[llvm-bugs] [Bug 39488] New: Tune lsr for the wasm backend

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=39488

            Bug ID: 39488
           Summary: Tune lsr for the wasm backend
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: WebAssembly
          Assignee: unassignedbugs at nondot.org
          Reporter: alonzakai at gmail.com
                CC: llvm-bugs at lists.llvm.org

We've noticed (e.g. [1]) that lsr is often unhelpful in the wasm backend and
leads to slower and larger code. sunfish suggests the pass may need to be tuned
for wasm.

Some data:

* The most extreme example is fannkuch, 5x slower in v8, 2x slower in sm, and
the crucial fannkuch_worker method is 45% larger, attaching.

* Many other minor examples too, primes-nocheck is 33% slower in v8 and 7%
slower in sm, and the main() function is 7% bigger with lsr, also attaching.

* This also affects real-world code, e.g. lua and box2d are 1% bigger.

[1] https://github.com/kripken/emscripten/pull/7386

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20181029/98c865ae/attachment.html>