Aman Gupta [Tue, 17 Apr 2018 21:28:00 +0000 (14:28 -0700)]
net: document caveats for (*syscall.RawConn).Write on Windows
Change-Id: I6e1fa67dc9d4d151c90eb19a6f736e4daa7d4fb3
Reviewed-on: https://go-review.googlesource.com/107615
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
cmd/internal/obj/x86: disallow PC/FP/SB scaled index
Reject to compile I386/AMD64 asm code that contains
(Register)(PseudoReg*scale) forms of memory operands.
Example of such program: "CALL (AX)(PC*2)".
PseudoReg is one of the PC, FP, SB (but not SP).
When pseudo-register is used in register indirect as
scaled index base, x86 backend will panic because
its register file misses SB/FP/PC registers.
Matthew Dempsky [Tue, 17 Apr 2018 21:40:56 +0000 (14:40 -0700)]
cmd/compile: refactor how declarations are imported
This CL moves all of the logic for wiring up imported declarations
into export.go, so that it can be reused by the indexed importer
code. While here, increase symmetry across routines.
Passes toolstash-check.
Change-Id: I1ccec5c3999522b010e4d04ed56b632fd4d712d9
Reviewed-on: https://go-review.googlesource.com/107621
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Add a new DWARF attribute, DW_AT_go_runtime_type, that gives the offset
of the runtime type structure, if any, for a DWARF type. This should
allow debuggers to decode interface content without having to do awkward
name matching.
Currently, each architecture lowers OpConvert to an arch-specific
OpXXXconvert. This is silly because OpConvert means the same thing on
all architectures and is logically a no-op that exists only to keep
track of conversions to and from unsafe.Pointer. Furthermore, lowering
it makes it harder to recognize in other analyses, particularly
liveness analysis.
This CL eliminates the lowering of OpConvert, leaving it as the
generic op until code generation time.
The main complexity here is that we still need to register-allocate
OpConvert operations. Currently, each arch's lowered OpConvert
specifies all GP registers in its register mask. Ideally, OpConvert
wouldn't affect value homing at all, and we could just copy the home
of OpConvert's source, but this can potentially home an OpConvert in a
LocalSlot, which neither regalloc nor stackalloc expect. Rather than
try to disentangle this assumption from regalloc and stackalloc, we
continue to register-allocate OpConvert, but teach regalloc that
OpConvert can be allocated to any allocatable GP register.
cmd/compile: generate load without DS relocation for go.string on ppc64le
Due to some recent optimizations related to the compare
instruction, DS-form load instructions started to be used
to load 8-byte go.strings. This can cause link time errors
if the go.string is not aligned to 4 bytes.
For DS-form instructions, the value in the offset field must
be a multiple of 4. If the offset is known at the time the
rules are processed, a DS-form load will not be chosen. But for
go.strings, the offset is not known at that time, but a
relocation is generated indicating that the linker should fill
in the DS relocation. When the linker tries to fill in the
relocation, if the offset is not aligned properly, a link error
will occur.
To fix this, when loading a go.string using MOVDload, the full
address of the go.string is generated and loaded into the base
register. Then the go.string is loaded with a 0 offset field.
fanzha02 [Thu, 15 Mar 2018 12:16:48 +0000 (12:16 +0000)]
cmd/internal/obj/arm64: summarize the Go assembly syntax and the GNU syntax mapping rules
The patch rewrites the content of doc.go file. The file describes some
general rules of the mapping between Go assembly syntax and GNU syntax.
And it gives some Go assembly examples and corresponding GNU assembly
examples.
The patch changes the doc.go to use standard doc comment format so that
the link https://golang.org/cmd/internal/obj/arm64/ can display it.
Assembly document framework is mainly contributed by Eric Fang <Eric.Fang@arm.com>
Documentation work is contributed by Eric Fang and Fannie Zhang <Fannie.Zhang@arm.com>
Change-Id: I8b3f6d6c6b91afdc2c44602e8f796beea905085e
Reviewed-on: https://go-review.googlesource.com/102055 Reviewed-by: Rob Pike <r@golang.org>
Matthew Dempsky [Thu, 19 Apr 2018 06:22:26 +0000 (23:22 -0700)]
cmd/compile: make generated function code more consistent
There are a bunch of places where we generate functions: equality and
hash functions; method expression and promoted method wrappers; and
print/delete wrappers for defer/go statements.
This CL brings them in sync by:
1) Always using dclfunc and funcbody. Most were already using this,
but makepartialcall needed some changes.
2) Removing duplicate types.Markdcl/types.Popdcl calls. These are
already handled by dclfunc and funcbody.
3) Using structargs (already used by genwrapper) to construct new
param/result lists from existing types.
4) Always accessing the parameter ONAME nodes through Field.Nname
instead of poking into the ODCLFIELD. Also, since creating a slice of
the entire parameter list is common, extract this out into a
paramNnames helper function.
5) Add a Type.IsVariadic method to simplify identifying variadic
function types.
Passes toolstash-check -gcflags=-dwarf=false. DWARF output changes
because using structargs in makepartialcall changes the generated
parameter names.
Change-Id: I6661d3699afdbe7852ad60db5a4ec6eeb2b696e4
Reviewed-on: https://go-review.googlesource.com/108216
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Michael Fraenkel [Thu, 19 Apr 2018 21:54:29 +0000 (17:54 -0400)]
net: calling File leaves the socket in nonblocking mode
On Unix systems, the underlying socket is no longer forced into blocking
mode.
Fixes #24942
Change-Id: I3e0c503c72df0844e30a63af298691dedacd1f46
Reviewed-on: https://go-review.googlesource.com/108297
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Jason A. Donenfeld [Thu, 19 Apr 2018 21:46:52 +0000 (23:46 +0200)]
syscall: avoid extra syscall on send/recvmsg on Linux
By simply rearranging the logic, we avoid the overhead of a superfluous
call to getsockopt. For, if p is already non empty, there's no point
in having to check if we need to attach dummy payload. This has
performance benefits when using send/recvmsg for high speed
communications.
Jakub Čajka [Fri, 5 Jan 2018 12:38:55 +0000 (13:38 +0100)]
cmd/go/internal/work: support pkgconf 1.4 and later
Fixes #23373
Fix interfacing with latest(1.4+) pkgconf versions, as they have change the
output format, by extending parsing function splitPkgConfigOutput to accommodate
more possible fragment escaping formats. Function is based on pkgconfigs own
implementation at
https://github.com/pkgconf/pkgconf/blob/master/libpkgconf/argvsplit.c. Along
with this change test case TestSplitPkgConfigOutput have been expanded. Thanks
to ignatenko for help on test cases and insights in to the pkgconfig.
Change-Id: I55301bb564b07128d5564ec1454dd247f84a95c3
Reviewed-on: https://go-review.googlesource.com/86541
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
runtime: specify behavior of SetMutexProfileFraction for negative values
Change-Id: Ie4da1a515d5405140d742bdcd55f54a73a7f71fe
Reviewed-on: https://go-review.googlesource.com/108175 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Kevin Burke [Wed, 18 Apr 2018 22:49:06 +0000 (15:49 -0700)]
sync/atomic: use package prefix in examples
Previously these examples declared "var v Value" but any caller would
need to write "var v atomic.Value", so we should use the external
package declaration form to avoid confusion about where Value comes
from.
Change-Id: Ic0b1a05fb6b700da61cfc8efca594c49a9bedb69
Reviewed-on: https://go-review.googlesource.com/107975 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Yuval Pavel Zholkover [Mon, 5 Feb 2018 15:11:27 +0000 (17:11 +0200)]
runtime: fast clock_gettime call on FreeBSD
Use AT_TIMEKEEP ELF aux entry to access a kernel mapped ring of timehands structs.
The timehands are updated by the kernel periodically, but for accurate measure the
timecounter still needs to be queried.
Currently the fast path is used only when kern.timecounter.hardware==TSC-low
or kern.timecounter.hardware=='ARM MPCore Timecounter',
other timecounters revert back to regular system call.
TODO: add support for HPET timecounter on 386/amd64.
Change-Id: I321ca4e92be63ba21a2574b758ef5c1e729086ad
Reviewed-on: https://go-review.googlesource.com/93156
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
cmd/internal/obj/arm64: refactor the extended/shifted register encoding to the backend
The current code encodes the register and the shift/extension into a.Offset
field and this is done in the frontend. The CL refactors it to have the
frontend record the register/shift/extension information in a.Reg or a.Index
and leave the encoding stuff for the backend.
cmd/internal/objfile: emit trailing tab outside of Disasm.Decode
Disasm.Decode currently always appends a tab to the formatted instruction,
although not to any relocations after it.
Decode has two clients: objdump and pprof.
pprof emits plain text, so it would be better not to have a trailing tab.
objdump wants the trailing tab for text/tabwriter,
but it is easy to add that to the Fprintf call.
Shifting the responsibility for the trailing tab to the caller
simplifies the code, increases correctness, and slightly improves
performance by reducing and coalescing string concatenations.
runtime: use saved state in SIGPROF handler for vDSO calls
VDSO calls do manual stack alignment, which doesn't get tracked in the
pcsp table. Without accurate pcsp information, backtracing them is
dangerous, and causes a crash in the SIGPROF handler. Fortunately,
https://golang.org/cl/97315 saves a clean state in m.vdsoPC/SP. Change
to use those if they're present, without attempting a normal backtrace.
Fixes #24925
Change-Id: I4b8501ae73a9d18209e22f839773c4fe6102a509
Reviewed-on: https://go-review.googlesource.com/107778
Run-TryBot: Heschi Kreinick <heschi@google.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Alex Brainman [Wed, 11 Apr 2018 09:43:39 +0000 (19:43 +1000)]
syscall: introduce Pointer type and use it instead of uintptr
Some syscall structures used by crypto/x509 have uintptr
fields that store pointers. These pointers are set with
a pointer to another Go structure. But the pointers are
not visible by garbage collector, and GC does not update
the fields after they were set. So when structure with
invalid uintptr pointers passed to Windows, we get
memory corruption.
This CL introduces CertInfo, CertTrustListInfo and
CertRevocationCrlInfo types. It uses pointers to new types
instead of uintptr in CertContext, CertSimpleChain and
CertRevocationInfo.
CertRevocationInfo, CertChainPolicyPara and
CertChainPolicyStatus types have uintptr field that can
be pointer to many different things (according to Windows
API). So this CL introduces Pointer type to be used for
those cases.
cmd/compile: in escape analysis, use element type for OIND of slice
The escape analysis models the flow of "content" of X with a
level of "indirection" (OIND node) of X. This content can be
pointer dereference, or slice/string element. For the latter
case, the type of the OIND node should be the element type of
the slice/string. This CL fixes this. In particular, this
matters when the element type is pointerless, where the data
flow should not cause any escape.
Fixes #15730.
Change-Id: Iba9f92898681625e7e3ddef76ae65d7cd61c41e0
Reviewed-on: https://go-review.googlesource.com/107597 Reviewed-by: David Chase <drchase@google.com>
The documentation for %v behavior for compound objects uses an ellipsis
to indicate indefinite lenght of elements. This is done for struct
fields as well as elements of arrays and slices. This adds the missing
ellipsis for maps.
Matthew Dempsky [Tue, 17 Apr 2018 21:56:29 +0000 (14:56 -0700)]
cmd/compile: use empty package name for runtime/{race,msan}
These fake imports are just so we can emit build dependencies for the
linker, so the package name isn't really necessary. Also, the package
import logic assumes that if we have the name for a package, then
we've already read some package data for it.
Using the empty string allows the importers to correctly populate it
the first time these packages are seen in package export data.
Michael Fraenkel [Tue, 17 Apr 2018 22:38:06 +0000 (18:38 -0400)]
encoding/json: simplify dominantField
Fixes #18037
Change-Id: I20e27bcc013b00b726eb348daf5ca86b138ddcc2
Reviewed-on: https://go-review.googlesource.com/107598
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Ilya Tocar [Thu, 22 Mar 2018 21:53:52 +0000 (16:53 -0500)]
compress/flate: optimize huffSym
By using local variables and assigning them back to decompressor
at the end of huffSym, we allow compiler to keep them in registers
and avoid reloading/storing them repeatedly. To archive this,
moreBits was inlined and specialized to work with local variables.
Also move EOF error conversion to helper function, to make inlined
part of moreBits more readable. Together this results in nice speed-up:
cmd/internal/obj/arm, runtime: delete old ARM softfloat code
CL 106735 changed to the new softfloat support on GOARM=5.
ARM assembly code that uses FP instructions not guarded on GOARM,
if any, will break. The easiest way to fix is probably to use Go
implementation on GOARM=5, like
MOVB runtime·goarm(SB), R11
CMP $5, R11
BEQ arm5
... FP instructions ...
RET
arm5:
CALL or JMP to Go implementation
Alberto Donizetti [Mon, 16 Apr 2018 11:35:35 +0000 (13:35 +0200)]
fmt: document that Scan etc. accept 'p' format floats
In the Scan functions documentation, clarify that for float/complex
literals in scientific notation both decimal (e) and binary (p)
exponents are accepted.
Fixes #24453
Change-Id: Ic6dcdb0c36e088ffb65177038aff7a57ab56b805
Reviewed-on: https://go-review.googlesource.com/107416 Reviewed-by: Rob Pike <r@golang.org>
Tim Cooper [Sat, 27 Jan 2018 02:29:55 +0000 (22:29 -0400)]
os/exec: remove "binary" when talking about executables
The use of binary was incorrect as executable files can also be scripts.
The docs for Error are also reworded. The old docs implied that Error was
returned when attempting to start an executable, which is not correct: it
was returned by LookPath when the file was not found or did not have the
attributes of an executable.
erifan01 [Tue, 28 Nov 2017 03:38:03 +0000 (03:38 +0000)]
math: add a testcase for Mod and Remainder respectively
One might try to implement the Mod or Remainder function with the expression
x - TRUNC(x/y + 0.5)*y, but in fact this method is wrong, because the rounding
of (x/y + 0.5) to initialize the argument of TRUNC may lose too much precision.
However, the current test cases can not detect this error. This CL adds two
test cases to prevent people from continuing to do such attempts.
Change-Id: I6690f5cffb21bf8ae06a314b7a45cafff8bcee13
Reviewed-on: https://go-review.googlesource.com/84275 Reviewed-by: Robert Griesemer <gri@golang.org>
Rob Pike [Mon, 16 Apr 2018 21:16:40 +0000 (07:16 +1000)]
doc/contribute.html: adjust wording from previous CL
The previous CL, 107197, overclarified the need for short subject
lines. Tweak the wording to be a guideline (keep it short) rather
than a limit (76 characters), which is more the Go way.
Michael Munday [Thu, 12 Apr 2018 14:24:08 +0000 (15:24 +0100)]
cmd/compile: generate constants for NeqPtr, EqPtr and IsNonNil ops
If both inputs are constant offsets from the same pointer then we
can evaluate NeqPtr and EqPtr at compile time. Triggers a few times
during all.bash. Removes a conditional branch in the following
code:
copy(x[1:], x[:])
This branch was recently added as an optimization in CL 94596. We
now skip the memmove if the pointers are equal. However, in the
above code we know at compile time that they are never equal.
Also, when the offset is variable, check if the offset is zero
rather than if the pointers are equal. For example:
copy(x[a:], x[:])
This would now skip the copy if a == 0, rather than if x + a == x.
Finally I've also added a rule to make IsNonNil true for pointers
to values on the stack. The nil check elimination pass will catch
these anyway, but eliminating them here might eliminate branches
earlier.
Change-Id: If72f436fef0a96ad0f4e296d3a1f8b6c3e712085
Reviewed-on: https://go-review.googlesource.com/106635
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
cmd/dist, os/user: test os/user in osusergo mode as well, fix plan9 & windows
Would've caught two regressions so far, and found two more.
Updates #24841
Updates #24845 (package net remains)
Change-Id: I57ad06eb54e04b8c99b5d2e7f24c77ad865224e8
Reviewed-on: https://go-review.googlesource.com/107300 Reviewed-by: Ian Lance Taylor <iant@golang.org>
David Url [Mon, 2 Apr 2018 10:57:59 +0000 (12:57 +0200)]
net/http: omit forbidden Trailer headers from response
Use the vendored ValidTrailerHeader function from x/net/http/httpguts to
check Trailer headers according to RFC 7230. The previous implementation
only omitted illegal Trailer headers defined in RFC 2616.
This CL adds x/net/http/httpguts from CL 104042 (git rev a35a21de97)
Change-Id: I71d0d7f75108bf4ad606733a45bb71baa66a4e91
Reviewed-on: https://go-review.googlesource.com/107197 Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com> Reviewed-by: Giovanni Bajo <rasky@develer.com>
Daniel Martí [Sun, 8 Apr 2018 12:39:10 +0000 (13:39 +0100)]
cmd/compile/internal/gc: add some Node methods
Focus on "isfoo" funcs that take a *Node, and conver them to isFoo
methods instead. This makes for more idiomatic Go code, and also more
readable func names.
Found candidates with grep, and applied most changes with sed. The funcs
chosen were isgoconst, isnil, and isblank. All had the same signature,
func(*Node) bool.
While at it, camelCase the isliteral and iszero function names. Don't
move these to methods, as they are only used in the backend part of gc,
which might one day be split into a separate package.
Passes toolstash -cmp on std cmd.
Change-Id: I4df081b12d36c46c253167c8841c5a841f1c5a16
Reviewed-on: https://go-review.googlesource.com/105555
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Giovanni Bajo [Sun, 15 Apr 2018 20:53:58 +0000 (22:53 +0200)]
test: small cleanup of code and comments in run.go
While writing CL 107315, I went back and forth for the syntax used for
constraints of build environments in which the architecture did not
support varitants ("plan9/amd64" vs "plan9/amd64/"). I eventually
settled for the latter because the code required less heuristics
(think parsing "plan9/386" vs "386/sse2") but there were a few
leftovers in code and comments.
Change-Id: I9d9a008f3814f9a1642609650eb571e7f1a675cf
Reviewed-on: https://go-review.googlesource.com/107338
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
runtime: use internal/cpu.X86.HasAVX2 instead of support_avx2
After CL 104636 cpu.X86.HasAVX is set early enough that it can be used
in runtime·memclrNoHeapPointers. Add an offset to use in assembly and
replace the only occurence of support_avx2.
Giovanni Bajo [Sun, 15 Apr 2018 17:00:27 +0000 (19:00 +0200)]
test: run codegen tests on all supported architecture variants
This CL makes the codegen testsuite automatically test all
architecture variants for architecture specified in tests. For
instance, if a test file specifies a "arm" test, it will be
automatically run on all GOARM variants (5,6,7), to increase
the coverage.
The CL also introduces a syntax to specify only a specific
variant (eg: "arm/7") in case the test makes sense only there.
The same syntax also allows to specify the operating system
in case it matters (eg: "plan9/386/sse2").
Fixes #24658
Change-Id: I2eba8b918f51bb6a77a8431a309f8b71af07ea22
Reviewed-on: https://go-review.googlesource.com/107315
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
CL 106463 introduced what seems to trim the "go:" prefix in pramas
comments twice, so "//go:go:foo" would be handled, too.
So either the strings.TrimPrefix or the [3:]-slicing is not needed.
I opted to remove the [3:]-slicing.
Change-Id: I1325bbc08a9be9ae100c5a7775b0a23f9ed0a419
Reviewed-on: https://go-review.googlesource.com/107256 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
text/tabwriter: reduce allocations from tracking cells
The tabwriter tracks cells on a line-by-line basis.
This can be memory-hungry when working with large input.
This change adds two optimizations.
First, when there's an existing cell slice for a line,
don't overwrite it by appending.
This helps when re-using a Writer,
or when the output is broken into groups,
e.g. by a blank line.
We now re-use that existing cell slice.
Second, we predict that the number of cells in a line
will probably match those of the previous line,
since tabwriter is most often used to format tables.
This has a noticeable impact on cmd/objdump (#24725).
It reduces allocated space by about 55%.
It also speeds it up some.
Using "benchcmd -n 10 Objdump go tool objdump `which go`":
name old time/op new time/op delta
ObjdumpCompile 9.03s ± 1% 8.51s ± 1% -5.81% (p=0.000 n=10+10)
It might also imaginably speed up gofmt on some
large machine-generated code.
Aman Gupta [Tue, 7 Nov 2017 07:02:21 +0000 (23:02 -0800)]
net: implement (*syscall.RawConn).Read/Write on Windows
RawRead assumes the callback will perform either (a) a blocking read
and always return true, (b) a blocking read with a SO_RCVTIMEO set
returning false on WSAETIMEDOUT, or (c) a non-blocking read
returning false on WSAEWOULDBLOCK. In the latter two cases, it uses
a 0-byte overlapped read for notifications from the IOCP runtime
when the socket becomes readable before trying again.
RawWrite assumes the callback will perform blocking write and will
always return true, and makes no effort to tie into the runtime loop.
Alex Brainman [Sat, 14 Apr 2018 04:14:28 +0000 (14:14 +1000)]
cmd/internal/obj/arm64: do not clear environment in TestLarge and TestNoRet
Windows process cannot run properly, if it only has
GOOS and GOARCH environment variables set. It needs
other environment variables. So adjust TestLarge and
TestNoRet to add GOOS and GOARCH to the existing
variables set instead of clearing environment.
Alex Brainman [Sat, 14 Apr 2018 02:48:55 +0000 (12:48 +1000)]
syscall: remove WSAEMSGSIZE
CL 92475 added WSAEMSGSIZE const to syscall package. But there
is already copy of WSAEMSGSIZE in internal/syscall/windows.
So delete syscall.WSAEMSGSIZE
cmd/compile/internal/ssa: prefer non-indexed stores on amd64
We sometimes generate code like this:
v473 = MOVQconst <uintptr> // constant..
v580 = MOVBstoreidx1 <mem> v1056 v473 v819 v491 // ..only used as an index
Rewrite indexed stores to non-indexed version, where possible.
This allows to eliminate const->register move, reducing codesize and lowering register pressure.
Ilya Tocar [Fri, 23 Mar 2018 20:45:03 +0000 (15:45 -0500)]
runtime: avoid division in growslice
Add a special case for power-of-2 sized elements.
We can replace div/mul with left/right shift and avoid expensive operation.
growslice is hotter for short slices of small elements, such as int16, so
add an int16 version for GrowSlice benchmark.