@ben that file won’t work but isn’t needed

does this error make sense to anyone?: ERROR: uncaught exception: (exn:fail uncaught exception: (exn:fail uncaught exception: (exn:fail uncaught exception: (exn:fail uncaught exception: (exn:fail uncaught exception: (exn:fail uncaught exception: (exn:fail uncaught exception: (exn:fail bad token in read: " #<continuation-mark-set>) #<continuation-mark-set>) #<continuation-mark-set>) #<continuation-mark-set>) #<continuation-mark-set>) #<continuation-mark-set>) #<continuation-mark-set>) #<continuation-mark-set>) while executing: (require /home/ben/code/racket/eidetic/vm/response6/benchmark-baseline/racket/share/pkgs/typed-racket-lib/typed/racket/base.rkt) in module: main RPython traceback: File "pycket_entry_point.c", line 47, in entry_point File "pycket_entry_point.c", line 1272, in old_pycket_actual_entry File "pycket_interpreter.c", line 81, in interpret_module File "pycket_interpreter.c", line 976, in Module__interpret_mod File "pycket_interpreter.c", line 22345, in interpret_one File "rpython_jit_metainterp.c", line 2635, in ll_portal_runner__pycket_AST_ASTPtr_pycket_AST_A File "pycket_interpreter.c", line 65438, in portal_1 File "pycket_prims.c", line 14998, in raise_exception Fatal RPython error: SchemeException zsh: abort pycket-c main.rkt

yes

I’m fixing that now

is that running the baseline or the collapsible branch?

because I only saw that on the collapsible branch

both

I tried baseline first & got the error

then I switched to collapsible by changing my PATH’s racket and running make setup with the collapsible raco

what benchmark/file/etc?

sieve, for all but the untyped config

also for fsm config 1000

unfortunately there’s at least one more bug triggered on the collapsible code

something about methods

@ben one thing I’ve already learned — pycket is very bad on the baseline version of fsm

I looked at configuration 0010 and changed the # of steps from 500 to 50

that takes racket about 1200 ms

it takes pycket about 30000 ms

almost all of that time is spent tracing

ie not doing actual work

now trying to get the collapsible version to work, but running into errors as mentioned

hmm. can you try baseline fsmoo?

ok, it looks like when I tried Pycket in June: - fsmoo +zombie ran faster on Pycket than Racket - fsm segfaulted for me

I’m pretty sure the segfault is OOM

at least for some configurations

@ben unfortunately I’m now having to implement all the impersonator changes that happened between 6.6 and 7.0

ok. Can you email Robby & Dan to let them know?

I don’t think collapsible + baseline make any extra changes that aren’t in 7.0, but they would know for sure

@ben woo, I successfully ran that configuration on the collapsible branch on pycket (in the pycket interpreter with only 1 iteration, but we’ll see how the compiled version goes soon)

turns out there were only two features missing for that — procedure-impersonator*? and the optional extra argument to impersonator property accessors

plus in that time Racket has grown a need to read prefab structs, literal vectors, and literal byte strings on startup

plus we had broken a few uses of syntax object operations

spoke too soon — still have to implement some things

apparently running 2 steps doesn’t exercise everything

just trying to understand: does “exercise” mean “pycket tries to trace + JIT compile”?

no, exercise means “call all the racket functions that I still need to implement”