[Libre-soc-bugs] [Bug 558] gcc SV intrinsics concept

bugzilla-daemon at libre-soc.org bugzilla-daemon at libre-soc.org
Thu Dec 31 02:14:43 GMT 2020


https://bugs.libre-soc.org/show_bug.cgi?id=558

--- Comment #34 from Alexandre Oliva <oliva at libre-soc.org> ---
so what you seem to be telling me is that I should disregard the knowledge I
have of what the compiler *will* do, and of the least that *needs* to be
modelled in order to get us the reliable functionality we want, and instead do
what you think is the bare minimum that will get us a high-level assembler as
long as you don't breathe too hard while compiling the code, and if the
compiler still breaks stuff because we haven't modelled the bare minimum I know
we should, then, well, screw the user...

can I do that?  sure.  is that what I think we should do?  definitely not,
unless you want it to not really work, and then to throw it away when we get to
the second stage.

now, one of us is a compiler engineer who wrote a report on all the GCC
internal transformations and optimizations not very long ago.

the other is someone who has a lot of exceptionally important pieces of
knowledge to make this project work, including of what we'd like to have for a
first-step implementation to apply for a grant.

what we have is a mismatch between the assumptions made by one person and the
reality of what needs to be done to make this first step a step in the right
direction, rather than too-fragile-to-use throw-away code.

knowing the internals as I do, I want to figure out where we're aiming to get
in the end, so that I can *then* figure out what we need to get the feature you
wish to apply for a grant on that will take us closer to the end goal.

arguing we don't wish the compiler to do things that it will normally do
doesn't make the job easier as you seem to think; it makes it harder, because
then we'll have to do the job AND stop the compiler from doing what it normally
does.  is that understood?

-- 
You are receiving this mail because:
You are on the CC list for the bug.


More information about the libre-soc-bugs mailing list