.com/v2/url?u=http-3A__reviews.llvm.org_D11345&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=EH36SVVrOANvPpWUfmtjHvcZene16JZYtYcOsOWHmmQ&s=q0g5ELtAiUzID-3xj21ZzkiASkbrFaYCWXDy-nBF0OI&e=>
>  
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150803/81dc5f33/attachment.html>