The bitLen function currently shifts out blocks of 8 bits at a time.
This change replaces this sorta-linear algorithm with a log(N)
one (shift out 16 bits, then 8, then 4, then 2, then 1).
I left the start of it linear at 16 bits at a time so that
the function continues to work with 32 or 64 bit values
without any funkiness.
The algorithm is similar to several of the nlz ("number of
leading zeros") algorithms from "Hacker's Delight" or the
"bit twiddling hacks" pages.
Doesn't make a big difference to the existing benchmarks, but
I'm using the code in a different context that calls bitLen
much more often, so it seemed worthwhile making the existing
codebase faster so that it's a better building block.
Microbenchmark results on a 64-bit Macbook Pro using 6g from weekly.2012-01-20:
benchmark old ns/op new ns/op delta
big.BenchmarkBitLen0 4 6 +50.12%
big.BenchmarkBitLen1 4 6 +33.91%
big.BenchmarkBitLen2 6 6 +3.05%
big.BenchmarkBitLen3 7 6 -19.05%
big.BenchmarkBitLen4 9 6 -30.19%
big.BenchmarkBitLen5 11 6 -42.23%
big.BenchmarkBitLen8 16 6 -61.78%
big.BenchmarkBitLen9 5 6 +18.29%
big.BenchmarkBitLen16 18 7 -60.99%
big.BenchmarkBitLen17 7 6 -4.64%
big.BenchmarkBitLen31 19 7 -62.49%
On an ARM machine (with the previous weekly):
benchmark old ns/op new ns/op delta
big.BenchmarkBitLen0 37 50 +36.56%
big.BenchmarkBitLen1 59 51 -13.69%
big.BenchmarkBitLen2 74 59 -20.40%
big.BenchmarkBitLen3 92 60 -34.89%
big.BenchmarkBitLen4 110 59 -46.09%
big.BenchmarkBitLen5 127 60 -52.68%
big.BenchmarkBitLen8 181 59 -67.24%
big.BenchmarkBitLen9 78 60 -23.05%
big.BenchmarkBitLen16 199 69 -65.13%
big.BenchmarkBitLen17 91 70 -23.17%
big.BenchmarkBitLen31 210 95 -54.43%
R=golang-dev, dave, edsrzf, gri
CC=golang-dev
https://golang.org/cl/5570044
Move error information into Package struct, so that
a package can be returned even if a dependency failed
to load or did not exist. This makes it possible to run
'go fix' or 'go fmt' on packages with broken dependencies
or missing imports. It also enables go get -fix.
The new go list -e flag lets go list process those package
errors as normal data.
Change p.Doc to be first sentence of package doc, not
entire package doc. Makes go list -json or
go list -f '{{.ImportPath}} {{.Doc}}' much more reasonable.
The go tool now depends on http, which means also
net and crypto/tls, both of which use cgo. Trying to
make the build scripts that build the go tool understand
and handle cgo is too much work. Instead, we build
a stripped down version of the go tool, compiled as go_bootstrap,
that substitutes an error stub for the usual HTTP code.
The buildscript builds go_bootstrap, go_bootstrap builds
the standard packages and commands, including the full
including-HTTP-support go tool, and then go_bootstrap
gets deleted.
Also handle the case where the buildscript needs updating
during all.bash: if it fails but a go command can be found on
the current $PATH, try to regenerate it. This gracefully
handles situations like adding a new file to a package
used by the go tool.
R=r, adg
CC=golang-dev
https://golang.org/cl/5553059
This lets the client of go/build specify additional tags that
can be recognized in a // +build directive. For example,
a build for a custom environment like App Engine might
include "appengine" in the BuildTags list, so that packages
can be written with some files saying
// +build appengine (build only on app engine)
or
// +build !appengine (build only when NOT on app engine)
App Engine here is just a hypothetical context. I plan to use
this in the cmd/go sources to distinguish the bootstrap version
of cmd/go (which will not use networking) from the full version
using a custom tag. It might also be useful in App Engine.
Also, delete Build and Script, which we did not end up using for
cmd/go and which never got turned on for real in goinstall.
R=r, adg
CC=golang-dev
https://golang.org/cl/5554079
Golden files have extension .d.golden where d is the mode value (0 or 1 for now)
(i.e., testdata/file.out is now testdata/file.0.golden, and there is a new file
testdata/file.1.golden for each testcase)
R=rsc
CC=golang-dev
https://golang.org/cl/5573046
Document that indirection through a nil pointer will panic.
Explain function invocation.
This section will need more work, but it's a start.
Fixes#1865.
Fixes#2252.
R=rsc, iant, r
CC=golang-dev
https://golang.org/cl/5532114
Marshalling of []byte in attributes and the general
marshalling of named []byte types was fixed.
A []byte field also won't be nil if an XML element
was mapped to it, even if the element is empty.
Tests were introduced to make sure that *struct{}
fields works correctly for element presence testing.
No changes to the logic made in that regard.
R=rsc
CC=golang-dev
https://golang.org/cl/5539070
It was 2^31, but that could cause overflow and trouble.
Reduce it to 2^30 and add a TODO.
R=golang-dev, iant
CC=golang-dev
https://golang.org/cl/5562049
I haven't looked at the source, but the gc compiler appears to
omit "not used" errors when there is an error in the
initializer. This is harder to do in gccgo, and frankly I
think the "not used" error is still useful even if the
initializer has a problem. This CL tweaks some tests to avoid
the error, which is not the point of these tests in any case.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5561059
It takes too much memory to be reliable and causes
trouble on 32-bit machines.
Sigh.
Fixes#2756.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/5567043
Make the panic detectable, and use that in ioutil.ReadFile to
give an error if the file is too big.
R=golang-dev, minux.ma, bradfitz
CC=golang-dev
https://golang.org/cl/5563045
This fixes the bug Rob ran into when editing package bytes.
Regexp imports regexp/syntax, which imports bytes, and
regexp/syntax was not being properly recompiled during a
test of a change to package bytes.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5555065
This should make it easier to add the zillion little changes coming.
No content change here beyond a couple of introductory sentences.
Sections have been moved wholesale without editing them.
R=golang-dev, rsc, gri
CC=golang-dev
https://golang.org/cl/5557074
Preserve test.
changeset: 11593:f1deaf35e1d1
user: Luuk van Dijk <lvd@golang.org>
date: Tue Jan 17 10:00:57 2012 +0100
summary: gc: fix infinite recursion for embedded interfaces
This is causing 'interface type loop' errors during compilation
of a complex program. I don't understand what's happening
well enough to boil it down to a simple test case, but undoing
this change fixes the problem.
The change being undone is fixing a corner case (uses of
pointer to interface in an interface definition) that basically
only comes up in erroneous Go programs. Let's not try to
fix this again until after Go 1.
Unfixes issue 1909.
TBR=lvd
CC=golang-dev
https://golang.org/cl/5555063
