CL 307010 for ppc64.
I spent a long time trying to figure out how to use the carry bit from
ADDCCC to further simplify this (like what we do on arm64), but gave
up after I couldn't figure out how to access the carry bit without
just adding more instructions.
Change-Id: I6cad51b93616865b203cb16554f16121375aabbc
Reviewed-on: https://go-review.googlesource.com/c/go/+/307149
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
For stack frames larger than StackBig, the stack split prologue needs
to guard against potential wraparound. Currently, it carefully
arranges to avoid underflow, but this is complicated and requires a
special check for StackPreempt. StackPreempt is no longer the only
stack poison value, so this check will incorrectly succeed if the
stack bound is poisoned with any other value.
This CL simplifies the logic of the check, reduces its length, and
accounts for any possible poison value by directly checking for
underflow.
Change-Id: I917a313102d6a21895ef7c4b0f304fb84b292c81
Reviewed-on: https://go-review.googlesource.com/c/go/+/307010
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
The CV add changes according to TODO in Go source-code.
Internal atomic set does not comply with sync/atomic library and has shortage
operations for signed integers.
This patch extend internal atomic set by Int32 and Int64 operations. It's
implemented new aliases and asm versions of operations. As a result Cas64 was
replaced by Casint64 in findRunnableGCWorker without type casting.
Another purpose is unified structure of internal atomics' source code. Before,
assembly impementations for different archs were in different files. For
example, filename for AMD64 was asm_amd64.s, but filename for RISC-V was
atomic_riscv64.s. Some arches have both files without any meaning. So, assembly
files were merged and renamed to atomic_{$ARCH}.s filenames.
Change-Id: I29a05a7cbf5f4a9cc146e8315536c038af545677
Reviewed-on: https://go-review.googlesource.com/c/go/+/289152
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Confusingly, the set of all methods of an interface is currently set in
Fields field of types.Interface. This is true, even though there is
already an allMethods field (and AllMethods method) of types.Type.
Change so the set of all methods of an interface are stored in
Type.allMethods, and Interface.Fields is removed. Update the comments
for Methods and AllMethods.
Change-Id: Ibc32bafae86831cba62606b079a855690612c759
Reviewed-on: https://go-review.googlesource.com/c/go/+/307209
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>
Extract queryLatestVersionIgnoringRetractions, which returns the
version we should load retractions and deprecations from. This will be
shared with CheckDeprecations.
Rename ShortRetractionRationale to ShortMessage. This will be used to
shorten deprecation warnings as well.
For #40357
Change-Id: Ic1e0c670396bdb3bd87c7a97cf2b14ca58ea1d80
Reviewed-on: https://go-review.googlesource.com/c/go/+/306332
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Instead of accepting bool flags, ListModules now accepts ListMode, a
set of bit flags.
Four flags are defined. listRetracted is split into ListRetracted and
ListRetractedVersion to avoid ambiguity with -u, -retracted, and
-versions.
For #40357
Change-Id: Ibbbe44dc1e285ed17f27a6581f3392679f2124fb
Reviewed-on: https://go-review.googlesource.com/c/go/+/306331
Trust: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Revert CL 306290 changes to TestRemoveAllLongPath. This breaks the test
on aix, illumos and solaris. We need to chdir out of startPath before
attempting to remove it.
Updates #45182
Change-Id: Ic14fa1962d6f2cc83238f6fc2c6932fd9a6e52a1
Reviewed-on: https://go-review.googlesource.com/c/go/+/307189
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Most of the time, the pprof tests are passing, except
for the builder. The reason is still unknown but I'd rather release
the builder to avoid missing other more important bugs.
Updates #45170
Change-Id: I667543ee1ae309b7319c5b3676a0901b4d0ecf2e
Reviewed-on: https://go-review.googlesource.com/c/go/+/306489
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
(*File).readdir allocates a fixed-size 8KiB buffer on unix systems that
cannot be reused. While it accounts for just a single allocation, it's
more than large enough to show up in profiles and make things quite a
bit slower.
Instead of allocating so often, use a sync.Pool to allow these buffers to
be reused. This has a large impact on readdir heavy workloads.
Package os benchmarks:
name old time/op new time/op delta
Readdirname-12 35.6µs ± 5% 18.1µs ± 4% -49.00% (p=0.000 n=10+10)
Readdir-12 142µs ± 1% 121µs ± 0% -14.87% (p=0.000 n=10+9)
ReadDir-12 44.0µs ± 6% 28.4µs ± 8% -35.58% (p=0.000 n=9+10)
name old alloc/op new alloc/op delta
Readdirname-12 14.4kB ± 0% 6.2kB ± 0% -57.08% (p=0.000 n=10+10)
Readdir-12 41.6kB ± 0% 33.4kB ± 0% -19.77% (p=0.000 n=10+9)
ReadDir-12 21.9kB ± 0% 13.7kB ± 0% -37.39% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Readdirname-12 131 ± 0% 130 ± 0% -0.76% (p=0.000 n=10+10)
Readdir-12 367 ± 0% 366 ± 0% -0.27% (p=0.000 n=10+10)
ReadDir-12 249 ± 0% 248 ± 0% -0.40% (p=0.000 n=10+10)
A clunky benchmark I threw together that calls filepath.WalkDir on $GOMODCACHE:
name old time/op new time/op delta
WalkDir-12 91.2ms ±19% 48.7ms ± 0% -46.54% (p=0.000 n=10+10)
name old alloc/op new alloc/op delta
WalkDir-12 54.0MB ± 0% 7.6MB ± 0% -85.92% (p=0.000 n=8+9)
name old allocs/op new allocs/op delta
WalkDir-12 136k ± 0% 130k ± 0% -4.15% (p=0.000 n=8+8)
Change-Id: I00e4d48726da0e46c528ab205409afd03127b844
Reviewed-on: https://go-review.googlesource.com/c/go/+/302169
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>
The switch case for fs.ModeDevice can only be reached for block devices
while character devices match fs.ModeDevice | fs.ModeCharDevice. This
would cause character devices to wrongly be reported as regular files.
This bug has existed since the switch was first introduced in CL 5624048.
Change-Id: Icdbedb015e5376b385b3115d2e4574daa052f796
Reviewed-on: https://go-review.googlesource.com/c/go/+/300891
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Calling fs.Sub with the result of fs.Sub multiple times creates a deep
call stack for Open and other methods. Enhance the fs.FS returned by
fs.Sub to implement fs.SubFS and reduce the call stack.
Fixes#45349
Change-Id: I10e10501e030176e10e2ae5ad260212e5c784bed
Reviewed-on: https://go-review.googlesource.com/c/go/+/306769
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Renamed newNamedTypeWithSym to newIncompleteNamedType. Added some extra
comments to types.NewNamed and types.SetUnderlying.
Change-Id: Idc5a6379991c26b429d91bae9fe1adef8457a75c
Reviewed-on: https://go-review.googlesource.com/c/go/+/307029
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Ingo Oeser <nightlyone@googlemail.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
This way, "go version" will report the "base version" or major version
that the tool corresponds to. This is the same version number that is
matched against build tags such as "go1.14" or "!go1.16".
Obtaining this version being built is non-trivial, since we can't just
import a package or query git. The added comments document the chosen
mechanism, based on a regular expression. It was chosen over AST parsing
as it would add a significant amount of code without much gain, given
how simple the goversion.go file is.
For #41116.
Change-Id: I653ae935e27c13267f23898f89c84020dcd6e194
Reviewed-on: https://go-review.googlesource.com/c/go/+/264938
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Trust: Daniel Martí <mvdan@mvdan.cc>
Trust: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
The changes between (equivalent, and reviewed) go/types/errors.go
and errors.go can be seen by comparing patchset 1 and 2. The actual
changes are removing the "// UNREVIEWED" marker.
The go/types version is significantly different as it handles
error codes but doesn't have some of the types2 changes.
Change-Id: I48f79ce31490db938c830df7d38e247d55d54f2d
Reviewed-on: https://go-review.googlesource.com/c/go/+/305577
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
The changes between (equivalent, and reviewed) go/types/stmt.go
and stmt.go can be seen by comparing patchset 1 and 2. The actual
changes are removing the "// UNREVIEWED" marker, and minor adjustments
to get the code slightly closer to go/types/stmt.go.
The primary differences compared to go/types are:
- use of syntax rather than go/ast package, with significant
differences in the representation of switch and select statements,
range clauses of for statements, and inc/dec statements.
- no reporting of error codes
- use or error_ for collecting addition error information
Change-Id: I4409f62ecafd0653e4c8ef087c2580d8f0544efc
Reviewed-on: https://go-review.googlesource.com/c/go/+/305576
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
The changes between (equivalent, and reviewed) go/types/check_test.go
and check_test.go can be seen by comparing patchset 1 and 2. The actual
changes are removing the "// UNREVIEWED" marker, and minor adjustments
to get the code slightly closer to go/types/check_test.go.
The primary differences compared to go/types are:
- use of syntax rather than go/ast package
- re-implemented mechanism for error matching and elimination
based on the syntax.ErrorMap mechanism (there's no exported
access to the syntax scanner)
- error matching permits for column tolerances because types2
column information doesn't match go/types column information
Change-Id: I8ae6bc93dfa2b517673b642064a1f09166755286
Reviewed-on: https://go-review.googlesource.com/c/go/+/305573
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>
The changes between (equivalent, and reviewed) go/types/resolver.go
and resolver.go can be seen by comparing patchset 1 and 2. The actual
changes are removing the "// UNREVIEWED" marker.
The primary differences compared to go/types are:
- use of syntax rather than go/ast package
- not using a walkDecl abstraction for const/var/type declarations
Change-Id: Id8d7b069813149ca4eadbb61d1124b22c56a91b1
Reviewed-on: https://go-review.googlesource.com/c/go/+/305572
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>
There are a few remaining places in obj6 where we hard-code
safe-on-entry registers. Fix those to use the consts.
For #40724.
Change-Id: Ie640521aa67d6c99bc057553dc122160049c6edc
Reviewed-on: https://go-review.googlesource.com/c/go/+/307009
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Change-Id: I2ed394fccec83aac09a49673554cbf504787965b
Reviewed-on: https://go-review.googlesource.com/c/go/+/306810
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Internally the compiler uses a0,a1,a3 string, not [3]string,
and this lead to different parameter passing (memory, versus registers)
which of course did not work.
Updates #40724.
Change-Id: I1dbf479b88134559ba54b4b00a042b9a0fd128b8
Reviewed-on: https://go-review.googlesource.com/c/go/+/306910
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Fixes a mistake from golang.org/cl/298670.
Change-Id: I2f789f9fe003c856a993f7d462a2e84936743a1a
Reviewed-on: https://go-review.googlesource.com/c/go/+/306930
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
This change removes two short-circuits for zero-sized types (zero-sized
structs and zero-sized struct fields) in the recursive cases of the ABI
algorithm, because this does not match the spec's algorithm, nor the
compiler's algorithm.
The failing case here is a struct with a field that is an array of
non-zero length but whose element type is zero-sized. This struct must
be stack-assigned because of the array, according to the algorithm.
The reflect package was register-assigning it.
Because there were two short-circuits, this can also appear if a struct
has a field that is a zero-sized struct but contains such an array,
also.
This change adds regression tests for both of these cases.
For #44816.
For #40724.
Change-Id: I956804170962448197a1c9853826e3436fc8b1ea
Reviewed-on: https://go-review.googlesource.com/c/go/+/306929
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Leftover values that have been replaced can cause problems in later
passes (within expandCalls). For example, a struct select that
itself yields a struct will have a problematic rewrite, if the chance
is presented.
Updates #40724.
Change-Id: I1b445c47c301c3705f7fc0a9d39f1f5c84f4e190
Reviewed-on: https://go-review.googlesource.com/c/go/+/306869
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>
Reviewed-by: Than McIntosh <thanm@google.com>
This change finishes off functionality register ABI for the reflect
package.
Specifically, it implements a call on a MakeFunc'd value by performing
the reverse process that reflect.Value.Call does, using the same ABI
steps. It implements a call on a method value created by reflect by
translating between the method value's ABI to the method's ABI.
Tests are added for both cases.
For #40724.
Change-Id: I302820b61fc0a8f94c5525a002bc02776aef41af
Reviewed-on: https://go-review.googlesource.com/c/go/+/298670
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
cgo_unsafe_args paragma indicates that the function (or its
callee) uses address and offsets to dispatch arguments, which
currently using ABI0 frame layout. Pin them to ABI0.
With this, "GOEXPERIMENT=regabi,regabiargs go run hello.go" works
on Darwin/AMD64.
Change-Id: I3eadd5a3646a9de8fa681fa0a7f46e7cdc217d24
Reviewed-on: https://go-review.googlesource.com/c/go/+/306609
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
asm_windows.s contains dummy references of syscall.loadlibrary
and syscall.getprocaddress, to trigger ABI wrapper/alias
generation to get ABI0 symbols for them. The comment says they
are called from assembly in other packages, but I couldn't find
where. They are defined in Go and only referenced in Go.
CL 179862 removed dummy references in the runtime. This CL
is similar, for the syscall package.
Also, with CL 306609, they will have ABI0 definitions anyway.
Change-Id: I5c7b0efb9762e4ad9c94f0beea8d053d8c8b2cd1
Reviewed-on: https://go-review.googlesource.com/c/go/+/306709
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Austin Clements <austin@google.com>
"internal/abi" package depends on runtime package and are supposed to
compile with "-+" option.Added internal/abi in the known list of package
that runtime depends on,so that "internal/abi" compiles with "-+".
Fixes#45144
Change-Id: Iad9e2589f2fbd014260cb8ef9f943d2126015fe1
Reviewed-on: https://go-review.googlesource.com/c/go/+/306289
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
It is never used. It is actually CompilationUnit.Lib.Pkg that
contains the package path.
Change-Id: I18189644ea080080868d144e81dfee02f4549133
Reviewed-on: https://go-review.googlesource.com/c/go/+/306710
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Remove a no-longer-used function.
Change-Id: Iad383a9d158d31d4c8b65dd39b71849b44c6b52e
Reviewed-on: https://go-review.googlesource.com/c/go/+/306829
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
The math to invert the input index was wrong.
Fixes#45323
Change-Id: I7c68cac280e8f01a9c806ecb0f195f169267437e
Reviewed-on: https://go-review.googlesource.com/c/go/+/306431
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: fannie zhang <Fannie.Zhang@arm.com>
Reviewed-by: David Chase <drchase@google.com>
Currently, gcTestMoveStackOnNextCall doubles the stack allocation on
each call because stack movement always doubles the stack. That's
rather unfortunate if you're doing a bunch of stack movement tests in
a row that don't actually have to grow the stack because you'll
quickly hit the stack size limit even though you're hardly using any
of the stack.
Fix this by adding a special stack poison value for
gcTestMoveStackOnNextCall that newstack recognizes and inhibits the
allocation doubling.
Change-Id: Iace7055a0f33cb48dc97b8f4b46e45304bee832c
Reviewed-on: https://go-review.googlesource.com/c/go/+/306672
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
gcTestMoveStackOnNextCall can fail to move the stack in very rare
cases if there's an unfortunately timed preemption that clobbers the
stack guard. This won't happen multiple times in quick succession, so
make the test just retry a few times.
Change-Id: I247dc0551514e269e7132cee7945291429b0e865
Reviewed-on: https://go-review.googlesource.com/c/go/+/306671
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
In CL 305672 we preserve the pointer type of a store by just not
decomposing it. But this can be problematic when the source of
the store is a direct interface aggregate type (e.g.
struct { x map[int]int }.
In this CL we take a different approach: we preserve the store
type when generating the new store, but also decompose the source.
Fixes#45344.
Change-Id: If5dd496458dee95aa649c6d106b96a6cdcf3e60d
Reviewed-on: https://go-review.googlesource.com/c/go/+/306669
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>
This change causes finalizers, reflect calls, and Windows syscall
callbacks to assume the register ABI when GOEXPERIMENT=regabiargs is
set. That is, when all Go functions are using the new ABI by default,
these features should assume the new ABI too.
For #40724.
Change-Id: Ie4ee66b8085b692e1ff675f01134c9a4703ae1b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/306571
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Currently when register assignment fails we roll back all the abiParts
that were generated in the process. However, the total number of
registers also increases, but does not get rolled back. The result is
a very incorrect register assignment.
For #40724.
For #44816.
Change-Id: I1934ea5f95f7608ff2067166255099dbc9135e8c
Reviewed-on: https://go-review.googlesource.com/c/go/+/306109
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Includes test.
Long term, need to make the offending code be more in terms
of official types package offsets, instead of duplicating that
logic.
For #40724.
Change-Id: Id33a153f10aed3289cc48d1f99a8e0f6ece9474d
Reviewed-on: https://go-review.googlesource.com/c/go/+/306469
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
For struct like { { a int64; b int16 }; c int32 }, on 64-bit
machines the offset of c is 16, as the inner struct takes 16
bytes because we round up type size to its alignment. Update the
abi package's offset calculation to include this.
We only need to do this for struct type, because for all other
types its size is naturally aligned.
TODO: add a test.
Change-Id: I0c661768cb1ed3cb409b20a88b7e23e059f8e3e1
Reviewed-on: https://go-review.googlesource.com/c/go/+/306449
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: David Chase <drchase@google.com>
The covers three kinds of uses:
1. Calls of closures from assembly. These are always ABIInternal calls
without wrappers. I went through every indirect call in the runtime
and I think mcall is the only case of assembly calling a Go closure in
a way that's affected by ABIInternal. systemstack also calls a
closure, but it takes no arguments.
2. Calls of Go functions that expect raw ABIInternal pointers. I also
only found one of these: callbackasm1 -> cgocallback on Windows. These
are trickier to find, though.
3. Finally, I found one case on NetBSD where new OS threads were
directly calling the Go runtime entry-point from assembly via a PC,
rather than going through a wrapper. This meant new threads may not
have special registers set up. In this case, a change on all other
OSes had already forced new thread entry to go through an ABI wrapper,
so I just caught NetBSD up with that change.
With this change, I'm able to run a "hello world" with
GOEXPERIMENT=regabi,regabiargs.
For #40724.
Change-Id: I2a6d0e530c4fd4edf13484d923891c6160d683aa
Reviewed-on: https://go-review.googlesource.com/c/go/+/305669
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
The logic for constructing calls in (*state).call is based around
targeted experiments with register-based calls. However, when the
register ABI is turned on everywhere, it currently doesn't account for
direct calls to non-ABIInternal functions. This CL adds a much simpler
path to (*state).call when regabiargs is turned on that looks at the
ABI of the target function.
For #40724.
Change-Id: I7f4f5fed8a5ec131bcf1ce5b9d94d45672a304cb
Reviewed-on: https://go-review.googlesource.com/c/go/+/306410
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
When regabiargs is enabled, a function's incoming ABI should be
determined solely by the function's own definition ABI (which is
usually ABIInternal, but can be ABI0 for ABI wrappers).
For example, the current code miscompiles ABI0 -> ABIInternal wrappers
when the experiment is enabled because it treats the wrapper itself as
being called as ABIInternal. This causes it to assume the incoming
arguments are already in registers, so usually the wrapper doesn't do
anything with the arguments because it thinks they're already in the
right place. With this fix, these wrappers now correctly load the
arguments from the stack and put them in registers.
For #40724.
Change-Id: Iec784e88ebc55d9e95e830ed7533aa336f3b1ca2
Reviewed-on: https://go-review.googlesource.com/c/go/+/306409
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Currently the prologue generated for WRAPPER assembly functions uses BX
and DI, but these are argument registers in the register-based calling
convention. Thus, these end up being clobbered when we want to have an
ABIInternal assembly function.
Define REGENTRYTMP0 and REGENTRYTMP1, aliases for the dedicated function
entry scratch registers R12 and R13, and use those instead.
For #40724.
Change-Id: Ica78c4ccc67a757359900a66b56ef28c83d88b3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/303314
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
For in-register arguments, it must have only a single copy of it
present in the function. If there are multiple copies, it confuses
the register allocator, as they are in the same register.
Change-Id: I55cb06746f08aa7c9168026d0f411bce0a9f93f4
Reviewed-on: https://go-review.googlesource.com/c/go/+/306330
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>
Coupling object resolution to parsing complicates the parsing code, and
is a barrier to improvement. It requires passing around context such as
'lhs' or 'keyOk', and even then sometimes requires guess-work, such as
whether to resolve the key in a composite literal.
In this CL we delay object resolution to a separate pass after the file
parse completes. This makes it easier to see logic of scoping, and
removes state from the parsing code. This can enable subsequent
improvements such as optionally skipping object resolution, aligning the
parser with cmd/compile/internal/syntax, and allowing alternative
parsers to reuse object resolution.
The additional AST traversal appears to slow down parsing by around 4%.
That seems small enough not to worry about, especially since performance
sensitive users may eventually be able to disable object resolution
entirely, saving around 18% off the previous baseline. I'll also mail a
speculative CL showing how we can significantly mitigate the cost of
object resolution by transposing scopes.
For #45104
Change-Id: I98d9143fd77ae29e84ec7c3ae2fdb1139510da37
Reviewed-on: https://go-review.googlesource.com/c/go/+/304455
Trust: Robert Findley <rfindley@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
This change modifies the system that allows Go functions to be set as
callbacks in various Windows systems to support the new register ABI.
For #40724.
Change-Id: Ie067f9e8a76c96d56177d7aa88f89cbe7223e12e
Reviewed-on: https://go-review.googlesource.com/c/go/+/300113
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
