When we pass these types by reference, we usually have to allocate
temporaries on the stack, initialize them, then pass their address
to the conversion functions. It's simpler to pass these types
directly by value.
This particularly applies to conversions needed for fmt.Printf
(to interface{} for constructing a [...]interface{}).
func f(a, b, c string) {
fmt.Printf("%s %s\n", a, b)
fmt.Printf("%s %s\n", b, c)
}
This function's stack frame shrinks from 200 to 136 bytes, and
its code shrinks from 535 to 453 bytes.
The go binary shrinks 0.3%.
Update #24286
Aside: for this function f, we don't really need to allocate
temporaries for the convT2E function. We could use the address
of a, b, and c directly. That might get similar (or maybe better?)
improvements. I investigated a bit, but it seemed complicated
to do it safely. This change was much easier.
Change-Id: I78cbe51b501fb41e1e324ce4203f0de56a1db82d
Reviewed-on: https://go-review.googlesource.com/c/135377
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Instead of
MOVB go.string."foo"(SB), AX
do
MOVB $102, AX
When we know the global we're loading from is readonly, we can
do that read at compile time.
I've made this arch-dependent mostly because the cases where this
happens often are memory->memory moves, and those don't get
decomposed until lowering.
Did amd64/386/arm/arm64. Other architectures could follow.
Update #26498
Change-Id: I41b1dc831b2cd0a52dac9b97f4f4457888a46389
Reviewed-on: https://go-review.googlesource.com/c/141118
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Do []byte(string) conversions more efficiently when the string
is a constant. Instead of calling stringtobyteslice, allocate
just the space we need and encode the initialization directly.
[]byte("foo") rewrites to the following pseudocode:
var s [3]byte // on heap or stack, depending on whether b escapes
s = *(*[3]byte)(&"foo"[0]) // initialize s from the string
b = s[:]
which generates this assembly:
0x001d 00029 (tmp1.go:9) LEAQ type.[3]uint8(SB), AX
0x0024 00036 (tmp1.go:9) MOVQ AX, (SP)
0x0028 00040 (tmp1.go:9) CALL runtime.newobject(SB)
0x002d 00045 (tmp1.go:9) MOVQ 8(SP), AX
0x0032 00050 (tmp1.go:9) MOVBLZX go.string."foo"+2(SB), CX
0x0039 00057 (tmp1.go:9) MOVWLZX go.string."foo"(SB), DX
0x0040 00064 (tmp1.go:9) MOVW DX, (AX)
0x0043 00067 (tmp1.go:9) MOVB CL, 2(AX)
// Then the slice is b = {AX, 3, 3}
The generated code is still not optimal, as it still does load/store
from read-only memory instead of constant stores. Next CL...
Update #26498Fixes#10170
Change-Id: I4b990b19f9a308f60c8f4f148934acffefe0a5bd
Reviewed-on: https://go-review.googlesource.com/c/140698
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This was missed as part of adding a top-level VARDEF
for stack tracing (CL 134156).
Fixes#28055
Change-Id: Id14748dfccb119197d788867d2ec6a3b3c9835cf
Reviewed-on: https://go-review.googlesource.com/c/140304
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>
Allocate a long linked list on the stack. This tests both
lots of live stack objects, and lots of intra-stack pointers
to those objects.
Change-Id: I169e067416455737774851633b1e5367e10e1cf2
Reviewed-on: https://go-review.googlesource.com/c/135296
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
When a function triggers a signal (like a segfault which translates to
a nil pointer exception) during execution, a sigpanic handler is just
below it on the stack. The function itself did not stop at a
safepoint, so we have to figure out what safepoint we should use to
scan its stack frame.
Previously we used the site of the most recent defer to get the live
variables at the signal site. That answer is not quite correct, as
explained in #27518. Instead, use the site of a deferreturn call.
It has all the right variables marked as live (no args, all the return
values, except those that escape to the heap, in which case the
corresponding PAUTOHEAP variables will be live instead).
This CL requires stack objects, so that all the local variables
and args referenced by the deferred closures keep the right variables alive.
Fixes#27518
Change-Id: Id45d8a8666759986c203181090b962e2981e48ca
Reviewed-on: https://go-review.googlesource.com/c/134637
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
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>
In some optimization rules the type of generated OffPtr was
incorrectly set to the type of the pointee, instead of the
pointer. When the OffPtr value is spilled, this may generate
a spill of the wrong type, e.g. a floating point spill of an
integer (pointer) value. On Wasm, this leads to invalid
bytecode.
Fixes#27961.
Change-Id: I5d464847eb900ed90794105c0013a1a7330756cc
Reviewed-on: https://go-review.googlesource.com/c/139257
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Richard Musiol <neelance@gmail.com>
During a call to a reflect-generated function or method (via
makeFuncStub or methodValueCall), when should we scan the return
values?
When we're starting a reflect call, the space on the stack for the
return values is not initialized yet, as it contains whatever junk was
on the stack of the caller at the time. The return space must not be
scanned during a GC.
When we're finishing a reflect call, the return values are
initialized, and must be scanned during a GC to make sure that any
pointers in the return values are found and their referents retained.
When the GC stack walk comes across a reflect call in progress on the
stack, it needs to know whether to scan the results or not. It doesn't
know the progress of the reflect call, so it can't decide by
itself. The reflect package needs to tell it.
This CL adds another slot in the frame of makeFuncStub and
methodValueCall so we can put a boolean in there which tells the
runtime whether to scan the results or not.
This CL also adds the args length to reflectMethodValue so the
runtime can restrict its scanning to only the args section (not the
results) if the reflect package says the results aren't ready yet.
Do a delicate dance in the reflect package to set the "results are
valid" bit. We need to make sure we set the bit only after we've
copied the results back to the stack. But we must set the bit before
we drop reflect's copy of the results. Otherwise, we might have a
state where (temporarily) no one has a live copy of the results.
That's the state we were observing in issue #27695 before this CL.
The bitmap used by the runtime currently contains only the args.
(Actually, it contains all the bits, but the size is set so we use
only the args portion.) This is safe for early in a reflect call, but
unsafe late in a reflect call. The test issue27695.go demonstrates
this unsafety. We change the bitmap to always include both args
and results, and decide at runtime which portion to use.
issue27695.go only has a test for method calls. Function calls were ok
because there wasn't a safepoint between when reflect dropped its copy
of the return values and when the caller is resumed. This may change
when we introduce safepoints everywhere.
This truncate-to-only-the-args was part of CL 9888 (in 2015). That
part of the CL fixed the problem demonstrated in issue27695b.go but
introduced the problem demonstrated in issue27695.go.
TODO, in another CL: simplify FuncLayout and its test. stack return
value is now identical to frametype.ptrdata + frametype.gcdata.
Fixes#27695
Change-Id: I2d49b34e34a82c6328b34f02610587a291b25c5f
Reviewed-on: https://go-review.googlesource.com/137440
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Also includes a small tweak to test/run.go to allow package names
with Unicode letters (as opposed to just ASCII chars).
Updates #27836
Change-Id: Idbf0bdea24174808cddcb69974dab820eb13e521
Reviewed-on: https://go-review.googlesource.com/138075
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Show a more specifc error message in the form of "%d variables but %v
returns %d values" if an assignment mismatch occurs with a function
or method call on the right.
Fixes#27595
Change-Id: Ibc97d070662b08f150ac22d686059cf224e012ab
Reviewed-on: https://go-review.googlesource.com/135575
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
CL 136855 removed the encoding/binary dependency from the checkbce.go
test by defining a local Uint64 to fix the noopt builder; then a more
general mechanism to skip tests on the noopt builder was introduced in
CL 136898, so we can now restore the binary.Uint64 calls in testbce.
Change-Id: I3efbb41be0bfc446a7e638ce6a593371ead2684f
Reviewed-on: https://go-review.googlesource.com/137056
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Adds a new build tag "gcflags_noopt" that can be used in test/*.go
tests.
Fixes#27833
Change-Id: I4ea0ccd9e9e58c4639de18645fec81eb24a3a929
Reviewed-on: https://go-review.googlesource.com/136898
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
&^ and << have equal precedence. Add some parentheses to make sure
we shift before we andnot.
Fixes#27829
Change-Id: Iba8576201f0f7c52bf9795aaa75d15d8f9a76811
Reviewed-on: https://go-review.googlesource.com/136899
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The existing URL in comment points to an Alioth page which was
deprecated (and not working), so use the new Benchmarks Game URL.
Change-Id: Ifd694382a44a24c44acbed3fe1b17bca6dab998f
Reviewed-on: https://go-review.googlesource.com/136835
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The noopt builder is configured by setting GO_GCFLAGS=-N -l, but the
test/run.go test harness doesn't look at GO_GCFLAGS when processing
"errorcheck" files, it just calls compile:
cmdline := []string{goTool(), "tool", "compile", /* etc */}
This is working as intended, since it makes the tests more robust and
independent from the environment; errorcheck files are supposed to set
additional building flags, when needed, like in:
// errorcheck -0 -N -l
The test/bcecheck.go test used to work on the noopt builder (even if
bce is not active on -N -l) because the test was auto-contained and
the file always compiled with optimizations enabled.
In CL 107355, a new bce test dependent on an external package
(encoding.binary) was added. On the noopt builder the external package
is built using -N -l, and this causes a test failure that broke the
noopt builder:
https://build.golang.org/log/b2be319536285e5807ee9d66d6d0ec4d57433768
To reproduce the failure, one can do:
$ go install -a -gcflags="-N -l" std
$ go run run.go -- checkbce.go
This change fixes the noopt builder breakage by removing the bce test
dependency on encoding/binary by defining a local Uint64() function to
be used in the test.
Change-Id: Ife71aab662001442e715c32a0b7d758349a63ff1
Reviewed-on: https://go-review.googlesource.com/136855
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
See the change and comment in typecheck.go for a detailed explanation.
Fixes#26855.
Change-Id: I7867f948490fc0873b1bd849048cda6acbc36e76
Reviewed-on: https://go-review.googlesource.com/136395
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Teach samesafeexpr to handle arithmetic unary and binary ops.
It makes map lookup optimization possible in
m[k+1] = append(m[k+1], ...)
m[-k] = append(m[-k], ...)
... etc
Does not cover "+" for strings (concatenation).
Change-Id: Ibbb16ac3faf176958da344be1471b06d7cf33a6c
Reviewed-on: https://go-review.googlesource.com/135795
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
That optimization is not valid if x == -0.
The test is a bit tricky because 0 == -0. We distinguish
0 from -0 with 1/0 == inf, 1/-0 == -inf.
This has been a bug since CL 24790 in Go 1.8. Probably doesn't
warrant a backport.
Fixes#27718
Note: the optimization x-0 -> x is actually valid.
But it's probably best to take it out, so as to not confuse readers.
Change-Id: I99f16a93b45f7406ec8053c2dc759a13eba035fa
Reviewed-on: https://go-review.googlesource.com/135701
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Instead of skipping all OSLICEARR, skip only ones with non-pointer
array type. For pointers to arrays, it's safe to apply the
self-assignment slicing optimizations.
Refactored the matching code into separate function for readability.
This is an extension to already existing optimization.
On its own, it does not improve any code under std, but
it opens some new optimization opportunities. One
of them is described in the referenced issue.
Updates #7921
Change-Id: I08ac660d3ef80eb15fd7933fb73cf53ded9333ad
Reviewed-on: https://go-review.googlesource.com/133375
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Makes go binary smaller by 0.2%.
I noticed this in autogenerated equal methods, and there are
probably a lot of those.
Change-Id: I4e04eb3653fbceb9dd6a4eee97ceab1fa4d10b72
Reviewed-on: https://go-review.googlesource.com/135379
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
This adds some improvements to the rules for PPC64 to eliminate
unnecessary zero or sign extends, and fix some rule for truncates
which were not always using the correct sign instruction.
This reduces of size of many functions by 1 or 2 instructions and
can improve performance in cases where the execution time depends
on small loops where at least 1 instruction was removed and where that
loop contributes a significant amount of the total execution time.
Included is a testcase for codegen to verify the sign/zero extend
instructions are omitted.
An example of the improvement (strings):
IndexAnyASCII/256:1-16 392ns ± 0% 369ns ± 0% -5.79% (p=0.000 n=1+10)
IndexAnyASCII/256:2-16 397ns ± 0% 376ns ± 0% -5.23% (p=0.000 n=1+9)
IndexAnyASCII/256:4-16 405ns ± 0% 384ns ± 0% -5.19% (p=1.714 n=1+6)
IndexAnyASCII/256:8-16 427ns ± 0% 403ns ± 0% -5.57% (p=0.000 n=1+10)
IndexAnyASCII/256:16-16 441ns ± 0% 418ns ± 1% -5.33% (p=0.000 n=1+10)
IndexAnyASCII/4096:1-16 5.62µs ± 0% 5.27µs ± 1% -6.31% (p=0.000 n=1+10)
IndexAnyASCII/4096:2-16 5.67µs ± 0% 5.29µs ± 0% -6.67% (p=0.222 n=1+8)
IndexAnyASCII/4096:4-16 5.66µs ± 0% 5.28µs ± 1% -6.66% (p=0.000 n=1+10)
IndexAnyASCII/4096:8-16 5.66µs ± 0% 5.31µs ± 1% -6.10% (p=0.000 n=1+10)
IndexAnyASCII/4096:16-16 5.70µs ± 0% 5.33µs ± 1% -6.43% (p=0.182 n=1+10)
Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500
Reviewed-on: https://go-review.googlesource.com/129875
Reviewed-by: David Chase <drchase@google.com>
math.RoundToEven can be done by one arm64 instruction FRINTND, intrinsify it to improve performance.
The current pure Go implementation of the function Abs is translated into five instructions on arm64:
str, ldr, and, str, ldr. The intrinsic implementation requires only one instruction, so in terms of
performance, intrinsify it is worthwhile.
Benchmarks:
name old time/op new time/op delta
Abs-8 3.50ns ± 0% 1.50ns ± 0% -57.14% (p=0.000 n=10+10)
RoundToEven-8 9.26ns ± 0% 1.50ns ± 0% -83.80% (p=0.000 n=10+10)
Change-Id: I9456b26ab282b544dfac0154fc86f17aed96ac3d
Reviewed-on: https://go-review.googlesource.com/116535
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The CL 132915 added the wrong codegen test for math.Copysign(c, -1),
it should test that AND is not emitted. This CL fixes this error.
Change-Id: Ida1d3d54ebfc7f238abccbc1f70f914e1b5bfd91
Reviewed-on: https://go-review.googlesource.com/134815
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Added some more cases that should be guarded against regression.
Change-Id: I9f1dda2fd0be9b6e167ef1cc018fc8cce55c066c
Reviewed-on: https://go-review.googlesource.com/134017
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Add some rules to match the Go code like:
y &= 63
x << y | x >> (64-y)
or
y &= 63
x >> y | x << (64-y)
as a ROR instruction. Make math/bits.RotateLeft faster on arm64.
Extends CL 132435 to arm64.
Benchmarks of math/bits.RotateLeftxxN:
name old time/op new time/op delta
RotateLeft-8 3.548750ns +- 1% 2.003750ns +- 0% -43.54% (p=0.000 n=8+8)
RotateLeft8-8 3.925000ns +- 0% 3.925000ns +- 0% ~ (p=1.000 n=8+8)
RotateLeft16-8 3.925000ns +- 0% 3.927500ns +- 0% ~ (p=0.608 n=8+8)
RotateLeft32-8 3.925000ns +- 0% 2.002500ns +- 0% -48.98% (p=0.000 n=8+8)
RotateLeft64-8 3.536250ns +- 0% 2.003750ns +- 0% -43.34% (p=0.000 n=8+8)
Change-Id: I77622cd7f39b917427e060647321f5513973232c
Reviewed-on: https://go-review.googlesource.com/122542
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Rationale: small buffer optimization does not work and it has
made things slower since 2014. Until we can make it work,
we should prefer simpler code that also turns out to be more
efficient.
With this change, it's possible to use
NewBuffer(make([]byte, 0, bootstrapSize)) to get the desired
stack-allocated initial buffer since escape analysis can
prove the created slice to be non-escaping.
New implementation key points:
- Zero value bytes.Buffer performs better than before
- You can have a truly stack-allocated buffer, and it's not even limited to 64 bytes
- The unsafe.Sizeof(bytes.Buffer{}) is reduced significantly
- Empty writes don't cause allocations
Buffer benchmarks from bytes package:
name old time/op new time/op delta
ReadString-8 9.20µs ± 1% 9.22µs ± 1% ~ (p=0.148 n=10+10)
WriteByte-8 28.1µs ± 0% 26.2µs ± 0% -6.78% (p=0.000 n=10+10)
WriteRune-8 64.9µs ± 0% 65.0µs ± 0% +0.16% (p=0.000 n=10+10)
BufferNotEmptyWriteRead-8 469µs ± 0% 461µs ± 0% -1.76% (p=0.000 n=9+10)
BufferFullSmallReads-8 108µs ± 0% 108µs ± 0% -0.21% (p=0.000 n=10+10)
name old speed new speed delta
ReadString-8 3.56GB/s ± 1% 3.55GB/s ± 1% ~ (p=0.165 n=10+10)
WriteByte-8 146MB/s ± 0% 156MB/s ± 0% +7.26% (p=0.000 n=9+10)
WriteRune-8 189MB/s ± 0% 189MB/s ± 0% -0.16% (p=0.000 n=10+10)
name old alloc/op new alloc/op delta
ReadString-8 32.8kB ± 0% 32.8kB ± 0% ~ (all equal)
WriteByte-8 0.00B 0.00B ~ (all equal)
WriteRune-8 0.00B 0.00B ~ (all equal)
BufferNotEmptyWriteRead-8 4.72kB ± 0% 4.67kB ± 0% -1.02% (p=0.000 n=10+10)
BufferFullSmallReads-8 3.44kB ± 0% 3.33kB ± 0% -3.26% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
ReadString-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
WriteByte-8 0.00 0.00 ~ (all equal)
WriteRune-8 0.00 0.00 ~ (all equal)
BufferNotEmptyWriteRead-8 3.00 ± 0% 3.00 ± 0% ~ (all equal)
BufferFullSmallReads-8 3.00 ± 0% 2.00 ± 0% -33.33% (p=0.000 n=10+10)
The most notable thing in go1 benchmarks is reduced allocs in HTTPClientServer (-1 alloc):
HTTPClientServer-8 64.0 ± 0% 63.0 ± 0% -1.56% (p=0.000 n=10+10)
For more explanations and benchmarks see the referenced issue.
Updates #7921
Change-Id: Ica0bf85e1b70fb4f5dc4f6a61045e2cf4ef72aa3
Reviewed-on: https://go-review.googlesource.com/133715
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The existing implementation causes a compiler panic if a function parameter shadows a built-in function, and then calling that shadowed name.
Fixes#27356
Change-Id: I1ffb6dc01e63c7f499e5f6f75f77ce2318f35bcd
Reviewed-on: https://go-review.googlesource.com/132876
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
For OINDEX and other Left+Right nodes, we want the whole
node to be considered as "may affect memory" if either
of Left or Right affect memory. Initial implementation
only considered node as such if both Left and Right were non-safe.
Change-Id: Icfb965a0b4c24d8f83f3722216db068dad2eba95
Reviewed-on: https://go-review.googlesource.com/133275
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
File is compiled with "-l" flag, so go:noinline is redundant.
Change-Id: Ia269f3b9de9466857fc578ba5164613393e82369
Reviewed-on: https://go-review.googlesource.com/133295
Reviewed-by: Cherry Zhang <cherryyz@google.com>
