On 05/02/2016 07:31 AM, Jeff Law wrote:
On 04/29/2016 07:32 AM, Bernd Schmidt wrote:
On 04/29/2016 03:02 PM, David Edelsohn wrote:
How has this show general benefit for all architectures to deserve
enabling it by default at -O2?
It should improve postreload scheduling in general, and it can also help
clear up bad code generation left behind by register allocation.
Right. ISTM the round-robin renaming reduces the false dependencies
that are inherently created by register allocation.
It should benefit any architecture that utilizes post-reload scheduling.
Following up to myself... As was seen on x86-64, renaming can cause
problems with partial register stalls.
FWIW, partial register stalls are not strictly limited to the x86-64.
In fact the PA 2.0 processor had partial register stalls -- they were so
bad that I ended up just turning off the ability to access both halves
of the 64bit FP registers independently. I wouldn't be totally
surprised if other targets have micro-architectures that would tickle
this issue.
jeff
