]> Cypherpunks repositories - gostls13.git/log
gostls13.git
3 years agocmd/compile: delay fillinMethods to deal with mutually-recursive types
Dan Scales [Wed, 18 Aug 2021 13:28:40 +0000 (06:28 -0700)]
cmd/compile: delay fillinMethods to deal with mutually-recursive types

We need to delay fillinMethods until we get to a top-level type, so we
know all the TFORW types have been filled in, and we can do the
substitutions required by fillinMethods.

Fixes #47710

Change-Id: I298de7e7753ed31a2c2b1ff04f35177a8afc7a66
Reviewed-on: https://go-review.googlesource.com/c/go/+/345149
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
3 years agocmd/compile: eliminate repetitive code
wangyuntao [Fri, 16 Jul 2021 10:29:05 +0000 (10:29 +0000)]
cmd/compile: eliminate repetitive code

Change-Id: I02c8b65f7c1c1606c9964ab6c54d5ab5f1b444a5
GitHub-Last-Rev: 3d740b9ac129d38981ee295456d2d7f803a79b77
GitHub-Pull-Request: golang/go#47242
Reviewed-on: https://go-review.googlesource.com/c/go/+/334990
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>

3 years agocmd/compile: add types.RecalcSize
Matthew Dempsky [Thu, 26 Aug 2021 22:54:44 +0000 (15:54 -0700)]
cmd/compile: add types.RecalcSize

This is the only case where Align is assigned outside of package
types. Rather than adding a SetAlign method, adding a RecalcSize
function is a bit more descriptive.

Change-Id: I1b3c01ebd0e41183665baa63c926592865bbbd0b
Reviewed-on: https://go-review.googlesource.com/c/go/+/345479
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: remove ssagen/pgen_test.go
Matthew Dempsky [Thu, 26 Aug 2021 22:30:40 +0000 (15:30 -0700)]
cmd/compile: remove ssagen/pgen_test.go

This test was written in 2015, and hasn't had any meaningful changes
to it since. However, it's very ad hoc and pokes into internals that
it shouldn't, and it keeps getting in the way of more meaningful
refactorings. It doesn't even catch bugs; e.g., it didn't catch that
cmpstackvarlt wasn't asymmetric (CL 19778), and a bunch of its test
cases are bogus because we never actually use it to sort ir.PFUNC
nodes (assert added in this CL).

The compiler is woefully lacking in unit tests like these, but I
strongly feel this test is more of a hinderance than a help at the
moment.

Change-Id: I3ac7a3e533947aa95f325626c6eec13a43789f9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/345478
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: use typecheck.InitUniverse in unit tests
Matthew Dempsky [Thu, 26 Aug 2021 21:16:24 +0000 (14:16 -0700)]
cmd/compile: use typecheck.InitUniverse in unit tests

Rather than ad hoc setting up the universe, just initialize it
properly.

Change-Id: I18484b952321f55eb3e1e48fd383068a4ee75f66
Reviewed-on: https://go-review.googlesource.com/c/go/+/345475
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: move types init code into package types
Matthew Dempsky [Thu, 26 Aug 2021 20:17:56 +0000 (13:17 -0700)]
cmd/compile: move types init code into package types

This moves the package types setup code from package typecheck into
package types itself. This is a prereq for making types.Type more
opaque, because some unit tests depend on being able to init the basic
universal types.

A few notable details of this CL:

1. Creating the builtin types requires being able to create the
ir.Name/ir.OTYPE that represents it, but package types can't depend on
package ir. So we add a callback function to handle creating the
ir.Name.

2. This CL moves ir.Pkgs.Unsafe to types.UnsafePkg. Package unsafe is
part of the language, not like the other ir.Pkgs packages that are
purely implementation details.

3. This CL also moves typecheck.FakeRecv to types.FakeRecv, addressing
an outstanding TODO.

Change-Id: I64de04ce82fbcd1bb59f547e2eea3cda52d89429
Reviewed-on: https://go-review.googlesource.com/c/go/+/345474
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agocmd/compile/internal/types2: do not declare new methods on instantiated types
Robert Griesemer [Thu, 26 Aug 2021 19:27:06 +0000 (12:27 -0700)]
cmd/compile/internal/types2: do not declare new methods on instantiated types

Report an error if an alias is used to declare a method on an
instantiated type.

Also, when resolving the receiver type, don't use asNamed to
avoid premature expansion of the type.

Fixes #47968.

Change-Id: Ie5acc4cfb1944deaaeeaee98707f31e256f8ef5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/345472
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile/internal/types2: implement TypeList.String (debugging support)
Robert Griesemer [Thu, 26 Aug 2021 18:56:52 +0000 (11:56 -0700)]
cmd/compile/internal/types2: implement TypeList.String (debugging support)

Change-Id: Iaa203def3dac94a7d5ff6120e89315c3d7977ee1
Reviewed-on: https://go-review.googlesource.com/c/go/+/345471
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile: fix reference to generic type needed by crawler
Dan Scales [Wed, 25 Aug 2021 00:35:51 +0000 (17:35 -0700)]
cmd/compile: fix reference to generic type needed by crawler

This problem happens when you create a new local type that uses an
imported generic type (maybe just by instantiating it), and then that
local type needed to be included as part of an export. In that case, the
imported generic type is does not have a declaration in the local
package, so it is not necessarily created in types1, so the
crawler/export doesn't work.

To fix this issue, we just need to add a call to g.obj() for the base
generic type, to make sure that it will exist if needed later in the
compilation or for the crawler during export.

Fixes #47514

Change-Id: Ie756578f07ad0007de8a88ae909cf7534a22936e
Reviewed-on: https://go-review.googlesource.com/c/go/+/345411
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>

3 years agocmd/compile: unexport Type.Vargen
Matthew Dempsky [Thu, 26 Aug 2021 19:05:45 +0000 (12:05 -0700)]
cmd/compile: unexport Type.Vargen

This field is only used outside of packages types in two places, and
they follow the same pattern. So this CL creates a Type.Setvargen
function that they can use instead, so that Type.Vargen can be
unexported.

A bit clumsy, but it works for now.

Change-Id: I7b4f33fac635e2464df2fbc0607ab40902f6f09f
Reviewed-on: https://go-review.googlesource.com/c/go/+/345469
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile/internal/types: unexport Type.Extra
Matthew Dempsky [Thu, 26 Aug 2021 18:46:24 +0000 (11:46 -0700)]
cmd/compile/internal/types: unexport Type.Extra

Not used outside of package types anymore. Let's keep it that.

Change-Id: I69b464ac94edaacd219da4210f7b8618e2beaf70
Reviewed-on: https://go-review.googlesource.com/c/go/+/345413
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
3 years agocmd/compile: change typecheck.iscmp into ir.Op.IsCmp
Matthew Dempsky [Thu, 26 Aug 2021 18:29:44 +0000 (11:29 -0700)]
cmd/compile: change typecheck.iscmp into ir.Op.IsCmp

Change-Id: If89089cbd79b7ff030d856df3a7e6b7862c0f4ec
Reviewed-on: https://go-review.googlesource.com/c/go/+/345412
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agoembed: remove reference to global variables in docs
Dan Kortschak [Mon, 23 Aug 2021 09:43:35 +0000 (19:13 +0930)]
embed: remove reference to global variables in docs

Refering to variable is both redundant since package scope is used, and
incorrect since global variables are not described in the spec.

Change-Id: Ib08a9f072fc800ee36549f758b68167d8f044878
Reviewed-on: https://go-review.googlesource.com/c/go/+/344214
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Alexander Rakoczy <alex@golang.org>
Run-TryBot: Alexander Rakoczy <alex@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile/internal/types2: remove need for instance (struct)
Robert Griesemer [Thu, 26 Aug 2021 04:48:21 +0000 (21:48 -0700)]
cmd/compile/internal/types2: remove need for instance (struct)

instance was only used to hold the instantiation position for
lazy instantiation (and encode the fact that we have a lazy
instantiation). Just use a (pointer to a) syntax.Pos instead.

We could use a syntax.Pos (no pointer) and rely on the fact
that we have a known position (or fake position, if need be)
to indicate lazy instantiation. But using a pointer leads to
a smaller Named struct.

Change-Id: I441a839a125f453ad6c501de1ce499b72a2f67a4
Reviewed-on: https://go-review.googlesource.com/c/go/+/345177
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile/internal/types2: address some TODOs (cleanup)
Robert Griesemer [Thu, 26 Aug 2021 01:13:28 +0000 (18:13 -0700)]
cmd/compile/internal/types2: address some TODOs (cleanup)

- Address some easy TODOs.
- Remove some TODOs that are not correct anymore or are unimportent.
- Simplify some code on the way.

Change-Id: I4d20de3725b3a735022afe022cbc002b2798936d
Reviewed-on: https://go-review.googlesource.com/c/go/+/345176
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocrypto/tls: fix typo in PreferServerCipherSuites comment
vinckr [Thu, 26 Aug 2021 10:59:02 +0000 (10:59 +0000)]
crypto/tls: fix typo in PreferServerCipherSuites comment

Fixing a typo, Deprected -> Deprecated.

Change-Id: Ie0ccc9a57ae6a935b4f67154ac097dba4c3832ec
GitHub-Last-Rev: 57337cc1bfa771111f229e7b899fdfdad3b1655e
GitHub-Pull-Request: golang/go#47745
Reviewed-on: https://go-review.googlesource.com/c/go/+/342791
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
3 years agocmd/go: pass -gcflags after other flags generated by the go command
Cuong Manh Le [Wed, 25 Aug 2021 07:56:01 +0000 (14:56 +0700)]
cmd/go: pass -gcflags after other flags generated by the go command

Otherwise, any gc flags set by user using "-gcflags" won't have effect.

Fixes #47682

Change-Id: Icd365577cba1f64f6c7b8320d0c9019de9f062f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/344909
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: Jay Conrod <jayconrod@google.com>
3 years agocmd/compile/internal/types2: fix type set printing and add test
Robert Griesemer [Wed, 25 Aug 2021 04:12:06 +0000 (21:12 -0700)]
cmd/compile/internal/types2: fix type set printing and add test

Change-Id: I44ca1f889b041467d5febacaf6037cfd75859175
Reviewed-on: https://go-review.googlesource.com/c/go/+/344873
Trust: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile/internal/types2: rename IsMethodSet to IsConstraint (cleanup)
Robert Griesemer [Wed, 25 Aug 2021 03:20:07 +0000 (20:20 -0700)]
cmd/compile/internal/types2: rename IsMethodSet to IsConstraint (cleanup)

Invert the boolean result to match the new name.

Change-Id: Ide6c649ed8ac3a5d263640309960e61a005c886e
Reviewed-on: https://go-review.googlesource.com/c/go/+/344872
Trust: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile: always accept 1.18 syntax but complain if not 1.18
Robert Griesemer [Wed, 25 Aug 2021 01:07:42 +0000 (18:07 -0700)]
cmd/compile: always accept 1.18 syntax but complain if not 1.18

This CL configures the parser to always accept 1.18 syntax
(type parameters, type instantiations, interface elements),
even when -lang is set to an earlier release.

Instead, the type checker looks for 1.18 operations and
complains if the language version is set to an earlier
release.

Doing these checks during type checking is necessary because it
it is possible to write "generic" code using pre-1.18 syntax;
for instance, an imported generic function may be implicitly
instantiated (as in imported.Max(2, 3)), or an imported constraint
interface may be embedded in an "ordinary" interface.

Fixes #47818.

Change-Id: I83ec302b3f4ba7196c0a4743c03670cfb901310d
Reviewed-on: https://go-review.googlesource.com/c/go/+/344871
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agogo/types, types2: don't re-evaluate context string for each function argument (optimi...
Robert Griesemer [Tue, 24 Aug 2021 20:52:07 +0000 (13:52 -0700)]
go/types, types2: don't re-evaluate context string for each function argument (optimization)

Change-Id: Ie1b4d5b64350ea42484adea14df84cacd1d2653b
Reviewed-on: https://go-review.googlesource.com/c/go/+/344576
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>
3 years agocmd/compile/internal/syntax: fix position of type parameter field
Robert Griesemer [Tue, 24 Aug 2021 19:23:28 +0000 (12:23 -0700)]
cmd/compile/internal/syntax: fix position of type parameter field

Change-Id: I8bca01b935301e7bd4efa55ed21921dbf31a75b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/344575
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agogo/types: implement NewTypeList and use it instead of composite literals
Robert Griesemer [Tue, 24 Aug 2021 16:28:43 +0000 (09:28 -0700)]
go/types: implement NewTypeList and use it instead of composite literals

Also, simplify a bit of code in predicates.go.

This is a backport of changes in CL 344615 that were made in addition
to the changes of the original CL 343933; it brings go/types in sync
with types2.

Change-Id: I14cd4d4704d29894d0fbb8d129744d65e332ad22
Reviewed-on: https://go-review.googlesource.com/c/go/+/344570
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>
3 years agocmd/compile: generic SSA rules for simplifying 2 and 3 operand integer arithmetic...
Jake Ciolek [Wed, 25 Aug 2021 10:36:17 +0000 (12:36 +0200)]
cmd/compile: generic SSA rules for simplifying 2 and 3 operand integer arithmetic expressions

This applies the following generic integer addition/subtraction transformations:

x - (x + y) = -y
(x - y) - x = -y
y + (x - y) = x
y + (z + (x - y) = x + z

There's over 40 unique functions matching in Go. Hits 2 funcs in the runtime itself:

runtime.stackfree()
runtime.runqdrain()

Go binary size reduced by 0.05% on Linux x86_64.

StackCopy bench (perflocked Cascade Lake x86):

name                old time/op  new time/op  delta
StackCopyPtr-8      87.3ms ± 1%  86.9ms ± 0%  -0.45%  (p=0.000 n=20+20)
StackCopy-8         77.6ms ± 1%  77.0ms ± 0%  -0.76%  (p=0.000 n=20+20)
StackCopyNoCache-8  2.28ms ± 2%  2.26ms ± 2%  -0.93%  (p=0.008 n=19+20)

test/bench/go1 benchmarks (perflocked Cascade Lake x86):

name                     old time/op    new time/op    delta
BinaryTree17-8              1.88s ± 1%     1.88s ± 0%    ~     (p=0.373 n=15+12)
Fannkuch11-8                2.31s ± 0%     2.35s ± 0%  +1.52%  (p=0.000 n=15+14)
FmtFprintfEmpty-8          26.6ns ± 0%    26.6ns ± 0%    ~     (p=0.081 n=14+13)
FmtFprintfString-8         48.6ns ± 0%    50.0ns ± 0%  +2.86%  (p=0.000 n=15+14)
FmtFprintfInt-8            56.9ns ± 0%    54.8ns ± 0%  -3.70%  (p=0.000 n=15+15)
FmtFprintfIntInt-8         90.4ns ± 0%    88.8ns ± 0%  -1.78%  (p=0.000 n=15+15)
FmtFprintfPrefixedInt-8     104ns ± 0%     104ns ± 0%    ~     (p=0.905 n=14+13)
FmtFprintfFloat-8           148ns ± 0%     144ns ± 0%  -2.19%  (p=0.000 n=14+15)
FmtManyArgs-8               389ns ± 0%     390ns ± 0%  +0.35%  (p=0.000 n=12+15)
GobDecode-8                3.90ms ± 1%    3.88ms ± 0%  -0.49%  (p=0.000 n=15+14)
GobEncode-8                2.73ms ± 0%    2.73ms ± 0%    ~     (p=0.425 n=15+14)
Gzip-8                      169ms ± 0%     168ms ± 0%  -0.52%  (p=0.000 n=13+13)
Gunzip-8                   24.7ms ± 0%    24.8ms ± 0%  +0.61%  (p=0.000 n=15+15)
HTTPClientServer-8         60.5µs ± 6%    60.4µs ± 7%    ~     (p=0.595 n=15+15)
JSONEncode-8               6.97ms ± 1%    6.93ms ± 0%  -0.69%  (p=0.000 n=14+14)
JSONDecode-8               31.2ms ± 1%    30.8ms ± 1%  -1.27%  (p=0.000 n=14+14)
Mandelbrot200-8            3.87ms ± 0%    3.87ms ± 0%    ~     (p=0.652 n=15+14)
GoParse-8                  2.65ms ± 2%    2.64ms ± 1%    ~     (p=0.202 n=15+15)
RegexpMatchEasy0_32-8      45.1ns ± 0%    45.9ns ± 0%  +1.68%  (p=0.000 n=14+15)
RegexpMatchEasy0_1K-8       140ns ± 0%     139ns ± 0%  -0.44%  (p=0.000 n=15+14)
RegexpMatchEasy1_32-8      40.9ns ± 3%    40.5ns ± 0%  -0.88%  (p=0.000 n=15+13)
RegexpMatchEasy1_1K-8       215ns ± 1%     220ns ± 1%  +2.27%  (p=0.000 n=15+15)
RegexpMatchMedium_32-8      783ns ± 7%     738ns ± 0%    ~     (p=0.361 n=15+15)
RegexpMatchMedium_1K-8     24.1µs ± 6%    23.4µs ± 6%  -2.94%  (p=0.004 n=15+15)
RegexpMatchHard_32-8       1.10µs ± 1%    1.09µs ± 1%  -0.40%  (p=0.006 n=15+14)
RegexpMatchHard_1K-8       33.0µs ± 0%    33.0µs ± 0%    ~     (p=0.535 n=12+14)
Revcomp-8                   354ms ± 0%     353ms ± 0%  -0.23%  (p=0.002 n=15+13)
Template-8                 42.0ms ± 1%    41.8ms ± 2%  -0.37%  (p=0.023 n=14+15)
TimeParse-8                 181ns ± 0%     180ns ± 1%  -0.18%  (p=0.014 n=12+13)
TimeFormat-8                240ns ± 0%     242ns ± 1%  +0.69%  (p=0.000 n=12+15)
[Geo mean]                 35.2µs         35.1µs       -0.43%

name                     old speed      new speed      delta
GobDecode-8               197MB/s ± 1%   198MB/s ± 0%  +0.49%  (p=0.000 n=15+14)
GobEncode-8               281MB/s ± 0%   281MB/s ± 0%    ~     (p=0.419 n=15+14)
Gzip-8                    115MB/s ± 0%   115MB/s ± 0%  +0.52%  (p=0.000 n=13+13)
Gunzip-8                  786MB/s ± 0%   781MB/s ± 0%  -0.60%  (p=0.000 n=15+15)
JSONEncode-8              278MB/s ± 1%   280MB/s ± 0%  +0.69%  (p=0.000 n=14+14)
JSONDecode-8             62.3MB/s ± 1%  63.1MB/s ± 1%  +1.29%  (p=0.000 n=14+14)
GoParse-8                21.9MB/s ± 2%  22.0MB/s ± 1%    ~     (p=0.205 n=15+15)
RegexpMatchEasy0_32-8     709MB/s ± 0%   697MB/s ± 0%  -1.65%  (p=0.000 n=14+15)
RegexpMatchEasy0_1K-8    7.34GB/s ± 0%  7.37GB/s ± 0%  +0.43%  (p=0.000 n=15+15)
RegexpMatchEasy1_32-8     783MB/s ± 2%   790MB/s ± 0%  +0.88%  (p=0.000 n=15+13)
RegexpMatchEasy1_1K-8    4.77GB/s ± 1%  4.66GB/s ± 1%  -2.23%  (p=0.000 n=15+15)
RegexpMatchMedium_32-8   41.0MB/s ± 7%  43.3MB/s ± 0%    ~     (p=0.360 n=15+15)
RegexpMatchMedium_1K-8   42.5MB/s ± 6%  43.8MB/s ± 6%  +3.07%  (p=0.004 n=15+15)
RegexpMatchHard_32-8     29.2MB/s ± 1%  29.3MB/s ± 1%  +0.41%  (p=0.006 n=15+14)
RegexpMatchHard_1K-8     31.1MB/s ± 0%  31.1MB/s ± 0%    ~     (p=0.495 n=12+14)
Revcomp-8                 718MB/s ± 0%   720MB/s ± 0%  +0.23%  (p=0.002 n=15+13)
Template-8               46.3MB/s ± 1%  46.4MB/s ± 2%  +0.38%  (p=0.021 n=14+15)
[Geo mean]                205MB/s        206MB/s       +0.57%

Change-Id: Ibd1afdf8b6c0b08087dcc3acd8f943637eb95ac0
Reviewed-on: https://go-review.googlesource.com/c/go/+/344930
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Josh Bleecher Snyder <josharian@gmail.com>

3 years agobytes, strings: optimize Trim for single byte cutsets
Joe Tsai [Sun, 30 May 2021 02:11:37 +0000 (19:11 -0700)]
bytes, strings: optimize Trim for single byte cutsets

Using the latest version of all modules known by the module proxy,
we determine that for all Trim usages (and related functionality):
* 76.6% have cutsets of len=1, and
* 13.4% have cutsets of len=2.

Given that a vast majority of usages only have a cutset of len=1,
we should more heavily optimize for that situation.
Previously, there was some optimization for cutsets of len=1,
but it's within the internal makeCutsetFunc function.
This is sub-optimal as it incurs an allocation in makeCutsetFunc
for the closure over that single byte.

This CL removes special-casing of one-byte cutsets from makeCutsetFunc
and instead distributes it directly in Trim, TrimRight, and TrimLeft.
Whether we should distribute the entire ASCII cutset logic into Trim
is a future CL that should be discussed and handled separately.
The evidence for multibyte cutsets is not as obviously compelling.

name                old time/op  new time/op  delta
bytes/TrimByte-4    84.1ns ± 2%   7.5ns ± 1%  -91.10%  (p=0.000 n=9+7)
strings/TrimByte-4  86.2ns ± 3%   8.3ns ± 1%  -90.33%  (p=0.000 n=9+10)

Fixes #46446

Change-Id: Ia0e31a8384c3ce111ae35465605bcec45df2ebec
Reviewed-on: https://go-review.googlesource.com/c/go/+/323318
Trust: Joe Tsai <joetsai@digital-static.net>
Run-TryBot: Joe Tsai <joetsai@digital-static.net>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: fix function contains no TParam in generic function
korzhao [Wed, 25 Aug 2021 08:01:49 +0000 (16:01 +0800)]
cmd/compile: fix function contains no TParam in generic function

Fixes #47948

Change-Id: I446a9548265d195ae4d88aff6b1361474d1b6214
Reviewed-on: https://go-review.googlesource.com/c/go/+/344910
Trust: Alexander Rakoczy <alex@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Alexander Rakoczy <alex@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile: allow embed into any byte slice type
korzhao [Tue, 17 Aug 2021 11:34:40 +0000 (19:34 +0800)]
cmd/compile: allow embed into any byte slice type

Fixes #47735

Change-Id: Ia21ea9a67f36a3edfef1b299ae4f3b00c306cd68
Reviewed-on: https://go-review.googlesource.com/c/go/+/342851
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Alexander Rakoczy <alex@golang.org>

3 years agotime/format: avoid growslice in time.String()/time.GoString()
korzhao [Fri, 20 Aug 2021 23:26:19 +0000 (07:26 +0800)]
time/format: avoid growslice in time.String()/time.GoString()

Pre-allocate the slice of buf with enough capacity
to avoid growslice calls.

benchmark                   old ns/op     new ns/op     delta
BenchmarkTimeString-4       493           409           -17.12%
BenchmarkTimeGoString-4     309           182           -41.30%

benchmark                   old allocs     new allocs     delta
BenchmarkTimeString-4       5              3              -40.00%
BenchmarkTimeGoString-4     4              1              -75.00%

benchmark                   old bytes     new bytes     delta
BenchmarkTimeString-4       152           128           -15.79%
BenchmarkTimeGoString-4     248           80            -67.74%

Change-Id: I64eabe2ab0b3d4a846453c2e8e548a831d720b8c
Reviewed-on: https://go-review.googlesource.com/c/go/+/343971
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Alexander Rakoczy <alex@golang.org>

3 years agocmd/compile: fix stencil call expression.
wdvxdr [Sun, 22 Aug 2021 16:48:10 +0000 (00:48 +0800)]
cmd/compile: fix stencil call expression.

Fixes: #47878
Change-Id: I369350813726fd518b4eab2b98f43bf031a6dee6
Reviewed-on: https://go-review.googlesource.com/c/go/+/344210
Reviewed-by: Dan Scales <danscales@google.com>
Trust: Dan Scales <danscales@google.com>
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: fix CheckSize() calculation for -G=3 and stencils
Dan Scales [Tue, 24 Aug 2021 22:34:52 +0000 (15:34 -0700)]
cmd/compile: fix CheckSize() calculation for -G=3 and stencils

Because the Align/Width of pointer types are always set when created,
CalcSize() never descends past a pointer. Therefore, we need to do
CheckSize() at every level when creating type. We need to do this for
types creates by types2-to-types1 conversion and also by type
substitution (mostly for stenciling). We also need to do
Defer/ResumeCheckSize() at the top level in each of these cases to deal
with potentially recursive types.

These changes fix issue #47929 and also allow us to remove the
special-case CheckSize() call that causes the problem for issue #47901.

Fixes #47901
Fixes #47929

Change-Id: Icd8192431c145009cd6df2f4ade6db7da0f4dd3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/344829
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
3 years agotest: add test that caused gofrontend compiler crash
Ian Lance Taylor [Fri, 20 Aug 2021 18:35:31 +0000 (11:35 -0700)]
test: add test that caused gofrontend compiler crash

Updates https://gcc.gnu.org/PR101994

Change-Id: I50dcb90e315792efd7d83b496034ad33b5f199e9
Reviewed-on: https://go-review.googlesource.com/c/go/+/343874
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>
3 years agotest: add test case that gofrontend miscompiled
Ian Lance Taylor [Sat, 21 Aug 2021 19:40:25 +0000 (12:40 -0700)]
test: add test case that gofrontend miscompiled

For #47771

Change-Id: I99dfdd48def756bde68445b50741afd6d86b6cf2
Reviewed-on: https://go-review.googlesource.com/c/go/+/344169
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>
3 years agoos/user: don't skip TestLookupGroup if supported
Kir Kolyshkin [Wed, 23 Jun 2021 03:26:31 +0000 (20:26 -0700)]
os/user: don't skip TestLookupGroup if supported

CL 37664 implemented this functionality, yet the tests were skipped.

Introduce and use additional variable groupListImplemented to
distinguish between these cases and enable TestLookupGroup for
supported configurations (which looks like all but plan9).

Change-Id: Iabaa7f08b4551dc67e67bdb6e715f15bb20d6218
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Reviewed-on: https://go-review.googlesource.com/c/go/+/330751
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
3 years agocmd/compile: fix checkptr false positive for (*[Big]T)(ptr)[:n:n] pattern
Cuong Manh Le [Fri, 20 Aug 2021 04:38:54 +0000 (11:38 +0700)]
cmd/compile: fix checkptr false positive for (*[Big]T)(ptr)[:n:n] pattern

The checkptr instrumentation is currently inserted before slice
operation has validated that n <= Big. So instead of panic, checkptr
have false positive throws.

To fix this, just insert the checkptr instrumentation after the bound
checking during SSA generation.

Fixes #46938

Change-Id: I9dbf84441c711842ccc883f3654ca8766ac696d8
Reviewed-on: https://go-review.googlesource.com/c/go/+/343972
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: Matthew Dempsky <mdempsky@google.com>
3 years agoreflect: add MapIter.SetKey and MapIter.SetValue
Josh Bleecher Snyder [Mon, 17 May 2021 19:31:11 +0000 (12:31 -0700)]
reflect: add MapIter.SetKey and MapIter.SetValue

These augment the existing MapIter.Key and MapIter.Value methods.
The existing methods return new Values.
Constructing these new Values often requires allocating.
These methods allow the caller to bring their own storage.

The naming is somewhat unfortunate, in that the spec
uses the word "element" instead of "value",
as do the reflect.Type methods.
In a vacuum, MapIter.SetElem would be preferable.
However, matching the existing methods is more important.

Fixes #32424
Fixes #46131

Change-Id: I19c4d95c432f63dfe52cde96d2125abd021f24fa
Reviewed-on: https://go-review.googlesource.com/c/go/+/320929
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
3 years agocmd/compile: simplify bad conversion check
Keith Randall [Tue, 24 Aug 2021 21:30:24 +0000 (14:30 -0700)]
cmd/compile: simplify bad conversion check

Now that we're using OCONVIDATA(x) everywhere we formerly used
OIDATA(OCONVIFACE(x)), there should be no OCONVIFACE operations that
take a shape type.

Change-Id: I4fb056456c60481c6dfe7bc111fca6223567e6a8
Reviewed-on: https://go-review.googlesource.com/c/go/+/344577
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile: mark ODYNAMICDOTTYPE as an expression that can panic
Keith Randall [Tue, 24 Aug 2021 17:21:16 +0000 (10:21 -0700)]
cmd/compile: mark ODYNAMICDOTTYPE as an expression that can panic

Fixes #47924

Change-Id: I4325b3c4ed9d369d9ea778478285436e1b2ab08a
Reviewed-on: https://go-review.googlesource.com/c/go/+/344571
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile: fix naming of types inside instantiations
Dan Scales [Sun, 22 Aug 2021 15:58:24 +0000 (08:58 -0700)]
cmd/compile: fix naming of types inside instantiations

Issues 47713 and 47877 were both due to problems with the names used for
instantiated functions/methods, which must be in sync with the names
used by types2.

 - Switched to using NameString() for writing out type arguments in
   instantiation names. This ensures that we are always adding the
   package to type names even for the local package. Previously, we were
   explicitly adding the package name for local packages, but that
   doesn't handle the case when the local type is embedded inside a
   pointer or slice type. By switching to NameString(), we fix #47713.

 - types1 and types2 write out 'interface {' differently (vs.
   'interface{') and we were already handling that. But we needed to add
   similar code to handle 'struct {' vs 'struct{'. This fixes issue
   #47877.

While fixing these bugs, I also moved some duplicated code (which
include some of the changes above) into a common function addTargs(). I
also moved InstType() name to subr.go, and renamed: MakeInstName ->
MakeFuncInstSym and MakeDictName -> MakeDictSym.

Also removed a couple of ".inst..inst." prefix checks which are
irrelvant now, since we don't add ".inst." anymore to function
instantiations.

Fixes #47713
Fixes #47877
Fixes #47922

Change-Id: I19e9a073451f3ababd8ec31b6608cd79ba8cba36
Reviewed-on: https://go-review.googlesource.com/c/go/+/344613
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
3 years agocmd/compile: change irgen to generate exprs/stmts after decls processed
Matthew Dempsky [Tue, 24 Aug 2021 07:32:30 +0000 (00:32 -0700)]
cmd/compile: change irgen to generate exprs/stmts after decls processed

This CL changes irgen to wait until all top-level declarations have
been processed before constructing any expressions or statements that
reference them. This is the same approach that typecheck used.

Mechanically, it splits varDecl and funcDecl (the two top-level
declarations that can generate/contain code) into a part that runs
immediately for constructing the ir.ONAME, and then a separate task
that runs later to handle the code.

It also adds an exprStmtOK flag to indicate when it's actually safe to
start constructing (non-trivial) expressions and statements.

Fixes #47928.

Change-Id: I51942af6823aa561d341e2ffc1142948da025fa2
Reviewed-on: https://go-review.googlesource.com/c/go/+/344649
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/link: guarantee "section .debug_gdb_scripts" is always "$GOROOT/src/runtime/runti...
crazycomputer [Sun, 22 Aug 2021 09:26:20 +0000 (17:26 +0800)]
cmd/link: guarantee "section .debug_gdb_scripts" is always "$GOROOT/src/runtime/runtime-gdb.py".

to guarantee the "section .debug_gdb_scripts" is  always "$GOROOT/src/runtime/runtime-gdb.py" , a check for package name is needed when search "runtime/proc.go".

Fixes #47881

Change-Id: Ib471314ca1a6777f625b488e5ae3790457540ef9
Reviewed-on: https://go-review.googlesource.com/c/go/+/344229
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Trust: Michael Knyszek <mknyszek@google.com>

3 years agocmd/compile/internal/types2: use TypeList in the Inferred struct
Robert Griesemer [Tue, 24 Aug 2021 04:04:56 +0000 (21:04 -0700)]
cmd/compile/internal/types2: use TypeList in the Inferred struct

This is a port of CL 343934 from go/types with the necessary
adjustments to the compiler.

Change-Id: I810144e6e2eb2bc8fa0d34dc206403c993cbbe7a
Reviewed-on: https://go-review.googlesource.com/c/go/+/344616
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile/internal/types2: use a TypeList type to hold type arguments
Robert Griesemer [Tue, 24 Aug 2021 03:43:57 +0000 (20:43 -0700)]
cmd/compile/internal/types2: use a TypeList type to hold type arguments

This is a port of CL 343933 from go/types with the necessary
adjustments in the compiler.

With this CL type parameters and type lists are now held in
TParamList and TypeList data types which don't expose the
internal representation.

Change-Id: I6d60881b5db995dbc04ed3f4a96e8b5d41f83969
Reviewed-on: https://go-review.googlesource.com/c/go/+/344615
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile/internal/types2: use []*TypeParam rather than []*TypeName for type param...
Robert Griesemer [Tue, 24 Aug 2021 01:29:38 +0000 (18:29 -0700)]
cmd/compile/internal/types2: use []*TypeParam rather than []*TypeName for type param lists

This is a port of CL 343932 from go/types, with the necessary adjustments
to the compiler.

This change improves type safety slightly, avoids many internal type
assertions, and simplifies some code paths.

Change-Id: Ie9c4734814f49cd248927152d7b3264d3578428c
Reviewed-on: https://go-review.googlesource.com/c/go/+/344614
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile/internal/types2: use an opaque environment for Instantiate
Robert Griesemer [Tue, 24 Aug 2021 00:38:55 +0000 (17:38 -0700)]
cmd/compile/internal/types2: use an opaque environment for Instantiate

This is a port of CL 343930 from go/types, adjusted to work for
the compiler: here Environment carries a *Checker, if available.

Change-Id: I44544fad7da870fa0c02832baa6abd2909d50304
Reviewed-on: https://go-review.googlesource.com/c/go/+/344612
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile/internal/types2: don't export TypeSet
Robert Griesemer [Mon, 23 Aug 2021 23:25:37 +0000 (16:25 -0700)]
cmd/compile/internal/types2: don't export TypeSet

For now don't export TypeSet in the interest of
keeping the types2 API surface small(er).

This is a clean port of CL 341289 from go/types.

Change-Id: I50c747629f25472f2ec5ba59d7f543ee3c1c423b
Reviewed-on: https://go-review.googlesource.com/c/go/+/344610
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agoembed: document the maximum file size supported
korzhao [Thu, 12 Aug 2021 15:39:29 +0000 (23:39 +0800)]
embed: document the maximum file size supported

Fixes #47627

Change-Id: Ia1edfb6249863ab055fab68a35666bc2bdf21dcb
Reviewed-on: https://go-review.googlesource.com/c/go/+/341689
Reviewed-by: Jay Conrod <jayconrod@google.com>
Trust: Michael Matloob <matloob@golang.org>

3 years agogo/token: match the implementation of index selection with sort.Search
subham sarkar [Tue, 20 Jul 2021 17:17:14 +0000 (17:17 +0000)]
go/token: match the implementation of index selection with sort.Search

name          old time/op    new time/op    delta
SearchInts-8    15.5ns ± 2%    13.7ns ± 4%  -11.87%  (p=0.008 n=5+5)

(see CL 36332 for the original change to sort.Search)

Change-Id: If452818185b92b8b3548b066f475e493d604ea29
GitHub-Last-Rev: 32dd3cffa6b54b332948ac6a2929458defd4838f
GitHub-Pull-Request: golang/go#47293
Reviewed-on: https://go-review.googlesource.com/c/go/+/335809
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Robert Griesemer <gri@golang.org>
Trust: Than McIntosh <thanm@google.com>

3 years agocmd/compile: reuse same node for global dictionaries
Dan Scales [Mon, 23 Aug 2021 22:45:10 +0000 (15:45 -0700)]
cmd/compile: reuse same node for global dictionaries

Change stencil.go:getDictionaryValue() and reflect.go:getDictionary() to
reuse any existing name node that has been created for the needed global
dictionary. Otherwise, these functions may set the Def on a specific
dictionary sym to two different name nodes, which means the first node
will not satisfy the invariant 'n.Sym().Def.(*ir.Name) == n' (which is
the assertion in this issue).

Fixes #47896

Change-Id: I1e7ae1efd077a83c7878b4342feb6d28d52476cc
Reviewed-on: https://go-review.googlesource.com/c/go/+/344609
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>

3 years agocmd/compile: fixes for non-constant Sizeof/Alignof/Offsetof
Dan Scales [Sun, 22 Aug 2021 18:50:58 +0000 (11:50 -0700)]
cmd/compile: fixes for non-constant Sizeof/Alignof/Offsetof

Includes Robert's suggested fix in validate.go to not fail on
non-constant alignof/offsetof/sizeof calls. Further changes to wait on
transforming these calls until stenciling time, when we can call
EvalConst() to evaluate them once all the relevant types are known.

Added a bunch of new tests for non-constant Sizeof/Alignof/Offsetof.

Fixes #47716

Change-Id: I469af888eb9ce3a853124d919eda753971009b3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/344250
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>

3 years agoall: replace runtime SSE2 detection with GO386 setting
Martin Möhrmann [Mon, 23 Aug 2021 09:34:51 +0000 (11:34 +0200)]
all: replace runtime SSE2 detection with GO386 setting

When GO386=sse2 we can assume sse2 to be present without
a runtime check. If GO386=softfloat is set we can avoid
the usage of SSE2 even if detected.

This might cause a memcpy, memclr and bytealg slowdown of Go
binaries compiled with softfloat on machines that support
SSE2. Such setups are rare and should use GO386=sse2 instead
if performance matters.

On targets that support SSE2 we avoid the runtime overhead of
dynamic cpu feature dispatch.

The removal of runtime sse2 checks also allows to simplify
internal/cpu further by removing handling of the required
feature option as a followup after this CL.

Change-Id: I90a853a8853a405cb665497c6d1a86556947ba17
Reviewed-on: https://go-review.googlesource.com/c/go/+/344350
Trust: Martin Möhrmann <martin@golang.org>
Run-TryBot: Martin Möhrmann <martin@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
3 years agoruntime: use RDTSCP for instruction stream serialized read of TSC
Martin Möhrmann [Mon, 23 Aug 2021 11:53:22 +0000 (13:53 +0200)]
runtime: use RDTSCP for instruction stream serialized read of TSC

To measure all instructions having been completed before reading
the time stamp counter with RDTSC an instruction sequence that
has instruction stream serializing properties which guarantee
waiting until all previous instructions have been executed is
needed. This does not necessary mean to wait for all stores to
be globally visible.

This CL aims to remove vendor specific logic for determining the
instruction sequence with CPU feature flag checks that are
CPU vendor independent.

For intel LFENCE has the wanted properties at least
since it was introduced together with SSE2 support.

On AMD instruction stream serializing LFENCE is supported by setting
an MSR C001_1029[1]=1 on AMD family 10h/12h/14h/15h/16h/17h processors.
AMD family 0Fh/11h processors support LFENCE as serializing always.
AMD plans support for this MSR and access to this bit for all future processors.
Source: https://developer.amd.com/wp-content/resources/Managing-Speculation-on-AMD-Processors.pdf

Reading the MSR to determine LFENCE properties is not always possible
or reliable (hypervisors). The Linux kernel is relying on serializing
LFENCE on AMD CPUs since a commit in July 2019: https://lkml.org/lkml/2019/7/22/295
and the MSR C001_1029 to enable serialization has been set by default
with the Spectre v1 mitigations.

Using an MFENCE on AMD is waiting on previous instructions having been executed
but in addition also flushes store buffers.

To align the serialization properties without runtime detection
of CPU manufacturers we can use the newer RDTSCP instruction which
waits until all previous instructions have been executed.

RDTSCP is available on Intel since around 2008 and on AMD CPUs since
around 2006. Support for RDTSCP can be checked independently
of manufacturer by checking CPUID bits.

Using RDTSCP is the default in Linux to read TSC in program order
when the instruction is available.
https://github.com/torvalds/linux/blob/e22ce8eb631bdc47a4a4ea7ecf4e4ba499db4f93/arch/x86/include/asm/msr.h#L231

Change-Id: Ifa841843b9abb2816f8f0754a163ebf01385306d
Reviewed-on: https://go-review.googlesource.com/c/go/+/344429
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Martin Möhrmann <martin@golang.org>
Run-TryBot: Martin Möhrmann <martin@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agointernal/buildcfg: change GOEXPERIMENT to always return non-empty string
Matthew Dempsky [Mon, 23 Aug 2021 18:40:56 +0000 (11:40 -0700)]
internal/buildcfg: change GOEXPERIMENT to always return non-empty string

Rather than returning "", we now return "," (which is a no-op). This
ensures that the returned string always overrides DefaultGOEXPERIMENT.

This fixes a bootstrapping issue where GOROOT_BOOTSTRAP was built with
"GOEXPERIMENT=fieldtrack ./make.bash". cmd/dist sets GOEXPERIMENT=none
during bootstrapping, which was causing cmd/go to set GOEXPERIMENT=""
when executing cmd/compile; but then cmd/compile ignores the
environment variable (because it's empty) and instead uses
DefaultGOEXPERIMENT.

Fixes #47921.

Change-Id: I657ff6cdfb294a94f6a2f58c306ceed7f104416b
Reviewed-on: https://go-review.googlesource.com/c/go/+/344511
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>
3 years agocmd/compile: do not mark arrays used for map initialization noalg
Martin Möhrmann [Tue, 6 Jul 2021 19:45:02 +0000 (21:45 +0200)]
cmd/compile: do not mark arrays used for map initialization noalg

Arrays marked noalg are created by the compiler to hold keys and values
to initialize map literals. The ssa backend creates a pointer type for
the array type when creating an OpAddr while processing the loop that
initializes the map from the arrays. The pointer type does not inherit
the noalg property but points to the noalg array type.

This causes values created through reflect of types that should be
equal to compare unequal because the noalg and alg type might be
compared and these are not the same.

A similar problem occurred in #32595 for argument arrays of defer structs.

Created #47904 to track improve noalg handling to be able to
reintroduce this optimization again.

Fixes #47068

Change-Id: I87549342bd404b98d71a3c0f33e3c169e9d4efc8
Reviewed-on: https://go-review.googlesource.com/c/go/+/344349
Trust: Martin Möhrmann <martin@golang.org>
Run-TryBot: Martin Möhrmann <martin@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile: don't emit write barriers for offsets of global addresses
zikaeroh [Fri, 30 Jul 2021 04:15:52 +0000 (21:15 -0700)]
cmd/compile: don't emit write barriers for offsets of global addresses

Currently, write barriers aren't emitted for global addresses, but they
are emitted for addresses offset of global addresses.

This CL changes IsGlobalAddr to recognize offsets of global addresses
as globals too, removing write barriers for staticuint64s based
addresses. The logic added is the same as used in IsStackAddr.

Updates #37612

Change-Id: I537579f85b9ad02987d94f3ee0b4508b90097959
Reviewed-on: https://go-review.googlesource.com/c/go/+/342129
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: always remove receiver type from instantiated method values
Keith Randall [Fri, 20 Aug 2021 23:15:53 +0000 (16:15 -0700)]
cmd/compile: always remove receiver type from instantiated method values

If a type T has a method foo, then

var t T
var i interface{} = t.foo

The type of foo is a method type, but the type of t.foo should be a
standard function type. Make sure we always do that conversion.

Fixes #47775

Change-Id: I464ec792196b050aba1914e070a4ede34bfd0bfa
Reviewed-on: https://go-review.googlesource.com/c/go/+/343881
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agocmd/compile: copy captured dictionary var to local var
Keith Randall [Fri, 20 Aug 2021 17:19:28 +0000 (10:19 -0700)]
cmd/compile: copy captured dictionary var to local var

When starting a closure that needs a dictionary, copy the closure
variable to a local variable. This lets child closures capture that
dictionary variable correctly.

This is a better fix for #47684, which does not cause problems
like #47723.

Fixes #47723
Update #47684

Change-Id: Ib5d9ffc68a5142e28daa7d0d75683e7a35508540
Reviewed-on: https://go-review.googlesource.com/c/go/+/343871
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agospec: add example for method value in case of embedded method
Cuong Manh Le [Sun, 22 Aug 2021 05:07:14 +0000 (12:07 +0700)]
spec: add example for method value in case of embedded method

So it's clear to the reader that if "M" is a promoted method from
embedded field "T", then "x.M" will be expanded to "x.T.M" during the
evaluation of the method value.

Fixes #47863

Change-Id: Id3b82127a2054584b6842c487f6e15c3102dc9fe
Reviewed-on: https://go-review.googlesource.com/c/go/+/344209
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: Rob Pike <r@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile: fixing 15.go for -G=3
Dan Scales [Sun, 22 Aug 2021 20:34:22 +0000 (13:34 -0700)]
cmd/compile: fixing 15.go for -G=3

Required two changes:

 - avoid creating a closure in the case where the actual receiver of an
   embedded method is not generic even though the base operand of the
   selector is generic. This is similar to the test suggested by wayne
   zuo - I thought it was clear in buildClosure, and easier to comment.

 - Propagate //go:nointerface to base generic methods and then to
   instantiations.

Change-Id: If30c834e4223c2639b7f7e74d44e6087aa9ccd76
Reviewed-on: https://go-review.googlesource.com/c/go/+/344251
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Wayne Zuo <wdvxdr1123@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>

3 years agoreflect: fix memmove for big endian cases with new ABI
Lynn Boger [Fri, 20 Aug 2021 12:01:50 +0000 (07:01 -0500)]
reflect: fix memmove for big endian cases with new ABI

Some memmoves in reflect/value.go for copying arguments
related to the new ABI were using the address of the target
or source instead of using IntArgRegAddr or FloatArgRegAddr
to adjust the address for big endian.

This was found when testing patches for ppc64 and fixes the
failures that were found.

Change-Id: I119aa090a2a8eb859020ff1a1736107a6d0b76f1
Reviewed-on: https://go-review.googlesource.com/c/go/+/343869
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>

3 years agogo/types: use TypeList in the Inferred struct
Robert Findley [Fri, 20 Aug 2021 14:19:12 +0000 (10:19 -0400)]
go/types: use TypeList in the Inferred struct

This is for consistency with how we report TArgs elsewhere, and in case
we ever want to share an internal slice with inference reporting.

Change-Id: Ia8b705a155f4f82bd8da8dc2457289810f875f5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/343934
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: use a TypeList type to hold type arguments
Robert Findley [Thu, 19 Aug 2021 18:06:08 +0000 (14:06 -0400)]
go/types: use a TypeList type to hold type arguments

This resolves an asymmetry between the TParams and TArgs APIs, and
reduces the size of the Named struct at the cost of some additional nil
checks.

While at it, move TParamList and TypeList to a new file:typelists.go,
and change TParamList to access the tparams slice directly in At. There
is no reason to guard against a nil receiver, as accessing an index on
the empty slice will panic anyway.

Change-Id: I9b65247e06c697a57a4efe40c3390e0faff91441
Reviewed-on: https://go-review.googlesource.com/c/go/+/343933
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: use []*TypeParam rather than []*TypeName type param lists
Robert Findley [Thu, 19 Aug 2021 17:31:36 +0000 (13:31 -0400)]
go/types: use []*TypeParam rather than []*TypeName type param lists

Making this change improves type safety slightly, and avoids many
internal type assertions.

Change-Id: I26519b0e57068e944e8243983ae90553d79e59c2
Reviewed-on: https://go-review.googlesource.com/c/go/+/343932
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: add the TypeParam.Obj method
Robert Findley [Wed, 18 Aug 2021 17:22:58 +0000 (13:22 -0400)]
go/types: add the TypeParam.Obj method

Users should be able to access the type name associated with a type
parameter.

Change-Id: I495c3b4377f9d4807b1e78ad341e573d4d3c7bff
Reviewed-on: https://go-review.googlesource.com/c/go/+/343931
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: move to an opaque environment for Instantiate
Robert Findley [Wed, 18 Aug 2021 17:22:38 +0000 (13:22 -0400)]
go/types: move to an opaque environment for Instantiate

To match the API proposal, switch the first argument to Instantiate to
an opaque Environment handle, though for now this handle is
unimplemented.

Change-Id: I6207f0beafdf8497587abdad37db92f927db29b6
Reviewed-on: https://go-review.googlesource.com/c/go/+/343930
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: return an error from Instantiate
Robert Findley [Mon, 16 Aug 2021 20:40:57 +0000 (16:40 -0400)]
go/types: return an error from Instantiate

This is a port of CL 342152 to go/types. Additionally, a panic was
removed from interface substitution, which is a fix from CL 333155 that
was previously missed.

A check for a nil Checker was also removed from types2.instantiate,
since check must not be nil in that method.

Change-Id: I4ea6bdccbd50ea2008ee6d870f702bee5cdd5a8e
Reviewed-on: https://go-review.googlesource.com/c/go/+/342671
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agoruntime: remove unused cpu architecture feature variables from binaries
Martin Möhrmann [Sat, 24 Oct 2020 03:54:46 +0000 (05:54 +0200)]
runtime: remove unused cpu architecture feature variables from binaries

On amd64 this reduces go binary sizes by 176 bytes due to not referencing
internal/cpu.ARM64 and internal/cpu.ARM.

Change-Id: I8e4f31e2b1939b05eec2148b44d7cff7e0aeb30e
Reviewed-on: https://go-review.googlesource.com/c/go/+/344329
Trust: Martin Möhrmann <martin@golang.org>
Run-TryBot: Martin Möhrmann <martin@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/go: fix long test builders
Cuong Manh Le [Mon, 23 Aug 2021 02:31:54 +0000 (09:31 +0700)]
cmd/go: fix long test builders

CL 343732 enabled -G=3 by default. The types2 typechecker uses slighly
different error message format for language feature constraint.

The old typechecker format:

vendor/example.net/need117/need117.go:5:16: cannot convert s (type []byte) to type *[4]byte:
conversion of slices to array pointers only supported as of -lang=go1.17

The new format:

vendor/example.net/need117/need117.go:5:17: conversion of slices to array pointers only supported as of -lang=go1.17

caused the long test builders failed.

This CL fixes the test by relaxing the regext pattern a bit, so it can
match both the format.

Change-Id: I1c4acaa9e34b6c08dccbbc3ce7a99d4cd79f748a
Reviewed-on: https://go-review.googlesource.com/c/go/+/344212
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agostrings: smarter growth of temporal buffer and avoid copying on return
Ignacio Hagopian [Tue, 17 Aug 2021 23:52:00 +0000 (23:52 +0000)]
strings: smarter growth of temporal buffer and avoid copying on return

The implementation for single strings had two optimization opportunities:
1. Grow the temporary buffer by known size before appending.
2. Avoid a full copy of the result since the underlying buffer won't be mutated afterward.
Both things were leveraged by using a Builder instead of a byte slice.

Relevant benchmark results:

        name           old time/op    new time/op    delta
        SingleMatch-8    32.0µs ± 3%    26.1µs ± 3%  -18.41%  (p=0.000 n=9+10)

        name           old speed      new speed      delta
        SingleMatch-8   469MB/s ± 3%   574MB/s ± 3%  +22.56%  (p=0.000 n=9+10)

        name           old alloc/op   new alloc/op   delta
        SingleMatch-8    81.3kB ± 0%    49.0kB ± 0%  -39.67%  (p=0.000 n=10+10)

        name           old allocs/op  new allocs/op  delta
        SingleMatch-8      19.0 ± 0%      11.0 ± 0%  -42.11%  (p=0.000 n=10+10)

Change-Id: I23af56a15875206c0ff4ce29a51bec95fd48bb11
GitHub-Last-Rev: 403cfc3c2794b5da27792c51999417a2a052b365
GitHub-Pull-Request: golang/go#47766
Reviewed-on: https://go-review.googlesource.com/c/go/+/343089
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Robert Griesemer <gri@golang.org>

3 years agogo/types: report argument type for unsafe.OffsetOf
Robert Griesemer [Sun, 22 Aug 2021 22:21:29 +0000 (15:21 -0700)]
go/types: report argument type for unsafe.OffsetOf

This is a clean port of CL 344252 to go/types.

For #47895.

Change-Id: I48cbb97ec28fcfb4fdf483594be9d29426c117ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/344254
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>
3 years agocmd/compile/internal/types2: enable TestSelection API test
Robert Griesemer [Sun, 22 Aug 2021 21:55:12 +0000 (14:55 -0700)]
cmd/compile/internal/types2: enable TestSelection API test

This test was never fully ported from go/types. Implement
a conversion function from syntax.Pos to string index so
that the test can be enabled again.

Also renamed the local variable syntax to segment to avoid
confusion with the syntax package.

Change-Id: I1b34e50ec138403798efb14c828545780f565507
Reviewed-on: https://go-review.googlesource.com/c/go/+/344253
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>
3 years agoos/user: simplify test skip for plan9
Kir Kolyshkin [Wed, 23 Jun 2021 03:19:57 +0000 (20:19 -0700)]
os/user: simplify test skip for plan9

There's no need to specifically check for runtime.GOOS as there's
already a generic mechanism for that.

Change-Id: I7125443ead456548bd503c5e71cd56e9eb30b446
Reviewed-on: https://go-review.googlesource.com/c/go/+/330750
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
3 years agocmd/compile/internal/types2: report argument type for unsafe.OffsetOf
Robert Griesemer [Sun, 22 Aug 2021 21:09:46 +0000 (14:09 -0700)]
cmd/compile/internal/types2: report argument type for unsafe.OffsetOf

Before parameterized types, unsafe.OffsetOf was always evaluating to
a constant. With parameterized types, the result may be a run-time
value, and unsafe.OffsetOf(x.f) is a call that is recorded. Also
record the argument x.f.

Fixes #47895.

Change-Id: Ia3da25028d4865d7295ce7990c7216bffe9e7c72
Reviewed-on: https://go-review.googlesource.com/c/go/+/344252
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>
3 years agomath/big: clarified doc string for SetMantExp
Robert Griesemer [Sun, 22 Aug 2021 19:11:02 +0000 (12:11 -0700)]
math/big: clarified doc string for SetMantExp

Fixes #47879.

Change-Id: I35efb5fc65c4f1eb1b45918f95bbe1ff4039950e
Reviewed-on: https://go-review.googlesource.com/c/go/+/344249
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
3 years agoruntime: use asmcgocall_no_g when calling sigprocmask on openbsd
Joel Sing [Sun, 30 May 2021 14:18:54 +0000 (00:18 +1000)]
runtime: use asmcgocall_no_g when calling sigprocmask on openbsd

sigprocmask is called from sigsave, which is called from needm. As such,
sigprocmask has to be able to run with no g. For some reason we do not
currently trip this on current libc platforms, but we do hit it on
openbsd/mips64 with external linking.

Updates #36435

Change-Id: I4dfae924245c5f68cc012755d6485939014898a3
Reviewed-on: https://go-review.googlesource.com/c/go/+/334879
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
3 years agoreflect: add example for FieldByIndex
Mostafa Solati [Mon, 8 Jun 2020 17:34:09 +0000 (22:04 +0430)]
reflect: add example for FieldByIndex

Change-Id: I539453e50ab85ec1b023bc9e329e6451c674e0c9
Reviewed-on: https://go-review.googlesource.com/c/go/+/236937
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agoruntime: fix buckHashSize duplication
Dmitry Vyukov [Sat, 12 Dec 2020 15:46:01 +0000 (16:46 +0100)]
runtime: fix buckHashSize duplication

We have a constant for 179999, don't duplicate it.

Change-Id: Iefb9c4746f6dda2e08b42e3c978963198469ee8c
Reviewed-on: https://go-review.googlesource.com/c/go/+/277375
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Pratt <mpratt@google.com>

3 years agostrconv: reject surrogate halves in Unquote
Joe Tsai [Fri, 20 Aug 2021 19:50:02 +0000 (12:50 -0700)]
strconv: reject surrogate halves in Unquote

Unquote implements unescaping a "single-quoted, doubled-quoted, or
backquoted Go string literal". Therefore, it should reject anything
that the Go specification explicitly forbids.

The section on "Rune literals" explicitly rejects rune values
"above 0x10FFFF and surrogate halves". We properly checked for
the previous condition, but were failing to check for the latter.

In general, "r > utf8.MaxRune" is probably the wrong check,
while !utf8.ValidRune(r) is the more correct check.
We make changes to both UnquoteChar and appendEscapedRune
to use the correct check. The change to appendEscapedRune
is technically a noop since callers of that function already
guarantee that the provided rune is valid.

Fixes #47853

Change-Id: Ib8977e56b91943ec8ada821b8d217b5e9a66f950
Reviewed-on: https://go-review.googlesource.com/c/go/+/343877
Trust: Joe Tsai <joetsai@digital-static.net>
Run-TryBot: Joe Tsai <joetsai@digital-static.net>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
3 years agocmd/compile: absorb NEG into branch when possible on riscv64
Joel Sing [Tue, 17 Aug 2021 09:09:33 +0000 (19:09 +1000)]
cmd/compile: absorb NEG into branch when possible on riscv64

We can end up with this situation due to our equality tests being based on
'SEQZ (SUB x y)' - if x is a zero valued constant, 'SUB x y' can be converted
to 'NEG x'. When used with a branch the SEQZ can be absorbed, leading to
'BNEZ (NEG x)' where the NEG is redundant.

Removes around 1700 instructions from the go binary on riscv64.

Change-Id: I947a080d8bf7d2d6378ab114172e2342ce2c51db
Reviewed-on: https://go-review.googlesource.com/c/go/+/342850
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: convert branch with zero to more optimal branch zero on riscv64
Joel Sing [Tue, 17 Aug 2021 09:01:52 +0000 (19:01 +1000)]
cmd/compile: convert branch with zero to more optimal branch zero on riscv64

Convert BLT and BGE with a zero valued constant to BGTZ/BLTZ/BLEZ/BGEZ as
appropriate.

Removes over 4,500 instructions from the go binary on riscv64.

Change-Id: Icc266e968b126ba04863ec88529630a9dd44498b
Reviewed-on: https://go-review.googlesource.com/c/go/+/342849
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: sort regalloc switch by architecture
Joel Sing [Thu, 15 Jul 2021 16:36:52 +0000 (02:36 +1000)]
cmd/compile: sort regalloc switch by architecture

Also tweak comment for the arm64 case.

Change-Id: I073405bd2acf901dcaaf33a034a84b6a09dd4a83
Reviewed-on: https://go-review.googlesource.com/c/go/+/334869
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Meng Zhuo <mzh@golangcn.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
3 years agogo/types: don't override x.mode before using it
Robert Griesemer [Fri, 20 Aug 2021 04:35:57 +0000 (21:35 -0700)]
go/types: don't override x.mode before using it

Changing the mode of x before using the old value is clearly wrong.
And x is not needed anymore afterward so besides being misplaced,
the assignment is not needed in the first place.

Tested manually as it's a bit complicated to set up a test.

Needs to be back-ported to 1.17.

Fixes #47777.

Change-Id: I06f1fa9443eb98009b4276f566d557fd52f1d6d3
Reviewed-on: https://go-review.googlesource.com/c/go/+/343809
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile: enable -G=3 by default
Matthew Dempsky [Thu, 19 Aug 2021 22:53:13 +0000 (15:53 -0700)]
cmd/compile: enable -G=3 by default

This CL changes cmd/compile's -G flag's default from 0 to 3, which
enables use of the new types2 type checker and support for type
parameters. The old type checker is still available with
-gcflags=all=-G=0.

The CL also updates the regress test harness to account for the change
in default behavior (e.g., to expect known types2 changes/failures).
However, the -G=0 mode is still being tested for now.

Copy of CL 340914 by danscales@, minus the cmd/internal/objabi.AbsFile
change (handled instead by CL 343731) and rebased to master branch.

Updates #43651.

Change-Id: I1f62d6c0a3ff245e15c5c0e8f3d922129fdd4f29
Reviewed-on: https://go-review.googlesource.com/c/go/+/343732
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>
3 years agotest/typeparam: add a test case for issue46591
korzhao [Fri, 20 Aug 2021 18:51:51 +0000 (02:51 +0800)]
test/typeparam: add a test case for issue46591

Fixes #46591

Change-Id: I4875092ecd7760b0cd487e793576ef7a9a569a0e
Reviewed-on: https://go-review.googlesource.com/c/go/+/343970
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/internal/buildid: reject empty id
Russ Cox [Fri, 20 Aug 2021 19:20:10 +0000 (15:20 -0400)]
cmd/internal/buildid: reject empty id

The loop that makes progress assumes that after matching an id
you should advance len(id) bytes in the file. If id is the empty string,
then it will match and advance 0 bytes repeatedly.

0-byte ids are not really build IDs, so just reject it outright.

Fixes #47852.

Change-Id: Ie44a3a51dec22e2f68fb72d54ead91be98000cfe
Reviewed-on: https://go-review.googlesource.com/c/go/+/344049
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agotest: enable regabi test on arm64
Cuong Manh Le [Thu, 19 Aug 2021 14:02:49 +0000 (21:02 +0700)]
test: enable regabi test on arm64

CL 324890 turned on register ABI by default on ARM64, causing neither
live.go nor live_regabi.go is run on ARM64.

This CL enables live_regabi.go test for ARM64.

Change-Id: I0c483a38b761c5a6f1fa9a5b3324b5da64907e61
Reviewed-on: https://go-review.googlesource.com/c/go/+/343531
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: Matthew Dempsky <mdempsky@google.com>
3 years agocmd/compile/internal/syntax: add PosBase.Trimmed
Matthew Dempsky [Thu, 19 Aug 2021 22:52:53 +0000 (15:52 -0700)]
cmd/compile/internal/syntax: add PosBase.Trimmed

With types2, some syntax.PosBases need to be constructed from export
data, which must only contain "trimmed" filenames (i.e., that they've
already been made absolute and undergone -trimpath processing).
However, it's not safe to apply trimming to a filename multiple times,
and in general we can't distinguish trimmed from untrimmed filenames.

This CL resolves this by adding a PosBase.Trimmed boolean so we can
distinguish whether the associated filename has been trimmed yet. This
is a bit hacky, but is the least bad solution I've come up with so
far.

This unblocks enabling -G=3 by default.

Change-Id: I7383becfb704680a36f7603e3246af38b21f100b
Reviewed-on: https://go-review.googlesource.com/c/go/+/343731
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agonet/http: fix typo in header.go
HuanCheng [Fri, 20 Aug 2021 16:26:26 +0000 (00:26 +0800)]
net/http: fix typo in header.go

Change-Id: Ia6df881badf9a704c7f56967404d37e230b88a09
Reviewed-on: https://go-review.googlesource.com/c/go/+/343969
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Damien Neil <dneil@google.com>

3 years agogo/types: change Checker.verify to return an error
Robert Findley [Mon, 16 Aug 2021 20:15:09 +0000 (16:15 -0400)]
go/types: change Checker.verify to return an error

This is a port of CL 342151 to go/types, adjusted for errors and
positions. Checker.sprintf was refactored to facilitate formatting
error messages with a nil Checker.

Change-Id: Ib2e5c942e55edaff7b5e77cf68a72bad70fea0b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/342670
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: no need to validate substituted instances
Robert Findley [Mon, 16 Aug 2021 20:03:59 +0000 (16:03 -0400)]
go/types: no need to validate substituted instances

This is a straightforward port of CL 342150 to go/types.

Change-Id: I7363e4642ade7ab30ca822a2be71f4d2804cc4a9
Reviewed-on: https://go-review.googlesource.com/c/go/+/342669
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: consolidate verification logic
Robert Findley [Mon, 16 Aug 2021 20:01:42 +0000 (16:01 -0400)]
go/types: consolidate verification logic

This is a straightforward port of CL 342149 to go/types.

Change-Id: I468c5154b7545b7816bb3f240b8db91e7a1fd3f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/342488
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agogo/types: clean up panics in instantiation
Robert Findley [Mon, 16 Aug 2021 19:59:08 +0000 (15:59 -0400)]
go/types: clean up panics in instantiation

This is a straightforward port of CL 341862 to go/types.

Change-Id: I4214c08d2889e2daf40254385656c6beed79571d
Reviewed-on: https://go-review.googlesource.com/c/go/+/342487
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agoarchive/zip: prevent preallocation check from overflowing
Roland Shoemaker [Wed, 18 Aug 2021 18:49:29 +0000 (11:49 -0700)]
archive/zip: prevent preallocation check from overflowing

If the indicated directory size in the archive header is so large that
subtracting it from the archive size overflows a uint64, the check that
the indicated number of files in the archive can be effectively
bypassed. Prevent this from happening by checking that the indicated
directory size is less than the size of the archive.

Thanks to the OSS-Fuzz project for discovering this issue and to
Emmanuel Odeke for reporting it.

Fixes #47801
Fixes CVE-2021-39293

Change-Id: Ifade26b98a40f3b37398ca86bd5252d12394dd24
Reviewed-on: https://go-review.googlesource.com/c/go/+/343434
Trust: Roland Shoemaker <roland@golang.org>
Run-TryBot: Roland Shoemaker <roland@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
3 years agocrypto/rand, internal/syscall/unix: don't use getentropy on iOS
Tobias Klauser [Thu, 19 Aug 2021 14:36:38 +0000 (16:36 +0200)]
crypto/rand, internal/syscall/unix: don't use getentropy on iOS

CL 302489 switched crypto/rand to use getentropy on darwin, however this
function is not available on iOS. Enable getentropy only on macOS and
disable it on iOS.

Fixes #47812

Change-Id: Ib7ba5d77346aee87904bb93d60cacc845f5c0089
Reviewed-on: https://go-review.googlesource.com/c/go/+/343609
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile: use typeAndStr directly in signatslice
Cuong Manh Le [Thu, 19 Aug 2021 10:01:26 +0000 (17:01 +0700)]
cmd/compile: use typeAndStr directly in signatslice

Currently, we store *types.Type in signatslice, then convert it to
typeAndStr during write runtime type process.

Instead, we can just store typeAndStr directly in signatslice when
adding type to signatset. Not a big win, but simplify the code a bit.

Change-Id: Ie1c8cfa5141da32b6ec3ce5844ba150d2765fe90
Reviewed-on: https://go-review.googlesource.com/c/go/+/343529
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>
3 years agocmd/asm/internal/arch: adds the missing type check for arm64 SXTB extension
eric fang [Mon, 16 Aug 2021 06:41:15 +0000 (06:41 +0000)]
cmd/asm/internal/arch: adds the missing type check for arm64 SXTB extension

Operands of memory type do not support SXTB extension. This CL adds this
missing check.

Change-Id: I1fa438dd314fc8aeb889637079cc67b538e83a89
Reviewed-on: https://go-review.googlesource.com/c/go/+/342769
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com>

3 years agocmd/internal/obj/arm64: disable the pre and post index formats for pseudo registers
eric fang [Mon, 16 Aug 2021 07:25:29 +0000 (07:25 +0000)]
cmd/internal/obj/arm64: disable the pre and post index formats for pseudo registers

When using the FP or SP pseudo-register to load or store, pre-index and post-index formats
are not supported because the RSP and pseudo registers are not allowed to be modified in this
way. This CL deletes the related entries in optab and adds a few test cases.

Change-Id: Ie30d27d0e7b959242f0e6298b950489669d07989
Reviewed-on: https://go-review.googlesource.com/c/go/+/342770
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: eric fang <eric.fang@arm.com>
Trust: Michael Knyszek <mknyszek@google.com>

3 years agocmd/dist: remove unused variables
Manlio Perillo [Fri, 21 May 2021 20:21:09 +0000 (22:21 +0200)]
cmd/dist: remove unused variables

Remove the unused defaultcflags and defaultldflags variables.

Reported by staticcheck.

Change-Id: Icc42f2e670496dbe2ffb26abe25128d8e53e2a6d
Reviewed-on: https://go-review.googlesource.com/c/go/+/321931
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
3 years agosyscall: add SyscallN
Changkun Ou [Thu, 22 Jul 2021 15:50:42 +0000 (17:50 +0200)]
syscall: add SyscallN

This CL adds a new syscall.SyscallN API.

The proposal discussion also suggests the API should not only for
Windows but other platforms. However, the existing API set already
contain differences between platforms, hence the CL only implements
the Windows platform.

Moreover, although the API offers variadic parameters, the permitted
parameters remains up to a limit, which is selected as 42, and arguably
large enough.

Fixes #46552

Change-Id: I66b49988a304d9fc178c7cd5de46d0b75e167a4f
Reviewed-on: https://go-review.googlesource.com/c/go/+/336550
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>

3 years agocmd/compile: prevent duplicated works in WriteRuntimeTypes
Cuong Manh Le [Tue, 8 Jun 2021 12:28:45 +0000 (19:28 +0700)]
cmd/compile: prevent duplicated works in WriteRuntimeTypes

While processing signatset, the entry is deleted immediately after being
pushed to signatslice. Then calling writeType may add the same type
to signatset again. That would add more works, though not a big impact
to the performace, since when writeType is guarded by s.Siggen() check.

Instead, we should keep the entry in signatset, so written type will
never be added again.

This change does not affect compiler performace, but help debugging
issue like one in #46386 easier.

Change-Id: Iddafe773885fa21cb7003ba27ddf9554fc3f297d
Reviewed-on: https://go-review.googlesource.com/c/go/+/326029
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>
3 years agoreflect: add test for invalid conversion
Keith Randall [Wed, 18 Aug 2021 16:38:19 +0000 (09:38 -0700)]
reflect: add test for invalid conversion

Conversion between slices with different element types is not allowed.
Previously (1.8 <= goversion <= 1.16), this conversion was allowed
if the base types were from different packages and had identical names.

Update #47785

Change-Id: I359de5b6fe3ff35bdbf9ab5a13902a0f820cac66
Reviewed-on: https://go-review.googlesource.com/c/go/+/343329
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>