status for week ending Jan 5, 2015
Reed.Kotler at imgtec.com
Mon Jan 5 20:07:45 PST 2015
Continued work on LLVM Mips fast-isel.
Currently there are 9 patches in the queue to be pushed.
Am debugging one regression in a new patch involving the intrinsic llvm.umul.with.overflow.i32 .
It unfortunately only occurs once in test-suite and in a very complicated set of code and
I'm trying to make a small test case for it.
The LLVM register allocator ends up causing bad code to be generated because it runs out of
registers (but it's an unusual situation because the register class it needs is the accumulator
and there is only one) and then just steals a register but
then the resulting code does not work.
Hopefully debug this issue with llvm.umul.with.overflow.i32 and continue on more patches.
Push patches as Daniel approves them and maybe make some changes to get them approved.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the llvm-commits