Currently, threads created by the runtime exist until the whole
program exits. For #14592 and #20395, we want to be able to exit and
clean up threads created by the runtime. This commit implements that
mechanism.
The main difficulty is how to clean up the g0 stack. In cgo mode and
on Solaris and Windows where the OS manages thread stacks, we simply
arrange to return from mstart and let the system clean up the thread.
If the runtime allocated the g0 stack, then we use a new exitThread
syscall wrapper that arranges to clear a flag in the M once the stack
can safely be reaped and call the thread termination syscall.
exitThread is based on the existing exit1 wrapper, which was always
meant to terminate the calling thread. However, exit1 has never been
used since it was introduced 9 years ago, so it was broken on several
platforms. exitThread also has the additional complication of having
to flag that the stack is unused, which requires some tricks on
platforms that use the stack for syscalls.
This still leaves the problem of how to reap the unused g0 stacks. For
this, we move the M from allm to a new freem list as part of the M
exiting. Later, allocm scans the freem list, finds Ms that are marked
as done with their stack, removes these from the list and frees their
g0 stacks. This also allows these Ms to be garbage collected.
This CL does not yet use any of this functionality. Follow-up CLs
will. Likewise, there are no new tests in this CL because we'll need
follow-up functionality to test it.
Change-Id: Ic851ee74227b6d39c6fc1219fc71b45d3004bc63
Reviewed-on: https://go-review.googlesource.com/46037
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Add a compiler intrinsic for getcallersp. So we are able to get
rid of the argument (not done in this CL).
Change-Id: Ic38fda1c694f918328659ab44654198fb116668d
Reviewed-on: https://go-review.googlesource.com/69350
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: David Chase <drchase@google.com>
Now that getcallerpc is a compiler intrinsic on x86 and non-x86
platforms don't need the argument, we can drop it.
Sadly, this doesn't let us remove any dummy arguments since all of
those cases also use getcallersp, which still takes the argument
pointer, but this is at least an improvement.
Change-Id: I9c34a41cf2c18cba57f59938390bf9491efb22d2
Reviewed-on: https://go-review.googlesource.com/65474
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
First step towards removing the mandatory argument for
getcallerpc, which solves certain problems for the runtime.
This might also slightly improve performance.
Intrinsic enabled on 386, amd64, amd64p32,
runtime asm implementation removed on those architectures.
Now-superfluous argument remains in getcallerpc signature
(for a future CL; non-386/amd64 asm funcs ignore it).
Added getcallerpc to the "not a real function" test
in dcl.go, that story is a little odd with respect to
unexported functions but that is not this CL.
Fixes#17327.
Change-Id: I5df1ad91f27ee9ac1f0dd88fa48f1329d6306c3e
Reviewed-on: https://go-review.googlesource.com/31851
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
The current generator is a simple LSFR, which showed strong
correlation in higher bits, as manifested by fastrandn().
Change it with xorshift64+, which is slightly more complex,
has a larger state, but has a period of 2^64-1 and is much better
at statistical tests. The version used here is capable of
passing Diehard and even SmallCrush.
Speed is slightly worse but is probably insignificant:
name old time/op new time/op delta
Fastrand-4 0.77ns ±12% 0.91ns ±21% +17.31% (p=0.048 n=5+5)
FastrandHashiter-4 13.6ns ±21% 15.2ns ±17% ~ (p=0.160 n=6+5)
Fastrandn/2-4 2.30ns ± 5% 2.45ns ±15% ~ (p=0.222 n=5+5)
Fastrandn/3-4 2.36ns ± 7% 2.45ns ± 6% ~ (p=0.222 n=5+5)
Fastrandn/4-4 2.33ns ± 8% 2.61ns ±30% ~ (p=0.126 n=6+5)
Fastrandn/5-4 2.33ns ± 5% 2.48ns ± 9% ~ (p=0.052 n=6+5)
Fixes#21806
Change-Id: I013bb37b463fdfc229a7f324df8fe2da8d286f33
Reviewed-on: https://go-review.googlesource.com/62530
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
eqstring is only called for strings with equal lengths.
Instead of pushing a pointer and length for each argument string
on the stack we can omit pushing one of the lengths on the stack.
Changing eqstrings signature to eqstring(*uint8, *uint8, int) bool
to implement the above optimization would make it very similar to the
existing memequal(*any, *any, uintptr) bool function.
Since string lengths are positive we can avoid code redundancy and
use memequal instead of using eqstring with an optimized signature.
go command binary size reduced by 4128 bytes on amd64.
name old time/op new time/op delta
CompareStringEqual 6.03ns ± 1% 5.71ns ± 1% -5.23% (p=0.000 n=19+18)
CompareStringIdentical 2.88ns ± 1% 3.22ns ± 7% +11.86% (p=0.000 n=20+20)
CompareStringSameLength 4.31ns ± 1% 4.01ns ± 1% -7.17% (p=0.000 n=19+19)
CompareStringDifferentLength 0.29ns ± 2% 0.29ns ± 2% ~ (p=1.000 n=20+20)
CompareStringBigUnaligned 64.3µs ± 2% 64.1µs ± 3% ~ (p=0.164 n=20+19)
CompareStringBig 61.9µs ± 1% 61.6µs ± 2% -0.46% (p=0.033 n=20+19)
Change-Id: Ice15f3b937c981f0d3bc8479a9ea0d10658ac8df
Reviewed-on: https://go-review.googlesource.com/53650
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Although mincore is declared in stubs.go, mincore isn't used by any
OSes except linux. Move it to os_linux.go and clean up unused code.
Change-Id: I6cfb0fed85c0317a4d091a2722ac55fa79fc7c9a
Reviewed-on: https://go-review.googlesource.com/54910
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The only non test user of the assembler prefetch functions is the
heapBits.prefetch function which is itself unused.
The runtime prefetch functions have no functionality on most platforms
and are not inlineable since they are written in assembler. The function
call overhead eliminates the performance gains that could be achieved with
prefetching and would degrade performance for platforms where the functions
are no-ops.
If prefetch functions are needed back again later they can be improved
by avoiding the function call overhead and implementing them as intrinsics.
Change-Id: I52c553cf3607ffe09f0441c6e7a0a818cb21117d
Reviewed-on: https://go-review.googlesource.com/44370
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Currently, we mix objects with pointers and objects without pointers
("noscan" objects) together in memory. As a result, for every object
we grey, we have to check that object's heap bits to find out if it's
noscan, which adds to the per-object cost of GC. This also hurts the
TLB footprint of the garbage collector because it decreases the
density of scannable objects at the page level.
This commit improves the situation by using separate spans for noscan
objects. This will allow a much simpler noscan check (in a follow up
CL), eliminate the need to clear the bitmap of noscan objects (in a
follow up CL), and improves TLB footprint by increasing the density of
scannable objects.
This is also a step toward eliminating dead bits, since the current
noscan check depends on checking the dead bit of the first word.
This has no effect on the heap size of the garbage benchmark.
We'll measure the performance change of this after the follow-up
optimizations.
This is a cherry-pick from dev.garbage commit d491e550c3. The only
non-trivial merge conflict was in updatememstats in mstats.go, where
we now have to separate the per-spanclass stats from the per-sizeclass
stats.
Change-Id: I13bdc4869538ece5649a8d2a41c6605371618e40
Reviewed-on: https://go-review.googlesource.com/41251
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
This commit reworks multiway select statements to use normal control
flow primitives instead of the previous setjmp/longjmp-like behavior.
This simplifies liveness analysis and should prevent issues around
"returns twice" function calls within SSA passes.
test/live.go is updated because liveness analysis's CFG is more
representative of actual control flow. The case bodies are the only
real successors of the selectgo call, but previously the selectsend,
selectrecv, etc. calls were included in the successors list too.
Updates #19331.
Change-Id: I7f879b103a4b85e62fc36a270d812f54c0aa3e83
Reviewed-on: https://go-review.googlesource.com/37661
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Now that we don't rescan stacks, stack barriers are unnecessary. This
removes all of the code and structures supporting them as well as
tests that were specifically for stack barriers.
Updates #17503.
Change-Id: Ia29221730e0f2bbe7beab4fa757f31a032d9690c
Reviewed-on: https://go-review.googlesource.com/36620
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Extend period of fastrand from (1<<31)-1 to (1<<32)-1 by
choosing other polynom and reacting on high bit before shift.
Polynomial is taken at https://users.ece.cmu.edu/~koopman/lfsr/index.html
from 32.dat.gz . It is referred as F7711115 cause this list of
polynomials is for LFSR with shift to right (and fastrand uses shift to
left). (old polynomial is referred in 31.dat.gz as 7BB88888).
There were couple of places with conversation of fastrand to int, which
leads to negative values on 32bit platforms. They are fixed.
Change-Id: Ibee518a3f9103e0aea220ada494b3aec77babb72
Reviewed-on: https://go-review.googlesource.com/36875
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
So it could be inlined.
Using bit-tricks it could be implemented without condition
(improved trick version by Minux Ma).
Simple benchmark shows it is faster on i386 and x86_64, though
I don't know will it be faster on other architectures?
benchmark old ns/op new ns/op delta
BenchmarkFastrand-3 2.79 1.48 -46.95%
BenchmarkFastrandHashiter-3 25.9 24.9 -3.86%
Change-Id: Ie2eb6d0f598c0bb5fac7f6ad0f8b5e3eddaa361b
Reviewed-on: https://go-review.googlesource.com/34782
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Fetch both monotonic and wall time together when possible.
Avoids skew and is cheaper.
Also shave a few ns off in conversion in package time.
Compared to current implementation (after monotonic changes):
name old time/op new time/op delta
Now 19.6ns ± 1% 9.7ns ± 1% -50.63% (p=0.000 n=41+49) darwin/amd64
Now 23.5ns ± 4% 10.6ns ± 5% -54.61% (p=0.000 n=30+28) windows/amd64
Now 54.5ns ± 5% 29.8ns ± 9% -45.40% (p=0.000 n=27+29) windows/386
More importantly, compared to Go 1.8:
name old time/op new time/op delta
Now 9.5ns ± 1% 9.7ns ± 1% +1.94% (p=0.000 n=41+49) darwin/amd64
Now 12.9ns ± 5% 10.6ns ± 5% -17.73% (p=0.000 n=30+28) windows/amd64
Now 15.3ns ± 5% 29.8ns ± 9% +94.36% (p=0.000 n=30+29) windows/386
This brings time.Now back in line with Go 1.8 on darwin/amd64 and windows/amd64.
It's not obvious why windows/386 is still noticeably worse than Go 1.8,
but it's better than before this CL. The windows/386 speed is not too
important; the changes just keep the two architectures similar.
Change-Id: If69b94970c8a1a57910a371ee91e0d4e82e46c5d
Reviewed-on: https://go-review.googlesource.com/36428
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Since barrier-less memclr is only safe in very narrow circumstances,
this commit renames memclr to avoid accidentally calling memclr on
typed memory. This can cause subtle, non-deterministic bugs, so it's
worth some effort to prevent. In the near term, this will also prevent
bugs creeping in from any concurrent CLs that add calls to memclr; if
this happens, whichever patch hits master second will fail to compile.
This also adds the other new memclr variants to the compiler's
builtin.go to minimize the churn on that binary blob. We'll use these
in future commits.
Updates #17503.
Change-Id: I00eead049f5bd35ca107ea525966831f3d1ed9ca
Reviewed-on: https://go-review.googlesource.com/31369
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
The hybrid barrier requires distinguishing typed and untyped memory
even when zeroing because the *current* contents of the memory matters
even when overwriting.
This commit introduces runtime.typedmemclr and runtime.memclrHasPointers
as a typed memory clearing functions parallel to runtime.typedmemmove.
Currently these simply call memclr, but with the hybrid barrier we'll
need to shade any pointers we're overwriting. These will provide us
with the necessary hooks to do so.
Updates #17503.
Change-Id: I74478619f8907825898092aaa204d6e4690f27e6
Reviewed-on: https://go-review.googlesource.com/31366
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
This commit fixes two bizarrely related bugs:
1. The signatures for the call* functions were wrong, indicating that
they had only two pointer arguments instead of three. We didn't notice
because the call* functions are defined by a macro expansion, which go
vet doesn't see.
2. deferArgs on a defer object with a zero-sized frame returned a
pointer just past the end of the allocated object, which is illegal in
Go (and can cause the "sweep increased allocation count" crashes).
In a fascinating twist, these two bugs canceled each other out, which
is why I'm fixing them together. The pointer returned by deferArgs is
used in only two ways: as an argument to memmove and as an argument to
reflectcall. memmove is NOSPLIT, so the argument was unobservable.
reflectcall immediately tail calls one of the call* functions, which
are not NOSPLIT, but the deferArgs pointer just happened to be the
third argument that was accidentally marked as a scalar. Hence, when
the garbage collector scanned the stack, it didn't see the bad
pointer as a pointer.
I believe this was all ultimately benign. In principle, stack growth
during the reflectcall could fail to update the args pointer, but it
never points to the stack, so it never needs to be updated. Also in
principle, the garbage collector could fail to mark the args object
because of the incorrect call* signatures, but in all calls to
reflectcall (including the ones spelled "call" in the reflect package)
the args object is kept live by the calling stack.
Change-Id: Ic932c79d5f4382be23118fdd9dba9688e9169e28
Reviewed-on: https://go-review.googlesource.com/31654
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
This makes it possible to inline getcallersp. getcallersp is on the
hot path of defers, so this slightly speeds up defer:
name old time/op new time/op delta
Defer-4 78.3ns ± 2% 75.1ns ± 1% -4.00% (p=0.000 n=9+8)
Updates #14939.
Change-Id: Icc1cc4cd2f0a81fc4c8344432d0b2e783accacdd
Reviewed-on: https://go-review.googlesource.com/29655
TryBot-Result: Gobot Gobot <gobot@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
noescape is now 0 instructions with the SSA backend.
fast atomics are no longer a TODO (at least for amd64).
Change-Id: Ib6e06f7471bef282a47ba236d8ce95404bb60a42
Reviewed-on: https://go-review.googlesource.com/28087
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Add missing function prototypes.
Fix function prototypes.
Use FP references instead of SP references.
Fix variable names.
Update comments.
Clean up whitespace. (Not for vet.)
All fairly minor fixes to make vet happy.
Updates #11041
Change-Id: Ifab2cdf235ff61cdc226ab1d84b8467b5ac9446c
Reviewed-on: https://go-review.googlesource.com/27713
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The cgocallback function picked up a ctxt parameter in CL 22508.
That CL updated the assembler implementation, but there are a few
mentions in Go code that were not updated. This CL fixes that.
Fixes#16326
Change-Id: I5f68e23565c6a0b11057aff476d13990bff54a66
Reviewed-on: https://go-review.googlesource.com/24848
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
They do the same thing, except memequal also has the short-circuit
check if the two pointers are equal.
A) We might as well always do the short-circuit check, it is only 2 instructions.
B) The extra function call (memequal->memeq) is expensive.
benchmark old ns/op new ns/op delta
BenchmarkArrayEqual-8 8.56 5.31 -37.97%
No noticeable affect on the former memeq user (maps).
Fixes#14302
Change-Id: I85d1ada59ed11e64dd6c54667f79d32cc5f81948
Reviewed-on: https://go-review.googlesource.com/19843
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
We need a runtime check because the original issue is encountered
when running cross compiled windows program from linux. It's better
to give a meaningful crash message earlier than to segfault later.
The added test should not impose any measurable overhead to Go
programs.
For #12415.
Change-Id: Ib4a24ef560c09c0585b351d62eefd157b6b7f04c
Reviewed-on: https://go-review.googlesource.com/14207
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
runtime/internal/sys will hold system-, architecture- and config-
specific constants.
Updates #11647
Change-Id: I6db29c312556087a42e8d2bdd9af40d157c56b54
Reviewed-on: https://go-review.googlesource.com/16817
Reviewed-by: Russ Cox <rsc@golang.org>
This change breaks out most of the atomics functions in the runtime
into package runtime/internal/atomic. It adds some basic support
in the toolchain for runtime packages, and also modifies linux/arm
atomics to remove the dependency on the runtime's mutex. The mutexes
have been replaced with spinlocks.
all trybots are happy!
In addition to the trybots, I've tested on the darwin/arm64 builder,
on the darwin/arm builder, and on a ppc64le machine.
Change-Id: I6698c8e3cf3834f55ce5824059f44d00dc8e3c2f
Reviewed-on: https://go-review.googlesource.com/14204
Run-TryBot: Michael Matloob <matloob@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Currently, runtime.Goexit() calls goexit()—the goroutine exit stub—to
terminate the goroutine. This *mostly* works, but can cause a
"leftover stack barriers" panic if the following happens:
1. Goroutine A has a reasonably large stack.
2. The garbage collector scan phase runs and installs stack barriers
in A's stack. The top-most stack barrier happens to fall at address X.
3. Goroutine A unwinds the stack far enough to be a candidate for
stack shrinking, but not past X.
4. Goroutine A calls runtime.Goexit(), which calls goexit(), which
calls goexit1().
5. The garbage collector enters mark termination.
6. Goroutine A is preempted right at the prologue of goexit1() and
performs a stack shrink, which calls gentraceback.
gentraceback stops as soon as it sees goexit on the stack, which is
only two frames up at this point, even though there may really be many
frames above it. More to the point, the stack barrier at X is above
the goexit frame, so gentraceback never sees that stack barrier. At
the end of gentraceback, it checks that it saw all of the stack
barriers and panics because it didn't see the one at X.
The fix is simple: call goexit1, which actually implements the process
of exiting a goroutine, rather than goexit, the exit stub.
To make sure this doesn't happen again in the future, we also add an
argument to the stub prototype of goexit so you really, really have to
want to call it in order to call it. We were able to reliably
reproduce the above sequence with a fair amount of awful code inserted
at the right places in the runtime, but chose to change the goexit
prototype to ensure this wouldn't happen again rather than pollute the
runtime with ugly testing code.
Change-Id: Ifb6fb53087e09a252baddadc36eebf954468f2a8
Reviewed-on: https://go-review.googlesource.com/13323
Reviewed-by: Russ Cox <rsc@golang.org>
Otherwise the GC may see uninitialized memory there,
which might be old pointers that are retained, or it might
trigger the invalid pointer check.
Fixes#11907.
Change-Id: I67e306384a68468eef45da1a8eb5c9df216a77c0
Reviewed-on: https://go-review.googlesource.com/12852
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
This is mostly Russ's https://golang.org/cl/12145 but with some extra fixes to
account for the fact that function declarations without implementations now
break shared libraries, and including my test case.
Fixes#11480.
Change-Id: Iabdc2934a0378e5025e4e7affadb535eaef2c8f1
Reviewed-on: https://go-review.googlesource.com/12340
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Currently its possible for the garbage collector to observe
uninitialized memory or stale heap bitmap bits on weakly ordered
architectures such as ARM and PPC. On such architectures, the stores
that zero newly allocated memory and initialize its heap bitmap may
move after a store in user code that makes the allocated object
observable by the garbage collector.
To fix this, add a "publication barrier" (also known as an "export
barrier") before returning from mallocgc. This is a store/store
barrier that ensures any write done by user code that makes the
returned object observable to the garbage collector will be ordered
after the initialization performed by mallocgc. No barrier is
necessary on the reading side because of the data dependency between
loading the pointer and loading the contents of the object.
Fixes one of the issues raised in #9984.
Change-Id: Ia3d96ad9c5fc7f4d342f5e05ec0ceae700cd17c8
Reviewed-on: https://go-review.googlesource.com/11083
Reviewed-by: Rick Hudson <rlh@golang.org>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Martin Capitanio <capnm9@gmail.com>
Reviewed-by: Russ Cox <rsc@golang.org>
In preparation for rename of cgocall_errno into cgocall and
asmcgocall_errno into asmcgocall in the fllowinng CL.
rsc requested CL 9387 to be split into two parts. This is first part.
Change-Id: I7434f0e4b44dd37017540695834bfcb1eebf0b2f
Reviewed-on: https://go-review.googlesource.com/11166
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This commit implements stack barriers to minimize the amount of
stack re-scanning that must be done during mark termination.
Currently the GC scans stacks of active goroutines twice during every
GC cycle: once at the beginning during root discovery and once at the
end during mark termination. The second scan happens while the world
is stopped and guarantees that we've seen all of the roots (since
there are no write barriers on writes to local stack
variables). However, this means pause time is proportional to stack
size. In particularly recursive programs, this can drive pause time up
past our 10ms goal (e.g., it takes about 150ms to scan a 50MB heap).
Re-scanning the entire stack is rarely necessary, especially for large
stacks, because usually most of the frames on the stack were not
active between the first and second scans and hence any changes to
these frames (via non-escaping pointers passed down the stack) were
tracked by write barriers.
To efficiently track how far a stack has been unwound since the first
scan (and, hence, how much needs to be re-scanned), this commit
introduces stack barriers. During the first scan, at exponentially
spaced points in each stack, the scan overwrites return PCs with the
PC of the stack barrier function. When "returned" to, the stack
barrier function records how far the stack has unwound and jumps to
the original return PC for that point in the stack. Then the second
scan only needs to proceed as far as the lowest barrier that hasn't
been hit.
For deeply recursive programs, this substantially reduces mark
termination time (and hence pause time). For the goscheme example
linked in issue #10898, prior to this change, mark termination times
were typically between 100 and 500ms; with this change, mark
termination times are typically between 10 and 20ms. As a result of
the reduced stack scanning work, this reduces overall execution time
of the goscheme example by 20%.
Fixes#10898.
The effect of this on programs that are not deeply recursive is
minimal:
name old time/op new time/op delta
BinaryTree17 3.16s ± 2% 3.26s ± 1% +3.31% (p=0.000 n=19+19)
Fannkuch11 2.42s ± 1% 2.48s ± 1% +2.24% (p=0.000 n=17+19)
FmtFprintfEmpty 50.0ns ± 3% 49.8ns ± 1% ~ (p=0.534 n=20+19)
FmtFprintfString 173ns ± 0% 175ns ± 0% +1.49% (p=0.000 n=16+19)
FmtFprintfInt 170ns ± 1% 175ns ± 1% +2.97% (p=0.000 n=20+19)
FmtFprintfIntInt 288ns ± 0% 295ns ± 0% +2.73% (p=0.000 n=16+19)
FmtFprintfPrefixedInt 242ns ± 1% 252ns ± 1% +4.13% (p=0.000 n=18+18)
FmtFprintfFloat 324ns ± 0% 323ns ± 0% -0.36% (p=0.000 n=20+19)
FmtManyArgs 1.14µs ± 0% 1.12µs ± 1% -1.01% (p=0.000 n=18+19)
GobDecode 8.88ms ± 1% 8.87ms ± 0% ~ (p=0.480 n=19+18)
GobEncode 6.80ms ± 1% 6.85ms ± 0% +0.82% (p=0.000 n=20+18)
Gzip 363ms ± 1% 363ms ± 1% ~ (p=0.077 n=18+20)
Gunzip 90.6ms ± 0% 90.0ms ± 1% -0.71% (p=0.000 n=17+18)
HTTPClientServer 51.5µs ± 1% 50.8µs ± 1% -1.32% (p=0.000 n=18+18)
JSONEncode 17.0ms ± 0% 17.1ms ± 0% +0.40% (p=0.000 n=18+17)
JSONDecode 61.8ms ± 0% 63.8ms ± 1% +3.11% (p=0.000 n=18+17)
Mandelbrot200 3.84ms ± 0% 3.84ms ± 1% ~ (p=0.583 n=19+19)
GoParse 3.71ms ± 1% 3.72ms ± 1% ~ (p=0.159 n=18+19)
RegexpMatchEasy0_32 100ns ± 0% 100ns ± 1% -0.19% (p=0.033 n=17+19)
RegexpMatchEasy0_1K 342ns ± 1% 331ns ± 0% -3.41% (p=0.000 n=19+19)
RegexpMatchEasy1_32 82.5ns ± 0% 81.7ns ± 0% -0.98% (p=0.000 n=18+18)
RegexpMatchEasy1_1K 505ns ± 0% 494ns ± 1% -2.16% (p=0.000 n=18+18)
RegexpMatchMedium_32 137ns ± 1% 137ns ± 1% -0.24% (p=0.048 n=20+18)
RegexpMatchMedium_1K 41.6µs ± 0% 41.3µs ± 1% -0.57% (p=0.004 n=18+20)
RegexpMatchHard_32 2.11µs ± 0% 2.11µs ± 1% +0.20% (p=0.037 n=17+19)
RegexpMatchHard_1K 63.9µs ± 2% 63.3µs ± 0% -0.99% (p=0.000 n=20+17)
Revcomp 560ms ± 1% 522ms ± 0% -6.87% (p=0.000 n=18+16)
Template 75.0ms ± 0% 75.1ms ± 1% +0.18% (p=0.013 n=18+19)
TimeParse 358ns ± 1% 364ns ± 0% +1.74% (p=0.000 n=20+15)
TimeFormat 360ns ± 0% 372ns ± 0% +3.55% (p=0.000 n=20+18)
Change-Id: If8a9bfae6c128d15a4f405e02bcfa50129df82a2
Reviewed-on: https://go-review.googlesource.com/10314
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
These currently use portable implementations in terms of their uint64
counterparts.
Change-Id: Icba5f7134cfcf9d0429edabcdd73091d97e5e905
Reviewed-on: https://go-review.googlesource.com/8831
Reviewed-by: Rick Hudson <rlh@golang.org>
readyExecute passes a closure to mcall that captures an argument to
readyExecute. Since mcall is marked noescape, this closure lives on
the stack of the calling goroutine. However, the closure puts the
calling goroutine on the run queue (and switches to a new
goroutine). If the calling goroutine gets scheduled before the mcall
returns, this stack-allocated closure will become invalid while it's
still executing. One consequence of this we've observed is that the
captured gp variable can get overwritten before the call to
execute(gp), causing execute(gp) to segfault.
Fix this by passing the currently captured gp variable through a field
in the calling goroutine's g struct so that the func is no longer a
closure.
To prevent problems like this in the future, this change also removes
the go:noescape annotation from mcall. Due to a compiler bug, this
will currently cause a func closure passed to mcall to be implicitly
allocated rather than refusing the implicit allocation. However, this
is okay because there are no other closures passed to mcall right now
and the compiler bug will be fixed shortly.
Fixes#10428.
Change-Id: I49b48b85de5643323b89e9eaa4df63854e968c32
Reviewed-on: https://go-review.googlesource.com/8866
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Move code from malloc1.go, malloc2.go, mem.go, mgc0.go into
appropriate locations.
Factor mgc.go into mgc.go, mgcmark.go, mgcsweep.go, mstats.go.
A lot of this code was in certain files because the right place was in
a C file but it was written in Go, or vice versa. This is one step toward
making things actually well-organized again.
Change-Id: I6741deb88a7cfb1c17ffe0bcca3989e10207968f
Reviewed-on: https://go-review.googlesource.com/5300
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
Call frame allocations can account for significant portion
of all allocations in a program, if call is executed
in an inner loop (e.g. to process every line in a log).
On the other hand, the allocation is easy to remove
using sync.Pool since the allocation is strictly scoped.
benchmark old ns/op new ns/op delta
BenchmarkCall 634 338 -46.69%
BenchmarkCall-4 496 167 -66.33%
benchmark old allocs new allocs delta
BenchmarkCall 1 0 -100.00%
BenchmarkCall-4 1 0 -100.00%
Update #7818
Change-Id: Icf60cce0a9be82e6171f0c0bd80dee2393db54a7
Reviewed-on: https://go-review.googlesource.com/1954
Reviewed-by: Keith Randall <khr@golang.org>
printf, vprintf, snprintf, gc_m_ptr, gc_g_ptr, gc_itab_ptr, gc_unixnanotime.
These were called from C.
There is no more C.
Now that vprintf is gone, delete roundup, which is unsafe (see CL 2814).
Change-Id: If8a7b727d497ffa13165c0d3a1ed62abc18f008c
Reviewed-on: https://go-review.googlesource.com/2824
Reviewed-by: Austin Clements <austin@google.com>
Use typedmemmove, typedslicecopy, and adjust reflect.call
to execute the necessary write barriers.
Found with GODEBUG=wbshadow=2 mode.
Eventually that will run automatically, but right now
it still detects other missing write barriers.
Change-Id: Iec5b5b0c1be5589295e28e5228e37f1a92e07742
Reviewed-on: https://go-review.googlesource.com/2312
Reviewed-by: Keith Randall <khr@golang.org>
Add write barrier to atomic operations manipulating pointers.
In general an atomic write of a pointer word may indicate racy accesses,
so there is no strictly safe way to attempt to keep the shadow copy
in sync with the real one. Instead, mark the shadow copy as not used.
Redirect sync/atomic pointer routines back to the runtime ones,
so that there is only one copy of the write barrier and shadow logic.
In time we might consider doing this for most of the sync/atomic
functions, but for now only the pointer routines need that treatment.
Found with GODEBUG=wbshadow=1 mode.
Eventually that will run automatically, but right now
it still detects other missing write barriers.
Change-Id: I852936b9a111a6cb9079cfaf6bd78b43016c0242
Reviewed-on: https://go-review.googlesource.com/2066
Reviewed-by: Rick Hudson <rlh@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Rename "gothrow" to "throw" now that the C version of "throw"
is no longer needed.
This change is purely mechanical except in panic.go where the
old version of "throw" has been deleted.
sed -i "" 's/[[:<:]]gothrow[[:>:]]/throw/g' runtime/*.go
Change-Id: Icf0752299c35958b92870a97111c67bcd9159dc3
Reviewed-on: https://go-review.googlesource.com/2150
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
Replace with uses of //go:linkname in Go files, direct use of name in .s files.
The only one that really truly needs a jump is reflect.call; the jump is now
next to the runtime.reflectcall assembly implementations.
Change-Id: Ie7ff3020a8f60a8e4c8645fe236e7883a3f23f46
Reviewed-on: https://go-review.googlesource.com/1962
Reviewed-by: Austin Clements <austin@google.com>
