On 2/12/14 10:58 PM, Jim Blandy wrote:
I think folks are being a little optimistic about the impact of having
MIPS code in tree. My experience has been that it usually does end up
being a distraction.
If we truly treat MIPS as a tier-3 platform - for example, if we don't
hesitate to remove, refactor, or even just fiddle with the types of
stuff that a grep tells us the MIPS back end uses - then in effect, the
MIPS port will only work after one of its maintainers has just landed a
merge. People interested in MIPS will only find those changesets useful.
If that's the case, then it doesn't seem any better than a
separately-maintained branch --- and that, at least, will be a repo in
which MIPS users can trust that tip works.
In other words, if the presence of the MIPS code in the tree isn't going
to affect our tier-1 decisions; and if MIPS users will find the branch
repo more valuable anyway; then what is the value of having the code in
tree?
This is largely the reason why I am slow to add the PowerPC Baseline
Compiler work back to the main tree; it's likely just to get broken by
other things, and then I have to go through the code review process to
get it fixed. PPC OS X ABI codegen is of only marginal use to Linux/ppc
and other SysV ABI-based operating systems anyway.
Once it gets mature, I'd probably do so, but we'll probably be a dead
port by then.
Cameron Kaiser
_______________________________________________
dev-tech-js-engine-internals mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-tech-js-engine-internals
