feat: parenthetical syntax for `cycles`, `timeout` etc. #4608

ggreif · 2024-07-10T11:41:15Z

ICCallPrim (and friends) should carry the fragment to set the SystemCyclesAddPrim while the call is being set up towards the replica.

Parentheticals

now: (with cycles = 42_000_000) Actor.call(param)
later: (with timeout = 10)
ic0.call_with_best_effort_response : (timeout_seconds : i32) -> ()
maybe: (with receiveMax = 50_000) (to limit the response size)
maybe: (with resend = true; resendDelay = 3) when SYS_TRANSIENT reject response (like Unix EINTR)
support feat(system-api): add call_cycles_add128_up_to ic#1158
usecase (with memoryLimit = 1G) ActorClass(<args>)

TODOs:

test one-ways
typecheck
- warn when an attribute is moot
- cycles : Nat for canister sends (self and raw sends too)
async blocks
ICCallPrim, see top
ICCallRawPrim, not directly annotatable (desisting for now)
FIXMEs, TODOs
warn on queries?
best-effort: new error code when deadline passed?

ggreif · 2024-07-10T13:51:26Z

src/ir_def/construct.ml

@@ -312,14 +312,16 @@ let funcE name sort ctrl typ_binds args typs exp =
    note = Note.{ def with typ; eff = T.Triv };
  }

+let recordE' = ref (fun _ -> nullE ()) (* gets correctly filled below *)


use and to allow mutual recursion

ggreif · 2024-07-10T13:54:32Z

src/ir_passes/tailcall.ml

    begin match e1.it, env with
    | VarE f1, { tail_pos = true;
                 info = Some { func; typ_binds; temps; label; tail_called } }
         when f1 = func && are_generic_insts typ_binds insts  ->
      tail_called := true;
      (blockE (assignEs temps (exp env e2)) (breakE label (unitE ()))).it
-    | _,_-> PrimE (CallPrim insts, [exp env e1; exp env e2])
+    | _,_-> PrimE (CallPrim (insts, pars), [exp env e1; exp env e2])


Suggested change

| _,_-> PrimE (CallPrim (insts, pars), [exp env e1; exp env e2])

| _,_-> PrimE (CallPrim (insts, exp env pars), [exp env e1; exp env e2])

src/mo_frontend/definedness.ml

unfortunately it doesn't arrive in the `async.ml`

at least now I get > ingress Err: IC0504: Canister rwlgt-iiaaa-aaaaa-aaaaa-cai violated contract: ic0_call_cycles_add128 called when no call is under construction.

we should use the system call though, instead of assigning to `@cycles`, as that will go away

github-actions · 2024-12-03T18:22:33Z

Comparing from 76f9087 to dc37044:
In terms of gas, 1 tests improved and the mean change is -0.0%.
In terms of size, 5 tests regressed and the mean change is +0.0%.

crusso

Some interim comments

crusso · 2024-11-29T16:43:03Z

src/codegen/compile_classical.ml

@@ -2541,7 +2551,7 @@ module Closure = struct
      I32Type :: Lib.List.make n_args I32Type,
      FakeMultiVal.ty (Lib.List.make n_res I32Type))) in
    (* get the table index *)
-    Tagged.load_forwarding_pointer env ^^
+    (*Tagged.load_forwarding_pointer env ^^ FIXME: NOT needed, accessing immut slots*)


Are you sure this is ok? I'd first verify with @luc. Also, not related to this PR at all.

crusso · 2024-12-03T15:28:35Z

src/ir_interpreter/interpret_ir.ml

    assert env.flavor.has_await;
    async env
      exp.at
      (fun k' r ->
        let env' = { env with labs = V.Env.empty; rets = Some k'; throws = Some r }
        in interpret_exp env' exp1 k')
      k
-  | AsyncE (Type.Cmp, _, exp1, _) ->
+  | AsyncE (_, Type.Cmp, _, exp1, _) ->


Suggested change

| AsyncE (_, Type.Cmp, _, exp1, _) ->

| AsyncE (_FIXME, Type.Cmp, _, exp1, _) ->

crusso · 2024-12-05T14:20:50Z

default.nix

+      ## FOR NOW drun-eop-release = enhanced_orthogonal_persistence_subdir "run-drun" [ moc nixpkgs.drun ] ;
+      ## FOR NOW drun-eop-debug = snty_enhanced_orthogonal_persistence_subdir "run-drun" [ moc nixpkgs.drun ] ;


don't forget to revert

crusso · 2024-12-05T14:29:51Z

src/ir_def/ir.ml

@@ -116,7 +116,7 @@ and lexp' =
 all call-by-value. Many passes can treat them uniformly, so they are unified
 using the PrimE node. *)
 and prim =
-  | CallPrim of Type.typ list         (* function call *)
+  | CallPrim of Type.typ list * exp   (* function call *)


I makes me nervous that prim now as subexpressions since we might miss them in traversals. For example, is closure conversion in the backend considering these? It might be safer to put these arguments in the prim argument list.

crusso · 2024-12-05T14:30:14Z

src/ir_def/ir.ml

@@ -167,12 +168,12 @@ and prim =
  (* backend stuff *)
  | CPSAwait of Type.async_sort * Type.typ
                                      (* typ is the current continuation type of cps translation *)
-  | CPSAsync of Type.async_sort * Type.typ
+  | CPSAsync of Type.async_sort * Type.typ * exp


crusso · 2024-12-05T14:30:21Z

src/ir_def/ir.ml

  | ICPerformGC
  | ICReplyPrim of Type.typ list
  | ICRejectPrim
  | ICCallerPrim
-  | ICCallPrim
+  | ICCallPrim of exp option


crusso · 2024-12-05T14:33:11Z

src/ir_interpreter/interpret_ir.ml

@@ -310,7 +310,7 @@ and interpret_exp_mut env exp (k : V.value V.cont) =
  | PrimE (p, es) ->
    interpret_exps env es [] (fun vs ->
      match p, vs with
-      | CallPrim typs, [v1; v2] ->
+      | CallPrim (typs, _), [v1; v2] ->


Yep, needs fixing

crusso · 2024-12-05T14:47:35Z

src/mo_frontend/parser.mly

+        { e with it = CallE (Some (ObjE (Option.(to_list base0_opt @ to_list base), fs) @? e.at), f, is, args) }
+      | AsyncE (base0_opt, Type.Fut, binds, exp) ->
+        { e with it = AsyncE (Some (ObjE (Option.(to_list base0_opt @ to_list base), fs) @? e.at), Type.Fut, binds, exp) }
+      | _ -> e (* FIXME: meh, can we emit a warning? *)


Err - this will just drop the parenthetical on the floor, including its side-effects/non-termination. Would it not be possible to only allow these on calls and asyncs, syntactically?

Should be possible. I'll look into it.

and observe the deadline

…elect fields in parentheticals

ggreif · 2024-12-10T05:42:56Z

src/lang_utils/error_codes.ml

@@ -203,5 +203,8 @@ let error_codes : (string * string option) list =
    "M0197", Some([%blob "lang_utils/error_codes/M0197.md"]); (* `system` capability required *)
    "M0198", Some([%blob "lang_utils/error_codes/M0198.md"]); (* Unused field pattern warning *)
    "M0199", Some([%blob "lang_utils/error_codes/M0199.md"]); (* Deprecate experimental stable memory *)
-    "M0200", Some([%blob "lang_utils/error_codes/M0200.md"]) (* Cannot determine subtyping or equality *)
+    "M0200", Some([%blob "lang_utils/error_codes/M0200.md"]); (* Cannot determine subtyping or equality *)
+    "M0202", None; (* parenthetical note must be applied to a message send *)


make this contiguous

ggreif · 2024-12-10T10:19:31Z

src/ir_passes/async.ml

+      let cyclesSetup = if hasCycles
+        then Some (thenE
+                     (natE Mo_values.Numerics.Nat.zero |> assignVarE "@cycles")
+                     (primE SystemCyclesAddPrim [dotE pars "cycles" T.nat]))


Bind pars to an object variable? Otherwise we get duplicated effects!

this probably doesn't make much sense

ggreif · 2024-12-11T14:37:46Z

src/codegen/compile_classical.ml

@@ -12187,6 +12215,29 @@ and compile_prim_invocation (env : E.t) ae p es at =
    SR.Vanilla, Cycles.refunded env
  | SystemCyclesBurnPrim, [e1] ->
    SR.Vanilla, compile_exp_vanilla env ae e1 ^^ Cycles.burn env
+  | ICCyclesPrim, [] ->


This is not an idempotent operation, so we have to be careful to not call it twice. E.g. it fails for paired up environment+parenthetical.

This should return two options. Possibly just the Nat (cycles) and Nat32 (timeout). We definitely have to search the attributes.

ggreif · 2024-12-11T14:38:39Z

src/codegen/compile_classical.ml

+            Arr.load_field env 1l ^^
+            G.i (LocalSet (nr 0l))
+          ; Tagged.Object,
+            Opt.inject_simple env G.nop


Don't care storing back anything, as there is no captured environment.

ggreif · 2024-12-11T14:40:51Z

src/codegen/compile_classical.ml

+        Tagged.branch_with env [I32Type]
+          [ Tagged.Closure,
+            G.i Drop ^^
+            Opt.null_lit env


use br 1 to fall out?

ggreif added 3 commits July 10, 2024 13:39

WIP: surface syntax for parentheticals

d4811e2

WIP: first AST modifications

b6d32ec

WIP: augment IR too

d3b9415

ggreif commented Jul 10, 2024

View reviewed changes

src/mo_frontend/definedness.ml Outdated Show resolved Hide resolved

ggreif added 10 commits July 10, 2024 15:55

Update src/mo_frontend/definedness.ml

3f57777

WIP: fill the parenthetical

7e23ac8

unfortunately it doesn't arrive in the `async.ml`

tweaks

b0516b0

define and use tupVarsP helper

dc5a72f

teach about SystemCyclesAddPrim

8f7df27

examine all exprs

e5f3ca9

WIP: doing naughty stuff

5eb79a3

at least now I get > ingress Err: IC0504: Canister rwlgt-iiaaa-aaaaa-aaaaa-cai violated contract: ic0_call_cycles_add128 called when no call is under construction.

WIP: this starts working

9672a96

we should use the system call though, instead of assigning to `@cycles`, as that will go away

accept

00b2507

cleanup

3d4bb5a

ggreif changed the title ~~WIP: surface syntax for parentheticals~~ feat: parenthetical syntax for cycles etc. Jul 10, 2024

compress

d989397

ggreif self-assigned this Jul 10, 2024

ggreif added the language design Requires design work label Jul 10, 2024

ggreif added 10 commits July 11, 2024 11:52

WIP: prepare ICCallPrim to carry setup code

892ea54

WIP: compile the setup code

42a0471

elim a FIXME

e43b1cc

minor refactor

43e816c

Merge branch 'master' into gabor/parentheticals

272870b

explain more cycles

2f22db2

remove because redundant

b6ee8dd

fix IR renaming

812d78d

cleanup

e7f13f6

tweaks

ccd03db

explicitly zero cycles when user doesn't specify any

64696a1

crusso reviewed Dec 5, 2024

View reviewed changes

ggreif force-pushed the gabor/parentheticals branch from d3085f0 to 57e213b Compare December 6, 2024 11:51

ggreif added 2 commits December 6, 2024 15:25

Merge branch 'master' into gabor/parentheticals

339a98e

add new API

a0d475a

ggreif force-pushed the gabor/parentheticals branch from 770d5c5 to a0d475a Compare December 6, 2024 14:42

ggreif added 2 commits December 6, 2024 19:30

WIP: prepare for timeout

c9bd84f

WIP: replyDeadline is still 0

3bed728

ggreif force-pushed the gabor/parentheticals branch from 0b40189 to 3bed728 Compare December 6, 2024 19:16

ggreif added 7 commits December 6, 2024 20:30

check for timeout type error

cbfacdf

intro nat32

f1cc907

WIP: begin checking timeout

2edbcd9

WIP: invoke call_with_best_effort_response

83e70ab

access at the correct type

0878d90

and observe the deadline

break out infer_check_bases_fields and use it to check just a few s…

ec31258

…elect fields in parentheticals

some errors are now emitted in a centralised fashion

ca66302

ggreif commented Dec 10, 2024

View reviewed changes

ggreif added 5 commits December 10, 2024 06:52

tweaks

4c37385

intro and use SystemTimeoutPrim

e79c48d

accept

b5eb704

fix warning

f80ffa8

unfake, but beware of the duplication!

2e58517

ggreif commented Dec 10, 2024

View reviewed changes

ggreif added 2 commits December 10, 2024 14:05

WIP: timeout for one-shot calls

7cb6cf0

this probably doesn't make much sense

oneshot tests actually work!

dc37044

ggreif changed the title ~~feat: parenthetical syntax for cycles etc.~~ feat: parenthetical syntax for cycles, timeout etc. Dec 11, 2024

ggreif commented Dec 11, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: parenthetical syntax for `cycles`, `timeout` etc. #4608

feat: parenthetical syntax for `cycles`, `timeout` etc. #4608

ggreif commented Jul 10, 2024 •

edited

Loading

ggreif Jul 10, 2024

ggreif Jul 10, 2024

github-actions bot commented Dec 3, 2024 •

edited

Loading

crusso left a comment

crusso Nov 29, 2024

crusso Dec 3, 2024

crusso Dec 5, 2024

crusso Dec 5, 2024

crusso Dec 5, 2024

crusso Dec 5, 2024

crusso Dec 5, 2024

crusso Dec 5, 2024 •

edited by ggreif

Loading

ggreif Dec 6, 2024

ggreif Dec 10, 2024

ggreif Dec 10, 2024

ggreif Dec 11, 2024

ggreif Dec 11, 2024

ggreif Dec 11, 2024 •

edited

Loading

ggreif Dec 11, 2024 •

edited

Loading

	\| _,_-> PrimE (CallPrim (insts, pars), [exp env e1; exp env e2])
	\| _,_-> PrimE (CallPrim (insts, exp env pars), [exp env e1; exp env e2])

	\| AsyncE (_, Type.Cmp, _, exp1, _) ->
	\| AsyncE (_FIXME, Type.Cmp, _, exp1, _) ->

		## FOR NOW drun-eop-release = enhanced_orthogonal_persistence_subdir "run-drun" [ moc nixpkgs.drun ] ;
		## FOR NOW drun-eop-debug = snty_enhanced_orthogonal_persistence_subdir "run-drun" [ moc nixpkgs.drun ] ;

feat: parenthetical syntax for cycles, timeout etc. #4608

Are you sure you want to change the base?

feat: parenthetical syntax for cycles, timeout etc. #4608

Conversation

ggreif commented Jul 10, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

github-actions bot commented Dec 3, 2024 • edited Loading

crusso 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

crusso Dec 5, 2024 • edited by ggreif Loading

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

ggreif Dec 11, 2024 • edited Loading

Choose a reason for hiding this comment

ggreif Dec 11, 2024 • edited Loading

Choose a reason for hiding this comment

feat: parenthetical syntax for `cycles`, `timeout` etc. #4608

feat: parenthetical syntax for `cycles`, `timeout` etc. #4608

ggreif commented Jul 10, 2024 •

edited

Loading

github-actions bot commented Dec 3, 2024 •

edited

Loading

crusso Dec 5, 2024 •

edited by ggreif

Loading

ggreif Dec 11, 2024 •

edited

Loading

ggreif Dec 11, 2024 •

edited

Loading