|Assigning multiple contiguous registers to a virtual register email@example.com (shrey) (2010-03-13)|
|Re: Assigning multiple contiguous registers to a virtual register firstname.lastname@example.org (Fernando) (2010-03-15)|
|Re: Assigning multiple contiguous registers to a virtual register email@example.com (shrey) (2010-03-16)|
|Date:||Sat, 13 Mar 2010 11:59:32 -0800 (PST)|
|Posted-Date:||13 Mar 2010 15:06:07 EST|
I have seen this being done in multiple compilers. However, am
not sure abt why the compilers choose to do this way.
Usually the register allocation does a one to one mapping
between virtual registers and physical registers. If you have virtual
registers that needs multiple physical registers, then you define new
pseudo-names that represent a set of physical registers. Now, you
assign this new pseudo-physical register instead.
My question is what is the motivation here ?
1. Is it that there is no need to do it otherwise? Most program dont
see virtual registers of very different sizes ?
2. Makes it easier. If this is the case, I am curious abt how ? Is it
that dependency information is likely to be too cluttered if there are
no pseudo names ?
[I expect this is for architectures where some of the instructions need
register pairs. -John]
Return to the
Search the comp.compilers archives again.