| Age | Commit message (Collapse) | Author |
|---|
|
When -clobberdead compiler flag is set, the compiler inserts
instructions that set dead slots a specific value. If the GC sees
this value as a live pointer, something is probably wrong. Crash.
Only do this on AMD64 for now, as it is the only platform where
compiler's clobberdead mode is implemented. And on AMD64 the
clobberdead address can never be a valid address.
Change-Id: Ica687b132b5d3ba2a062500d13264fa730405d11
Reviewed-on: https://go-review.googlesource.com/c/go/+/310330
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
It's both simpler and faster to just unconditionally do two 32-bit
multiplies rather than a bunch of branching to try to avoid them.
This is safe thanks to the tight bounds derived in [1] and verified
during mksizeclasses.go.
Benchstat results below for compilebench benchmarks on my P920. See
also [2] for micro benchmarks comparing the new functions against the
originals (as well as several basic attempts at optimizing them).
name old time/op new time/op delta
Template 295ms ± 3% 290ms ± 1% -1.95% (p=0.000 n=20+20)
Unicode 113ms ± 3% 110ms ± 2% -2.32% (p=0.000 n=21+17)
GoTypes 1.78s ± 1% 1.76s ± 1% -1.23% (p=0.000 n=21+20)
Compiler 119ms ± 2% 117ms ± 4% -1.53% (p=0.007 n=20+20)
SSA 14.3s ± 1% 13.8s ± 1% -3.12% (p=0.000 n=17+20)
Flate 173ms ± 2% 170ms ± 1% -1.64% (p=0.000 n=20+19)
GoParser 278ms ± 2% 273ms ± 2% -1.92% (p=0.000 n=20+19)
Reflect 686ms ± 3% 671ms ± 3% -2.18% (p=0.000 n=19+20)
Tar 255ms ± 2% 248ms ± 2% -2.90% (p=0.000 n=20+20)
XML 335ms ± 3% 327ms ± 2% -2.34% (p=0.000 n=20+20)
LinkCompiler 799ms ± 1% 799ms ± 1% ~ (p=0.925 n=20+20)
ExternalLinkCompiler 1.90s ± 1% 1.90s ± 0% ~ (p=0.327 n=20+20)
LinkWithoutDebugCompiler 385ms ± 1% 386ms ± 1% ~ (p=0.251 n=18+20)
[Geo mean] 512ms 504ms -1.61%
name old user-time/op new user-time/op delta
Template 286ms ± 4% 282ms ± 4% -1.42% (p=0.025 n=21+20)
Unicode 104ms ± 9% 102ms ±14% ~ (p=0.294 n=21+20)
GoTypes 1.75s ± 3% 1.72s ± 2% -1.36% (p=0.000 n=21+20)
Compiler 109ms ±11% 108ms ± 8% ~ (p=0.187 n=21+19)
SSA 14.0s ± 1% 13.5s ± 2% -3.25% (p=0.000 n=16+20)
Flate 166ms ± 4% 164ms ± 4% -1.34% (p=0.032 n=19+19)
GoParser 268ms ± 4% 263ms ± 4% -1.71% (p=0.011 n=18+20)
Reflect 666ms ± 3% 654ms ± 4% -1.77% (p=0.002 n=18+20)
Tar 245ms ± 5% 236ms ± 6% -3.34% (p=0.000 n=20+20)
XML 320ms ± 4% 314ms ± 3% -2.01% (p=0.001 n=19+18)
LinkCompiler 744ms ± 4% 747ms ± 3% ~ (p=0.627 n=20+19)
ExternalLinkCompiler 1.71s ± 3% 1.72s ± 2% ~ (p=0.149 n=20+20)
LinkWithoutDebugCompiler 345ms ± 6% 342ms ± 8% ~ (p=0.355 n=20+20)
[Geo mean] 484ms 477ms -1.50%
[1] Daniel Lemire, Owen Kaser, Nathan Kurz. 2019. "Faster Remainder by
Direct Computation: Applications to Compilers and Software Libraries."
https://arxiv.org/abs/1902.01961
[2] https://github.com/mdempsky/benchdivmagic
Change-Id: Ie4d214e7a908b0d979c878f2d404bd56bdf374f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/300994
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
This change modifies mheap's span allocation API to have each caller
declare a purpose, defined as a new enum called spanAllocType.
The purpose behind this change is two-fold:
1. Tight control over who gets to allocate heap memory is, generally
speaking, a good thing. Every codepath that allocates heap memory
places additional implicit restrictions on the allocator. A notable
example of a restriction is work bufs coming from heap memory: write
barriers are not allowed in allocation paths because then we could
have a situation where the allocator calls into the allocator.
2. Memory statistic updating is explicit. Instead of passing an opaque
pointer for statistic updating, which places restrictions on how that
statistic may be updated, we use the spanAllocType to determine which
statistic to update and how.
We also take this opportunity to group all the statistic updating code
together, which should make the accounting code a little easier to
follow.
Change-Id: Ic0b0898959ba2a776f67122f0e36c9d7d60e3085
Reviewed-on: https://go-review.googlesource.com/c/go/+/246970
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
|
|
This CL introduces a 24 byte allocation size class which
fits 3 pointers on 64 bit and 6 pointers on 32 bit architectures.
Notably this new size class fits a slice header on 64 bit
architectures exactly while previously a 32 byte size class
would have been used for allocating a slice header on the heap.
The main complexity added with this CL is that heapBitsSetType
needs to handle objects that aren't 16-byte aligned but contain
more than a single pointer on 64-bit architectures.
Due to having a non 16 byte aligned size class on 32 bit a
h.shift of 2 is now possible which means a heap bitmap byte might
only be partially written. Due to this already having been
possible on 64 bit before the heap bitmap code only needed
minor adjustments for 32 bit doublecheck code paths.
Note that this CL changes the slice capacity allocated by append
for slice growth to a target capacity of 17 to 24 bytes.
On 64 bit architectures the capacity of the slice returned by
append([]byte{}, make([]byte, 24)...)) is 32 bytes before and
24 bytes after this CL. Depending on allocation patterns of the
specific Go program this can increase the number of total
alloctions as subsequent appends to the slice can trigger slice
growth earlier than before. On the other side if the slice is
never appended to again above its capacity this will lower heap
usage by 8 bytes.
This CL changes the set of size classes reported in the
runtime.MemStats.BySize array due to it being limited to a
total of 61 size classes. The new 24 byte size class is now
included and the 20480 byte size class is not included anymore.
Fixes #8885
name old time/op new time/op delta
Template 196ms ± 3% 194ms ± 2% ~ (p=0.247 n=10+10)
Unicode 85.6ms ±16% 88.1ms ± 1% ~ (p=0.165 n=10+10)
GoTypes 673ms ± 2% 668ms ± 2% ~ (p=0.258 n=9+9)
Compiler 3.14s ± 6% 3.08s ± 1% ~ (p=0.243 n=10+9)
SSA 6.82s ± 1% 6.76s ± 1% -0.87% (p=0.006 n=9+10)
Flate 128ms ± 7% 127ms ± 3% ~ (p=0.739 n=10+10)
GoParser 154ms ± 3% 153ms ± 4% ~ (p=0.730 n=9+9)
Reflect 404ms ± 1% 412ms ± 4% +1.99% (p=0.022 n=9+10)
Tar 172ms ± 4% 170ms ± 4% ~ (p=0.065 n=10+9)
XML 231ms ± 4% 230ms ± 3% ~ (p=0.912 n=10+10)
LinkCompiler 341ms ± 1% 339ms ± 1% ~ (p=0.243 n=9+10)
ExternalLinkCompiler 1.72s ± 1% 1.72s ± 1% ~ (p=0.661 n=9+10)
LinkWithoutDebugCompiler 221ms ± 2% 221ms ± 2% ~ (p=0.529 n=10+10)
StdCmd 18.4s ± 3% 18.2s ± 1% ~ (p=0.515 n=10+8)
name old user-time/op new user-time/op delta
Template 238ms ± 4% 243ms ± 6% ~ (p=0.661 n=9+10)
Unicode 116ms ± 6% 113ms ± 3% -3.37% (p=0.035 n=9+10)
GoTypes 854ms ± 2% 848ms ± 2% ~ (p=0.604 n=9+10)
Compiler 4.10s ± 1% 4.11s ± 1% ~ (p=0.481 n=8+9)
SSA 9.49s ± 1% 9.41s ± 1% -0.92% (p=0.001 n=9+10)
Flate 149ms ± 6% 151ms ± 7% ~ (p=0.481 n=10+10)
GoParser 189ms ± 2% 190ms ± 2% ~ (p=0.497 n=9+10)
Reflect 511ms ± 2% 508ms ± 2% ~ (p=0.211 n=9+10)
Tar 215ms ± 4% 212ms ± 3% ~ (p=0.105 n=10+10)
XML 288ms ± 2% 288ms ± 2% ~ (p=0.971 n=10+10)
LinkCompiler 559ms ± 4% 557ms ± 1% ~ (p=0.968 n=9+10)
ExternalLinkCompiler 1.78s ± 1% 1.77s ± 1% ~ (p=0.055 n=8+10)
LinkWithoutDebugCompiler 245ms ± 3% 245ms ± 2% ~ (p=0.684 n=10+10)
name old alloc/op new alloc/op delta
Template 34.8MB ± 0% 34.4MB ± 0% -0.95% (p=0.000 n=9+10)
Unicode 28.6MB ± 0% 28.3MB ± 0% -0.95% (p=0.000 n=10+10)
GoTypes 115MB ± 0% 114MB ± 0% -1.02% (p=0.000 n=10+9)
Compiler 554MB ± 0% 549MB ± 0% -0.86% (p=0.000 n=9+10)
SSA 1.28GB ± 0% 1.27GB ± 0% -0.83% (p=0.000 n=10+10)
Flate 21.8MB ± 0% 21.6MB ± 0% -0.87% (p=0.000 n=8+10)
GoParser 26.7MB ± 0% 26.4MB ± 0% -0.97% (p=0.000 n=10+9)
Reflect 75.0MB ± 0% 74.1MB ± 0% -1.18% (p=0.000 n=10+10)
Tar 32.6MB ± 0% 32.3MB ± 0% -0.94% (p=0.000 n=10+7)
XML 41.5MB ± 0% 41.2MB ± 0% -0.90% (p=0.000 n=10+8)
LinkCompiler 105MB ± 0% 104MB ± 0% -0.94% (p=0.000 n=10+10)
ExternalLinkCompiler 153MB ± 0% 152MB ± 0% -0.69% (p=0.000 n=10+10)
LinkWithoutDebugCompiler 63.7MB ± 0% 63.6MB ± 0% -0.13% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Template 336k ± 0% 336k ± 0% +0.02% (p=0.002 n=10+10)
Unicode 332k ± 0% 332k ± 0% ~ (p=0.447 n=10+10)
GoTypes 1.16M ± 0% 1.16M ± 0% +0.01% (p=0.001 n=10+10)
Compiler 4.92M ± 0% 4.92M ± 0% +0.01% (p=0.000 n=10+10)
SSA 11.9M ± 0% 11.9M ± 0% +0.02% (p=0.000 n=9+10)
Flate 214k ± 0% 214k ± 0% +0.02% (p=0.032 n=10+8)
GoParser 270k ± 0% 270k ± 0% +0.02% (p=0.004 n=10+9)
Reflect 877k ± 0% 877k ± 0% +0.01% (p=0.000 n=10+10)
Tar 313k ± 0% 313k ± 0% ~ (p=0.075 n=9+10)
XML 387k ± 0% 387k ± 0% +0.02% (p=0.007 n=10+10)
LinkCompiler 455k ± 0% 456k ± 0% +0.08% (p=0.000 n=10+9)
ExternalLinkCompiler 670k ± 0% 671k ± 0% +0.06% (p=0.000 n=10+10)
LinkWithoutDebugCompiler 113k ± 0% 113k ± 0% ~ (p=0.149 n=10+10)
name old maxRSS/op new maxRSS/op delta
Template 34.1M ± 1% 34.1M ± 1% ~ (p=0.853 n=10+10)
Unicode 35.1M ± 1% 34.6M ± 1% -1.43% (p=0.000 n=10+10)
GoTypes 72.8M ± 3% 73.3M ± 2% ~ (p=0.724 n=10+10)
Compiler 288M ± 3% 295M ± 4% ~ (p=0.393 n=10+10)
SSA 630M ± 1% 622M ± 1% -1.18% (p=0.001 n=10+10)
Flate 26.0M ± 1% 26.2M ± 2% ~ (p=0.493 n=10+10)
GoParser 28.6M ± 1% 28.5M ± 2% ~ (p=0.256 n=10+10)
Reflect 55.5M ± 2% 55.4M ± 1% ~ (p=0.436 n=10+10)
Tar 33.0M ± 1% 32.8M ± 2% ~ (p=0.075 n=10+10)
XML 38.7M ± 1% 39.0M ± 1% ~ (p=0.053 n=9+10)
LinkCompiler 164M ± 1% 164M ± 1% -0.27% (p=0.029 n=10+10)
ExternalLinkCompiler 174M ± 0% 173M ± 0% -0.33% (p=0.002 n=9+10)
LinkWithoutDebugCompiler 137M ± 0% 136M ± 2% ~ (p=0.825 n=9+10)
Change-Id: I9ecf2a10024513abef8fbfbe519e44e0b29b6167
Reviewed-on: https://go-review.googlesource.com/c/go/+/242258
Trust: Martin Möhrmann <moehrmann@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Martin Möhrmann <martisch@uos.de>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
Currently, the GC stores the object marks for checkmarks mode in the
heap bitmap using a rather complex encoding: for one word objects, the
checkmark is stored in the pointer/scalar bit since one word objects
must be pointers; for larger objects, the checkmark is stored in what
would be the scan/dead bit for the second word of the object. This
encoding made more sense when the runtime used the first scan/dead bit
as the regular mark bit, but we moved away from that long ago.
This encoding and overloading of the heap bitmap bits causes a great
deal of complexity in many parts of the allocator and garbage
collector and leads to some subtle bugs like #15903.
This CL moves the checkmarks mark bits into their own per-arena bitmap
and reclaims the second scan/dead bit as a regular scan/dead bit.
I tested this by enabling doubleCheck mode in heapBitsSetType and
running in both regular and GODEBUG=gccheckmark=1 mode.
Fixes #15903.
No performance degradation. (Very slight improvement on a few
benchmarks, but it's probably just noise.)
name old time/op new time/op delta
BiogoIgor 16.6s ± 1% 16.4s ± 1% -0.94% (p=0.000 n=25+24)
BiogoKrishna 19.2s ± 3% 19.2s ± 3% ~ (p=0.638 n=23+25)
BleveIndexBatch100 6.12s ± 5% 6.17s ± 4% ~ (p=0.170 n=25+25)
CompileTemplate 206ms ± 1% 205ms ± 1% -0.43% (p=0.005 n=24+24)
CompileUnicode 82.2ms ± 2% 81.5ms ± 2% -0.95% (p=0.001 n=22+22)
CompileGoTypes 755ms ± 3% 754ms ± 4% ~ (p=0.715 n=25+25)
CompileCompiler 3.73s ± 1% 3.73s ± 1% ~ (p=0.445 n=25+24)
CompileSSA 8.67s ± 1% 8.66s ± 1% ~ (p=0.836 n=24+22)
CompileFlate 134ms ± 2% 133ms ± 1% -0.66% (p=0.001 n=24+23)
CompileGoParser 164ms ± 1% 163ms ± 1% -0.85% (p=0.000 n=24+24)
CompileReflect 466ms ± 5% 466ms ± 3% ~ (p=0.863 n=25+25)
CompileTar 182ms ± 1% 182ms ± 1% -0.31% (p=0.048 n=24+24)
CompileXML 249ms ± 1% 248ms ± 1% -0.32% (p=0.031 n=21+25)
CompileStdCmd 10.3s ± 1% 10.3s ± 1% ~ (p=0.459 n=23+23)
FoglemanFauxGLRenderRotateBoat 8.66s ± 1% 8.62s ± 1% -0.47% (p=0.000 n=23+24)
FoglemanPathTraceRenderGopherIter1 20.3s ± 3% 20.2s ± 2% ~ (p=0.893 n=25+25)
GopherLuaKNucleotide 29.7s ± 1% 29.8s ± 2% ~ (p=0.421 n=24+25)
MarkdownRenderXHTML 246ms ± 1% 247ms ± 1% ~ (p=0.558 n=25+24)
Tile38WithinCircle100kmRequest 779µs ± 4% 779µs ± 3% ~ (p=0.954 n=25+25)
Tile38IntersectsCircle100kmRequest 1.02ms ± 3% 1.01ms ± 4% ~ (p=0.658 n=25+25)
Tile38KNearestLimit100Request 984µs ± 4% 986µs ± 4% ~ (p=0.627 n=24+25)
[Geo mean] 552ms 551ms -0.19%
https://perf.golang.org/search?q=upload:20200723.6
Change-Id: Ic703f26a83fb034941dc6f4788fc997d56890dec
Reviewed-on: https://go-review.googlesource.com/c/go/+/244539
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
|
|
The heapBitsSetType function has a slow doubleCheck debugging mode
that checks the bitmap written out by the rest of the function using
far more obvious logic. But even this has some surprisingly complex
logic in it. Simplify it a bit. This also happens to fix the logic on
32-bit.
Fixes #40335.
Change-Id: I5cee482ad8adbd01cf5b98e35a270fe941ba4940
Reviewed-on: https://go-review.googlesource.com/c/go/+/244538
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
This change modifies the implementation of (*mspan).countAlloc by
using OnesCount64 (which on many systems is intrinsified). It does so by
using an unsafe pointer cast, but in this case we don't care about
endianness because we're just counting bits set.
This change means we no longer need the popcnt table which was redundant
in the runtime anyway. We can also simplify the logic here significantly
by observing that mark bits allocations are always 8-byte aligned, so we
don't need to handle any edge-cases due to the fact that OnesCount64
operates on 64 bits at a time: all irrelevant bits will be zero.
Overall, this implementation is significantly faster than the old one on
amd64, and should be similarly faster (or better!) on other systems
which support the intrinsic. On systems which do not, it should be
roughly the same performance because OnesCount64 is implemented using a
table in the general case.
Results on linux/amd64:
name old time/op new time/op delta
MSpanCountAlloc/bits=64-4 16.8ns ± 0% 12.7ns ± 0% -24.40% (p=0.000 n=5+4)
MSpanCountAlloc/bits=128-4 23.5ns ± 0% 12.8ns ± 0% -45.70% (p=0.000 n=4+5)
MSpanCountAlloc/bits=256-4 43.5ns ± 0% 12.8ns ± 0% -70.67% (p=0.000 n=4+5)
MSpanCountAlloc/bits=512-4 59.5ns ± 0% 15.4ns ± 0% -74.12% (p=0.008 n=5+5)
MSpanCountAlloc/bits=1024-4 116ns ± 1% 23ns ± 0% -79.84% (p=0.000 n=5+4)
Change-Id: Id4c994be22224653af5333683a69b0937130ed04
Reviewed-on: https://go-review.googlesource.com/c/go/+/216558
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
materializeGCProg allocates a temporary buffer for unrolling a GC
program. Unfortunately, when computing the size of the buffer, it
rounds *down* the number of bytes needed to store bitmap before
rounding up the number of pages needed to store those bytes. The fact
that it rounds up to pages usually mitigates the rounding down, but
the type from #37470 exists right on the boundary where this doesn't
work:
type Sequencer struct {
htable [1 << 17]uint32
buf []byte
}
On 64-bit, this GC bitmap is exactly 8 KiB of zeros, followed by three
one bits. Hence, this needs 8193 bytes of storage, but the current
math in materializeGCProg rounds *down* the three one bits to 8192
bytes. Since this is exactly pageSize, the next step of rounding up to
the page size doesn't mitigate this error, and materializeGCProg
allocates a buffer that is one byte too small. runGCProg then writes
one byte past the end of this buffer, causing either a segfault (if
you're lucky!) or memory corruption.
Fixes #37470.
Change-Id: Iad24c463c501cd9b1dc1924bc2ad007991a094a0
Reviewed-on: https://go-review.googlesource.com/c/go/+/221197
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
When everything is working correctly, any pointer the garbage
collector encounters can only point into a fully initialized heap
span, since the span must have been initialized before that pointer
could escape the heap allocator and become visible to the GC.
However, in various cases, we try to be defensive against bad
pointers. In findObject, this is just a sanity check: we never expect
to find a bad pointer, but programming errors can lead to them. In
spanOfHeap, we don't necessarily trust the pointer and we're trying to
check if it really does point to the heap, though it should always
point to something. Conservative scanning takes this to a new level,
since it can only guess that a word may be a pointer and verify this.
In all of these cases, we have a problem that the span lookup and
check can race with span initialization, since the span becomes
visible to lookups before it's fully initialized.
Furthermore, we're about to start initializing the span without the
heap lock held, which is going to introduce races where accesses were
previously protected by the heap lock.
To address this, this CL makes accesses to mspan.state atomic, and
ensures that the span is fully initialized before setting the state to
mSpanInUse. All loads are now atomic, and in any case where we don't
trust the pointer, it first atomically loads the span state and checks
that it's mSpanInUse, after which it will have synchronized with span
initialization and can safely check the other span fields.
For #10958, #24543, but a good fix in general.
Change-Id: I518b7c63555b02064b98aa5f802c92b758fef853
Reviewed-on: https://go-review.googlesource.com/c/go/+/203286
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
Currently, several important fields of a heap span are set by
heapBits.initSpan, which happens after the span has already been
published and returned from the locked region of alloc_m. In
particular, allocBits is set very late, which makes mspan.isFree
unsafe even if you were to lock the heap because it tries to access
allocBits.
This CL fixes this by populating these fields in alloc_m. The next CL
builds on this to only publish the span once it is fully initialized.
Together, they'll make it safe to check allocBits even if there is a
race with alloc_m.
For #10958, #24543, but a good fix in general.
Change-Id: I7fde90023af0f497e826b637efa4d19c32840c08
Reviewed-on: https://go-review.googlesource.com/c/go/+/203285
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
|
|
findObject takes the pointer argument as uintptr. If the pointer is to
the local stack and calling findObject happens to require the stack to
be reallocated, then spanOf is called for the old pointer.
Marking findObject as nosplit fixes the issue.
Fixes #35068
Change-Id: I029d36f9c23f91812f18f98839edf02e0ba4082e
Reviewed-on: https://go-review.googlesource.com/c/go/+/202798
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
|
|
This helps keeping findObject's frame small.
Updates #35068
Change-Id: I1b8c1fcc5831944c86f1a30ed2f2d867a5f2b242
Reviewed-on: https://go-review.googlesource.com/c/go/+/202797
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
|
|
Factor out case s == nil, make the code cleaner and easier to read.
Change-Id: I63f52e14351c0a0d20a611b1fe10fdc0d4947d96
Reviewed-on: https://go-review.googlesource.com/c/go/+/202498
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
Those print statements are not a good debug helpers
and only clutter the code.
Change-Id: Ifbf450a04e6fa538af68e6352c016728edb4119a
Reviewed-on: https://go-review.googlesource.com/c/go/+/160537
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
|
|
We already have the ptrdata field in a type, which encodes exactly
the same information that kindNoPointers does.
My problem with kindNoPointers is that it often leads to
double-negative code like:
t.kind & kindNoPointers != 0
Much clearer is:
t.ptrdata == 0
Update #27167
Change-Id: I92307d7f018a6bbe3daca4a4abb4225e359349b1
Reviewed-on: https://go-review.googlesource.com/c/go/+/169157
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
|
|
Follow-up for CL 147037 and after Brad noticed the "returns whether"
pattern during the review of CL 150621.
Go documentation style for boolean funcs is to say:
// Foo reports whether ...
func Foo() bool
(rather than "returns whether")
Created with:
$ perl -i -npe 's/returns whether/reports whether/' $(git grep -l "returns whether" | grep -v vendor)
Change-Id: I15fe9ff99180ad97750cd05a10eceafdb12dc0b4
Reviewed-on: https://go-review.googlesource.com/c/150918
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
|
|
Go documentation style for boolean funcs is to say:
// Foo reports whether ...
func Foo() bool
(rather than "returns true if")
This CL also replaces 4 uses of "iff" with the same "reports whether"
wording, which doesn't lose any meaning, and will prevent people from
sending typo fixes when they don't realize it's "if and only if". In
the past I think we've had the typo CLs updated to just say "reports
whether". So do them all at once.
(Inspired by the addition of another "returns true if" in CL 146938
in fd_plan9.go)
Created with:
$ perl -i -npe 's/returns true if/reports whether/' $(git grep -l "returns true iff" | grep -v vendor)
$ perl -i -npe 's/returns true if/reports whether/' $(git grep -l "returns true if" | grep -v vendor)
Change-Id: Ided502237f5ab0d25cb625dbab12529c361a8b9f
Reviewed-on: https://go-review.googlesource.com/c/147037
Reviewed-by: Ian Lance Taylor <iant@golang.org>
|
|
Since atomic.Or8 is now an intrinsic (and has been for some time),
markBits.setMarked is inlinable. Undo the manual inlining of it.
Change-Id: I8e37ccf0851ad1d3088d9c8ae0f6f0c439d7eb2d
Reviewed-on: https://go-review.googlesource.com/c/138659
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
|
|
The previous CL introduced stack objects. This CL removes the old
ambiguously live liveness analysis. After this CL we're relying
on stack objects exclusively.
Update a bunch of liveness tests to reflect the new world.
Fixes #22350
Change-Id: I739b26e015882231011ce6bc1a7f426049e59f31
Reviewed-on: https://go-review.googlesource.com/c/134156
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
|
|
Rework how the compiler+runtime handles stack-allocated variables
whose address is taken.
Direct references to such variables work as before. References through
pointers, however, use a new mechanism. The new mechanism is more
precise than the old "ambiguously live" mechanism. It computes liveness
at runtime based on the actual references among objects on the stack.
Each function records all of its address-taken objects in a FUNCDATA.
These are called "stack objects". The runtime then uses that
information while scanning a stack to find all of the stack objects on
a stack. It then does a mark phase on the stack objects, using all the
pointers found on the stack (and ancillary structures, like defer
records) as the root set. Only stack objects which are found to be
live during this mark phase will be scanned and thus retain any heap
objects they point to.
A subsequent CL will remove all the "ambiguously live" logic from
the compiler, so that the stack object tracing will be required.
For this CL, the stack tracing is all redundant with the current
ambiguously live logic.
Update #22350
Change-Id: Ide19f1f71a5b6ec8c4d54f8f66f0e9a98344772f
Reviewed-on: https://go-review.googlesource.com/c/134155
Reviewed-by: Austin Clements <austin@google.com>
|
|
We already aliased mSpanInUse to _MSpanInUse. The dual constants are
getting annoying, so fix all of these to use the mSpan* naming
convention.
This was done automatically with:
sed -i -re 's/_?MSpan(Dead|InUse|Manual|Free)/mSpan\1/g' *.go
plus deleting the existing definition of mSpanInUse.
Change-Id: I09979d9d491d06c10689cea625dc57faa9cc6767
Reviewed-on: https://go-review.googlesource.com/137875
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
|
|
The new slice created in growslice is cleared during malloc for
element types containing pointers and therefore can only contain
nil pointers. This change avoids executing write barriers for these
nil pointers by adding and using a special bulkBarrierPreWriteSrcOnly
function that does not enqueue pointers to slots in dst to the write
barrier buffer.
Change-Id: If9b18248bfeeb6a874b0132d19520adea593bfc4
Reviewed-on: https://go-review.googlesource.com/115996
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
When an object spans heap arenas, its bitmap is discontiguous, so
heapBitsSetType unrolls the bitmap into the object itself and then
copies it out to the real heap bitmap. Unfortunately, since this code
path is rare, it had two unnoticed bugs related to the head and tail
of the bitmap:
1. At the head of the object, we were using hbitp as the destination
bitmap pointer rather than h.bitp, but hbitp points into the
*temporary* bitmap space (that is, the object itself), so we were
failing to copy the partial bitmap byte at the head of an object.
2. The core copying loop copied all of the full bitmap bytes, but
always drove the remaining word count down to 0, even if there was a
partial bitmap byte for the tail of the object. As a result, we never
wrote partial bitmap bytes at the tail of an object.
I found these by enabling out-of-place unrolling all the time. To
improve our chances of detecting these sorts of bugs in the future,
this CL mimics this by enabling out-of-place mode 50% of the time when
doubleCheck is enabled so that we test both in-place and out-of-place
mode.
Change-Id: I69e5d829fb3444be4cf11f4c6d8462c26dc467e8
Reviewed-on: https://go-review.googlesource.com/110995
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
Currently we have two nearly identical copies of the code that fetches
the locals and arguments liveness maps for a frame, plus a third
that's a poor knock-off. Unify these all into a single function.
Change-Id: Ibce7926a0b0e3d23182112da4e25df899579a585
Reviewed-on: https://go-review.googlesource.com/109698
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
There are several things combined in this change.
First, eliminate the gobitvector type in favor
of adding a ptrbit method to bitvector.
In non-performance-critical code, use that method.
In performance critical code, though, load the bitvector data
one byte at a time and iterate only over set bits.
To support that, add and use sys.Ctz8.
name old time/op new time/op delta
StackCopyPtr-8 81.8ms ± 5% 78.9ms ± 3% -3.58% (p=0.000 n=97+96)
StackCopy-8 65.9ms ± 3% 62.8ms ± 3% -4.67% (p=0.000 n=96+92)
StackCopyNoCache-8 105ms ± 3% 102ms ± 3% -3.38% (p=0.000 n=96+95)
Change-Id: I00b80f45612708bd440b1a411a57fa6dfa24aa74
Reviewed-on: https://go-review.googlesource.com/109716
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
|
|
Currently, when the runtime looks up the stack map for a frame, it
uses frame.continpc - 1 unless continpc is the function entry PC, in
which case it uses frame.continpc. As a result, if continpc is the
function entry point (which happens for deferred frames), it will
actually look up the stack map *following* the first instruction.
I think, though I am not positive, that this is always okay today
because the first instruction of a function can never change the stack
map. It's usually not a CALL, so it doesn't have PCDATA. Or, if it is
a CALL, it has to have the entry stack map.
But we're about to start emitting stack maps at every instruction that
changes them, which means the first instruction can have PCDATA
(notably, in leaf functions that don't have a prologue).
To prepare for this, tweak how the runtime looks up stack map indexes
so that if continpc is the function entry point, it directly uses the
entry stack map.
For #24543.
Change-Id: I85aa818041cd26aff416f7b1fba186e9c8ca6568
Reviewed-on: https://go-review.googlesource.com/109349
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
This was a typo mistake according to if cond and runtime/mheap.go:323
Change-Id: Id046d4afbfe0ea43cb29e1a9f400e1f130de221d
Reviewed-on: https://go-review.googlesource.com/102575
Reviewed-by: Austin Clements <austin@google.com>
|
|
Change-Id: I289af4884583537639800e37928c22814d38cba9
Reviewed-on: https://go-review.googlesource.com/98115
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>
|
|
Currently, the heap arena map is a single, large array that covers
every possible arena frame in the entire address space. This is
practical up to about 48 bits of address space with 64 MB arenas.
However, there are two problems with this:
1. mips64, ppc64, and s390x support full 64-bit address spaces (though
on Linux only s390x has kernel support for 64-bit address spaces).
On these platforms, it would be good to support these larger
address spaces.
2. On Windows, processes are charged for untouched memory, so for
processes with small heaps, the mostly-untouched 32 MB arena map
plus a 64 MB arena are significant overhead. Hence, it would be
good to reduce both the arena map size and the arena size, but with
a single-level arena, these are inversely proportional.
This CL adds support for a two-level arena map. Arena frame numbers
are now divided into arenaL1Bits of L1 index and arenaL2Bits of L2
index.
At the moment, arenaL1Bits is always 0, so we effectively have a
single level map. We do a few things so that this has no cost beyond
the current single-level map:
1. We embed the L2 array directly in mheap, so if there's a single
entry in the L2 array, the representation is identical to the
current representation and there's no extra level of indirection.
2. Hot code that accesses the arena map is structured so that it
optimizes to nearly the same machine code as it does currently.
3. We make some small tweaks to hot code paths and to the inliner
itself to keep some important functions inlined despite their
now-larger ASTs. In particular, this is necessary for
heapBitsForAddr and heapBits.next.
Possibly as a result of some of the tweaks, this actually slightly
improves the performance of the x/benchmarks garbage benchmark:
name old time/op new time/op delta
Garbage/benchmem-MB=64-12 2.28ms ± 1% 2.26ms ± 1% -1.07% (p=0.000 n=17+19)
(https://perf.golang.org/search?q=upload:20180223.2)
For #23900.
Change-Id: If5164e0961754f97eb9eca58f837f36d759505ff
Reviewed-on: https://go-review.googlesource.com/96779
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
There are too many places where I want to talk about "indexing into
the arena index". Make this less awkward and ambiguous by calling it
the "arena map" instead.
Change-Id: I726b0667bb2139dbc006175a0ec09a871cdf73f9
Reviewed-on: https://go-review.googlesource.com/96777
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
Now that we support the full non-contiguous virtual address space of
amd64 hardware, some of the comments and constants related to this are
out of date.
This renames memLimitBits to heapAddrBits because 1<<memLimitBits is
no longer the limit of the address space and rewrites the comment to
focus first on hardware limits (which span OSes) and then discuss
kernel limits.
Second, this eliminates the memLimit constant because there's no
longer a meaningful "highest possible heap pointer value" on amd64.
Updates #23862.
Change-Id: I44b32033d2deb6b69248fb8dda14fc0e65c47f11
Reviewed-on: https://go-review.googlesource.com/95498
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
Accessing the arena index is about to get slightly more complicated.
Abstract this away into a set of functions for going back and forth
between addresses and arena slice indexes.
For #23862.
Change-Id: I0b20e74ef47a07b78ed0cf0a6128afe6f6e40f4b
Reviewed-on: https://go-review.googlesource.com/95496
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
Currently, bulkBarrierPreWrite uses inheap to decide whether the
destination is in the heap or whether to check for stack or global
data. However, this isn't the best question to ask.
Instead, get the span directly and query its state. This lets us
directly determine whether this might be a global, or is stack memory,
or is heap memory.
At this point, inheap is no longer used in the hot path, so drop it
from the must-be-inlined list and substitute spanOf.
This will help in a circuitous way with #23862, since fixing that is
going to push inheap very slightly over the inline-able threshold on a
few platforms.
Change-Id: I5360fc1181183598502409f12979899e1e4d45f7
Reviewed-on: https://go-review.googlesource.com/95495
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
This replaces all uses of the mheap_.arena_* fields outside of
mallocinit and sysAlloc. These fields fundamentally assume a
contiguous heap between two bounds, so eliminating these is necessary
for a sparse heap.
Many of these are replaced with checks for non-nil spans at the test
address (which in turn checks for a non-nil entry in the heap arena
array). Some of them are just for debugging and somewhat meaningless
with a sparse heap, so those we just delete.
Updates #10460.
Change-Id: I8345b95ffc610aed694f08f74633b3c63506a41f
Reviewed-on: https://go-review.googlesource.com/85886
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
This splits the heap bitmap into separate chunks for every 64MB of the
heap and introduces an index mapping from virtual address to metadata.
It modifies the heapBits abstraction to use this two-level structure.
Finally, it modifies heapBitsSetType to unroll the bitmap into the
object itself and then copy it out if the bitmap would span
discontiguous bitmap chunks.
This is a step toward supporting general sparse heaps, which will
eliminate address space conflict failures as well as the limit on the
heap size.
It's also advantageous for 32-bit. 32-bit already supports
discontiguous heaps by always starting the arena at address 0.
However, as a result, with a contiguous bitmap, if the kernel chooses
a high address (near 2GB) for a heap mapping, the runtime is forced to
map up to 128MB of heap bitmap. Now the runtime can map sections of
the bitmap for just the parts of the address space used by the heap.
Updates #10460.
This slightly slows down the x/garbage and compilebench benchmarks.
However, I think the slowdown is acceptably small.
name old time/op new time/op delta
Template 178ms ± 1% 180ms ± 1% +0.78% (p=0.029 n=10+10)
Unicode 85.7ms ± 2% 86.5ms ± 2% ~ (p=0.089 n=10+10)
GoTypes 594ms ± 0% 599ms ± 1% +0.70% (p=0.000 n=9+9)
Compiler 2.86s ± 0% 2.87s ± 0% +0.40% (p=0.001 n=9+9)
SSA 7.23s ± 2% 7.29s ± 2% +0.94% (p=0.029 n=10+10)
Flate 116ms ± 1% 117ms ± 1% +0.99% (p=0.000 n=9+9)
GoParser 146ms ± 1% 146ms ± 0% ~ (p=0.193 n=10+7)
Reflect 399ms ± 0% 403ms ± 1% +0.89% (p=0.001 n=10+10)
Tar 173ms ± 1% 174ms ± 1% +0.91% (p=0.013 n=10+9)
XML 208ms ± 1% 210ms ± 1% +0.93% (p=0.000 n=10+10)
[Geo mean] 368ms 371ms +0.79%
name old time/op new time/op delta
Garbage/benchmem-MB=64-12 2.17ms ± 1% 2.21ms ± 1% +2.15% (p=0.000 n=20+20)
Change-Id: I037fd283221976f4f61249119d6b97b100bcbc66
Reviewed-on: https://go-review.googlesource.com/85883
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
There are various places that assume the heap bitmap is contiguous and
scan it sequentially. We're about to split up the heap bitmap. This
commit modifies all of these except heapBitsSetType to use the
heapBits abstractions so they can transparently switch to a
discontiguous bitmap.
Updates #10460. This is a step toward supporting sparse heaps.
Change-Id: I2f3994a5785e4dccb66602fb3950bbd290d9392c
Reviewed-on: https://go-review.googlesource.com/85882
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
Currently the heap bitamp is laid in reverse order in memory relative
to the heap itself. This was originally done out of "excessive
cleverness" so that computing a bitmap pointer could load only the
arena_start field and so that heaps could be more contiguous by
growing the arena and the bitmap out from a common center point.
However, this appears to have no actual performance benefit, it
complicates nearly every use of the bitmap, and it makes already
confusing code more confusing. Furthermore, it's still possible to use
a single field (the new bitmap_delta) for the bitmap pointer
computation by employing slightly different excessive cleverness.
Hence, this CL puts the bitmap into forward order.
This is a (very) updated version of CL 9404.
Change-Id: I743587cc626c4ecd81e660658bad85b54584108c
Reviewed-on: https://go-review.googlesource.com/85881
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
The logic in the spanOf* functions is open-coded in a lot of places
right now. Replace these with calls to the spanOf* functions.
Change-Id: I3cc996aceb9a529b60fea7ec6fef22008c012978
Reviewed-on: https://go-review.googlesource.com/85880
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
heapBitsForObject does two things: it finds the base of the object and
it creates the heapBits for the base of the object. There are several
places where we just care about the base of the object. Furthermore,
greyobject only needs the heapBits in the checkmark path and can
easily compute them only when needed. Once we eliminate passing the
heap bits to grayobject, almost all uses of heapBitsForObject don't
need the heap bits.
Hence, this splits heapBitsForObject into findObject and
heapBitsForAddr (the latter already exists), removes the hbits
argument to grayobject, and replaces all heapBitsForObject calls with
calls to findObject.
In addition to making things cleaner overall, heapBitsForAddr is going
to get more expensive shortly, so it's important that we don't do it
needlessly.
Note that there's an interesting performance pitfall here. I had
originally moved findObject to mheap.go, since it made more sense
there. However, that leads to a ~2% slow down and a whopping 11%
increase in L1 icache misses on both the x/garbage and compilebench
benchmarks. This suggests we may want to be more principled about
this, but, for now, let's just leave findObject in mbitmap.go.
(I tried to make findObject small enough to inline by splitting out
the error case, but, sadly, wasn't quite able to get it under the
inlining budget.)
Change-Id: I7bcb92f383ade565d22a9f2494e4c66fd513fb10
Reviewed-on: https://go-review.googlesource.com/85878
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
These functions all serve essentially the same purpose. mlookup is
used in only one place and findObject in only three. Use
heapBitsForObject instead, which is the most optimized implementation.
(This may seem slightly silly because none of these uses care about
the heap bits, but we're about to split up the functionality of
heapBitsForObject anyway. At that point, findObject will rise from the
ashes.)
Change-Id: I906468c972be095dd23cf2404a7d4434e802f250
Reviewed-on: https://go-review.googlesource.com/85877
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
Calls to writebarrierptr can simply be actual pointer writes. Calls to
writebarrierptr_prewrite need to go through the write barrier buffer.
Updates #22460.
Change-Id: I92cee4da98c5baa499f1977563757c76f95bf0ca
Reviewed-on: https://go-review.googlesource.com/92704
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
heapBits.bits is used during bulkBarrierPreWrite via
heapBits.isPointer, which means it must not be preempted. If it is
preempted, several bad things can happen:
1. This could allow a GC phase change, and the resulting shear between
the barriers and the memory writes could result in a lost pointer.
2. Since bulkBarrierPreWrite uses the P's local write barrier buffer,
if it also migrates to a different P, it could try to append to the
write barrier buffer concurrently with another write barrier. This can
result in the buffer's next pointer skipping over its end pointer,
which results in a buffer overflow that can corrupt arbitrary other
fields in the Ps (or anything in the heap, really, but it'll probably
crash from the corrupted P quickly).
Fix this by marking heapBits.bits go:nosplit. This would be the
perfect use for a recursive no-preempt annotation (#21314).
This doesn't actually affect any binaries because this function was
always inlined anyway. (I discovered it when I was modifying heapBits
and make h.bits() no longer inline, which led to rampant crashes from
problem 2 above.)
Updates #22987 and #22988 (but doesn't fix because it doesn't actually
change the generated code).
Change-Id: I60ebb928b1233b0613361ac3d0558d7b1cb65610
Reviewed-on: https://go-review.googlesource.com/83015
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
|
|
This modifies bulkBarrierPreWrite to use the buffered write barrier
instead of the eager write barrier. This reduces the number of system
stack switches and sanity checks by a factor of the buffer size
(currently 256). This affects both typedmemmove and typedmemclr.
Since this is purely a runtime change, it applies to all arches
(unlike the pointer write barrier).
name old time/op new time/op delta
BulkWriteBarrier-12 7.33ns ± 6% 4.46ns ± 9% -39.10% (p=0.000 n=20+19)
Updates #22460.
Change-Id: I6a686a63bbf08be02b9b97250e37163c5a90cdd8
Reviewed-on: https://go-review.googlesource.com/73832
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@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>
|
|
Replace int division with (cheaper) byte division in heapBitsSetType.
Provides noticeable speed-up:
GrowSlicePtr-6 181ns ± 3% 169ns ± 3% -6.85% (p=0.000 n=10+10)
Change-Id: I4064bb72e8e692023783b8f58d19491844c39382
Reviewed-on: https://go-review.googlesource.com/42290
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
|
|
This adds debugging information when we panic with "heapBitsForSpan:
base out of range".
Updates #20259.
Change-Id: I0dc1a106aa9e9531051c7d08867ace5ef230eb3f
Reviewed-on: https://go-review.googlesource.com/43310
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
|
|
It's no longer necessary to maintain the bitmap of noscan objects
since we now use the span metadata to determine that they're noscan
instead of the bitmap.
The combined effect of segregating noscan spans and the follow-on
optimizations is roughly a 1% improvement in performance across the
go1 benchmarks and the x/benchmarks, with no increase in heap size.
Benchmark details: https://perf.golang.org/search?q=upload:20170420.1
name old time/op new time/op delta
Garbage/benchmem-MB=64-12 2.27ms ± 0% 2.25ms ± 1% -0.96% (p=0.000 n=15+18)
name old time/op new time/op delta
BinaryTree17-12 2.53s ± 2% 2.55s ± 1% +0.68% (p=0.001 n=17+16)
Fannkuch11-12 3.02s ± 0% 3.01s ± 0% -0.15% (p=0.000 n=16+16)
FmtFprintfEmpty-12 47.1ns ± 7% 47.0ns ± 5% ~ (p=0.886 n=20+17)
FmtFprintfString-12 73.6ns ± 3% 73.8ns ± 1% +0.30% (p=0.026 n=19+17)
FmtFprintfInt-12 80.3ns ± 2% 80.2ns ± 1% ~ (p=0.994 n=20+18)
FmtFprintfIntInt-12 124ns ± 0% 124ns ± 0% ~ (all samples are equal)
FmtFprintfPrefixedInt-12 172ns ± 1% 171ns ± 1% -0.72% (p=0.003 n=20+18)
FmtFprintfFloat-12 217ns ± 1% 216ns ± 1% -0.27% (p=0.019 n=18+19)
FmtManyArgs-12 490ns ± 1% 488ns ± 0% -0.36% (p=0.014 n=18+18)
GobDecode-12 6.71ms ± 1% 6.73ms ± 1% +0.42% (p=0.000 n=20+20)
GobEncode-12 5.25ms ± 0% 5.24ms ± 0% -0.20% (p=0.001 n=18+20)
Gzip-12 227ms ± 0% 226ms ± 1% ~ (p=0.107 n=20+19)
Gunzip-12 38.8ms ± 0% 38.8ms ± 0% ~ (p=0.221 n=19+18)
HTTPClientServer-12 75.4µs ± 1% 76.3µs ± 1% +1.26% (p=0.000 n=20+19)
JSONEncode-12 14.7ms ± 0% 14.7ms ± 1% -0.14% (p=0.002 n=18+17)
JSONDecode-12 57.6ms ± 0% 55.2ms ± 0% -4.13% (p=0.000 n=19+19)
Mandelbrot200-12 3.73ms ± 0% 3.73ms ± 0% -0.09% (p=0.000 n=19+17)
GoParse-12 3.18ms ± 1% 3.15ms ± 1% -0.90% (p=0.000 n=18+20)
RegexpMatchEasy0_32-12 73.3ns ± 2% 73.2ns ± 1% ~ (p=0.994 n=20+18)
RegexpMatchEasy0_1K-12 236ns ± 2% 234ns ± 1% -0.70% (p=0.002 n=19+17)
RegexpMatchEasy1_32-12 69.7ns ± 2% 69.9ns ± 2% ~ (p=0.416 n=20+20)
RegexpMatchEasy1_1K-12 366ns ± 1% 365ns ± 1% ~ (p=0.376 n=19+17)
RegexpMatchMedium_32-12 109ns ± 1% 108ns ± 1% ~ (p=0.461 n=17+18)
RegexpMatchMedium_1K-12 35.2µs ± 1% 35.2µs ± 3% ~ (p=0.238 n=19+20)
RegexpMatchHard_32-12 1.77µs ± 1% 1.77µs ± 1% +0.33% (p=0.007 n=17+16)
RegexpMatchHard_1K-12 53.2µs ± 0% 53.3µs ± 0% +0.26% (p=0.001 n=17+17)
Revcomp-12 1.13s ±117% 0.87s ±184% ~ (p=0.813 n=20+19)
Template-12 63.9ms ± 1% 64.6ms ± 1% +1.18% (p=0.000 n=19+20)
TimeParse-12 313ns ± 5% 312ns ± 0% ~ (p=0.114 n=20+19)
TimeFormat-12 336ns ± 0% 333ns ± 0% -0.97% (p=0.000 n=18+16)
[Geo mean] 50.6µs 50.1µs -1.04%
This is a cherry-pick of dev.garbage commit edb54c300f, with updated
benchmark results.
Change-Id: Ic77faaa15cdac3bfbbb0032dde5c204e05a0fd8e
Reviewed-on: https://go-review.googlesource.com/41253
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
|
This is no longer necessary now that we can more efficiently consult
the span's noscan bit.
This is a cherry-pick of dev.garbage commit 312aa09996.
Change-Id: Id0b00b278533660973f45eb6efa5b00f373d58af
Reviewed-on: https://go-review.googlesource.com/41252
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@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 introduces a new type, *gcBits, to use for alloc/mark bitmap
allocations instead of *uint8. This type is marked go:notinheap, so
uses of it correctly eliminate write barriers. Since we now have a
type, this also extracts some common operations to methods both for
convenience and to avoid (*uint8) casts at most use sites.
For #19325.
Change-Id: Id51f734fb2e96b8b7715caa348c8dcd4aef0696a
Reviewed-on: https://go-review.googlesource.com/38580
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
|
