Reduce size of InterpErrorInfo to 8 bytes #82116

tmiasko · 2021-02-14T20:26:50Z

r? @ghost

tmiasko · 2021-02-14T20:49:19Z

@bors try @rust-timer queue

rust-timer · 2021-02-14T20:49:20Z

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

bors · 2021-02-14T20:49:31Z

⌛ Trying commit c7393c762a516a2151cfbe27f7096f01cbb19933 with merge e868f5cd6ef13410be12eacb2cdf41f68cb1c226...

bors · 2021-02-14T21:49:52Z

☀️ Try build successful - checks-actions
Build commit: e868f5cd6ef13410be12eacb2cdf41f68cb1c226 (e868f5cd6ef13410be12eacb2cdf41f68cb1c226)

rust-timer · 2021-02-14T21:49:54Z

Queued e868f5cd6ef13410be12eacb2cdf41f68cb1c226 with parent 5fa22fe, future comparison URL.

rust-timer · 2021-02-14T23:28:11Z

Finished benchmarking try commit (e868f5cd6ef13410be12eacb2cdf41f68cb1c226): comparison url.

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. Please note that if the perf results are neutral, you should likely undo the rollup=never given below by specifying rollup- to bors.

Importantly, though, if the results of this run are non-neutral do not roll this PR up -- it will mask other regressions or improvements in the roll up.

@bors rollup=never
@rustbot label: +S-waiting-on-review -S-waiting-on-perf

tmiasko · 2021-02-14T23:36:05Z

r? @oli-obk

pickfire · 2021-02-15T09:30:53Z

compiler/rustc_middle/src/mir/interpret/error.rs

+
+#[derive(Debug)]
+struct InterpErrorInfoInner<'tcx> {
+    kind: InterpError<'tcx>,
    backtrace: Option<Box<Backtrace>>,


Since this is already in a Box do we need a second Box here?

Since this will be None in most cases (unless debugging the const evaluator), I'd leave this box in in order to not increase the regular size of the InterpErrorInfoInner.

But if kind is used won't backtrace be used as well? Do we need a second indirection?

backtrace is only used if RUSTC_CTFE_BACKTRACE is set. In all other cases there will be no Box allocated. I'm not sure how large a Backtracestruct is, but if it's larger than8bytes, for all normal cases we don't waste space, as noBoxis allocated. It's fine to have arbitrary costs ifRUSTC_CTFE_BACKTRACE` is used, as that is already expensive by itself

oli-obk · 2021-02-15T09:56:50Z

Amazing! Really good find.

@bors r+

bors · 2021-02-15T09:56:51Z

📌 Commit c7393c762a516a2151cfbe27f7096f01cbb19933 has been approved by oli-obk

bors · 2021-02-17T02:18:56Z

⌛ Testing commit c7393c762a516a2151cfbe27f7096f01cbb19933 with merge d8ac472db5025de359865f5779c9d6b15b893692...

bors · 2021-02-17T02:44:41Z

💔 Test failed - checks-actions

The test case fails with either one error or two errors. Use a single code generation unit to avoid nondeterminism.

tmiasko · 2021-02-17T09:10:13Z

Used -Ccodegen-units=1 to avoid nondeterminism in failed test case ui/issues/issue-23458.rs.

oli-obk · 2021-02-17T13:28:51Z

@bors r+

bors · 2021-02-17T13:28:52Z

📌 Commit 614b0cc has been approved by oli-obk

bors · 2021-02-17T16:47:23Z

⌛ Testing commit 614b0cc with merge 5ef2106...

bors · 2021-02-17T19:39:52Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing 5ef2106 to master...

rust-log-analyzer · 2021-02-17T19:41:28Z

A job failed! Check out the build log: (web) (plain)

Click to see the possible cause of the failure (guessed by this bot)

  CACHE_DOMAIN: ci-caches.rust-lang.org
  TOOLSTATE_REPO_ACCESS_TOKEN: ***
##[endgroup]
Cloning into 'rust-toolstate'...
/home/runner/work/rust/rust/src/tools/publish_toolstate.py:121: DeprecationWarning: 'U' mode is deprecated
📣 Toolstate changed by rust-lang/rust#82116!
  with open(path, 'rU') as f:

Tested on commit rust-lang/rust@5ef21063f0c0fd5b973bfa8cb88c0b70982da977.
Direct link to PR: <https://github.com/rust-lang/rust/pull/82116>

💔 miri on windows: test-fail → build-fail (cc @oli-obk @RalfJung @eddyb).
💔 miri on linux: test-fail → build-fail (cc @oli-obk @RalfJung @eddyb).
Traceback (most recent call last):
Traceback (most recent call last):
  File "/home/runner/work/rust/rust/src/tools/publish_toolstate.py", line 338, in <module>
    response = urllib2.urlopen(urllib2.Request(
  File "/usr/lib/python3.8/urllib/request.py", line 222, in urlopen
    return opener.open(url, data, timeout)
  File "/usr/lib/python3.8/urllib/request.py", line 522, in open
    req = meth(req)
  File "/usr/lib/python3.8/urllib/request.py", line 1281, in do_request_
    raise TypeError(msg)
TypeError: POST data should be bytes, an iterable of bytes, or a file object. It cannot be of type str.
##[error]Process completed with exit code 1.

RalfJung · 2021-02-18T09:29:11Z

compiler/rustc_mir/src/interpret/intern.rs

@@ -356,7 +356,7 @@ where
                // an allocation, which we should avoid. When that happens,
                // dedicated error variants should be introduced instead.
                assert!(
-                    !error.kind.allocates(),
+                    !error.kind().allocates(),


Doesn't this stop making any sense now?
The point of this allocates thing was to make sure we don't allocate an error and then "catch" it (for perf reasons we wanted to avoid those allocations). But with this change we now allocate for all errors, so (a) it seems silly to also separately Box some error kinds again, and (b) this check here makes no sense since we always allocate. So shouldn't we get rid of the extra nested Box and the allocates stuff?

(Btw I'd appreciate a Cc for changes like this that affect the Miri/CTFE engine.)

Unboxing of UndefinedBehaviorInfo::InvalidUninitBytes content would be one thing to consider, but generally this check makes as much sense now as it made before. Heap allocation inside & outside cannot be easily merged together.

this check makes as much sense now as it made before

If that means it didn't make any sense before, that's okay and we should still remove it. ;)

But I feel like the difference between 0 and 1 allocation is much bigger than between 1 and 2, so I am not sure this check is worth keeping. OTOH, maybe the expensive bit we should protect against isn't the allocation itself but the string formatting in the format!-using errors. In that case we should probably rename allocates to pre_formatted or so (and unbox InvalidUninitBytes as you said).

all InterpError allocate now, so adjust alloc-error-check Cc rust-lang#82116 (comment) r? `@oli-obk`

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Feb 14, 2021

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Feb 14, 2021

rust-highfive assigned oli-obk Feb 14, 2021

pickfire reviewed Feb 15, 2021

View reviewed changes

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Feb 15, 2021

This comment has been minimized.

Sign in to view

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Feb 17, 2021

tmiasko added 2 commits February 17, 2021 08:57

Reduce size of InterpErrorInfo to 8 bytes

e190f0d

Use -Ccodegen-units=1 to make issue-23458 test deterministic

614b0cc

The test case fails with either one error or two errors. Use a single code generation unit to avoid nondeterminism.

tmiasko force-pushed the box-error branch from c7393c7 to 614b0cc Compare February 17, 2021 09:05

bors removed the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Feb 17, 2021

bors added the S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. label Feb 17, 2021

bors added the merged-by-bors This PR was explicitly merged by bors. label Feb 17, 2021

bors merged commit 5ef2106 into rust-lang:master Feb 17, 2021

rustbot added this to the 1.52.0 milestone Feb 17, 2021

tmiasko deleted the box-error branch February 17, 2021 19:40

RalfJung reviewed Feb 18, 2021

View reviewed changes

RalfJung mentioned this pull request Feb 20, 2021

all InterpError allocate now, so adjust alloc-error-check #82338

Merged

bors added a commit to rust-lang-ci/rust that referenced this pull request Feb 25, 2021

Auto merge of rust-lang#82338 - RalfJung:interp-error-allocs, r=oli-obk

89d32eb

all InterpError allocate now, so adjust alloc-error-check Cc rust-lang#82116 (comment) r? `@oli-obk`

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Reduce size of InterpErrorInfo to 8 bytes #82116

Reduce size of InterpErrorInfo to 8 bytes #82116

tmiasko commented Feb 14, 2021

tmiasko commented Feb 14, 2021

rust-timer commented Feb 14, 2021

bors commented Feb 14, 2021

bors commented Feb 14, 2021

rust-timer commented Feb 14, 2021

rust-timer commented Feb 14, 2021

tmiasko commented Feb 14, 2021

pickfire Feb 15, 2021

oli-obk Feb 15, 2021

pickfire Feb 15, 2021 •

edited

Loading

oli-obk Feb 15, 2021

oli-obk commented Feb 15, 2021

bors commented Feb 15, 2021

bors commented Feb 17, 2021

This comment has been minimized.

bors commented Feb 17, 2021

tmiasko commented Feb 17, 2021

oli-obk commented Feb 17, 2021

bors commented Feb 17, 2021

bors commented Feb 17, 2021

bors commented Feb 17, 2021

rust-log-analyzer commented Feb 17, 2021

RalfJung Feb 18, 2021

tmiasko Feb 18, 2021

RalfJung Feb 18, 2021

Reduce size of InterpErrorInfo to 8 bytes #82116

Reduce size of InterpErrorInfo to 8 bytes #82116

Conversation

tmiasko commented Feb 14, 2021

tmiasko commented Feb 14, 2021

rust-timer commented Feb 14, 2021

bors commented Feb 14, 2021

bors commented Feb 14, 2021

rust-timer commented Feb 14, 2021

rust-timer commented Feb 14, 2021

tmiasko commented Feb 14, 2021

pickfire Feb 15, 2021

Choose a reason for hiding this comment

oli-obk Feb 15, 2021

Choose a reason for hiding this comment

pickfire Feb 15, 2021 • edited Loading

Choose a reason for hiding this comment

oli-obk Feb 15, 2021

Choose a reason for hiding this comment

oli-obk commented Feb 15, 2021

bors commented Feb 15, 2021

bors commented Feb 17, 2021

This comment has been minimized.

bors commented Feb 17, 2021

tmiasko commented Feb 17, 2021

oli-obk commented Feb 17, 2021

bors commented Feb 17, 2021

bors commented Feb 17, 2021

bors commented Feb 17, 2021

rust-log-analyzer commented Feb 17, 2021

RalfJung Feb 18, 2021

Choose a reason for hiding this comment

tmiasko Feb 18, 2021

Choose a reason for hiding this comment

RalfJung Feb 18, 2021

Choose a reason for hiding this comment

pickfire Feb 15, 2021 •

edited

Loading