Conversation
This helps with things like :long-float that are determined by build parameters, but more importantly things like :long-float/binary80 that are controlled by the C++ runtime. Thanks yitzi for help with the script.
there are far too many of them, god damn
This should let us clear out way more of the explicit definition of CORE, but I don't want to bother figuring out what's actually needed, at the moment
and add the CLEAVIR-ENV nickname. The common-macros defpackage doesn't add nicknames if the package already exists, so the first definition of the package we give it needs to have all the nicks.
it's used in clasp-cleavir for type derivers
why did i think this had to be done at compile time? i dunno anyway this makes cross-clasp easier
it's useless, but so it goes
makes things easier for cross-clasp
dunno why we would need it in init
It might be smarter to just not use CL, but we do actually have some code that's read in the LLVM-SYS package, like defaults in lambda lists defined in C++.
for debugging. Probably should hook these in to clos:g-f-methods eventually
probably would be even better to make %make-package use and unuse appropriately
not hooked up yet but does at least load. waow
this lets infinities be printed in primitive clasp.
No idea how this works in existing clasp - I guess it must load the code treating with-irbuilder as a function and so on, and then that gets overriden in the compiled code? That's pretty stupid
definitely beats keeping track of internal functions
So that READ when compiling in an alternate environment works.
Give up if something goes wrong because who cares. Also continue better if something's missing from the fmap. no big
i.e. actually use it everywhere
With the new compile-file-pathname we were using .faso but putting a bytecode fasl in it. oops
since it supports first class environments now, mostly. Optimization is still technically off (it uses main env identities and inline definitions in Cleavir, I think?) but that doesn't matter since the FCGE definitions have the same properties as the main env's. I can actually do a native self-build now. incredible shit. go me
I think this remove all use of #+bytecode #-bytecode, which is good since relying on build parameters to tell us what compiler is being used is not compatible with a b2b world.
though it doesn't seem to have a ton of practical effect, so I should probably bark up a different tree. Oh well. I think doing less work is nice regardless.
native compilation still takes a noticeable chunk of time, so do it in threads if we can swing it.
avoid having MESSAGE quit. that's a separate concern
This caused numerous errors in native compilation, in particular ruining the CLOS code walker.
constant-form-value needs fixing up for both cmp:lexenv and cleavir environments
This way it matches *compile-file-parallel* and we don't need to worry about values other than :parallel and :serial. We could restore :execution in the future if we ever have other methods.
problems happened with e.g. (min (floor ...)) because lo and behold FLOOR returns two values, conflicting with the &rest nil in the type declaration. I think putting in VALUES is ok, though it's slightly suboptimal in bytecode
Preliminary merge before merging cross into main, hopefully.
temporary, hopefully
These were replaced with C++ code when yitzi added long-float, but that change somehow didn't make it fully into the cross branch. There's no point in keeping the code around commented out so deletion it is.
no longer required as we are moving to native code in FASLs
just :native t and a default. keeping it simple
if the modules are out of order, the loader messes up finding the literals. no good at all
instead of cross-clasp. this silences some warnings during build. Still not an exported interface, though.
catch will be more of a journey This could probably be used for the bytecode-skipping compiler, presuming we even want to retain it.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Overhauls Clasp's build procedure. Instead of bootstrapping itself, another Lisp like SBCL loads a bytecode compiler (Maclina) and a specialized first class environment, and uses these to compile the Clasp library code into a bytecode FASL. This FASL can then be loaded by the Clasp C++ runtime which is mostly unchanged.
A "native" build in which Clasp itself is compiled by clasp-cleavir can be done by taking this bytecode FASL, loading that same specialized environment, and then using it to natively compile the Clasp library code. This produces a bytecode FASL with native versions of most functions through the b2b compiler; the native code is kept as attributes in the FASL.
This is mostly done but still needs a bit more work. In particular a few little things are probably still broken.
Other changes include:
ext:deoptimizecan be used to temporarily replace native-compiled functions with more debuggable bytecode versions