Keith Randall [Sat, 15 Mar 2025 00:52:33 +0000 (17:52 -0700)]
cmd/compile: allow pointer-containing elements in stack allocations
For variable-sized allocations.
Turns out that we already implement the correct escape semantics
for this case. Even when the result of the "make" does not escape,
everything assigned into it does.
Change-Id: Ia123c538d39f2f1e1581c24e4135a65af3821c5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/657937 Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Robert Griesemer <gri@google.com>
The runtime/poll package has just gained support for overlapped IO,
see CL 660595 and CL 661955. The only remaining piece was making it
visible to user code via os.NewFile.
Some of the poll.FD.Init responsibility has been moved to os.NewFile
to avoid unnecessary syscalls for the common case of using os.Open,
os.Create, os.OpenFile, and os.Pipe, where we know that the file
is not opened for overlapped IO.
Some internal/poll tests have been moved to the os package to exercise
public APIs rather than internal ones.
The os.NewFile function definition has been moved into an OS-agnostic
file to avoid having duplicated documentation and ensure that the
caller is aware of its behavior across all platforms.
Closes #19098.
Cq-Include-Trybots: luci.golang.try:gotip-windows-amd64-longtest,gotip-windows-amd64-race,gotip-windows-arm64
Change-Id: If043f8b34d588cd4b481777203107ed92d660fd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/662236
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Damien Neil <dneil@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
Auto-Submit: Damien Neil <dneil@google.com>
Keith Randall [Wed, 12 Feb 2025 02:58:13 +0000 (18:58 -0800)]
cmd/compile: aggregate scalar allocations for heap escapes
If multiple small scalars escape to the heap, allocate them together
with a single allocation. They are going to be aggregated together
in the tiny allocator anyway, might as well do just one runtime call.
Change-Id: I4317e29235af63de378a26436a18d7fb0c39e41f
Reviewed-on: https://go-review.googlesource.com/c/go/+/648536 Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
internal/syscall/windows: use unsafe.Pointer instead of uintptr
Some functions accept a uintptr when they should accept an
unsafe.Pointer, else the compiler won't know that the pointer should
be kept alive across the call, potentially causing undefined behavior.
Fixes #73156 (potentially)
Change-Id: I29c847eb8ffbb785fabf217e9f3718d10cfb5047
Reviewed-on: https://go-review.googlesource.com/c/go/+/662855
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Damien Neil <dneil@google.com> Reviewed-by: Damien Neil <dneil@google.com>
Instead of always allocating variable-sized "make" calls on the heap,
allocate a small, constant-sized array on the stack and use that array
as the backing store if it is big enough.
Requires the result of the "make" doesn't escape.
if cap <= K {
var arr [K]E
slice = arr[:len:cap]
} else {
slice = makeslice(E, len, cap)
}
Pretty conservatively for now, K = 32/sizeof(E). The slice header is
already 24 bytes, so wasting 32 bytes of stack if the requested size
is too big isn't that bad. Larger would waste more stack space but
maybe avoid more allocations.
This CL also requires the element type be pointer-free. Maybe we
could relax that at some point, but it is hard. If the element type
has pointers we can get heap->stack pointers (in the case where the
requested size is too big and the slice is heap allocated).
Note that this only handles the case of makeslice called directly from
compiler-generated code. It does not handle slices built in the
runtime on behalf of the program (e.g. in growslice). Some of those
are currently handled by passing in a tmpBuf (e.g. concatstrings),
but we could probably do more.
Change-Id: I8378efad527cd00d25948a80b82a68d88fbd93a1
Reviewed-on: https://go-review.googlesource.com/c/go/+/653856 Reviewed-by: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Alexander Musman [Tue, 1 Apr 2025 15:43:38 +0000 (18:43 +0300)]
cmd/compile: improve store-to-load forwarding with compatible types
Improve the compiler's store-to-load forwarding optimization by relaxing the
type comparison condition. Instead of requiring exact type equality (CMPeq),
we now use copyCompatibleType which allows forwarding between compatible
types where safe.
Fix several size comparison bugs in the nested store patterns. Previously,
we were comparing the size of the outer store with the load type,
rather than comparing with the size of the actual store being forwarded
from.
Skip OpConvert in dead store elimination to help get rid of dead stores such
as zeroing slices. OpConvert, like OpInlMark, doesn't really use the memory.
This optimization is particularly beneficial for code that creates slices with
computed pointers, such as the runtime's heapBitsSlice function, where
intermediate calculations were previously causing the compiler to miss
store-to-load forwarding opportunities.
Change-Id: I031552623e6d5a3b1b5be8325e6314706e45534f
Reviewed-on: https://go-review.googlesource.com/c/go/+/662075 Reviewed-by: Carlos Amedee <carlos@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Keith Randall <khr@golang.org>
Change-Id: I2a941275dd64ef858cbf02d31a759fdc5c082ceb
Reviewed-on: https://go-review.googlesource.com/c/go/+/662635
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Carlos Amedee <carlos@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org>
execIO has multiple return paths and multiple places where error is
mangled. This CL simplifies the function by just having one return
path.
Some more tests have been added to ensure that the error handling
is done correctly.
Updates #19098.
Change-Id: Ida0b1e85d4d123914054306e5bef8da94408b91c
Reviewed-on: https://go-review.googlesource.com/c/go/+/662215
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Damien Neil <dneil@google.com>
Auto-Submit: Carlos Amedee <carlos@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org>
Michael Pratt [Thu, 3 Apr 2025 15:16:36 +0000 (15:16 +0000)]
runtime: add thread exit plus vgetrandom stress test
Add a regression test similar to the reproducer from #73141 to try to
help catch future issues with vgetrandom and thread exit. Though the
test isn't very precise, it just hammers thread exit.
When the test reproduces #73141, it simply crashes with a SIGSEGV and no
output or stack trace, which would be very unfortunate on a builder.
https://go.dev/issue/49165 tracks collecting core dumps from builders,
which would make this more tractable to debug.
For #73141.
Change-Id: I6a6a636c7d7b41e2729ff6ceb30fd7f979aa9978
Reviewed-on: https://go-review.googlesource.com/c/go/+/662636 Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
Mark Freeman [Thu, 3 Apr 2025 21:26:55 +0000 (14:26 -0700)]
go/types, types2: better error messages for invalid qualified identifiers
This change borrows code from CL 631356 by Emmanuel Odeke (thanks!).
Fixes #70549.
Change-Id: Id6f794ea2a95b4297999456f22c6e02890fce13b
Reviewed-on: https://go-review.googlesource.com/c/go/+/662775
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Robert Griesemer <gri@google.com> Reviewed-by: Robert Griesemer <gri@google.com> Reviewed-by: Mark Freeman <mark@golang.org>
testing: clarify how B.Loop avoids optimizing away all the useful work
As discussed in #73137, we want to clarify the description of how
B.Loop avoids surprising optimizations, while also hinting that
the exact approach might change in the future.
Updates #73137
Change-Id: I8536540cd5d79804a47fba8cd6eec3821864309d
Reviewed-on: https://go-review.googlesource.com/c/go/+/662356
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
Auto-Submit: Alan Donovan <adonovan@google.com> Reviewed-by: Alan Donovan <adonovan@google.com>
internal/syscall/windows: define NtQueryInformationFile buffer as unsafe.Pointer
The unsafe.Pointer -> uintptr conversion must happen when calling
syscall.Syscall, not when calling the auto-generated wrapper function,
else the Go compiler doesn't know that it has to keep the pointer alive.
This can cause undefined behavior and stack corruption.
Change-Id: I1fc246783ebb215882d7144d05dbe2433dc66751
Reviewed-on: https://go-review.googlesource.com/c/go/+/662415 Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>
Change-Id: I5af1be2b117d9fb8d570739637499923de62251c
Reviewed-on: https://go-review.googlesource.com/c/go/+/662395
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com>
Commit-Queue: Keith Randall <khr@golang.org> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Keith Randall <khr@golang.org>
0x2b3bfa0 [Fri, 28 Feb 2025 09:05:52 +0000 (10:05 +0100)]
make.bat: fix GOROOT_BOOTSTRAP detection
Due to a flaw in the %GOROOT_BOOTSTRAP% detection logic, the last Go
executable found by `where go` was taking precedence over the first one.
In batch scripts, environment variable expansion happens when each line
of the script is read, not when it is executed. Thus, the check in the
loop for GOROOT_BOOTSTRAP being unset would always be true, even when
the variable had been set in a previous loop iteration.
See SET /? for more information.
Change-Id: I15ddcbe771a902acb47a1f07ba7f4cb8a311e0dc
Reviewed-on: https://go-review.googlesource.com/c/go/+/653535
Auto-Submit: Carlos Amedee <carlos@golang.org> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Quim Muntal <quimmuntal@gmail.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Alan Donovan [Thu, 3 Apr 2025 14:04:02 +0000 (10:04 -0400)]
api: move go1.25 to next/70250
My CL 645115 added the new entries in the wrong place,
prematurely creating the go1.25 file.
Also, add the missing release note.
Change-Id: Ib5b5ccfb42757a9ea9dc93e33b3e3ed8e8bd7d3f
Reviewed-on: https://go-review.googlesource.com/c/go/+/662615
Auto-Submit: Alan Donovan <adonovan@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Joel Sing [Thu, 20 Feb 2025 11:55:37 +0000 (22:55 +1100)]
cmd/internal/obj/arm64: return a bit shift from movcon
Return the shift in bits from movcon, rather than returning an index.
This allows a number of multiplications to be removed, making the code
more readable. Scale down to an index only when encoding.
Change-Id: I1be91eb526ad95d389e2f8ce97212311551790df
Reviewed-on: https://go-review.googlesource.com/c/go/+/650939
Auto-Submit: Joel Sing <joel@sing.id.au> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Joel Sing [Thu, 20 Feb 2025 11:54:56 +0000 (22:54 +1100)]
cmd/internal/obj/arm64: deduplicate con32class
Teach conclass how to handle 32 bit values and deduplicate the code
between con32class and conclass.
Change-Id: I9c5eea31d443fd4c2ce700c6ea21e1d0bef665b0
Reviewed-on: https://go-review.googlesource.com/c/go/+/650938
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Auto-Submit: Joel Sing <joel@sing.id.au>
Michael Pratt [Thu, 3 Apr 2025 03:26:25 +0000 (03:26 +0000)]
runtime: cleanup M vgetrandom state before dropping P
When an M is destroyed, we put its vgetrandom state back on the shared
list for another M to reuse. This list is simply a slice, so appending
to the slice may allocate. Currently this operation is performed in
mdestroy, after the P is released, meaning allocation is not allowed.
More the cleanup earlier in mdestroy when allocation is still OK.
Also add //go:nowritebarrierrec to mdestroy since it runs without a P,
which would have caught this bug.
Fixes #73141.
Change-Id: I6a6a636c3fbf5c6eec09d07a260e39dbb4d2db12
Reviewed-on: https://go-review.googlesource.com/c/go/+/662455 Reviewed-by: Jason Donenfeld <Jason@zx2c4.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com>
Julia Lapenko [Thu, 6 Mar 2025 14:54:17 +0000 (17:54 +0300)]
cmd/compile/internal/devirtualize: do not select a zero-weight edge as the hottest one
When both a direct call and an interface call appear on the same line,
PGO devirtualization may make a suboptimal decision. In some cases,
the directly called function becomes a candidate for devirtualization
if no other relevant outgoing edges with non-zero weight exist for the
caller's IRNode in the WeightedCG. The edge to this candidate is
considered the hottest. Despite having zero weight, this edge still
causes the interface call to be devirtualized.
This CL prevents devirtualization when the weight of the hottest edge
is 0.
Fixes #72092
Change-Id: I06c0c5e080398d86f832e09244aceaa4aeb98721
Reviewed-on: https://go-review.googlesource.com/c/go/+/655475 Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Sergey Slukin [Sat, 22 Mar 2025 18:15:18 +0000 (21:15 +0300)]
cmd/compile: changed variable name due to shadowing of package name min
Change-Id: I52e5de04d137238d6f6779edcc662f5c7433c61e
Reviewed-on: https://go-review.googlesource.com/c/go/+/660195 Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Alexander Musman [Fri, 7 Mar 2025 18:49:56 +0000 (21:49 +0300)]
cmd/compile: Remove unused 'NoInline' field from CallExpr stucture
Remove the 'NoInline' field from CallExpr stucture, as it's no longer
used after enabling of tail call inlining.
Change-Id: Ief3ada9938589e7a2f181582ef2758ebc4d03aad
Reviewed-on: https://go-review.googlesource.com/c/go/+/655816 Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Keith Randall <khr@golang.org>
Alan Donovan [Fri, 28 Mar 2025 22:04:16 +0000 (18:04 -0400)]
cmd/vet: add waitgroup analyzer
+ relnote
Fixes #18022
Change-Id: I92d1939e9d9f16824655c6c909a5f58ed9500014
Reviewed-on: https://go-review.googlesource.com/c/go/+/661519 Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Commit-Queue: Alan Donovan <adonovan@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Auto-Submit: Alan Donovan <adonovan@google.com>
Change-Id: I15a6d1979cc1e71d3065bc50f09dc8d3f6c6cdc0
Reviewed-on: https://go-review.googlesource.com/c/go/+/661518
Auto-Submit: Alan Donovan <adonovan@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Commit-Queue: Alan Donovan <adonovan@google.com>
internal/poll: defer IOCP association until first IO operation
Defer the association of the IOCP to the handle until the first
I/O operation is performed.
A handle can only be associated with one IOCP at a time, so this allows
external code to associate the handle with their own IOCP and still be
able to use a FD (through os.NewFile) to pass the handle around
(e.g. to a child process standard input, output, and error) without
having to worry about the IOCP association.
This CL doesn't change any user-visible behavior, as os.NewFile still
initializes the FD as non-pollable.
For #19098.
Change-Id: Id22a49846d4fda3a66ffcc0bc1b48eb39b395dc5
Reviewed-on: https://go-review.googlesource.com/c/go/+/661955 Reviewed-by: Damien Neil <dneil@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Joel Sing [Sun, 5 Jan 2025 13:30:30 +0000 (00:30 +1100)]
cmd/internal/obj/riscv: add support for vector fixed-point arithmetic instructions
Add support for vector fixed-point arithmetic instructions to the
RISC-V assembler. This includes single width saturating addition
and subtraction, averaging addition and subtraction and scaling
shift instructions.
Change-Id: I9aa27e9565ad016ba5bb2b479e1ba70db24e4ff5
Reviewed-on: https://go-review.googlesource.com/c/go/+/646776 Reviewed-by: Mark Ryan <markdryan@rivosinc.com> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Russ Cox [Tue, 11 Mar 2025 15:23:24 +0000 (11:23 -0400)]
runtime: fix plan9 monotonic time, crypto randomness
Open /dev/bintime at process start on Plan 9,
marked close-on-exec, hold it open for the duration of the
process, and use it for obtaining time.
The change to using /dev/bintime also sets up for an upcoming
Plan 9 change to add monotonic time to that file. If the monotonic
field is available, then nanotime1 and time.now use that field.
Otherwise they fall back to using Unix nanoseconds as "monotonic",
as they always have.
Before this CL, monotonic time went backward any time
aux/timesync decided to adjust the system's time-of-day backward.
Also use /dev/random for randomness (once at startup).
Before this CL, there was no real randomness in the runtime
on Plan 9 (the crypto/rand package still had some). Now there will be.
qmuntal [Mon, 31 Mar 2025 09:43:12 +0000 (11:43 +0200)]
os,internal/poll: support I/O on overlapped handles not added to the poller
Calling syscall.ReadFile and syscall.WriteFile on overlapped handles
always need to be passed a valid *syscall.Overlapped structure, even if
the handle is not added to a IOCP (like the Go runtime poller). Else,
the syscall will fail with ERROR_INVALID_PARAMETER.
We also need to handle ERROR_IO_PENDING errors when the overlapped
handle is not added to the poller, in which case we need to block until
the operation completes.
Previous CLs already added support for overlapped handles to the poller,
mostly to keep track of the file offset independently of the file
pointer (which is not supported for overlapped handles).
Fixed #15388.
Updates #19098.
Change-Id: I2103ab892a37d0e326752ae8c2771a43c13ba42e
Reviewed-on: https://go-review.googlesource.com/c/go/+/661795
Auto-Submit: Quim Muntal <quimmuntal@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Carlos Amedee <carlos@golang.org> Reviewed-by: Damien Neil <dneil@google.com>
Mateusz Poliwczak [Sat, 29 Mar 2025 18:49:25 +0000 (19:49 +0100)]
cmd/compile/internal/ssa: replace uses of interface{} with Sym/Aux
Change-Id: I0a3ce2e823697eee5bb5e7d5ea0ef025132c0689
Reviewed-on: https://go-review.googlesource.com/c/go/+/661655 Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Keith Randall <khr@google.com>
thepudds [Sat, 7 Dec 2024 21:17:33 +0000 (16:17 -0500)]
internal/runtime/maps: speed up small map lookups ~1.7x for unpredictable keys
On master, lookups on small Swiss Table maps (<= 8 elements) for
non-specialized key types are seemingly a performance regression
compared to the Go 1.23 map implementation (reported in #70849).
Currently, a linear scan is used for gets in these cases.
This CL changes (*Map).getWithKeySmall to instead use the SIMD or SWAR
match on the control bytes to then jump to candidate matching slots,
with sample results below for a 16-byte key. This especially helps the
hit case when the key is unpredictable, which previously had to scan an
unpredictable number of control bytes to find a candidate slot when the
key is unpredictable.
Separately, other CLs in this stack modify the main Swiss Table
benchmarks to randomize lookup key order (vs. previously most of the
benchmarks had a repeating lookup key ordering, which likely is
predictable until the map is too big). We have sample results for the
randomized key order benchmarks followed by results from the older
benchmarks.
The first table below is with randomized key order. For hits, the older
results get slower as there are more elements. With this CL, we see hits
for unpredictable key ordering (sizes 2-8) get a ~1.7x speedup from
~25ns to ~14ns, with a now consistent lookup time for the different
sizes. (The 1 element size map has a predictable key ordering because
there is only one key, and that reports a modest ~0.5ns or ~3%
performance penalty). Misses for unpredictable key order get a ~1.3x
speedup, from ~13ns to ~10ns, with similar results for the 1 element
size.
The next table uses the original benchmarks from just before this CL
stack (i.e., without shuffling lookup keys).
With this CL, we see improvement that is directionally similar to the
above results but not as large, presumably because the branches in the
linear scan are fairly predictable with predictable keys. (The numbers
here also include the time from a mod in the benchmark code, which
seemed to take around ~1/3 of CPU time based on spot checking a couple
of examples, vs. the modified benchmarks shown above have removed that
mod).
Joel Sing [Tue, 18 Jun 2024 13:36:43 +0000 (23:36 +1000)]
cmd/internal/obj/riscv: add support for vector integer arithmetic instructions
Add support for vector integer arithmetic instructions to the RISC-V
assembler. This includes vector addition, subtraction, integer
extension, add-with-carry, subtract-with-borrow, bitwise logical
operations, comparison, min/max, integer division and multiplication
instructions.
Change-Id: I8c191ef8e31291e13743732903e4f12356133a46
Reviewed-on: https://go-review.googlesource.com/c/go/+/646775 Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Cherry Mui [Tue, 25 Mar 2025 19:09:31 +0000 (15:09 -0400)]
cmd/link: handle Mach-O X86_64_RELOC_SUBTRACTOR in internal linking
With recent LLVM toolchain, on macOS/AMD64, the race detector syso
file built from it contains X86_64_RELOC_SUBTRACTOR relocations,
which the Go linker currently doesn't handle in internal linking
mode. To ensure internal linking mode continue to work with the
race detector syso, this CL adds support of X86_64_RELOC_SUBTRACTOR
relocations.
X86_64_RELOC_SUBTRACTOR is actually a pair of relocations that
resolves to the difference between two symbol addresses (each
relocation specifies a symbol). For the cases we care (the race
syso), the symbol being subtracted out is always in the current
section, so we can just convert it to a PC-relative relocation,
with the addend adjusted. If later we need the more general form,
we can introduce a new mechanism (say, objabi.R_DIFF) that works
as a pair of relocations like the Mach-O one.
As we expect the pair of relocations be consecutive, don't reorder
(sort) relocation records when loading Mach-O objects.
Change-Id: I757456b07270fb4b2a41fd0fef67a2b39dd6b238
Reviewed-on: https://go-review.googlesource.com/c/go/+/660715
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Than McIntosh <thanm@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
qmuntal [Fri, 28 Mar 2025 19:38:34 +0000 (20:38 +0100)]
internal/poll: honor ERROR_OPERATION_ABORTED if pipe is not closed
FD.Read converts a syscall.ERROR_OPERATION_ABORTED error to
ErrFileClosing. It does that in case the pipe operation was aborted by
a CancelIoEx call in FD.Close.
It doesn't take into account that the operation might have been
aborted by a CancelIoEx call in external code. In that case, the
operation should return the error as is.
Change-Id: I75dcf0edaace8b57dc47b398ea591ca9f116112b
Reviewed-on: https://go-review.googlesource.com/c/go/+/661555
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Damien Neil <dneil@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Michael Anthony Knyszek [Thu, 23 Jan 2025 15:51:35 +0000 (15:51 +0000)]
weak: clarify Pointer equality semantics
The docs currently are imprecise about comparisons. This could lead
users to believe that objects of the same type, allocated at the same
address, could produce weak pointers that are equal to
previously-created weak pointers. This is not the case. Weak pointers
map to objects, not addresses.
Update the documentation to state precisely that if two pointers do not
compare equal, then two weak pointers created from those two pointers
are guaranteed not to compare equal. Since a future pointer pointing to
the same address is not comparable with a pointer produced *before* an
object at that address has been reclaimed, this is sufficient to explain
that weak pointers map 1:1 with object offsets, not addresses.
(An object slot cannot be reused unless that slot is unreachable, so
by construction, there's never an opportunity to compare an "old" and
"new" pointer unless one uses unsafe tricks that violate the
unsafe.Pointer rules.)
Fixes #71381.
Change-Id: I5509fd433cde013926d725694d480c697a8bc911
Reviewed-on: https://go-review.googlesource.com/c/go/+/643935 Reviewed-by: Carlos Amedee <carlos@golang.org>
Auto-Submit: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Filippo Valsorda <filippo@golang.org>
Cherry Mui [Tue, 25 Mar 2025 20:55:54 +0000 (16:55 -0400)]
cmd/link: choose one with larger size for duplicated BSS symbols
When two packages declare a variable with the same name (with
linkname at least on one side), the linker will choose one as the
actual definition of the symbol if one has content (i.e. a DATA
symbol) and the other does not (i.e. a BSS symbol). When both have
content, it is redefinition error. When neither has content,
currently the choice is sort of arbitrary (depending on symbol
loading order, etc. which are subject to change).
One use case for that is that one wants to reference a symbol
defined in another package, and the reference side just wants to
see some of the fields, so it may be declared with a smaller type.
In this case, we want to choose the one with the larger size as
the true definition. Otherwise the code accessing the larger
sized one may read/write out of bounds, corrupting the next
variable. This CL makes the linker do so.
Fixes #72032.
Change-Id: I160aa9e0234702066cb8f141c186eaa89d0fcfed
Reviewed-on: https://go-review.googlesource.com/c/go/+/660696
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Than McIntosh <thanm@golang.org>
qmuntal [Fri, 28 Mar 2025 09:39:49 +0000 (10:39 +0100)]
internal/poll: don't skip empty writes on Windows
Empty writes might be important for some protocols. Let Windows decide
what do with them rather than skipping them on our side. This is inline
with the behavior of other platforms.
While here, refactor the Read/Write/Pwrite methods to reduce one
indentation level and make the code easier to read.
Fixes #73084.
Change-Id: Ic5393358e237d53b8be6097cd7359ac0ff205309
Reviewed-on: https://go-review.googlesource.com/c/go/+/661435 Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Joel Sing [Mon, 19 Aug 2024 13:57:56 +0000 (23:57 +1000)]
cmd/compile/internal/ssa: optimise more branches with zero on riscv64
Optimise more branches with zero on riscv64. In particular, BLTU with
zero occurs with IsInBounds checks for index zero. This currently results
in two instructions and requires an additional register:
li t2, 0
bltu t2, t1, 0x174b4
This is equivalent to checking if the bounds is not equal to zero. With
this change:
bnez t1, 0x174c0
This removes more than 500 instructions from the Go binary on riscv64.
Change-Id: I6cd861d853e3ef270bd46dacecdfaa205b1c4644
Reviewed-on: https://go-review.googlesource.com/c/go/+/606715
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Keith Randall [Fri, 21 Mar 2025 23:07:20 +0000 (16:07 -0700)]
maps: implement faster clone
│ base │ experiment │
│ sec/op │ sec/op vs base │
MapClone-24 66.802m ± 7% 3.348m ± 2% -94.99% (p=0.000 n=10)
Fixes #70836
Change-Id: I9e192b1ee82e18f5580ff18918307042a337fdcc
Reviewed-on: https://go-review.googlesource.com/c/go/+/660175 Reviewed-by: Michael Pratt <mpratt@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Keith Randall <khr@google.com>
Mark Freeman [Wed, 26 Mar 2025 20:13:16 +0000 (20:13 +0000)]
cmd/compile: rename some test packages in codegen
All other files here use the codegen package.
Change-Id: I714162941b9fa9051dacc29643e905fe60b9304b
Reviewed-on: https://go-review.googlesource.com/c/go/+/661135 Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>
Junyang Shao [Sun, 9 Mar 2025 20:14:32 +0000 (20:14 +0000)]
cmd/compile: match more patterns for shortcircuit
This CL tries to generalize the pattern matching of certain
shortcircuit-able CFGs a bit more:
For a shortcircuit-able CFG:
p q
\ /
b
/ \
t u
Where the constant branch is t, and b has multiple phi values
other than the control phi.
For the non-constant branch target u, we try to match the
"diamond" shape CFG:
p q
\ /
b
/ \
t u
\ /
m
or
p q
\ /
b
|\
| u
|/
m
Instead of matching u as a single block, we know try to
generalize it as a subgraph that satisfy condition:
it's a DAG that has a single entry point u, and has a
path to m.
Mateusz Poliwczak [Wed, 26 Mar 2025 14:46:30 +0000 (15:46 +0100)]
cmd/compile: remove references to *gc.Node in docs
ssa.Sym is only implemented by *ir.Name or *obj.LSym.
Change-Id: Ia171db618abd8b438fcc2cf402f40f3fe3ec6833
Reviewed-on: https://go-review.googlesource.com/c/go/+/660995
Auto-Submit: Keith Randall <khr@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Keith Randall <khr@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
qmuntal [Thu, 27 Mar 2025 09:39:40 +0000 (10:39 +0100)]
internal/syscall/windows: run go generate
CL 660595 manually edited zsyscall_windows.go, making it be out of sync
with its associated //sys directive. Longtest builders are failing
due to that.
Fixes #73069.
Change-Id: If7256ef4b831423e4925fb6e5656fe3f7ef77fea
Reviewed-on: https://go-review.googlesource.com/c/go/+/661275 Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Provide a synthesized version of the MIN/MAX/MINU/MAXU instructions
if they're not natively available. This allows these instructions to
be used in assembly unconditionally.
Use MIN in internal/bytealg.compare.
Cq-Include-Trybots: luci.golang.try:gotip-linux-riscv64
Change-Id: I8a5a3a59f0a9205e136fc3d673b23eaf3ca469f8
Reviewed-on: https://go-review.googlesource.com/c/go/+/653295 Reviewed-by: Mark Ryan <markdryan@rivosinc.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Joel Sing [Thu, 27 Feb 2025 11:54:51 +0000 (22:54 +1100)]
cmd/internal/obj/riscv: improve constant construction
Attempt to construct large constants that have a consecutive sequence
of ones from a small negative constant, with a logical right and/or
left shift. This allows for a large range of mask like constants to be
constructed with only two or three instructions, avoiding the need to
load from memory.
Change-Id: I35a77fecdd2df0ed3f33b772d518f85119d4ff66
Reviewed-on: https://go-review.googlesource.com/c/go/+/652778
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Mark Ryan <markdryan@rivosinc.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Change-Id: I2b9fe834c31d786d2e30cc02c65a6f9c455c4e8d
Reviewed-on: https://go-review.googlesource.com/c/go/+/657835
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Meidan Li <limeidan@loongson.cn>
qmuntal [Tue, 25 Mar 2025 15:42:51 +0000 (16:42 +0100)]
internal/poll: support async file operations on Windows
This CL adds support for async file operations on Windows. The affected
functions are Read, Write, Pread, and Pwrite.
The code has been slightly refactored to avoid duplication. Both the
async and sync variants follow the same code path, with the exception of
the async variant passes an overlapped structure to the syscalls
and supports the use of a completion port.
This doesn't change any user-facing behavior, as the os package still
sets the pollable parameter to false when calling FD.Init.
For #19098.
Change-Id: Iead6e51fa8f57e83456eb5ccdce28c2ea3846cc2
Reviewed-on: https://go-review.googlesource.com/c/go/+/660595 Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Damien Neil <dneil@google.com>
Jonathan Amsterdam [Wed, 26 Mar 2025 15:05:29 +0000 (11:05 -0400)]
log/slog: Handler doc points to handler guide
There's a link in the package doc, but there should be one here too.
For #73057.
Change-Id: I8f8fe73f20bb6dd49cdf23b5f7634a92d4f7add9
Reviewed-on: https://go-review.googlesource.com/c/go/+/661015 Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org> Reviewed-by: Sean Liao <sean@liao.dev>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Joel Sing [Fri, 14 Jun 2024 15:29:02 +0000 (01:29 +1000)]
cmd/internal/obj/arm64: add support for BTI instruction
Add support for the `BTI' instruction to the arm64 assembler. This
instruction provides Branch Target Identification for targets of
indirect branches. A BTI can be marked with a target type of
'C' (call), 'J' (jump) or 'JC' (jump or call).
Updates #66054
Change-Id: I1cf31a0382207bb75b9b2deb49ac298a59c00d8a
Reviewed-on: https://go-review.googlesource.com/c/go/+/646781
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Marvin Drees <marvin.drees@9elements.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Michael Matloob [Tue, 25 Mar 2025 18:10:31 +0000 (14:10 -0400)]
cmd/go/internal/modload: improve ErrNoModRoot error text
Before this change, in several cases where HasModRoot() returned false,
we'd return ErrNoModRoot. ErrNoModRoot would say that there was no
go.mod file but would not mention workspaces. With this change,
ErrNoModRoot will return error text that's different if we're in a
workspace, saying that there are no modules in the workspace.
Fixes #54419
Change-Id: I77c94d0011947bf8e33c066416ab3762502fd2e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/660675
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Alan Donovan <adonovan@google.com>
qmuntal [Tue, 25 Mar 2025 09:29:22 +0000 (10:29 +0100)]
internal/poll,net: set SIO_UDP_CONNRESET in net
Setting the SIO_UDP_CONNRESET option in internal/poll.FD.Init
adds unnecessary complexity to the FD.Init signature and
implementation. Better to set it in the net package when initializing
the UDP connection, which is where conceptually it belongs.
While here, update an outdated comment in FD.Init that said the runtime
poller doesn't support I/O operations initialized by the user
outside the internal/poll package. It does support those operations
since CL 561895.
For #19098.
Updates #21172.
Change-Id: I9a70b0deafdb4619830abe2147e2d366b4c2b890
Reviewed-on: https://go-review.googlesource.com/c/go/+/660496
Auto-Submit: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Damien Neil <dneil@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
qmuntal [Tue, 25 Mar 2025 09:42:10 +0000 (10:42 +0100)]
internal/poll: always use SetFileCompletionNotificationModes on non-socket handles
SetFileCompletionNotificationModes can be unconditionally called on
non-socket handles.
The Windows poll.FD implementation still doesn't support non-socket
pollable handles yet, so this CL doesn't change any behavior.
Support for pollable non-socket handles will come in subsequent CLs.
For #19098.
Change-Id: I811a61497cfbb26acb566c20367d212335b9d551
Reviewed-on: https://go-review.googlesource.com/c/go/+/660495 Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Auto-Submit: Damien Neil <dneil@google.com>
Joel Sing [Mon, 9 Sep 2024 15:16:14 +0000 (01:16 +1000)]
test/codegen: add combined conversion and shift tests
This adds tests for type conversion and shifts, detailing various
poor bad code generation that currently exists for riscv64. This
will be addressed in future CLs.
Change-Id: Ie1d366dfe878832df691600f8500ef383da92848
Reviewed-on: https://go-review.googlesource.com/c/go/+/615678 Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com> Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Carlos Amedee <carlos@golang.org>
Joel Sing [Mon, 27 Feb 2023 07:23:42 +0000 (18:23 +1100)]
cmd/internal/obj/arm64: move register encoding into opldrr/opstrr
Rather than having register encoding knowledge in each caller of opldrr/opstrr
(and in a separate olsxrr function), pass the registers into opldrr/opstrr and
let them handle the encoding. This reduces duplication and improves readability.
Change-Id: I50a25263f305d01454f3ff95e8b6e7c76e760ab0
Reviewed-on: https://go-review.googlesource.com/c/go/+/471521 Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Joel Sing [Sat, 7 Jan 2023 07:46:37 +0000 (18:46 +1100)]
cmd/internal/obj/arm64: move register encoding into oprrr
Rather than having register encoding knowledge in each caller of oprrr,
pass the registers into oprrr and let it handle the encoding. This reduces
duplication and improves readability.
Change-Id: Iab6c70f7796b7a8c071419654b8a5686aeee8c1b
Reviewed-on: https://go-review.googlesource.com/c/go/+/471518 Reviewed-by: Fannie Zhang <Fannie.Zhang@arm.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Michael Pratt [Mon, 24 Mar 2025 08:24:35 +0000 (04:24 -0400)]
runtime: rename runfinq to runFinalizersAndCleanups
Users see this frame in tracebacks and it serves as a hint that what is
running here is a finalizer or cleanup. But runfinq is a rather dense
name. We can give it a more obvious name to help users realize what it
is.
For #73011.
Change-Id: I6a6a636ce9a493fd00d4b4c60c23f2b1c96d3568
Reviewed-on: https://go-review.googlesource.com/c/go/+/660296
Auto-Submit: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Michael Knyszek <mknyszek@google.com>
Michael Pratt [Mon, 24 Mar 2025 07:08:33 +0000 (03:08 -0400)]
runtime: always show runfinq in traceback
Today, runtime.runfinq is hidden whenever runtime frames are hidden.
However this frame serves as a hint that this goroutine is running
finalizers, which is otherwise unclear, but can be useful when debugging
issues with finalizers.
Fixes #73011.
Change-Id: I6a6a636cb63951fbe1fefc3554fe9cea5d0a0fb6
Reviewed-on: https://go-review.googlesource.com/c/go/+/660295
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Michael Pratt <mpratt@google.com> Reviewed-by: Michael Knyszek <mknyszek@google.com>
Austin Clements [Thu, 20 Mar 2025 16:44:07 +0000 (12:44 -0400)]
testing: streamline logic in loopSlowPath
There's a fair amount of duplication of logic between various return
branches of loopSlowPath and stopOrScaleBLoop. Restructure these so
there's a single "keep going" path and a single "we're done" path.
Change-Id: I38e4c7a616f8bd7707f3ca886f38ff21dbd78b6b
Reviewed-on: https://go-review.googlesource.com/c/go/+/659658
Auto-Submit: Austin Clements <austin@google.com> Reviewed-by: Junyang Shao <shaojunyang@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Michael Pratt <mpratt@google.com>
Austin Clements [Thu, 20 Mar 2025 16:16:17 +0000 (12:16 -0400)]
testing: detect a stopped timer in B.Loop
Currently, if the user stops the timer in a B.Loop benchmark loop, the
benchmark will run until it hits the timeout and fails.
Fix this by detecting that the timer is stopped and failing the
benchmark right away. We avoid making the fast path more expensive for
this check by "poisoning" the B.Loop iteration counter when the timer
is stopped so that it falls back to the slow path, which can check the
timer.
This causes b to escape from B.Loop, which is totally harmless because
it was already definitely heap-allocated. But it causes the
test/inline_testingbloop.go errorcheck test to fail. I don't think the
escape messages actually mattered to that test, they just had to be
matched. To fix this, we drop the debug level to -m=1, since -m=2
prints a lot of extra information for escaping parameters that we
don't want to deal with, and change one error check to allow b to
escape.
Austin Clements [Thu, 20 Mar 2025 14:26:54 +0000 (10:26 -0400)]
testing: detect early return from B.Loop
Currently, if a benchmark function returns prior to B.Loop() returning
false, we'll report a bogus result. While there was no way to detect
this with b.N-style benchmarks, one way b.Loop()-style benchmarks are
more robust is that we *can* detect it.
This CL adds a flag to B that tracks if B.Loop() has finished and
checks it after the benchmark completes. If there was an early exit
(not caused by another error), it reports a B.Error.
Fixes #72933.
Updates #72971.
Change-Id: I731c1350e6df938c0ffa08fcedc11dc147e78854
Reviewed-on: https://go-review.googlesource.com/c/go/+/659656
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Austin Clements <austin@google.com> Reviewed-by: Michael Pratt <mpratt@google.com> Reviewed-by: Junyang Shao <shaojunyang@google.com>
Sean Liao [Fri, 21 Mar 2025 23:06:54 +0000 (23:06 +0000)]
go/build: prioritize build constraints in docs
Now that modules is the primary way of building go code,
GOPATH shouldn't be the first topic.
Change-Id: Icaf6c651bdcfbe69068c5980845f8eb5e40ead99
Reviewed-on: https://go-review.googlesource.com/c/go/+/660135
Auto-Submit: Ian Lance Taylor <iant@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com> Reviewed-by: Ian Lance Taylor <iant@google.com>
Jake Bailey [Sat, 22 Mar 2025 05:34:52 +0000 (22:34 -0700)]
testing: remove redundant variable capture from subtest example
Loop variable capturing hasn't been needed like this since Go 1.21;
remove it from the example.
Change-Id: I231dcfdb57832e32d524f156a605ba36d1c9d6d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/660176 Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Alan Donovan <adonovan@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Michael Matloob [Wed, 5 Mar 2025 20:17:20 +0000 (15:17 -0500)]
cmd/go: upgrade go install pkg@version's go using local mod, work
This CL changes the toolchain selection behavior for go install pkg@v
and go run pkg@v to also take into account the go and toolchain version
lines in the containing go.mod and go.work file.
Before this change, the go command would detect that go install
pkg@version or go run pkg@version was being run and skip the standard
behavior that would select the toolchain based on the go version in the
go.mod or go.work file. It would instead check the go line of the module
being downloaded and switch to that version if necessary.
With this change, the go command does not skip the standard behavior. It
proceeds to determine if an upgrade is required based on the containing
go.mod or go.work file's go and toolchain lines. Then, it checks the
module being installed to see if it would require a higher version than
the determined upgrade (or the local version if no upgrade was
determined). If it does require a higher version, then a switch happens
to that version, and if not the upgrade logic proceeds as usual doing
the upgrade if one was determined.
Sean Liao [Sat, 22 Mar 2025 01:15:22 +0000 (01:15 +0000)]
net/url: document referenced RFCs
Fixes #30611
Change-Id: If933c2a7e63d89402d2034618057ad546cf9641b
Reviewed-on: https://go-review.googlesource.com/c/go/+/660077
Auto-Submit: Ian Lance Taylor <iant@google.com> Reviewed-by: Ian Lance Taylor <iant@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Damien Neil <dneil@google.com>
Robert Griesemer [Thu, 20 Mar 2025 23:12:15 +0000 (16:12 -0700)]
go/types, types2: align trace output a bit better for easier debugging
Compute alignment padding rather than using a tab in trace output.
This aligns the ":" after the file position in typical cases (files
shorter than 1000 lines, lines shorter than 100 columns), resulting
in nicer trace output for easier debugging.
Before this CL (example trace):
x.go:8:2: -- checking type A (white, objPath = )
x.go:8:11: . -- type B
x.go:9:2: . . -- checking type B (white, objPath = A)
x.go:9:14: . . . -- type C[D]
x.go:9:13: . . . . -- instantiating type C with [D]
x.go:9:13: . . . . . -- type C
x.go:10:2: . . . . . . -- checking type C (white, objPath = A->B)
x.go:10:6: . . . . . . . -- type any
x.go:10:6: . . . . . . . => any (under = any) // *Alias
x.go:10:11: . . . . . . . -- type struct{}
x.go:10:11: . . . . . . . => struct{} // *Struct
x.go:10:2: . . . . . . => type C[_ any] struct{} (black)
With this CL:
x.go:8:2 : -- checking type A (white, objPath = )
x.go:8:11 : . -- type B
x.go:9:2 : . . -- checking type B (white, objPath = A)
x.go:9:14 : . . . -- type C[D]
x.go:9:13 : . . . . -- instantiating type C with [D]
x.go:9:13 : . . . . . -- type C
x.go:10:2 : . . . . . . -- checking type C (white, objPath = A->B)
x.go:10:6 : . . . . . . . -- type any
x.go:10:6 : . . . . . . . => any (under = any) // *Alias
x.go:10:11 : . . . . . . . -- type struct{}
x.go:10:11 : . . . . . . . => struct{} // *Struct
x.go:10:2 : . . . . . . => type C[_ any] struct{} (black)
Change-Id: Ibcf346737f57ec5351d1e1e65178e2c3c155d766
Reviewed-on: https://go-review.googlesource.com/c/go/+/659755
Auto-Submit: Robert Griesemer <gri@google.com> Reviewed-by: Robert Findley <rfindley@google.com> Reviewed-by: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Change-Id: I0dda684713dbdf5336af393f5ccbdae861c4f694
Reviewed-on: https://go-review.googlesource.com/c/go/+/652321 Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Mark Ryan <markdryan@rivosinc.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Joel Sing [Tue, 18 Feb 2025 07:23:27 +0000 (18:23 +1100)]
cmd/internal/obj/riscv: factor out shift constant code
Move the code that tests to see if a constant can be represented by a
32 bit signed integer and a logical left shift. This reduces duplication
and increases readability. Also add test coverage now that this is an
independent function.
Change-Id: Id25395b1380b00cf5b69ca201b7715ef84f7ade6
Reviewed-on: https://go-review.googlesource.com/c/go/+/652777
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Joel Sing [Tue, 18 Feb 2025 06:53:19 +0000 (17:53 +1100)]
cmd/asm/internal/asm: add additional tests for constant loads on riscv64
This improves test coverage around the various constant load edge cases.
Change-Id: Ibafeec78e76d95e9f56b48fa6bd012772bf505c5
Reviewed-on: https://go-review.googlesource.com/c/go/+/652776
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com> Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Austin Clements [Wed, 19 Mar 2025 15:46:41 +0000 (11:46 -0400)]
testing: separate b.Loop counter from b.N
Currently, b.Loop uses b.N as the iteration count target. However,
since it updates the target as it goes, the behavior is quite
different from a b.N-style benchmark. To avoid user confusion, this CL
gives b.Loop a separate, unexported iteration count target. It ensures
b.N is 0 within the b.Loop loop to help catch misuses, and commits the
final iteration count to b.N only once the loop is done (as the
documentation states "After Loop returns false, b.N contains the total
number of iterations that ran, so the benchmark may use b.N to compute
other average metrics.")
Since there are now two variables used by b.Loop, we put them in an
unnamed struct. Also, we rename b.loopN to b.loop.i because this
variable tracks the current iteration index (conventionally "i"), not
the target (conventionally "n").
Unfortunately, a simple renaming causes B.Loop to be too large for the
inliner. Thus, we make one simplification to B.Loop to keep it under
the threshold. We're about to lean into that simplification anyway in
a follow-up CL, so this is just temporary.
Derek Parker [Fri, 21 Mar 2025 19:43:32 +0000 (19:43 +0000)]
cmd/compile/internal/abi: fix ComputePadding
Fixes the ComputePadding calculation to take into account the padding
added for the current offset. This fixes an issue where padding can be
added incorrectly for certain structs.
qmuntal [Fri, 14 Mar 2025 10:56:20 +0000 (11:56 +0100)]
runtime: deduplicate context call injection on Windows
Injecting a call to a thread context is complex enough to warrant
a dedicated function so that we don't repeat the same code in multiple
places. Note that the unix sigctxt struct also follows the
same approach.
The behavior is unchanged, but the implementation semantics are now
clearer by using goarch.StackAlign instead of a mix of goarch.PtrSize,
goarch.StackAlign and hardcoded values.
While here, fix #68552.
Cq-Include-Trybots: luci.golang.try:gotip-windows-arm64
Change-Id: Ic29cd2bf322b520127fecccafd61577076945758
Reviewed-on: https://go-review.googlesource.com/c/go/+/657815 Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: David Chase <drchase@google.com>
Damien Neil [Mon, 17 Mar 2025 20:48:50 +0000 (13:48 -0700)]
runtime: fix interactions between synctest, race detector, and timers
When an AfterFunc executes in a synctest bubble, there is a series of
happens-before relationships:
1. The AfterFunc is created.
2. The AfterFunc goroutine executes.
3. The AfterFunc goroutine returns.
4. A subsequent synctest.Wait call returns.
We were failing to correctly establish the happens-before relationship
between the AfterFunc goroutine and the AfterFunc itself being created.
When an AfterFunc executes, the G running the timer temporarily switches
to the timer heap's racectx. It then calls time.goFunc, which starts a
new goroutine to execute the timer. time.goFunc relies on the new goroutine
inheriting the racectx of the G running the timer.
Normal, non-synctest timers, execute with m.curg == nil, which causes
new goroutines to inherit the g0 racectx. We were running synctest
timers with m.curg set (to the G executing synctest.Run), so the new
AfterFunc goroutine was created using m.curg's racectx. This resulted
in us not properly establishing the happens-before relationship between
AfterFunc being called and the AfterFunc goroutine starting.
Fix this by setting m.curg to nil while executing timers.
As one additional fix, when waking a blocked bubble, wake the root
goroutine rather than a goroutine blocked in Wait if there is a
timer that can fire.
Fixes #72750
Change-Id: I2b2d6b0f17f64649409adc93c2603f720494af89
Reviewed-on: https://go-review.googlesource.com/c/go/+/658595
Auto-Submit: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Michael Pratt <mpratt@google.com>
Mark Ryan [Wed, 11 Dec 2024 17:09:32 +0000 (18:09 +0100)]
cmd/internal/obj/riscv: fix the encoding for REV8 and ORCB
The instructions are currently encoded and validated using an
iIIEncoding which is incorrect as these instructions do not
take an immediate operand. Encode them instead using an
rIIEncoding as is done for the other two register argument bitmanip
instructions.
Change-Id: Ia4d9c6f6ebd2dfc381935ebc11afa8fc3664232b
Reviewed-on: https://go-review.googlesource.com/c/go/+/637317 Reviewed-by: David Chase <drchase@google.com> Reviewed-by: Joel Sing <joel@sing.id.au> Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com> Reviewed-by: Cherry Mui <cherryyz@google.com>
Damien Neil [Thu, 20 Mar 2025 22:03:43 +0000 (15:03 -0700)]
runtime/race: fix reversed conditional in test for fatal map access
Change-Id: Icc4d90355d8af07fdec852b2adf720f7cfd1edd6
Reviewed-on: https://go-review.googlesource.com/c/go/+/659735 Reviewed-by: Ian Lance Taylor <iant@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Derek Parker [Wed, 19 Mar 2025 01:54:41 +0000 (01:54 +0000)]
cmd/compile/internal/abi: fix ComputePadding
Fixes the ComputePadding calculation to take into account
the padding added for the current offset. This fixes an issue
where padding can be added incorrectly for certain structs.