I previously used a clumsy hack to copy Closgen back and forth while
inlining, to handle when an inlined function contains closures, which
need to each be uniquely numbered.
The real solution was to name the closures using r.inlCaller, rather
than r.curfn. This CL adds a helper method to do exactly this.
Change-Id: I510553b5d7a8f6581ea1d21604e834fd6338cb06
Reviewed-on: https://go-review.googlesource.com/c/go/+/520339
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Auto-Submit: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
In Go 1.17, cmd/compile gained the ability to inline calls to
functions that contain function literals (aka "closures"). This was
implemented by duplicating the function literal body and emitting a
second LSym, because in general it might be optimized better than the
original function literal.
However, the second LSym was named simply as any other function
literal appearing literally in the enclosing function would be named.
E.g., if f has a closure "f.funcX", and f is inlined into g, we would
create "g.funcY" (N.B., X and Y need not be the same.). Users then
have no idea this function originally came from f.
With this CL, the inlined call stack is incorporated into the clone
LSym's name: instead of "g.funcY", it's named "g.f.funcY".
In the future, it seems desirable to arrange for the clone's name to
appear exactly as the original name, so stack traces remain the same
as when -l or -d=inlfuncswithclosures are used. But it's unclear
whether the linker supports that today, or whether any downstream
tooling would be confused by this.
Updates #60324.
Change-Id: Ifad0ccef7e959e72005beeecdfffd872f63982f8
Reviewed-on: https://go-review.googlesource.com/c/go/+/497137
Reviewed-by: Michael Pratt <mpratt@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
[This is a roll-forward of CL 479095, which was reverted due to a bad
interaction between inlining and escape analysis, then later fixed
first with an attempt in CL 482355, then again in CL 484859, and then
one more time with CL 492135.]
Currently, when the inliner is determining if a function is
inlineable, it descends into the bodies of closures constructed by
that function. This has several unfortunate consequences:
- If the closure contains a disallowed operation (e.g., a defer), then
the outer function can't be inlined. It makes sense that the
*closure* can't be inlined in this case, but it doesn't make sense
to punish the function that constructs the closure.
- The hairiness of the closure counts against the inlining budget of
the outer function. Since we currently copy the closure body when
inlining the outer function, this makes sense from the perspective
of export data size and binary size, but ultimately doesn't make
much sense from the perspective of what should be inlineable.
- Since the inliner walks into every closure created by an outer
function in addition to starting a walk at every closure, this adds
an n^2 factor to inlinability analysis.
This CL simply drops this behavior.
In std, this makes 57 more functions inlinable, and disallows inlining
for 10 (due to the basic instability of our bottom-up inlining
approach), for an net increase of 47 inlinable functions (+0.6%).
This will help significantly with the performance of the functions to
be added for #56102, which have a somewhat complicated nesting of
closures with a performance-critical fast path.
The downside of this seems to be a potential increase in export data
and text size, but the practical impact of this seems to be
negligible:
│ before │ after │
│ bytes │ bytes vs base │
Go/binary 15.12Mi ± 0% 15.14Mi ± 0% +0.16% (n=1)
Go/text 5.220Mi ± 0% 5.237Mi ± 0% +0.32% (n=1)
Compile/binary 22.92Mi ± 0% 22.94Mi ± 0% +0.07% (n=1)
Compile/text 8.428Mi ± 0% 8.435Mi ± 0% +0.08% (n=1)
Change-Id: I5f75fcceb177f05853996b75184a486528eafe96
Reviewed-on: https://go-review.googlesource.com/c/go/+/492017
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
[This is a roll-forward of CL 479095, which was reverted due to a bad
interaction between inlining and escape analysis, then later fixed
fist with an attempt in CL 482355, then again in 484859 .]
Currently, when the inliner is determining if a function is
inlineable, it descends into the bodies of closures constructed by
that function. This has several unfortunate consequences:
- If the closure contains a disallowed operation (e.g., a defer), then
the outer function can't be inlined. It makes sense that the
*closure* can't be inlined in this case, but it doesn't make sense
to punish the function that constructs the closure.
- The hairiness of the closure counts against the inlining budget of
the outer function. Since we currently copy the closure body when
inlining the outer function, this makes sense from the perspective
of export data size and binary size, but ultimately doesn't make
much sense from the perspective of what should be inlineable.
- Since the inliner walks into every closure created by an outer
function in addition to starting a walk at every closure, this adds
an n^2 factor to inlinability analysis.
This CL simply drops this behavior.
In std, this makes 57 more functions inlinable, and disallows inlining
for 10 (due to the basic instability of our bottom-up inlining
approach), for an net increase of 47 inlinable functions (+0.6%).
This will help significantly with the performance of the functions to
be added for #56102, which have a somewhat complicated nesting of
closures with a performance-critical fast path.
The downside of this seems to be a potential increase in export data
and text size, but the practical impact of this seems to be
negligible:
│ before │ after │
│ bytes │ bytes vs base │
Go/binary 15.12Mi ± 0% 15.14Mi ± 0% +0.16% (n=1)
Go/text 5.220Mi ± 0% 5.237Mi ± 0% +0.32% (n=1)
Compile/binary 22.92Mi ± 0% 22.94Mi ± 0% +0.07% (n=1)
Compile/text 8.428Mi ± 0% 8.435Mi ± 0% +0.08% (n=1)
Updates #56102.
Change-Id: I6e938d596992ffb473cf51e7e598f372ce08deb0
Reviewed-on: https://go-review.googlesource.com/c/go/+/484860
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
This reverts commit http://go.dev/cl/c/482356.
Reason for revert: Reverting this change again, since it is causing additional failures in google-internal testing.
Change-Id: I9234946f62e5bb18c2f873a65e8b298d04af0809
Reviewed-on: https://go-review.googlesource.com/c/go/+/484735
Reviewed-by: Florian Zenker <floriank@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
Auto-Submit: Than McIntosh <thanm@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
[This is a roll-forward of CL 479095, which was reverted due to a bad
interaction between inlining and escape analysis since fixed in CL 482355.]
Currently, when the inliner is determining if a function is
inlineable, it descends into the bodies of closures constructed by
that function. This has several unfortunate consequences:
- If the closure contains a disallowed operation (e.g., a defer), then
the outer function can't be inlined. It makes sense that the
*closure* can't be inlined in this case, but it doesn't make sense
to punish the function that constructs the closure.
- The hairiness of the closure counts against the inlining budget of
the outer function. Since we currently copy the closure body when
inlining the outer function, this makes sense from the perspective
of export data size and binary size, but ultimately doesn't make
much sense from the perspective of what should be inlineable.
- Since the inliner walks into every closure created by an outer
function in addition to starting a walk at every closure, this adds
an n^2 factor to inlinability analysis.
This CL simply drops this behavior.
In std, this makes 57 more functions inlinable, and disallows inlining
for 10 (due to the basic instability of our bottom-up inlining
approach), for an net increase of 47 inlinable functions (+0.6%).
This will help significantly with the performance of the functions to
be added for #56102, which have a somewhat complicated nesting of
closures with a performance-critical fast path.
The downside of this seems to be a potential increase in export data
and text size, but the practical impact of this seems to be
negligible:
│ before │ after │
│ bytes │ bytes vs base │
Go/binary 15.12Mi ± 0% 15.14Mi ± 0% +0.16% (n=1)
Go/text 5.220Mi ± 0% 5.237Mi ± 0% +0.32% (n=1)
Compile/binary 22.92Mi ± 0% 22.94Mi ± 0% +0.07% (n=1)
Compile/text 8.428Mi ± 0% 8.435Mi ± 0% +0.08% (n=1)
Updates #56102.
Change-Id: I1f4fc96c71609c8feb59fecdb92b69ba7e3b5b41
Reviewed-on: https://go-review.googlesource.com/c/go/+/482356
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Currently, when the inliner is determining if a function is
inlineable, it descends into the bodies of closures constructed by
that function. This has several unfortunate consequences:
- If the closure contains a disallowed operation (e.g., a defer), then
the outer function can't be inlined. It makes sense that the
*closure* can't be inlined in this case, but it doesn't make sense
to punish the function that constructs the closure.
- The hairiness of the closure counts against the inlining budget of
the outer function. Since we currently copy the closure body when
inlining the outer function, this makes sense from the perspective
of export data size and binary size, but ultimately doesn't make
much sense from the perspective of what should be inlineable.
- Since the inliner walks into every closure created by an outer
function in addition to starting a walk at every closure, this adds
an n^2 factor to inlinability analysis.
This CL simply drops this behavior.
In std, this makes 57 more functions inlinable, and disallows inlining
for 10 (due to the basic instability of our bottom-up inlining
approach), for an net increase of 47 inlinable functions (+0.6%).
This will help significantly with the performance of the functions to
be added for #56102, which have a somewhat complicated nesting of
closures with a performance-critical fast path.
The downside of this seems to be a potential increase in export data
and text size, but the practical impact of this seems to be
negligible:
│ before │ after │
│ bytes │ bytes vs base │
Go/binary 15.12Mi ± 0% 15.14Mi ± 0% +0.16% (n=1)
Go/text 5.220Mi ± 0% 5.237Mi ± 0% +0.32% (n=1)
Compile/binary 22.92Mi ± 0% 22.94Mi ± 0% +0.07% (n=1)
Compile/text 8.428Mi ± 0% 8.435Mi ± 0% +0.08% (n=1)
Change-Id: Ie9e38104fed5689a94c368288653fd7cb4b7a35e
Reviewed-on: https://go-review.googlesource.com/c/go/+/479095
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Auto-Submit: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
The mismatch between Unified IR and the old frontend is not about how
they number the closures, but how they name them. For nested closure,
the old frontend use the immediate function which contains the closure
as the outer function, while Unified IR uses the outer most function as
the outer for all closures.
That said, what important is matching the number of closures, not their
name prefix. So this CL relax the test to match both "main.func1.func2"
and "main.func1.2" to satisfy both Unified IR and the old frontend.
Updates #53058
Change-Id: I66ed816d1968aa68dd3089a4ea5850ba30afd75b
Reviewed-on: https://go-review.googlesource.com/c/go/+/437216
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Auto-Submit: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
This patch revises the fix for issue 46234, fixing a bug that was
accidentally introduced by CL 320913. When inlining a chunk of code
with a closure expression, we want to avoid updating the source
positions in the function being closed over, but we do want to update
the position for the ClosureExpr itself (since it is part of the
function we are inlining). CL 320913 unintentionally did away with the
closure expr source position update; here we restore it again.
Updates #46234.
Fixes#49171.
Change-Id: Iaa51bc498e374b9e5a46fa0acd7db520edbbbfca
Reviewed-on: https://go-review.googlesource.com/c/go/+/366494
Trust: Than McIntosh <thanm@google.com>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
When inlining functions with closures, ensure that we don't mark the
body of the closure with a src.Pos marker that reflects the inline,
since this will result in the generation of an inltree table for the
closure itself (as opposed to the routine that the func-with-closure
was inlined into).
Fixes#46234.
Change-Id: I348296de6504fc4745d99adab436640f50be299a
Reviewed-on: https://go-review.googlesource.com/c/go/+/320913
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Than McIntosh <thanm@google.com>
For the new export/import of node types, we were just missing setting
the types of the closure variables (which have the same types as the
captured variables) and the OCLOSURE node itself (which has the same
type as the Func node).
Re-enabled inlining of functions with closures.
Change-Id: I687149b061f3ffeec3244ff02dc6e946659077a9
Reviewed-on: https://go-review.googlesource.com/c/go/+/308974
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Include type information on exported function bodies, so that the
importer does not have to re-typecheck the body. This involves
including type information in the encoded output, as well as
avoiding some of the opcode rewriting and other changes that the
old exporter did assuming there would be a re-typechecking pass.
This CL could be considered a cleanup, but is more important than that
because it is an enabling change for generics. Without this CL, we'd
have to upgrade the current typechecker to understand generics. With
this CL, the current typechecker can mostly go away in favor of the
types2 typechecker.
For now, inlining of functions that contain closures is turned off.
We will hopefully resolve this before freeze.
Object files are only 0.07% bigger.
Change-Id: I85c9da09f66bfdc910dc3e26abb2613a1831634d
Reviewed-on: https://go-review.googlesource.com/c/go/+/301291
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
I have exporting, importing, and inlining of functions with closures
working in all cases (issue #28727). all.bash runs successfully without
errors.
Approach:
- Write out the Func type, Dcls, ClosureVars, and Body when exporting
an OCLOSURE.
- When importing an OCLOSURE, read in the type, dcls, closure vars,
and body, and then do roughly equivalent code to (*noder).funcLit
- During inlining of a closure within inlined function, create new
nodes for all params and local variables (including closure
variables), so they can have a new Curfn and some other field
values. Must substitute not only on the Nbody of the closure, but
also the Type, Cvars, and Dcl fields.
Fixes#28727
Change-Id: I4da1e2567c3fa31a5121afbe82dc4e5ee32b3170
Reviewed-on: https://go-review.googlesource.com/c/go/+/283112
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Currently, typecheck leaves arguments to OPANIC as their original
type. This CL changes it to insert implicit OCONVIFACE operations to
convert arguments to `interface{}` like how any other function call
would be handled.
No immediate benefits, other than getting to remove a tiny bit of
special-case logic in order.go's handling of OPANICs. Instead, the
generic code path for handling OCONVIFACE is used, if necessary.
Longer term, this should be marginally helpful for #43753, as it
reduces the number of cases where we need values to be addressable for
runtime calls.
However, this does require adding some hacks to appease existing
tests:
1. We need yet another kludge in inline budgeting, to ensure that
reflect.flag.mustBe stays inlinable for cmd/compile/internal/test's
TestIntendedInlining.
2. Since the OCONVIFACE expressions are now being introduced during
typecheck, they're now visible to escape analysis. So expressions like
"panic(1)" are now seen as "panic(interface{}(1))", and escape
analysis warns that the "interface{}(1)" escapes to the heap. These
have always escaped to heap, just now we're accurately reporting about
it.
(Also, unfortunately fmt.go hides implicit conversions by default in
diagnostics messages, so instead of reporting "interface{}(1) escapes
to heap", it actually reports "1 escapes to heap", which is
confusing. However, this confusing messaging also isn't new.)
Change-Id: Icedf60e1d2e464e219441b8d1233a313770272af
Reviewed-on: https://go-review.googlesource.com/c/go/+/284412
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Matthew Dempsky <mdempsky@google.com>
We already allow inlining "if" and "goto" statements, so we might as
well allow "for" loops too. The majority of frontend support is
already there too.
The critical missing feature at the moment is that inlining doesn't
properly reassociate OLABEL nodes with their control statement (e.g.,
OFOR) after inlining. This eventually causes SSA construction to fail.
As a workaround, this CL only enables inlining for unlabeled "for"
loops. It's left to a (yet unplanned) future CL to add support for
labeled "for" loops.
The increased opportunity for inlining leads to a small growth in
binary size. For example:
$ size go.old go.new
text data bss dec hex filename
9740163 320064 230656 10290883 9d06c3 go.old
9793399 320064 230656 10344119 9dd6b7 go.new
Updates #14768.
Fixes#41474.
Change-Id: I827db0b2b9d9fa2934db05caf6baa463f0cd032a
Reviewed-on: https://go-review.googlesource.com/c/go/+/256459
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
For most nodes (e.g., OPTRLIT, OMAKESLICE, OCONVIFACE), escape
analysis prints "escapes to heap" or "does not escape" to indicate
whether that node's allocation can be heap or stack allocated.
These messages are also emitted for OADDR, even though OADDR does not
actually allocate anything itself. Moreover, it's redundant because
escape analysis already prints "moved to heap" diagnostics when an
OADDR node like "&x" causes x to require heap allocation.
Because OADDR nodes don't allocate memory, my escape analysis rewrite
doesn't naturally emit the "escapes to heap" / "does not escape"
diagnostics for them. It's also non-trivial to replicate the exact
semantics esc.go uses for OADDR.
Since there are so many of these messages, I'm disabling them in this
CL by themselves. I modified esc.go to suppress the Warnl calls
without any other behavior changes, and then used a shell script to
automatically remove any ERROR messages mentioned by run.go in
"missing error" or "no match for" lines.
Fixes#16300.
Updates #23109.
Change-Id: I3993e2743c3ff83ccd0893f4e73b366ff8871a57
Reviewed-on: https://go-review.googlesource.com/c/go/+/170319
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
This is a simple tweak to allow a bit more mid-stack inlining.
In cases like this:
func f() {
g()
}
We'd really like to inline f into its callers. It can't hurt.
We implement this optimization by making calls a bit cheaper, enough
to afford a single call in the function body, but not 2.
The remaining budget allows for some argument modification, or perhaps
a wrapping conditional:
func f(x int) {
g(x, 0)
}
func f(x int) {
if x > 0 {
g()
}
}
Update #19348
Change-Id: Ifb1ea0dd1db216c3fd5c453c31c3355561fe406f
Reviewed-on: https://go-review.googlesource.com/c/147361
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: David Chase <drchase@google.com>
This line of the inlining tuning experiment
https://go-review.googlesource.com/c/go/+/109918/1/src/cmd/compile/internal/gc/inl.go#347
was incorrectly rewritten in a later patch to use the call
cost, not the panic cost, and thus the inlining of panic
didn't occur when it should. I discovered this when I
realized that tests should have failed, but didn't.
Fix is to make the correct change, and also to modify the
tests that this causes to fail. One test now asserts the
new normal, the other calls "ppanic" instead which is
designed to behave like panic but not be inlined.
Change-Id: I423bb7f08bd66a70d999826dd9b87027abf34cdf
Reviewed-on: https://go-review.googlesource.com/116656
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
CL 65071 enabled inlining for local closures with no captures.
To determine safety of inlining a call sites, we check whether the
variable holding the closure has any assignments after its original
definition.
Unfortunately, that check did not catch OAS2MAPR and OAS2DOTTYPE,
leading to incorrect inlining when a variable holding a closure was
subsequently reassigned through a type conversion or a 2-valued map
access.
There was another more subtle issue wherein reassignment check would
always return a false positive for closure calls inside other
closures. This was caused by the Name.Curfn field of local variables
pointing to the OCLOSURE node instead of the corresponding ODCLFUNC,
which resulted in reassigned walking an empty Nbody and thus never
seeing any reassignments.
This CL fixes these oversights and adds many more tests for closure
inlining which ensure not only that inlining triggers but also the
correctness of the resulting code.
Updates #15561
Change-Id: I74bdae849c4ecfa328546d6d62b512e8d54d04ce
Reviewed-on: https://go-review.googlesource.com/75770
Reviewed-by: Hugues Bruant <hugues.bruant@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
