Optimising ARM DMBs

Reinoud Elhorst re302 at cam.ac.uk
Tue Apr 1 01:17:23 PDT 2014


I just recently joined, and this will be my first patch. It finds places in
ARM machine code where 2 dmbs follow one another, and eliminates one of
them. In some cases this can result in considerable (30%) speedups, as I
will argue in my talk at EuroLLVM next week.

I discussed the patch off-list with several people, but am more than happy
to get some more feedback.

Thanks,


Reinoud

M       lib/Target/ARM/ARM.h
A       lib/Target/ARM/ARMOptimizeBarriersPass.cpp
M       lib/Target/ARM/ARMTargetMachine.cpp
M       lib/Target/ARM/CMakeLists.txt
A       test/CodeGen/ARM/optimize-dmbs-v7.ll
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140401/65e4ea2a/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ARMBarriers.patch
Type: application/octet-stream
Size: 6805 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140401/65e4ea2a/attachment.obj>


More information about the llvm-commits mailing list