Add generic rule to rewrite the single-precision square root expression
with one single-precision instruction. The optimization will reduce two
times of precision converting between double-precision and single-precision.
On arm64 flatform.
previous:
FCVTSD F0, F0
FSQRTD F0, F0
FCVTDS F0, F0
optimized:
FSQRTS S0, S0
And this patch adds the test case to check the correctness.
This patch refers to CL 241877, contributed by Alice Xu
(dianhong.xu@arm.com)
Change-Id: I6de5d02281c693017ac4bd4c10963dd55989bd7e
Reviewed-on: https://go-review.googlesource.com/c/go/+/276873
Trust: fannie zhang <Fannie.Zhang@arm.com>
Run-TryBot: fannie zhang <Fannie.Zhang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
On import, make sure that an empty closure is represented as a single
empty block statement. Otherwise, the closure is dropped. Block
statements are not exported explicitly, so must recreate on import.
Fixes#44330
Change-Id: I061598f0f859dd71d2d0cbd10c77cdd81525d1f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/297569
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
One of escape analysis's responsibilities is to summarize whether/how
each function parameter flows to the heap so we can correctly
incorporate those flows into callers' escape analysis data flow
graphs.
As an optimization, we separately record when parameters flow to
result parameters, so that we can more precisely analyze parameter
flows based on how the results are used at the call site. However, if
a named result parameter itself needs to be heap allocated, this
optimization isn't safe and the parameter needs to be recorded as
flowing to heap rather than flowing to result.
Escape analysis used to get this correct because it conservatively
rewalked the data-flow graph multiple times. So even though it would
incorrectly record the result parameter flow, it would separately find
a flow to the heap. However, CL 196811 (specifically, case 3)
optimized the walking logic to reduce unnecessary rewalks causing us
to stop finding the extra heap flow.
This CL fixes the issue by correcting location.leakTo to be sensitive
to sink.escapes and not record result-flows when the result parameter
escapes to the heap.
Fixes#44614.
Change-Id: I48742ed35a6cab591094e2d23a439e205bd65c50
Reviewed-on: https://go-review.googlesource.com/c/go/+/297289
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
A function returning multiple results, some of them zero-width,
will have more than one result present at an offset. Be sure
that offset AND type match.
Includes test.
Change-Id: I3eb1f56116d989b4e73f533fefabb1bf554c901b
Reviewed-on: https://go-review.googlesource.com/c/go/+/297169
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
- Deal with closures in generic functions by fixing the stenciling code
- Deal with instantiated function values (instantiated generic
functions that are not immediately called) during stenciling. This
requires changing the OFUNCINST node to an ONAME node for the
appropriately instantiated function. We do this in a second pass,
since this is uncommon, but requires editing the tree at multiple
levels.
- Check global assignments (as well as functions) for generic function
instantiations.
- Fix a bug in (*subst).typ where a generic type in a generic function
may definitely not use all the type args of the function, so we need
to translate the rparams of the type based on the tparams/targs of
the function.
- Added new test combine.go that tests out closures in generic
functions and instantiated function values.
- Added one new variant to the settable test.
- Enabling inlining functions with closures for -G=3. (For now, set
Ntype on closures in -G=3 mode to keep compatibility with later parts
of compiler, and allow inlining of functions with closures.)
Change-Id: Iea63d5704c322e42e2f750a83adc8b44f911d4ec
Reviewed-on: https://go-review.googlesource.com/c/go/+/296269
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
In CL 253457, we did the same fix for direct function calls. But for
method calls, the receiver argument also need to be passed through the
wrapper function, which we are not doing so the compiler crashes with
the code in #44415.
It will be nicer if we can rewrite OCALLMETHOD to normal OCALLFUNC, but
that will be for future CL. The passing receiver argument to wrapper
function is easier for backporting to go1.16 branch.
Fixes#44415
Change-Id: I03607a64429042c6066ce673931db9769deb3124
Reviewed-on: https://go-review.googlesource.com/c/go/+/296490
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
ComputeAddrtaken needs to descend into closures, now that imported
bodies can include closures. The bug was that we weren't properly
setting Addrtaken for a variable inside a closure inside a function that
we were importing.
For now, still disable inlining of functions with closures for -G mode.
I'll enable it in a later change -- there are just a few fixes related
to the fact that we don't need to set Ntype for closure functions.
Added a test derived from the cilium repro in the issue.
Fixes#44370
Change-Id: Ida2a403636bf8740b471b3ad68b5474951811e19
Reviewed-on: https://go-review.googlesource.com/c/go/+/296649
Run-TryBot: Dan Scales <danscales@google.com>
Trust: 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>
This reverts commit CL 294849.
Reason for revert: this doesn't actually fix the issue, as revealed
by the noopt builder's failures.
Change-Id: Ib4ea9ceb4d75e46b3b91ec348b365fd8c83316ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/296629
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
The code for delayed declaration of inlined result parameters only
handles non-empty return statements. This is generally okay, because
we already early declare if there are any (non-blank) named result
parameters.
But if a user writes a function with only blank result parameters and
with exactly one return statement, which is empty, then they could end
up hitting the dreaded "Value live at entry" ICE.
This CL fixes the issue by ensuring we always early declare inlined
result parameters if there are any empty return statements.
Fixes#44355.
Change-Id: I315f3853be436452883b1ce31da1bdffdf24d506
Reviewed-on: https://go-review.googlesource.com/c/go/+/293293
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: David Chase <drchase@google.com>
In CL 253457, we did the same fix for direct function calls. But for
method calls, the receiver argument also need to be passed through the
wrapper function, which we are not doing so the compiler crashes with
the code in #44415.
As we already rewrite t.M(...) into T.M(t, ...) during walkCall1, to fix
this, we can do the same trick in wrapCall, so the receiver argument
will be treated as others.
Fixes#44415
Change-Id: I396182983c85d9c5e4494657da79d25636e8a079
Reviewed-on: https://go-review.googlesource.com/c/go/+/294849
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
bug511.dir/b.go:10:14: error: reference to undefined field or method 'M'
Change-Id: I9f96dc5c7254b310bc3e15b0bc588d62718cb4b2
Reviewed-on: https://go-review.googlesource.com/c/go/+/292009
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
math.Float32bits was not being inlined across package boundaries.
Create a private func that can be inlined with -l.
Change-Id: Ic50bf4727dd8ade09d011eb204006b7ee88db34a
Reviewed-on: https://go-review.googlesource.com/c/go/+/295989
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Added a flag '-d=inlfuncswithclosures=1' to allow inlining functions with
closures, and change the default to off for now, until #44370 is fixed.
Updates #44370.
Change-Id: Ic17723aa5c091d91f5f5004d8b63ec7125257acf
Reviewed-on: https://go-review.googlesource.com/c/go/+/296049
Run-TryBot: Dan Scales <danscales@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
The code generated when storing eight bytes loaded from memory created a
series of small writes instead of a single, large one. The specific
pattern of instructions generated stored 1 byte, then 2 bytes, then 4
bytes, and finally 1 byte.
The new rules match this specific pattern both for amd64 and for s390x,
and convert it into a single instruction to store the 8 bytes. arm64 and
ppc64le already generated the right code, but the new codegen test
covers also those architectures.
Fixes#41663
Change-Id: Ifb9b464be2d59c2ed5034acf7b9c3e473f344030
Reviewed-on: https://go-review.googlesource.com/c/go/+/280456
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Plumb abi information into ssa/ssagen for plain calls
and plain functions (not methods). Does not extend all the
way through the compiler (yet).
One test disabled because it extends far enough to break the test.
Normalized all the compiler's register args TODOs to
// TODO(register args) ...
For #40724.
Change-Id: I0173a4579f032ac3c9db3aef1749d40da5ea01ff
Reviewed-on: https://go-review.googlesource.com/c/go/+/293389
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
OCONVOP doesn't have effect in the compiled code so, it can be safely
excluded from inline cost calculation.
Also make sequence ODEREF OCONVNOP* OADDR cost 1. This is rather common
conversion, such as *(*uint32)(unsafe.Pointer(&x)).
Fixes#42788
Change-Id: I5001f7e89d985c198b6405694cdd5b819cf3f47a
Reviewed-on: https://go-review.googlesource.com/c/go/+/281232
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Elias Naur <mail@eliasnaur.com>
Some bit test instruction generation stopped triggering after
the change to addressing modes. I suspect this was just because
ANDQload was being generated before the rewrite rules could discover
the BTQ. Fix that by decomposing the ANDQload when it is surrounded
by a TESTQ (thus re-enabling the BTQ rules).
Fixes#44228
Change-Id: I489b4a5a7eb01c65fc8db0753f8cec54097cadb2
Reviewed-on: https://go-review.googlesource.com/c/go/+/291749
Trust: Keith Randall <khr@golang.org>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Currently, we keep track of slice len by mapping from slice ID to
len/cap SSA value. However, slice len/cap can have multiple SSA values,
so when updating fact table for slice len/cap, we only update in one
place.
Instead, we can take advantage of the transitive relations provided by
poset. So all duplicated slice lens are set as equal to one another.
When updating fact table for one, that fact will be reflected to all
others. The same mechanism is applied for slice cap.
Removes 15 bounds checks from std/cmd.
Fixes#42603
Change-Id: I32c07968824cc33765b1e441b3ae2c4b5f5997c3
Reviewed-on: https://go-review.googlesource.com/c/go/+/273670
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
CL 274294 improved findTypeLoop but also fixed a new found bug on master
branch. This Cl adds test cases for this.
Updates #44266
Change-Id: Ie4a07a3487758a1e4ad2f2847dcde975b10d2a77
Reviewed-on: https://go-review.googlesource.com/c/go/+/292889
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
For import of functions with closures, the connections among closure
variables are constructed on-the-fly via CaptureName(). For multiple
nested closures, we need to temporarily set r.curfn to each closure we
construct, so that the processing of closure variables will be correct
for any nested closure inside that closure.
Fixes#44335
Change-Id: I34f99e2822250542528ff6b2232bf36756140868
Reviewed-on: https://go-review.googlesource.com/c/go/+/294212
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
One CFGs that shortcircuit looks for is:
p q
\ /
b
/ \
t u
The test case creates a CFG like that in which p == t.
That caused the compiler to generate a (short-lived) invalid phi value.
Fix this with a relatively big hammer: Disallow single-length loops entirely.
This is probably overkill, but it such loops are very rare.
This doesn't change the generated code for anything in std.
It generates worse code for the test case:
It no longer compiles the entire function away.
Fixes#44465
Change-Id: Ib8cdcd6cc9d7f48b4dab253652038ace24eae152
Reviewed-on: https://go-review.googlesource.com/c/go/+/295130
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
It was fixed by CL 294289, for #44378.
This is a different style of test that uses
line directives instead of extremely long lines.
Fixes#38698.
Change-Id: I50a1585030978b35fffa9981d6ed96b99216dc3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/295129
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
DWARF generation uses variable source positions (file/line/col) as a
way to uniquely identify locals and parameters, as part of the process
of matching up post-optimization variables with the corresponding
pre-optimization versions (since the DWARF needs to be in terms of the
original source constructs).
This strategy can run into problems when compiling obfuscated or
machine-generated code, where you can in some circumstances wind up
with two local variables that appear to have the same name, file,
line, and column. This patch changes DWARF generation to skip over
such duplicates as opposed to issuing a fatal error (if an
obfuscation tool is in use, it is unlikely that a human being will be
able to make much sense of DWARF info in any case).
Fixes#44378.
Change-Id: I198022d184701aa9ec3dce42c005d29b72d2e321
Reviewed-on: https://go-review.googlesource.com/c/go/+/294289
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
By default, when typechecking a closure, tcClosure() creates a new
closure function. This should really be done separate from typechecking.
For now, we explicitly avoid creating a new closure function when
typechecking an inline body (in ImportedBody). However, the heuristic
for determining when we are typechecking an inline body was not correct
for double nested closures in an inline body, since CurFunc will then be
the inner closure, which has a body.
So, use a simple global variable to indicate when we typechecking an
inline body. The global variable is fine (just like ir.CurFunc), since
the front-end runs serially.
Fixes#44325
Change-Id: If2829fe1ebb195a7b1a240192b57fe6f04d1a36b
Reviewed-on: https://go-review.googlesource.com/c/go/+/294211
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
When generating DWARF inlined info records, it's possible to have a
local function whose only callsites are inlined away, meaning that we
emit an abstract function DIE but no regular subprogram DIE. When
emitting DWARF scope info we need to handle this case (specifically
when scoping PCs, check for the case that the func in question has
been entirely deleted).
Fixes#44344.
Change-Id: I9f5bc692f225aa4c5c23f7bd2e50bcf7fe4fc5f3
Reviewed-on: https://go-review.googlesource.com/c/go/+/293309
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
The address space starts at 4GB, so dummy is too far out.
This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
Change-Id: I5f67e268ce729086d9f9fc8541722fabccfd0145
Reviewed-on: https://go-review.googlesource.com/c/go/+/288823
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
A type may now have a type param in it, either because it has been
composed from a function type param, or it has been declared as or
derived from a reference to a generic type. No objects or types with
type params can be exported yet. No generic type has a runtime
descriptor (but will likely eventually be associated with a dictionary).
types.Type now has an RParam field, which for a Named type can specify
the type params (in order) that must be supplied to fully instantiate
the type. Also, there is a new flag HasTParam to indicate if there is
a type param (TTYPEPARAM) anywhere in the type.
An instantiated generic type (whether fully instantiated or
re-instantiated to new type params) is a defined type, even though there
was no explicit declaration. This allows us to handle recursive
instantiated types (and improves printing of types).
To avoid the need to transform later in the compiler, an instantiation
of a method of a generic type is immediately represented as a function
with the method as the first argument.
Added 5 tests on generic types to test/typeparams, including list.go,
which tests recursive generic types.
Change-Id: Ib7ff27abd369a06d1c8ea84edc6ca1fd74bbb7c2
Reviewed-on: https://go-review.googlesource.com/c/go/+/292652
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
This was a mostly clean merge, with the exception of codereview.cfg and
changes in src/go/types.
codereview.cfg for dev.typeparams is preserved in this CL. It should be
deleted before merging back to master.
The go/types changes were merged manually. For the most part this
involved taking the union of patches, with the following exceptions:
+ declInfo.aliasPos is removed, as it is not necessary in
dev.typeparams where we have access to the full TypeSpec.
+ Checker.overflow is updated to use the asBasic converter.
+ A TODO is added to errorcodes.go to ensure that go1.16 error codes
are preserved.
Change-Id: If9595196852e2163e27a9478df1e7b2c3704947d
Currently, we call Warnl in SSA backend when we see a function
(defined or called) with regparams pragma. Calling Warnl in
concurrent environment is racy. As the debugging output is
temporary, for testing purposes we just pass -c=1. We'll remove
the pragma and the debugging print some time soon.
Change-Id: I6f925a665b953259453fc458490c5ff91f67c91a
Reviewed-on: https://go-review.googlesource.com/c/go/+/291710
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Conflicts:
- src/cmd/compile/internal/gc/walk.go
gc/walk.go is changed in CL 290950 on the master branch but
deleted in the dev.regabi branch and moved over to the walk
package. This CL includes porting CL 290950 over to the new
walk.
Merge List:
+ 2021-02-12 ff0e93ea31 doc/go1.16: note that package path elements beginning with '.' are disallowed
+ 2021-02-11 249da7ec02 CONTRIBUTORS: update for the Go 1.16 release
+ 2021-02-11 864d4f1c6b cmd/go: multiple small 'go help' fixes
+ 2021-02-11 26ceae85a8 spec: More precise wording in section on function calls.
+ 2021-02-11 930c2c9a68 cmd/go: reject embedded files that can't be packed into modules
+ 2021-02-11 e5b08e6d5c io/fs: allow backslash in ValidPath, reject in os.DirFS.Open
+ 2021-02-10 ed8079096f cmd/compile: mark concrete call of reflect.(*rtype).Method as REFLECTMETHOD
+ 2021-02-09 e9c9683597 cmd/go: suppress errors from 'go get -d' for packages that only conditionally exist
+ 2021-02-09 e0ac989cf3 archive/tar: detect out of bounds accesses in PAX records resulting from padded lengths
+ 2021-02-09 c9d6f45fec runtime/metrics: fix a couple of documentation typpos
+ 2021-02-09 cea4e21b52 io/fs: backslash is always a glob meta character
+ 2021-02-08 dc725bfb3c doc/go1.16: mention new vet check for asn1.Unmarshal
Change-Id: Ib28fffa7dfbff7f6cdbfaf4a304757fead7bbf19
- Create the stencil name using targ.Type.String(), which handles cases
where, for example, a type argument is a pointer to a named type,
etc. *obj.
- Set name.Def properly for a new stenciled func (have the symbol point
back to the associated function node). Will be required when exporting.
- Add missing copying of Func field when making copies of Name nodes.
(On purpose (it seems), Name nodes don't have a copy() function, so
we have to copy all the needed fields explicitly.)
- Deal with nil type in subster.node(), which is the type of the return
value for a function that doesn't return anything.
- Fix min to match standard want/go form, and add in float tests. Changed
Got -> got in bunch of other typeparam tests.
- Add new tests index.go, settable.go, and smallest.go (similar to
examples in the type param proposal), some of which need the above
changes.
Change-Id: I09a72302bc1fd3635a326da92405222afa222e85
Reviewed-on: https://go-review.googlesource.com/c/go/+/291109
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
For functions that call reflect.Type.Method (or MethodByName), we
mark it as REFLECTMETHOD, which tells the linker that methods
can be retrieved via reflection and the linker keeps all exported
methods live. Currently, this marking expects exactly the
interface call reflect.Type.Method (or MethodByName). But now the
compiler can devirtualize that call to a concrete call
reflect.(*rtype).Method (or MethodByName), which is not handled
and causing the linker to discard methods too aggressively.
Handle the latter in this CL.
Fixes#44207.
Change-Id: Ia4060472dbff6ab6a83d2ca8e60a3e3f180ee832
Reviewed-on: https://go-review.googlesource.com/c/go/+/290950
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Add respective check to type checker.
Enables another excluded test in test/run.go.
This CL completes the currently required checks for
language compatibility in types2.
Updates #31793.
Change-Id: Icececff9e6023d38f600c93bcb54cdcafcf501b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/290911
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
When doing a type conversion using a type param, delay the
transformation to OCONV/OCONVNOP until stenciling, since the nodes
created depend on the actual type.
Re-enable the fact.go test.
Change-Id: I3d5861aab3dd0e781d767f67435afaf951dfe451
Reviewed-on: https://go-review.googlesource.com/c/go/+/290752
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
- Have to delay the extra transformation on methods invoked on a type
param, since the actual transformation (including path through
embedded fields) will depend on the instantiated type. I am currently
doing the transformation during the stencil substitution phase. We
probably should have a separate pass after noder2 and stenciling,
which drives the extra transformations that were in the old
typechecker.
- We handle method values (that are not called) and method calls. We
don't currently handle method expressions.
- Handle type substitution in function types, which is needed for
function args in generic functions.
- Added stringer.go and map.go tests, testing the above changes
(including constraints with embedded interfaces).
Change-Id: I3831a937d2b8814150f75bebf9f23ab10b93fa00
Reviewed-on: https://go-review.googlesource.com/c/go/+/290550
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
This is a port of CL 287494 to go/types. The additional checks in
test/fixedbugs are included, though they won't be executed by go/types.
Support for errorcheckdir checks will be added to go/types in a later
CL.
Change-Id: I37e202ea5daf7d7b8fc6ae93a4c4dbd11762480f
Reviewed-on: https://go-review.googlesource.com/c/go/+/290570
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
This involved a couple non-trivial fixes in go/types:
- move the check for main function signature to resolver.go, to be
consistent with init. Also, update uses of _InvalidInitSig to
_InvalidInitDecl, consistent with what we decided for dev.regabi.
- Update some tests in api_test.go which newly fail after CL 289715
(fixing reporting of untyped nil) In all cases but one, these updates
were consistent with types2. However, in one case types2 seems to be
able to resolve more type information than go/types for a broken
package. I left a TODO to investigate this further.
Change-Id: I8244b7c81654194edd5af8de689a13c262117dff
Disabled test/typeparam/fact.go for now as there's an issue
with stenciling.
Change-Id: Ie328a217de6d7b6695737f08ef5c944bcdaabd39
Reviewed-on: https://go-review.googlesource.com/c/go/+/290471
Trust: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
- Handle generic function calling itself or another generic function in
stenciling. This is easy - after it is created, just scan an
instantiated generic function for function instantiations (that may
needed to be stenciled), just like non-generic functions. The types
in the function instantiation will already have been set by the
stenciling.
- Handle OTYPE nodes in subster.node() (allows for generic type
conversions).
- Eliminated some duplicated work in subster.typ().
- Added new test case fact.go that tests a generic function calling
itself, and simple generic type conversions.
- Cause an error if a generic function is to be exported (which we
don't handle yet).
- Fixed some suggested changes in the add.go test case that I missed in
the last review.
Change-Id: I5d61704254c27962f358d5a3d2e0c62a5099f148
Reviewed-on: https://go-review.googlesource.com/c/go/+/290469
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Allow full compilation and running of simple programs with generic
functions by stenciling on the fly the needed generic functions. Deal
with some simple derived types based on type params.
Include a few new typeparam tests min.go and add.go which involve
fully compiling and running simple generic code.
Change-Id: Ifc2a64ecacdbd860faaeee800e2ef49ffef9df5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/289630
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
This enables another test.
Change-Id: I80763b97d939e225158a083299b2e0d189268bc7
Reviewed-on: https://go-review.googlesource.com/c/go/+/289569
Trust: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
This matches the compiler's existing limitations and thus ensures
that types2 reports the same errors for oversize integer constants.
Change-Id: I4fb7c83f3af69098d96f7b6c53dbe3eaf6ea9ee4
Reviewed-on: https://go-review.googlesource.com/c/go/+/288633
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
In ABIInternal, reserve X15 as constant zero, and use it to zero
memory. (Maybe there can be more use of it?)
The register is zeroed when transition to ABIInternal from ABI0.
Caveat: using X15 generates longer instructions than using X0.
Maybe we want to use X0?
Change-Id: I12d5ee92a01fc0b59dad4e5ab023ac71bc2a8b7d
Reviewed-on: https://go-review.googlesource.com/c/go/+/288093
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Change-Id: I6290bc4921ef17586b5028d3f40a88372b175014
Reviewed-on: https://go-review.googlesource.com/c/go/+/289269
Trust: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
The compiler uses 512 bit of precision for untyped constant
arithmetic but didn't restrict the length of incoming constant
literals in any way, possibly opening the door for excessively
long constants that could bring compilation to a crawl.
Add a simple check that refuses excessively long constants.
Add test.
Change-Id: I797cb2a8e677b8da2864eb92d686d271ab8a004d
Reviewed-on: https://go-review.googlesource.com/c/go/+/289049
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
