Propagate to Lambda type information for function parameters and return #1029

chambart · 2017-02-03T16:31:10Z

Type information about function parameter and return can be useful for various optimizations on flambda and clambda. In particular it helps function parameter unboxing and post-inlining primitive specialization.

@alainfrisch you might be concerned by that.

mshinwell

I'm not completely certain about some of the changes in Translcore (from line 1100 onwards) as I've mentioned in the comments, but the rest looked fine to me modulo a couple of minor things.

mshinwell · 2017-02-06T08:42:07Z

bytecomp/translcore.ml

                body = Matching.inline_lazy_force (Lvar parm) Location.none;
                loc = loc;
-                attr = default_stub_attribute }
+                attr = default_function_attribute }


Is this an unrelated change? (Same on line 452 and in other places below.)

Oups. good catch.

This patch was extracted from a sequence started before the merge of the stub patch.
I will fix this

mshinwell · 2017-02-06T08:43:15Z

bytecomp/simplif.ml

        when kind = Curried && optimize ->
-          Lfunction{kind; params = params @ params'; body; attr; loc}
+          Lfunction{kind; params = params @ params'; return; body; attr; loc}


It might just be worth a comment as to the fact that the [return] type from the second function is being used (correctly), rather than the shadowed [return].

mshinwell · 2017-02-06T09:34:57Z

bytecomp/translcore.ml

@@ -1160,15 +1167,21 @@ and transl_apply ?(should_be_tailcall=false) ?(inlined = Default_inline)
        let body =
          match build_apply handle ((Lvar id_arg, optional)::args') l with
            Lfunction{kind = Curried; params = ids; body = lam; attr; loc} ->


The workings of this whole function are rather obscure. Are some of the Pgenval annotations here over-specific? For example maybe it's the case that return should be bound on line 1169 and then used on line 1172? (Even if it turns out to always be Pgenval this may be clearer.) I think it might be worth adding some comments.

I made the propagation explicit. I tried propagating enough information for this not to be always Pgenval, but this seems too complex for the benefit.

Also I am not knowledgeable enough to be confident commenting this function.

mshinwell · 2017-02-06T09:35:16Z

bytecomp/translcore.ml

@@ -1202,16 +1218,22 @@ and transl_function loc untuplify_fn repr partial param cases =
              (Matching.flatten_pattern size c_lhs, c_guard, c_rhs))
            cases in
        let params = List.map (fun _ -> Ident.create "param") pl in
-        ((Tupled, params),
+        let tparams = List.map (fun p -> p, Pgenval) params in


Don't we know something about the kind of the patterns here?

We do, and the following case was wrong also.

This only makes things more obvious, nothing is gained as some other information is missing.

…ctions

bobzhang · 2017-02-08T20:35:30Z

@chambart this looks cool, would you also propogate that which argument is an optional?
I am working on a pass in bucklescript to unbox optional argument, with this information directly
from the compiler, the optimization could be running faster(faster compilation matters)
(people like optional argument, it is a nice feature to js people, but it comes with a non-trivial cost)

chambart · 2017-02-24T14:33:09Z

@bobzhang I plan to add other types to the value_kind type: for instance I have this patch around for array type propagation. chambart@c388c4d
You should create a PR to add the types you need in there.

@poll

b11eea1 flambda-backend: Introduce Import_info (ocaml#1036) bc5b135 flambda-backend: Fix `ocamlobjinfo` on flambda2 .cmx files (ocaml#1029) c8babbd flambda-backend: Compilation_unit optimisations (ocaml#1035) e8d3e22 flambda-backend: Use 4.14.0 opam switch for building (includes upgrading ocamlformat to 0.24.1) (ocaml#1030) eb14a86 flambda-backend: Port PR81 from ocaml-jst (ocaml#1024) 131bc12 flambda-backend: Merge ocaml-jst 2022-12-13 (ocaml#1022) 06c189a flambda-backend: Make stack allocation the default (ocaml#1013) 98debd5 flambda-backend: Initial support for value slots not of value kind (ocaml#946) deb1714 flambda-backend: Add is_last flag to closinfo words (ocaml#938) d07fce1 flambda-backend: Disable poll insertion in Configure (ocaml#967) 0f1ce0e flambda-backend: Regenerate ocaml/configure autoconf 2.69 (instead of 2.71) (ocaml#1012) 27132d8 flambda-backend: Fix for spurious typing error related to expanding through functor arguments (ocaml#997) 724fb68 flambda-backend: Use `Compilation_unit.t` instead of `Ident.t` for globals (ocaml#871) 396d5b8 flambda-backend: Add a test for frametable setup in natdynlinked libraries (ocaml#983) b73ab12 flambda-backend: Fix invocation of `caml_shared_startup` in native dynlink (ocaml#980) 7c7d75a flambda-backend: Fix split_default_wrapper which did not trigger anymore with flambda2 (ocaml#970) 8fb75bd flambda-backend: Port ocaml#11727 and ocaml#11732 (ocaml#965) fdb7987 flambda-backend: Fix include functor issue after 4.14 merge. (ocaml#948) 9745cdb flambda-backend: Print -dprofile/-dtimings output to stdout like 4.12 (ocaml#943) 5f51f21 flambda-backend: Merge pull request ocaml#932 from mshinwell/4.14-upgrade 841687d flambda-backend: Run make alldepend in ocaml/ (ocaml#936) 72a7658 flambda-backend: Remove reformatting changes only in dynlink/dune (preserving PR889 and adjusting to minimise diff) 6d758cd flambda-backend: Revert whitespace changes in dune files, to match upstream c86bf6e flambda-backend: Remove duplicate tests for polling 971dbeb flambda-backend: Testsuite fixes 32f8356 flambda-backend: Topeval fix for symbols patch befea01 flambda-backend: Compilation fixes / rectify merge faults a84543f flambda-backend: Merge ocaml-jst 8e65056 flambda-backend: Merge ocaml-jst 4d70045 flambda-backend: Remove filename from system frametable (amd64) (ocaml#920) 5e57b7d flambda-backend: Bugfix for runtime frame_descr logic for C frames (ocaml#918) 6423d5e flambda-backend: Merge pull request ocaml#914 from mshinwell/merge-ocaml-jst-2022-10-24 ead605c flambda-backend: Add a missing Extract_exception (ocaml#916) c8f1481 flambda-backend: Resolve conflicts and add specialise/specialised attributes to Builtin_attributes cf4d0d3 flambda-backend: Merge fixes (ocaml#21) c2f742f flambda-backend: Re-enable some tests for Flambda2 (ocaml#881) 3d38d13 flambda-backend: Long frames in frametable (ocaml#797) 85aec7b flambda-backend: Add loop attribute to Builtin_attributes c0f16e3 flambda-backend: Compilation fixes 90dea23 flambda-backend: Merge flambda-backend/main 5acc6ea flambda-backend: Fixes after merge e501946 flambda-backend: Merge ocaml-jst 115083b flambda-backend: Merge ocaml-jst 9943b2e flambda-backend: Revert "Revert "Transform tail-recursive functions into recursive continuations (ocaml#893)"" (ocaml#909) ce339f1 flambda-backend: Fix alloc modes and call kinds for overapplications (ocaml#902) e6a317c flambda-backend: Revert "Transform tail-recursive functions into recursive continuations (ocaml#893)" 853c488 flambda-backend: Transform tail-recursive functions into recursive continuations (ocaml#893) 5a977e4 flambda-backend: Fix missing End_region primitives on switch arms (ocaml#898) 7fa7f9d flambda-backend: Add missing dependencies to Dune files (ocaml#889) 3cd36f0 flambda-backend: Have Lambda `Pgetglobal` and `Psetglobal` take `Compilation_unit.t` (ocaml#896) 7565915 flambda-backend: [@poll error] attribute (ocaml#745) 9eb9448 flambda-backend: Backport the main safepoints PRs (ocaml#740) 689bdda flambda-backend: Add strict mode for ocamldep (ocaml#892) git-subtree-dir: ocaml git-subtree-split: b11eea1

* Remove trailing whitespace * Amazing, missing trailing whitespace leads to a syntax error! --------- Co-authored-by: Cuihtlauac ALVARADO <cuihtmlauac@tarides.com>

Propagate type information about function parameters and return

cad7701

mshinwell requested changes Feb 6, 2017

View reviewed changes

mshinwell requested a review from alainfrisch February 6, 2017 09:37

chambart added 5 commits February 6, 2017 13:41

Fix merge problems with GPR ocaml#957

4955f92

Comment an ambiguous variable shadowing in Simplif

b74df21

Propagate type information on stub return in the presence on labels

e5a2f00

This only makes things more obvious, nothing is gained as some other information is missing.

Fix parameter type annotation in case of GADT and annotate tupled fun…

efe1da0

…ctions

Update dlambda tests with type annotations

5bdb896

chambart mentioned this pull request May 29, 2017

Fix float array specialisation in flambda #1169

Closed

chambart mentioned this pull request Sep 8, 2017

Unboxed calling convention #1322

Closed

4 tasks

damiendoligez added this to the 4.07-or-later milestone Sep 29, 2017

damiendoligez removed this from the consider-for-4.07 milestone Jun 4, 2018

alainfrisch removed their request for review July 12, 2018 14:04

alainfrisch mentioned this pull request Nov 20, 2018

Keep more type information in Lambda #2156

Merged

alainfrisch closed this Nov 20, 2018

gasche mentioned this pull request Apr 26, 2021

[patch] Avoid boxing float/int32/int64 when doing direct call #5894

Closed

sadiqj pushed a commit to sadiqj/ocaml that referenced this pull request Feb 21, 2023

flambda-backend: Fix ocamlobjinfo on flambda2 .cmx files (ocaml#1029)

bc5b135

stedolan pushed a commit to stedolan/ocaml that referenced this pull request Mar 21, 2023

Fix ocamlobjinfo on flambda2 .cmx files (ocaml#1029)

b03bf0b

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Propagate to Lambda type information for function parameters and return #1029

Propagate to Lambda type information for function parameters and return #1029

chambart commented Feb 3, 2017

mshinwell left a comment

mshinwell Feb 6, 2017

chambart Feb 6, 2017

mshinwell Feb 6, 2017

mshinwell Feb 6, 2017

chambart Feb 6, 2017

chambart Feb 6, 2017

mshinwell Feb 6, 2017

chambart Feb 6, 2017

bobzhang commented Feb 8, 2017 •

edited

Loading

chambart commented Feb 24, 2017

Propagate to Lambda type information for function parameters and return #1029

Propagate to Lambda type information for function parameters and return #1029

Conversation

chambart commented Feb 3, 2017

mshinwell left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bobzhang commented Feb 8, 2017 • edited Loading

chambart commented Feb 24, 2017

bobzhang commented Feb 8, 2017 •

edited

Loading