Lazy deserialization of cmi files#1322
Merged
rleshchinskiy merged 4 commits intooxcaml:mainfrom Apr 28, 2023
Merged
Conversation
poechsel
reviewed
Apr 21, 2023
poechsel
approved these changes
Apr 26, 2023
lpw25
approved these changes
Apr 27, 2023
mshinwell
approved these changes
Apr 27, 2023
mshinwell
added a commit
to mshinwell/oxcaml
that referenced
this pull request
Apr 28, 2023
a7d005a flambda-backend: Lazy deserialization of cmi files (oxcaml#1322) aa83fa3 flambda-backend: Reinstate previous API for Env.lookup_value (oxcaml#1323) e4007a4 flambda-backend: Lazy substitution into value_declaration (oxcaml#1320) 634b607 flambda-backend: Merge Types.* and Subst.Lazy.* types (oxcaml#1312) cf82708 flambda-backend: Bump magic numbers for 4.14.1-7 (oxcaml#1317) 6470400 flambda-backend: zero_alloc attribute payload "assert all" and "ignore" (oxcaml#1296) bba5248 flambda-backend: Teach `ocamldep` about all the language extensions (oxcaml#1303) 33e97b0 flambda-backend: Change Includemod to work on lazy modtypes (oxcaml#1228) 16e5002 flambda-backend: zero_alloc new warning for unchecked functions (oxcaml#1302) 36b4626 flambda-backend: Attribute [@@@zero_alloc check] to turn the check on (oxcaml#1294) 3b524c6 flambda-backend: Cmm.value_kind cleanup (oxcaml#1091) ec99505 flambda-backend: Fix failure of `check_all_arches` on RISCV (oxcaml#1300) 450bc58 flambda-backend: Backend changes for multiple returns (oxcaml#1268) 84a7a26 flambda-backend: Static check for zero_alloc: ignore allocation that lead to exceptional return (oxcaml#1157) 1723728 flambda-backend: Re-enable parallel build of the runtime (oxcaml#1287) 26ea7f3 flambda-backend: Fix closure marshalling when not in NNP mode (oxcaml#1286) 9b91f2e flambda-backend: Reduce number of caml_apply functions taking/returning "I" and "V" (oxcaml#1272) 1686928 flambda-backend: Restore Cmm unboxing behaviour inside regions (oxcaml#1285) cf9be42 flambda-backend: Fix all the no-naked-pointers problems (oxcaml#1282) 8fe089e flambda-backend: Unrevert oxcaml#1131 and fix a Cmm unboxing bug (oxcaml#1284) c4143c3 flambda-backend: Revert "Make Selectgen treat region boundaries more precisely" (oxcaml#1283) 2078dce flambda-backend: Add some -dtimings output for the typechecker (oxcaml#1245) 273a7f9 flambda-backend: Make Selectgen treat region boundaries more precisely (oxcaml#1131) 47610e6 flambda-backend: Bump magic numbers for 4.14.1-6 (oxcaml#1274) fd53d38 flambda-backend: Generate *.cms files for merlin (oxcaml#1232) 853f95f flambda-backend: Add tail_mod_const to builtin_attrs (oxcaml#1265) f9ef051 flambda-backend: Fix issue caused by effects of gadt expansion in mode cross check (oxcaml#1263) e9ffcf8 flambda-backend: Fix dependencies for regenerating Flambda2 parser, tests (oxcaml#1255) 6f1cd1f flambda-backend: Restore a lost location, needed for merlin (oxcaml#1242) 009332b flambda-backend: Fix merge from ocaml-jst git-subtree-dir: ocaml git-subtree-split: a7d005a
ccasin
added a commit
to ccasin/oxcaml
that referenced
this pull request
Apr 29, 2023
a7d005a flambda-backend: Lazy deserialization of cmi files (oxcaml#1322) aa83fa3 flambda-backend: Reinstate previous API for Env.lookup_value (oxcaml#1323) e4007a4 flambda-backend: Lazy substitution into value_declaration (oxcaml#1320) 634b607 flambda-backend: Merge Types.* and Subst.Lazy.* types (oxcaml#1312) cf82708 flambda-backend: Bump magic numbers for 4.14.1-7 (oxcaml#1317) 6470400 flambda-backend: zero_alloc attribute payload "assert all" and "ignore" (oxcaml#1296) bba5248 flambda-backend: Teach `ocamldep` about all the language extensions (oxcaml#1303) 33e97b0 flambda-backend: Change Includemod to work on lazy modtypes (oxcaml#1228) 16e5002 flambda-backend: zero_alloc new warning for unchecked functions (oxcaml#1302) 36b4626 flambda-backend: Attribute [@@@zero_alloc check] to turn the check on (oxcaml#1294) 3b524c6 flambda-backend: Cmm.value_kind cleanup (oxcaml#1091) ec99505 flambda-backend: Fix failure of `check_all_arches` on RISCV (oxcaml#1300) 450bc58 flambda-backend: Backend changes for multiple returns (oxcaml#1268) 84a7a26 flambda-backend: Static check for zero_alloc: ignore allocation that lead to exceptional return (oxcaml#1157) 1723728 flambda-backend: Re-enable parallel build of the runtime (oxcaml#1287) 26ea7f3 flambda-backend: Fix closure marshalling when not in NNP mode (oxcaml#1286) 9b91f2e flambda-backend: Reduce number of caml_apply functions taking/returning "I" and "V" (oxcaml#1272) 1686928 flambda-backend: Restore Cmm unboxing behaviour inside regions (oxcaml#1285) cf9be42 flambda-backend: Fix all the no-naked-pointers problems (oxcaml#1282) 8fe089e flambda-backend: Unrevert oxcaml#1131 and fix a Cmm unboxing bug (oxcaml#1284) c4143c3 flambda-backend: Revert "Make Selectgen treat region boundaries more precisely" (oxcaml#1283) 2078dce flambda-backend: Add some -dtimings output for the typechecker (oxcaml#1245) 273a7f9 flambda-backend: Make Selectgen treat region boundaries more precisely (oxcaml#1131) 47610e6 flambda-backend: Bump magic numbers for 4.14.1-6 (oxcaml#1274) fd53d38 flambda-backend: Generate *.cms files for merlin (oxcaml#1232) 853f95f flambda-backend: Add tail_mod_const to builtin_attrs (oxcaml#1265) f9ef051 flambda-backend: Fix issue caused by effects of gadt expansion in mode cross check (oxcaml#1263) e9ffcf8 flambda-backend: Fix dependencies for regenerating Flambda2 parser, tests (oxcaml#1255) 6f1cd1f flambda-backend: Restore a lost location, needed for merlin (oxcaml#1242) 009332b flambda-backend: Fix merge from ocaml-jst git-subtree-dir: ocaml git-subtree-split: a7d005a
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
4 participants
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.
We now deserialise cmi files lazily. This is implemented by piggy-backing on the already existing mechanism for lazy substitutions. Now, a
Subst.Lazytype can contain things which will be deserialised on demand.Rather than just serializing the entire module signature, the new cmi format now stores a data block with serialised bits, identified by their offsets within that block. For details, see the comment in
cmi_format.ml. Note that we still read the entire .cmi file, we just don't deserialise everything immediately. Not reading all of it is left as future work.Due to the piecemeal serialization, we necessarily lose some sharing between individual signature components. This doesn't seem to be a problem in practice.
The size of .cmi files increases by ~15% on average. However, major heap allocations drop by ~44% and peak major heap size decreases by 32% at p90 and by 26% at p99. Compile times might improve by ~10%, with major GC times dropping by over 30%.