compress/bzip2: remove bit-tricks

Since compiler is now able to generate conditional moves, we can replace
bit-tricks with simple if/else. This even results in slightly better performance:

name            old time/op    new time/op    delta
DecodeDigits-6    13.4ms ± 4%    13.0ms ± 2%  -2.63%  (p=0.003 n=10+10)
DecodeTwain-6     37.5ms ± 1%    36.3ms ± 1%  -3.03%  (p=0.000 n=10+9)
DecodeRand-6      4.23ms ± 1%    4.07ms ± 1%  -3.67%  (p=0.000 n=10+9)

name            old speed      new speed      delta
DecodeDigits-6  7.47MB/s ± 4%  7.67MB/s ± 2%  +2.69%  (p=0.002 n=10+10)
DecodeTwain-6   10.4MB/s ± 1%  10.7MB/s ± 1%  +3.25%  (p=0.000 n=10+8)
DecodeRand-6    3.87MB/s ± 1%  4.03MB/s ± 2%  +4.08%  (p=0.000 n=10+10)
diff --git a/src/compress/bzip2/huffman.go b/src/compress/bzip2/huffman.go

Change-Id: Ie96ef1a9e07013b07e78f22cdccd531f3341caca
Reviewed-on: https://go-review.googlesource.com/102015
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
Reviewed-by: Joe Tsai <joetsai@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

all: enable c-shared/c-archive support for freebsd/amd64

Fixes #14327
Much of the code is based on the linux/amd64 code that implements these
build modes, and code is shared where possible.

Change-Id: Ia510f2023768c0edbc863aebc585929ec593b332
Reviewed-on: https://go-review.googlesource.com/93875
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

runtime,sync/atomic: replace asm BYTEs with insts for x86

For each replacement, test case is added to new 386enc.s file
with exception of EMMS, SYSENTER, MFENCE and LFENCE as they
are already covered in amd64enc.s (same on amd64 and 386).

The replacement became less obvious after go vet suggested changes
Before:
BYTE $0x0f; BYTE $0x7f; BYTE $0x44; BYTE $0x24; BYTE $0x08
Changed to MOVQ (this form is being tested):
MOVQ M0, 8(SP)
Refactored to FP-relative access (go vet advice):
MOVQ M0, val+4(FP)

Change-Id: I56b87cf3371b6ad81ad0cd9db2033aee407b5818
Reviewed-on: https://go-review.googlesource.com/101475
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>

net/url: fix contradiction in PathUnescape docs

Change-Id: If35e3faa738c5d7d72cf77d14b276690579180a1
Reviewed-on: https://go-review.googlesource.com/101921
Run-TryBot: Ross Light <light@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

runtime: parse auxv on freebsd

Decode AT_PAGESZ to determine physPageSize on freebsd/{386,amd64,arm}
and AT_HWCAP for hwcap and hardDiv on freebsd/arm. Also use hwcap to
perform the FP checks in checkgoarm akin to the linux/arm
implementation.

Change-Id: I532810a1581efe66277e4305cb234acdc79ee91e
Reviewed-on: https://go-review.googlesource.com/99780
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/doc: use empty GOPATH when running the tests

Otherwise, a populated GOPATH might result in failures such as:

$ go test
[...] no buildable Go source files in [...]/gopherjs/compiler/natives/src/crypto/rand
exit status 1

Move the initialization of the dirs walker out of the init func, so that
we can control its behavior in the tests.

Updates #24464.

Change-Id: I4b26a7d3d6809bdd8e9b6b0556d566e7855f80fe
Reviewed-on: https://go-review.googlesource.com/101836
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/trace: remove unused variable in tests

Unused variables in closures are currently not diagnosed by the
compiler (this is Issue #3059), while go/types catches them.

One unused variable in the cmd/trace tests is causing the go/types
test that typechecks the whole standard library to fail:

  FAIL: TestStdlib (8.05s)
    stdlib_test.go:223: cmd/trace/annotations_test.go:241:6: gcTime
    declared but not used
  FAIL

Remove it.

Updates #24464

Change-Id: I0f1b9db6ae1f0130616ee649bdbfdc91e38d2184
Reviewed-on: https://go-review.googlesource.com/101815
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>

go/internal/srcimporter: simplify and fix package file lookup

The old code was a blend of (copied) code that existed before go/build,
and incorrect adjustments made when go/build was introduced. This change
leaves package path determination entirely to go/build and in the process
fixes issues with relative import paths.

Fixes #23092
Fixes #24392

Change-Id: I9e900538b365398751bace56964495c5440ac4ae
Reviewed-on: https://go-review.googlesource.com/83415
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

net/http: remove extraneous call to VerifyHostname

VerifyHostname is called by tls.Conn during Handshake and does not need to be called explicitly.

Change-Id: I22b7fa137e76bb4be3d0018813a571acfb882219
Reviewed-on: https://go-review.googlesource.com/98618
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>

crypto/x509: support the PSS certificates that OpenSSL 1.1.0 generates.

It serialises optional parameters as empty rather than NULL. It's
probably technically correct, although ASN.1 has a long history of doing
this different ways.

But OpenSSL is likely common enough that we want to support this
encoding.

Fixes #23847

Change-Id: I81c60f0996edfecf59467dfdf75b0cf8ba7b1efb
Reviewed-on: https://go-review.googlesource.com/96417
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

cmd/compile/internal/ssa: update regalloc in loops

Currently we don't lift spill out of loop if loop contains call.
However often we have code like this:

for .. {
    if hard_case {
call()
    }
    // simple case, without call
}

So instead of checking for any call, check for unavoidable call.
For #22698 cases I see:
mime/quotedprintable/Writer-6                   10.9µs ± 4%      9.2µs ± 3%   -15.02%  (p=0.000 n=8+8)
And:
compress/flate/Encode/Twain/Huffman/1e4-6       99.4µs ± 6%     90.9µs ± 0%    -8.57%  (p=0.000 n=8+8)
compress/flate/Encode/Twain/Huffman/1e5-6       760µs ± 1%      725µs ± 1%     -4.56%  (p=0.000 n=8+8)
compress/flate/Encode/Twain/Huffman/1e6-6       7.55ms ± 0%      7.24ms ± 0%     -4.07%  (p=0.000 n=8+7)

There are no significant changes on go1 benchmarks.
But for cases with runtime arch checks, where we call generic version on old hardware,
there are respectable performance gains:
math/RoundToEven-6                             1.43ns ± 0%     1.25ns ± 0%   -12.59%  (p=0.001 n=7+7)
math/bits/OnesCount64-6                        1.60ns ± 1%     1.42ns ± 1%   -11.32%  (p=0.000 n=8+8)

Also on some runtime benchmarks loops have less loads and higher performance:
runtime/RuneIterate/range1/ASCII-6             15.6ns ± 1%     13.9ns ± 1%   -10.74%  (p=0.000 n=7+8)
runtime/ArrayEqual-6                           3.22ns ± 0%     2.86ns ± 2%   -11.06%  (p=0.000 n=7+8)

Fixes #22698
Updates #22234

Change-Id: I0ae2f19787d07a9026f064366dedbe601bf7257a
Reviewed-on: https://go-review.googlesource.com/84055
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>

test/codegen: port comparisons tests to codegen

And delete them from asm_test.

Change-Id: I64c512bfef3b3da6db5c5d29277675dade28b8ab
Reviewed-on: https://go-review.googlesource.com/101595
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>

cmd/compile: fix regression in DWARF inlined routine variable tracking

Fix a bug in the code that generates the pre-inlined variable
declaration table used as raw material for emitting DWARF inline
routine records. The fix for issue 23704 altered the recipe for
assigning file/line/col to variables in one part of the compiler, but
didn't update a similar recipe in the code for variable tracking.
Added a new test that should catch problems of a similar nature.

Fixes #24460.

Change-Id: I255c036637f4151aa579c0e21d123fd413724d61
Reviewed-on: https://go-review.googlesource.com/101676
Reviewed-by: Alessandro Arzilli <alessandro.arzilli@gmail.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>

cmd/compile: mark LAA and LAAG as clobbering flags on s390x

The atomic add instructions modify the condition code and so need to
be marked as clobbering flags.

Fixes #24449.

Change-Id: Ic69c8d775fbdbfb2a56c5e0cfca7a49c0d7f6897
Reviewed-on: https://go-review.googlesource.com/101455
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/asm: fix bug about VMOV instruction (move a vector element to another) on ARM64

This change fixes index error when encoding VMOV instruction which pattern
is vmov Vn.<T>[index], Vd.<T>[index]

Change-Id: I949166e6dfd63fb0a9365f183b6c50d452614f9d
Reviewed-on: https://go-review.googlesource.com/101335
Reviewed-by: Cherry Zhang <cherryyz@google.com>

cmd/asm: fix bug about VMOV instruction (move register to vector element) on ARM64

This change fixes index error when encoding VMOV instruction which pattern is
VMOV Rn, V.<T>[index]. For example VMOV R1, V1.S[1] is assembled as VMOV R1, V1.S[0]

Fixes #24400
Change-Id: I82b5edc8af4e06862bc4692b119697c6bb7dc3fb
Reviewed-on: https://go-review.googlesource.com/101297
Reviewed-by: Cherry Zhang <cherryyz@google.com>

cmd/compile: avoid mapaccess at m[k]=append(m[k]..

Currently rvalue m[k] is transformed during walk into:

        tmp1 := *mapaccess(m, k)
        tmp2 := append(tmp1, ...)
        *mapassign(m, k) = tmp2

However, this is suboptimal, as we could instead produce just:
        tmp := mapassign(m, k)
        *tmp := append(*tmp, ...)

Optimization is possible only if during Order it may tell that m[k] is
exactly the same at left and right part of assignment. It doesn't work:
1) m[f(k)] = append(m[f(k)], ...)
2) sink, m[k] = sink, append(m[k]...)
3) m[k] = append(..., m[k],...)

Benchmark:
name                           old time/op    new time/op    delta
MapAppendAssign/Int32/256-8      33.5ns ± 3%    22.4ns ±10%  -33.24%  (p=0.000 n=16+18)
MapAppendAssign/Int32/65536-8    68.2ns ± 6%    48.5ns ±29%  -28.90%  (p=0.000 n=20+20)
MapAppendAssign/Int64/256-8      34.3ns ± 4%    23.3ns ± 5%  -32.23%  (p=0.000 n=17+18)
MapAppendAssign/Int64/65536-8    65.9ns ± 7%    61.2ns ±19%   -7.06%  (p=0.002 n=18+20)
MapAppendAssign/Str/256-8         116ns ±12%      79ns ±16%  -31.70%  (p=0.000 n=20+19)
MapAppendAssign/Str/65536-8       134ns ±15%     111ns ±45%  -16.95%  (p=0.000 n=19+20)

name                           old alloc/op   new alloc/op   delta
MapAppendAssign/Int32/256-8       47.0B ± 0%     46.0B ± 0%   -2.13%  (p=0.000 n=19+18)
MapAppendAssign/Int32/65536-8     27.0B ± 0%     20.7B ±30%  -23.33%  (p=0.000 n=20+20)
MapAppendAssign/Int64/256-8       47.0B ± 0%     46.0B ± 0%   -2.13%  (p=0.000 n=20+17)
MapAppendAssign/Int64/65536-8     27.0B ± 0%     27.0B ± 0%     ~     (all equal)
MapAppendAssign/Str/256-8         94.0B ± 0%     78.0B ± 0%  -17.02%  (p=0.000 n=20+16)
MapAppendAssign/Str/65536-8       54.0B ± 0%     54.0B ± 0%     ~     (all equal)

Fixes #24364
Updates #5147

Change-Id: Id257d052b75b9a445b4885dc571bf06ce6f6b409
Reviewed-on: https://go-review.googlesource.com/100838
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

Revert "bytes: add optimized Compare for arm64"

This reverts commit bfa8b6f8ffa49b3e5796af45a2f0bf9714a37bb9.

Reason for revert: This depends on another CL which is not yet submitted.

Change-Id: I50e7594f1473c911a2079fe910849a6694ac6c07
Reviewed-on: https://go-review.googlesource.com/101496
Reviewed-by: Cherry Zhang <cherryyz@google.com>

bytes: add optimized Compare for arm64

Use LDP instructions to load 16 bytes per loop when the source length is long. Specially
process the 8 bytes length, 4 bytes length and 2 bytes length to get a better performance.

Benchmark result:
name                           old time/op   new time/op    delta
BytesCompare/1-8                21.0ns ± 0%    10.5ns ± 0%      ~     (p=0.079 n=4+5)
BytesCompare/2-8                11.5ns ± 0%    10.5ns ± 0%    -8.70%  (p=0.008 n=5+5)
BytesCompare/4-8                13.5ns ± 0%    10.0ns ± 0%   -25.93%  (p=0.008 n=5+5)
BytesCompare/8-8                28.8ns ± 0%     9.5ns ± 0%      ~     (p=0.079 n=4+5)
BytesCompare/16-8               40.5ns ± 0%    10.5ns ± 0%   -74.07%  (p=0.008 n=5+5)
BytesCompare/32-8               64.6ns ± 0%    12.5ns ± 0%   -80.65%  (p=0.008 n=5+5)
BytesCompare/64-8                112ns ± 0%      16ns ± 0%   -85.27%  (p=0.008 n=5+5)
BytesCompare/128-8               208ns ± 0%      24ns ± 0%   -88.22%  (p=0.008 n=5+5)
BytesCompare/256-8               400ns ± 0%      50ns ± 0%   -87.62%  (p=0.008 n=5+5)
BytesCompare/512-8               785ns ± 0%      82ns ± 0%   -89.61%  (p=0.008 n=5+5)
BytesCompare/1024-8             1.55µs ± 0%    0.14µs ± 0%      ~     (p=0.079 n=4+5)
BytesCompare/2048-8             3.09µs ± 0%    0.27µs ± 0%      ~     (p=0.079 n=4+5)
CompareBytesEqual-8             39.0ns ± 0%    12.0ns ± 0%   -69.23%  (p=0.008 n=5+5)
CompareBytesToNil-8             8.57ns ± 5%    8.23ns ± 2%    -3.99%  (p=0.016 n=5+5)
CompareBytesEmpty-8             7.37ns ± 0%    7.36ns ± 4%      ~     (p=0.690 n=5+5)
CompareBytesIdentical-8         7.39ns ± 0%    7.46ns ± 2%      ~     (p=0.667 n=5+5)
CompareBytesSameLength-8        17.0ns ± 0%    10.5ns ± 0%   -38.24%  (p=0.008 n=5+5)
CompareBytesDifferentLength-8   17.0ns ± 0%    10.5ns ± 0%   -38.24%  (p=0.008 n=5+5)
CompareBytesBigUnaligned-8      1.58ms ± 0%    0.19ms ± 0%   -88.31%  (p=0.016 n=4+5)
CompareBytesBig-8               1.59ms ± 0%    0.19ms ± 0%   -88.27%  (p=0.016 n=5+4)
CompareBytesBigIdentical-8      7.01ns ± 0%    6.60ns ± 3%    -5.91%  (p=0.008 n=5+5)

name                           old speed     new speed      delta
CompareBytesBigUnaligned-8     662MB/s ± 0%  5660MB/s ± 0%  +755.15%  (p=0.016 n=4+5)
CompareBytesBig-8              661MB/s ± 0%  5636MB/s ± 0%  +752.57%  (p=0.016 n=5+4)
CompareBytesBigIdentical-8     150TB/s ± 0%   159TB/s ± 3%    +6.27%  (p=0.008 n=5+5)

Change-Id: I70332de06f873df3bc12c4a5af1028307b670046
Reviewed-on: https://go-review.googlesource.com/90175
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

cmd/asm: add ARM64 assembler check for incorrect input

Current ARM64 assembler has no check for the invalid value of both
shift amount and post-index immediate offset of LD1/ST1. This patch
adds the check.

This patch also fixes the printing error of register number equals
to 31, which should be printed as ZR instead of R31. Test cases
are also added.

Change-Id: I476235f3ab3a3fc91fe89c5a3149a4d4529c05c7
Reviewed-on: https://go-review.googlesource.com/100255
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>

Revert "doc: remove non-prime from list of primes in spec"

This reverts commit 4b06d9d72747b0407e2cd9a423b689c7e9c8e92d.

Reason for revert: It's a reference to a legendary article
from the Journal of Irreproducible Results.

Updates golang/go#24451

Change-Id: I0288177f4e286bd6ace5774f2e5e0acb02370305
Reviewed-on: https://go-review.googlesource.com/101495
Reviewed-by: Andrew Bonventre <andybons@golang.org>

math/big: reduce amount of copying in Montgomery multiplication

Instead shifting the accumulator every iteration of the loop, shift
once in the end. This significantly improves performance on arm64.

On arm64:

name                  old time/op    new time/op    delta
RSA2048Decrypt          3.33ms ± 0%    2.63ms ± 0%  -20.94%  (p=0.000 n=11+11)
RSA2048Sign             4.22ms ± 0%    3.55ms ± 0%  -15.89%  (p=0.000 n=11+11)
3PrimeRSA2048Decrypt    1.95ms ± 0%    1.59ms ± 0%  -18.59%  (p=0.000 n=11+11)

On Skylake:

name                    old time/op  new time/op  delta
RSA2048Decrypt-8        1.73ms ± 2%  1.55ms ± 2%  -10.19%  (p=0.000 n=10+10)
RSA2048Sign-8           2.17ms ± 2%  2.00ms ± 2%   -7.93%  (p=0.000 n=10+10)
3PrimeRSA2048Decrypt-8  1.10ms ± 2%  0.96ms ± 2%  -13.03%  (p=0.000 n=10+9)

Change-Id: I5786191a1a09e4217fdb1acfd90880d35c5855f7
Reviewed-on: https://go-review.googlesource.com/99838
Run-TryBot: Vlad Krasnov <vlad@cloudflare.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

go/constant: add examples for working with constant.Value

Updates #24352.

Change-Id: Ie7c1de905154a483b7f7748de28e35b484cce6ea
Reviewed-on: https://go-review.googlesource.com/101279
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

doc: remove non-prime from list of primes in spec

Fixes golang/go#24451

Change-Id: Id9b4cbd1a1ff032f1cc4606e9734ddcc64892ae5
Reviewed-on: https://go-review.googlesource.com/101457
Reviewed-by: Filippo Valsorda <filippo@golang.org>

math/big: add 0 shift fastpath to shl and shr

One could expect calls like

  z.mant.shl(z.mant, shiftAmount)

(or higher-level-functions calls that use lhs/rhs) to be almost free
when shiftAmount = 0; and expect calls like

  z.mant.shl(x.mant, 0)

to have the same cost of a x.mant -> z.mant copy. Neither of this
things are currently true.

For an 800 words nat, the first kind of calls cost ~800ns for rigth
shifts and ~3.5µs for left shift; while the second kind of calls are
doing more work than necessary by calling shlVU/shrVU.

This change makes the first kind of calls ({Shl,Shr}Same) almost free,
and the second kind of calls ({Shl,Shr}) about 30% faster.

name                  old time/op  new time/op  delta
ZeroShifts/Shl-4      3.64µs ± 3%  2.49µs ± 1%  -31.55%  (p=0.000 n=10+10)
ZeroShifts/ShlSame-4  3.65µs ± 1%  0.01µs ± 1%  -99.85%  (p=0.000 n=9+9)
ZeroShifts/Shr-4      3.65µs ± 1%  2.49µs ± 1%  -31.91%  (p=0.000 n=10+10)
ZeroShifts/ShrSame-4   825ns ± 0%     6ns ± 1%  -99.33%  (p=0.000 n=9+10)

During go test math/big, the shl zeroshift fastpath is triggered 1380
times; while the shr fastpath is triggered 153334 times(!).

Change-Id: I5f92b304a40638bd8453a86c87c58e54b337bcdf
Reviewed-on: https://go-review.googlesource.com/87660
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

go/format: move example to external test file

Per #11257 all examples should be in external test files.

Additionally, doing so makes this example playable.

Updates #24352. (Albeit tangentially).

Change-Id: I77ab4655107f61db2e9d21a608b73ace3a230fb2
Reviewed-on: https://go-review.googlesource.com/101285
Reviewed-by: Robert Griesemer <gri@golang.org>

net: treat a nil *Resolver as a zero one, as documented

Add accessors that handle nil without crashing.

Fixes #24330

Change-Id: If5fbbb6015ca8d65f620a06bad6e52de8cd896ad
Reviewed-on: https://go-review.googlesource.com/101315
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>

test/codegen: port maps test to codegen

And delete them from asm_test.

Change-Id: I3cf0934706a640136cb0f646509174f8c1bf3363
Reviewed-on: https://go-review.googlesource.com/101395
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>

sort: fix typo, was a mixup between identifiers 'unsorted' and 'data'

Change-Id: If9ad8ae663f007efe43cc35631713565fa754e93
Reviewed-on: https://go-review.googlesource.com/93237
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

strconv: make code formatting more consistent in doc.go

Replaces " \t" code indentation with "\t".

Issues like this are easy to spot with editor that prints
whitespace charecters.

Change-Id: Ia82877e7c99121bf369fa76e46ba52dff84f36bf
Reviewed-on: https://go-review.googlesource.com/101355
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

test/codegen: port structs test to codegen

And delete them from asm_test.

Change-Id: Ia286239a3d8f3915f2ca25dbcb39f3354a4f8aea
Reviewed-on: https://go-review.googlesource.com/101138
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

vendor: update golang.org/x/net/dns/dnsmessage from upstream

Updates to x/net git rev 24dd378 for CL 100055

Fixes #10622
Updates #16218

Change-Id: I99e26da7b908b36585a0379d9381030c01819b54
Reviewed-on: https://go-review.googlesource.com/101278
Run-TryBot: Ian Gudger <igudger@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

crypto/x509: don't require C99 mode in Darwin cgo code

Fixes #24425

Change-Id: I2aacbced8cd14da67fe9a4cbd62b434c18b5fce2
Reviewed-on: https://go-review.googlesource.com/101215
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>

cmd/go: remove some unused parameters

Change-Id: I441b3045e76afc1c561914926c14efc8a116c8a7
Reviewed-on: https://go-review.googlesource.com/101195
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/compile: enable scopes unconditionally

This revives Alessandro Arzilli's CL to enable scopes
whenever any dwarf is emitted (with optimization or not),
adds a test that detects this changes and shows that it
creates more truthful debugging output.

Reverted change to ssa/debug_test tests made when
scopes were disabled during dwarflocationlist development.

Also included are updates to the Delve test output (it
had fallen out of sync; creating test output for one
updates it for all) and minor naming changes in
ssa/debug_test.

Compile-time/space changes (relative to tip including dwarflocationlists):

benchstat -geomean after.log scopes.log
name        old time/op     new time/op     delta
Template        182ms ± 1%      182ms ± 1%    ~     (p=0.666 n=9+9)
Unicode        82.8ms ± 1%     86.6ms ±14%    ~     (p=0.211 n=9+10)
GoTypes         611ms ± 1%      616ms ± 2%  +0.97%  (p=0.001 n=10+9)
Compiler        2.95s ± 1%      2.95s ± 0%    ~     (p=0.573 n=10+8)
SSA             6.70s ± 1%      6.81s ± 1%  +1.68%  (p=0.000 n=9+10)
Flate           117ms ± 1%      118ms ± 1%  +0.60%  (p=0.036 n=9+8)
GoParser        145ms ± 1%      145ms ± 1%    ~     (p=1.000 n=9+9)
Reflect         398ms ± 1%      396ms ± 1%    ~     (p=0.053 n=9+10)
Tar             171ms ± 1%      171ms ± 1%    ~     (p=0.356 n=9+10)
XML             214ms ± 1%      214ms ± 1%    ~     (p=0.605 n=9+9)
StdCmd          12.4s ± 2%      12.4s ± 1%    ~     (p=1.000 n=9+9)
[Geo mean]      506ms           509ms       +0.71%

name        old user-ns/op  new user-ns/op  delta
Template         254M ± 4%       249M ± 6%    ~     (p=0.155 n=10+10)
Unicode          121M ±11%       124M ± 6%    ~     (p=0.516 n=10+10)
GoTypes          824M ± 2%       869M ± 5%  +5.49%  (p=0.001 n=8+10)
Compiler        4.01G ± 2%      4.02G ± 1%    ~     (p=0.561 n=9+9)
SSA             10.0G ± 2%      10.2G ± 2%  +2.29%  (p=0.000 n=9+10)
Flate            154M ± 7%       154M ± 7%    ~     (p=0.960 n=10+9)
GoParser         190M ± 7%       196M ± 6%    ~     (p=0.064 n=9+10)
Reflect          528M ± 2%       517M ± 3%  -1.97%  (p=0.025 n=10+10)
Tar              227M ± 5%       232M ± 3%    ~     (p=0.061 n=9+10)
XML              286M ± 4%       283M ± 4%    ~     (p=0.343 n=9+9)
[Geo mean]       502M            508M       +1.09%

name        old text-bytes  new text-bytes  delta
HelloSize        672k ± 0%       672k ± 0%  +0.01%  (p=0.000 n=10+10)
CmdGoSize       7.21M ± 0%      7.21M ± 0%  -0.00%  (p=0.000 n=10+10)
[Geo mean]      2.20M           2.20M       +0.00%

name        old data-bytes  new data-bytes  delta
HelloSize       9.88k ± 0%      9.88k ± 0%    ~     (all equal)
CmdGoSize        248k ± 0%       248k ± 0%    ~     (all equal)
[Geo mean]      49.5k           49.5k       +0.00%

name        old bss-bytes   new bss-bytes   delta
HelloSize        125k ± 0%       125k ± 0%    ~     (all equal)
CmdGoSize        144k ± 0%       144k ± 0%  -0.04%  (p=0.000 n=10+10)
[Geo mean]       135k            135k       -0.02%

name        old exe-bytes   new exe-bytes   delta
HelloSize       1.30M ± 0%      1.34M ± 0%  +3.15%  (p=0.000 n=10+10)
CmdGoSize       13.5M ± 0%      13.9M ± 0%  +2.70%  (p=0.000 n=10+10)
[Geo mean]      4.19M           4.31M       +2.92%

Change-Id: Id53b8d57bd00440142ccbd39b95710e14e083fb5
Reviewed-on: https://go-review.googlesource.com/101217
Reviewed-by: Heschi Kreinick <heschi@google.com>

net: don't let cancelation of a DNS lookup affect another lookup

Updates #8602
Updates #20703
Fixes #22724

Change-Id: I27b72311b2c66148c59977361bd3f5101e47b51d
Reviewed-on: https://go-review.googlesource.com/100840
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

net: make Resolver.PreferGo work more as documented

Fixes #24393

Change-Id: I8bcee34cdf30472663d866ed6056301d8445215c
Reviewed-on: https://go-review.googlesource.com/100875
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

reflect: sort exported methods first

By moving exported methods to the front of method lists, filtering
down to only the exported methods just needs a count of how many
exported methods exist, which the compiler can statically
provide. This allows getting rid of the exported method cache.

For #22075.

Change-Id: I8eeb274563a2940e1347c34d673f843ae2569064
Reviewed-on: https://go-review.googlesource.com/100846
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/compile: sort method sets earlier

By sorting method sets earlier, we can change the interface
satisfaction problem from taking O(NM) time to O(N+M). This is the
same algorithm already used by runtime and reflect for dynamic
interface satisfaction testing.

For #22075.

Change-Id: I3d889f0227f37704535739bbde11f5107b4eea17
Reviewed-on: https://go-review.googlesource.com/100845
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

crypto/x509: clarify accepted keys for MarshalPKCS8PrivateKey

Fixes #24413.

Change-Id: I265088c9ddc624cb3b3132087cc3d4baf95d2777
Reviewed-on: https://go-review.googlesource.com/100839
Reviewed-by: Filippo Valsorda <filippo@golang.org>

cmd/compile: turn on DWARF locations lists for ssa vars

This changes the default setting for -dwarflocationlists
from false to true, removes the flag from ssa/debug_test.go,
and updates runtime/runtime-gdb_test.go to match a change
in debugging output for composite variables.

Current benchmarks (perflock, -count 10)

benchstat -geomean before.log after.log
name        old time/op     new time/op     delta
Template        175ms ± 0%      182ms ± 1%   +3.68%  (p=0.000 n=8+9)
Unicode        82.0ms ± 2%     82.8ms ± 1%   +0.96%  (p=0.019 n=9+9)
GoTypes         590ms ± 1%      611ms ± 1%   +3.42%  (p=0.000 n=9+10)
Compiler        2.85s ± 0%      2.95s ± 1%   +3.60%  (p=0.000 n=9+10)
SSA             6.42s ± 1%      6.70s ± 1%   +4.31%  (p=0.000 n=10+9)
Flate           113ms ± 2%      117ms ± 1%   +3.11%  (p=0.000 n=10+9)
GoParser        140ms ± 1%      145ms ± 1%   +3.47%  (p=0.000 n=10+9)
Reflect         384ms ± 0%      398ms ± 1%   +3.56%  (p=0.000 n=8+9)
Tar             165ms ± 1%      171ms ± 1%   +3.33%  (p=0.000 n=9+9)
XML             207ms ± 2%      214ms ± 1%   +3.41%  (p=0.000 n=9+9)
StdCmd          11.8s ± 2%      12.4s ± 2%   +4.41%  (p=0.000 n=10+9)
[Geo mean]      489ms           506ms        +3.38%

name        old user-ns/op  new user-ns/op  delta
Template         247M ± 4%       254M ± 4%   +2.76%  (p=0.040 n=10+10)
Unicode          118M ±16%       121M ±11%     ~     (p=0.364 n=10+10)
GoTypes          805M ± 2%       824M ± 2%   +2.37%  (p=0.003 n=9+8)
Compiler        3.92G ± 2%      4.01G ± 2%   +2.20%  (p=0.001 n=9+9)
SSA             9.63G ± 4%     10.00G ± 2%   +3.81%  (p=0.000 n=10+9)
Flate            155M ±10%       154M ± 7%     ~     (p=0.718 n=9+10)
GoParser         184M ±11%       190M ± 7%     ~     (p=0.220 n=10+9)
Reflect          506M ± 4%       528M ± 2%   +4.27%  (p=0.000 n=10+10)
Tar              224M ± 4%       227M ± 5%     ~     (p=0.207 n=10+9)
XML              272M ± 7%       286M ± 4%   +5.23%  (p=0.010 n=10+9)
[Geo mean]       489M            502M        +2.76%

name        old text-bytes  new text-bytes  delta
HelloSize        672k ± 0%       672k ± 0%     ~     (all equal)
CmdGoSize       7.21M ± 0%      7.21M ± 0%     ~     (all equal)
[Geo mean]      2.20M           2.20M        +0.00%

name        old data-bytes  new data-bytes  delta
HelloSize       9.88k ± 0%      9.88k ± 0%     ~     (all equal)
CmdGoSize        248k ± 0%       248k ± 0%     ~     (all equal)
[Geo mean]      49.5k           49.5k        +0.00%

name        old bss-bytes   new bss-bytes   delta
HelloSize        125k ± 0%       125k ± 0%     ~     (all equal)
CmdGoSize        144k ± 0%       144k ± 0%     ~     (all equal)
[Geo mean]       135k            135k        +0.00%

name        old exe-bytes   new exe-bytes   delta
HelloSize       1.10M ± 0%      1.30M ± 0%  +17.82%  (p=0.000 n=10+10)
CmdGoSize       11.6M ± 0%      13.5M ± 0%  +16.90%  (p=0.000 n=10+10)
[Geo mean]      3.57M           4.19M       +17.36%

Change-Id: I250055813cadd25cebee8da1f9a7f995a6eae432
Reviewed-on: https://go-review.googlesource.com/100738
Reviewed-by: Heschi Kreinick <heschi@google.com>

cmd/trace: filter tasks by log text

Add a search box to the top of the user task views that only displays
tasks containing a particular log message.

Change-Id: I92f4aa113f930954e8811416901e37824f0eb884
Reviewed-on: https://go-review.googlesource.com/100843
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>

internal/trace: fix GC time computation of short goroutines

Goroutine analysis reports the sum of all overlapping GC intervals as
the GCTime of a goroutine. The computation is done by adding the length
of a completed GC interval to 'active' goroutines when processing the
corresponding EvGCDone event. This change fixes the two corner cases
the current implementation ignores:

1) Goroutine that ends during GC. Previously, this goroutine was ignored
and GC time was undercounted. We handle this case by setting the
gcStartTime only when GC is active and handling non-zero gcStartTime
when processing EvGoStop and EvGoStart.

2) Goroutine that starts during GC. Previously, the entire GC interval
length was added to the Goroutine's GCTime which resulted in overcount
of GC time. We handle this case by computing the length of overlapped
period precisely.

Change-Id: Ifa8e82672ec341b5ff87837209f4311fa7262b7f
Reviewed-on: https://go-review.googlesource.com/100842
Reviewed-by: Heschi Kreinick <heschi@google.com>
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

test/codegen: port floats tests to codegen

And delete them from asm_test.

Change-Id: Ibdaca3496eefc73c731b511ddb9636a1f3dff68c
Reviewed-on: https://go-review.googlesource.com/100915
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

go/scanner: report errors for incorrect line directives

Based on decision for #24183. This makes the go/scanner behavior
match cmd/compile behavior. Adjusted a go/printer test that assumed
silent behavior for invalid line directive, and added more scanner
tests verifying the correct error position and message for invalid
line directives.

The filenames in line directives now remain untouched by the scanner;
there is no cleanup or conversion of relative into absolute paths
anymore, in sync with what the compiler's scanner/parser are doing.
Any kind of filename transformation has to be done by a client. This
makes the scanner code simpler and also more predictable.

For #24183.

Change-Id: Ia091548e1d3d89dfdf6e7d82dab50bea05742ce3
Reviewed-on: https://go-review.googlesource.com/100235
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

runtime: identify special functions by flag instead of address

When there are plugins, there may not be a unique copy of runtime
functions like goexit, mcall, etc.  So identifying them by entry
address is problematic.  Instead, keep track of each special function
using a field in the symbol table.  That way, multiple copies of
the same runtime function will be treated identically.

Fixes #24351
Fixes #23133

Change-Id: Iea3232df8a6af68509769d9ca618f530cc0f84fd
Reviewed-on: https://go-review.googlesource.com/100739
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/compile: cache sparse maps across ssa passes

This is done for sparse sets already, but it was missing for sparse
maps. Only affects deadstore and regalloc, as they're the only ones that
use sparse maps.

name                 old time/op    new time/op    delta
DSEPass-4               247µs ± 0%     216µs ± 0%  -12.75%  (p=0.008 n=5+5)
DSEPassBlock-4         3.05ms ± 1%    2.87ms ± 1%   -6.02%  (p=0.002 n=6+6)
CSEPass-4              2.30ms ± 0%    2.32ms ± 0%   +0.53%  (p=0.026 n=6+6)
CSEPassBlock-4         23.8ms ± 0%    23.8ms ± 0%     ~     (p=0.931 n=6+5)
DeadcodePass-4         51.7µs ± 1%    51.5µs ± 2%     ~     (p=0.429 n=5+6)
DeadcodePassBlock-4     734µs ± 1%     742µs ± 3%     ~     (p=0.394 n=6+6)
MultiPass-4             152µs ± 0%     149µs ± 2%     ~     (p=0.082 n=5+6)
MultiPassBlock-4       2.67ms ± 1%    2.41ms ± 2%   -9.77%  (p=0.008 n=5+5)

name                 old alloc/op   new alloc/op   delta
DSEPass-4              41.2kB ± 0%     0.1kB ± 0%  -99.68%  (p=0.002 n=6+6)
DSEPassBlock-4          560kB ± 0%       4kB ± 0%  -99.34%  (p=0.026 n=5+6)
CSEPass-4               189kB ± 0%     189kB ± 0%     ~     (all equal)
CSEPassBlock-4         3.10MB ± 0%    3.10MB ± 0%     ~     (p=0.444 n=5+5)
DeadcodePass-4         10.5kB ± 0%    10.5kB ± 0%     ~     (all equal)
DeadcodePassBlock-4     164kB ± 0%     164kB ± 0%     ~     (all equal)
MultiPass-4             240kB ± 0%     199kB ± 0%  -17.06%  (p=0.002 n=6+6)
MultiPassBlock-4       3.60MB ± 0%    2.99MB ± 0%  -17.06%  (p=0.002 n=6+6)

name                 old allocs/op  new allocs/op  delta
DSEPass-4                8.00 ± 0%      4.00 ± 0%  -50.00%  (p=0.002 n=6+6)
DSEPassBlock-4            240 ± 0%       120 ± 0%  -50.00%  (p=0.002 n=6+6)
CSEPass-4                9.00 ± 0%      9.00 ± 0%     ~     (all equal)
CSEPassBlock-4          1.35k ± 0%     1.35k ± 0%     ~     (all equal)
DeadcodePass-4           3.00 ± 0%      3.00 ± 0%     ~     (all equal)
DeadcodePassBlock-4      9.00 ± 0%      9.00 ± 0%     ~     (all equal)
MultiPass-4              11.0 ± 0%      10.0 ± 0%   -9.09%  (p=0.002 n=6+6)
MultiPassBlock-4          165 ± 0%       150 ± 0%   -9.09%  (p=0.002 n=6+6)

Change-Id: I43860687c88f33605eb1415f36473c5cfe8fde4a
Reviewed-on: https://go-review.googlesource.com/98449
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

cmd/compile: implement CMOV on amd64

This builds upon the branchelim pass, activating it for amd64 and
lowering CondSelect. Special care is made to FPU instructions for
NaN handling.

Benchmark results on Xeon E5630 (Westmere EP):

name                      old time/op    new time/op    delta
BinaryTree17-16              4.99s ± 9%     4.66s ± 2%     ~     (p=0.095 n=5+5)
Fannkuch11-16                4.93s ± 3%     5.04s ± 2%     ~     (p=0.548 n=5+5)
FmtFprintfEmpty-16          58.8ns ± 7%    61.4ns ±14%     ~     (p=0.579 n=5+5)
FmtFprintfString-16          114ns ± 2%     114ns ± 4%     ~     (p=0.603 n=5+5)
FmtFprintfInt-16             181ns ± 4%     125ns ± 3%  -30.90%  (p=0.008 n=5+5)
FmtFprintfIntInt-16          263ns ± 2%     217ns ± 2%  -17.34%  (p=0.008 n=5+5)
FmtFprintfPrefixedInt-16     230ns ± 1%     212ns ± 1%   -7.99%  (p=0.008 n=5+5)
FmtFprintfFloat-16           411ns ± 3%     344ns ± 5%  -16.43%  (p=0.008 n=5+5)
FmtManyArgs-16               828ns ± 4%     790ns ± 2%   -4.59%  (p=0.032 n=5+5)
GobDecode-16                10.9ms ± 4%    10.8ms ± 5%     ~     (p=0.548 n=5+5)
GobEncode-16                9.52ms ± 5%    9.46ms ± 2%     ~     (p=1.000 n=5+5)
Gzip-16                      334ms ± 2%     337ms ± 2%     ~     (p=0.548 n=5+5)
Gunzip-16                   64.4ms ± 1%    65.0ms ± 1%   +1.00%  (p=0.008 n=5+5)
HTTPClientServer-16          156µs ± 3%     155µs ± 3%     ~     (p=0.690 n=5+5)
JSONEncode-16               21.0ms ± 1%    21.8ms ± 0%   +3.76%  (p=0.016 n=5+4)
JSONDecode-16               95.1ms ± 0%    95.7ms ± 1%     ~     (p=0.151 n=5+5)
Mandelbrot200-16            6.38ms ± 1%    6.42ms ± 1%     ~     (p=0.095 n=5+5)
GoParse-16                  5.47ms ± 2%    5.36ms ± 1%   -1.95%  (p=0.016 n=5+5)
RegexpMatchEasy0_32-16       111ns ± 1%     111ns ± 1%     ~     (p=0.635 n=5+4)
RegexpMatchEasy0_1K-16       408ns ± 1%     411ns ± 2%     ~     (p=0.087 n=5+5)
RegexpMatchEasy1_32-16       103ns ± 1%     104ns ± 1%     ~     (p=0.484 n=5+5)
RegexpMatchEasy1_1K-16       659ns ± 2%     652ns ± 1%     ~     (p=0.571 n=5+5)
RegexpMatchMedium_32-16      176ns ± 2%     174ns ± 1%     ~     (p=0.476 n=5+5)
RegexpMatchMedium_1K-16     58.6µs ± 4%    57.7µs ± 4%     ~     (p=0.548 n=5+5)
RegexpMatchHard_32-16       3.07µs ± 3%    3.04µs ± 4%     ~     (p=0.421 n=5+5)
RegexpMatchHard_1K-16       89.2µs ± 1%    87.9µs ± 2%   -1.52%  (p=0.032 n=5+5)
Revcomp-16                   575ms ± 0%     587ms ± 2%   +2.12%  (p=0.032 n=4+5)
Template-16                  110ms ± 1%     107ms ± 3%   -3.00%  (p=0.032 n=5+5)
TimeParse-16                 463ns ± 0%     462ns ± 0%     ~     (p=0.810 n=5+4)
TimeFormat-16                538ns ± 0%     535ns ± 0%   -0.63%  (p=0.024 n=5+5)

name                      old speed      new speed      delta
GobDecode-16              70.7MB/s ± 4%  71.4MB/s ± 5%     ~     (p=0.452 n=5+5)
GobEncode-16              80.7MB/s ± 5%  81.2MB/s ± 2%     ~     (p=1.000 n=5+5)
Gzip-16                   58.2MB/s ± 2%  57.7MB/s ± 2%     ~     (p=0.452 n=5+5)
Gunzip-16                  302MB/s ± 1%   299MB/s ± 1%   -0.99%  (p=0.008 n=5+5)
JSONEncode-16             92.4MB/s ± 1%  89.1MB/s ± 0%   -3.63%  (p=0.016 n=5+4)
JSONDecode-16             20.4MB/s ± 0%  20.3MB/s ± 1%     ~     (p=0.135 n=5+5)
GoParse-16                10.6MB/s ± 2%  10.8MB/s ± 1%   +2.00%  (p=0.016 n=5+5)
RegexpMatchEasy0_32-16     286MB/s ± 1%   285MB/s ± 3%     ~     (p=1.000 n=5+5)
RegexpMatchEasy0_1K-16    2.51GB/s ± 1%  2.49GB/s ± 2%     ~     (p=0.095 n=5+5)
RegexpMatchEasy1_32-16     309MB/s ± 1%   307MB/s ± 1%     ~     (p=0.548 n=5+5)
RegexpMatchEasy1_1K-16    1.55GB/s ± 2%  1.57GB/s ± 1%     ~     (p=0.690 n=5+5)
RegexpMatchMedium_32-16   5.68MB/s ± 2%  5.73MB/s ± 1%     ~     (p=0.579 n=5+5)
RegexpMatchMedium_1K-16   17.5MB/s ± 4%  17.8MB/s ± 4%     ~     (p=0.500 n=5+5)
RegexpMatchHard_32-16     10.4MB/s ± 3%  10.5MB/s ± 4%     ~     (p=0.460 n=5+5)
RegexpMatchHard_1K-16     11.5MB/s ± 1%  11.7MB/s ± 2%   +1.57%  (p=0.032 n=5+5)
Revcomp-16                 442MB/s ± 0%   433MB/s ± 2%   -2.05%  (p=0.032 n=4+5)
Template-16               17.7MB/s ± 1%  18.2MB/s ± 3%   +3.12%  (p=0.032 n=5+5)

Change-Id: I6972e8f35f2b31f9a42ac473a6bf419a18022558
Reviewed-on: https://go-review.googlesource.com/100935
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

cmd/internal/obj/mips: load/store even float registers first

There is a bug in Octeon III processors where storing an odd floating
point register after it has recently been written to by a double
floating point operation will store the old value from before the double
operation (there are some extra details - the operation and store
must be a certain number of cycles apart). However, this bug does not
occur if the even register is stored first. Currently the bug only
happens on big endian because go always loads the even register first on
little endian.

Workaround the bug by always loading / storing the even floating point
register first. Since this is just an instruction reordering, it should
have no performance penalty. This follows other compilers like GCC which
will always store the even register first (although you do have to set
the ISA level to MIPS I to prevent it from using SDC1).

Change-Id: I5e73daa4d724ca1df7bf5228aab19f53f26a4976
Reviewed-on: https://go-review.googlesource.com/97735
Reviewed-by: Keith Randall <khr@golang.org>

cmd/compile/internal/ssa: add patterns for arm64 bitfield opcodes

Add patterns to match common idioms for EXTR, BFI, BFXIL, SBFIZ, SBFX,
UBFIZ and UBFX opcodes.

go1 benchmarks results on Amberwing:
name                   old time/op    new time/op    delta
FmtManyArgs               786ns ± 2%     714ns ± 1%  -9.20%  (p=0.000 n=10+10)
Gzip                      437ms ± 0%     402ms ± 0%  -7.99%  (p=0.000 n=10+10)
FmtFprintfIntInt          196ns ± 0%     182ns ± 0%  -7.28%  (p=0.000 n=10+9)
FmtFprintfPrefixedInt     207ns ± 0%     199ns ± 0%  -3.86%  (p=0.000 n=10+10)
FmtFprintfFloat           324ns ± 0%     316ns ± 0%  -2.47%  (p=0.000 n=10+8)
FmtFprintfInt             119ns ± 0%     117ns ± 0%  -1.68%  (p=0.000 n=10+9)
GobDecode                12.8ms ± 2%    12.6ms ± 1%  -1.62%  (p=0.002 n=10+10)
JSONDecode               94.4ms ± 1%    93.4ms ± 0%  -1.10%  (p=0.000 n=10+10)
RegexpMatchEasy0_32       247ns ± 0%     245ns ± 0%  -0.65%  (p=0.000 n=10+10)
RegexpMatchMedium_32      314ns ± 0%     312ns ± 0%  -0.64%  (p=0.000 n=10+10)
RegexpMatchEasy0_1K       541ns ± 0%     538ns ± 0%  -0.55%  (p=0.000 n=10+9)
TimeParse                 450ns ± 1%     448ns ± 1%  -0.42%  (p=0.035 n=9+9)
RegexpMatchEasy1_32       244ns ± 0%     243ns ± 0%  -0.41%  (p=0.000 n=10+10)
GoParse                  6.03ms ± 0%    6.00ms ± 0%  -0.40%  (p=0.002 n=10+10)
RegexpMatchEasy1_1K       779ns ± 0%     777ns ± 0%  -0.26%  (p=0.000 n=10+10)
RegexpMatchHard_32       2.75µs ± 0%    2.74µs ± 1%  -0.06%  (p=0.026 n=9+9)
BinaryTree17              11.7s ± 0%     11.6s ± 0%    ~     (p=0.089 n=10+10)
HTTPClientServer         89.1µs ± 1%    89.5µs ± 2%    ~     (p=0.436 n=10+10)
RegexpMatchHard_1K       78.9µs ± 0%    79.5µs ± 2%    ~     (p=0.469 n=10+10)
FmtFprintfEmpty          58.5ns ± 0%    58.5ns ± 0%    ~     (all equal)
GobEncode                12.0ms ± 1%    12.1ms ± 0%    ~     (p=0.075 n=10+10)
Revcomp                   669ms ± 0%     668ms ± 0%    ~     (p=0.091 n=7+9)
Mandelbrot200            5.35ms ± 0%    5.36ms ± 0%  +0.07%  (p=0.000 n=9+9)
RegexpMatchMedium_1K     52.1µs ± 0%    52.1µs ± 0%  +0.10%  (p=0.000 n=9+9)
Fannkuch11                3.25s ± 0%     3.26s ± 0%  +0.36%  (p=0.000 n=9+10)
FmtFprintfString          114ns ± 1%     115ns ± 0%  +0.52%  (p=0.011 n=10+10)
JSONEncode               20.2ms ± 0%    20.3ms ± 0%  +0.65%  (p=0.000 n=10+10)
Template                 91.3ms ± 0%    92.3ms ± 0%  +1.08%  (p=0.000 n=10+10)
TimeFormat                484ns ± 0%     495ns ± 1%  +2.30%  (p=0.000 n=9+10)

There are some opportunities to improve this change further by adding
patterns to match the "extended register" versions of ADD/SUB/CMP, but I
think that should be evaluated on its own.  The regressions in Template
and TimeFormat would likely be recovered by this, as they seem to be due
to generating:

    ubfiz x0, x0, #3, #8
    add x1, x2, x0

instead of

    add x1, x2, x0, lsl #3

Change-Id: I5644a8d70ac7a98e784a377a2b76ab47a3415a4b
Reviewed-on: https://go-review.googlesource.com/88355
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

test/codegen: port len/cap pow2 div tests to codegen

And delete them from asm_test.

Change-Id: I29c8d098a8893e6b669b6272a2f508985ac9d618
Reviewed-on: https://go-review.googlesource.com/100876
Reviewed-by: Cherry Zhang <cherryyz@google.com>

syscall: use Android O friendly fstatat syscall to implement Stat on linux/amd64

The Android O seccomp policy disallows the stat syscall on amd64, see
https://android.googlesource.com/platform/bionic/+/android-4.2.2_r1.2/libc/SYSCALLS.TXT

Use the fstatat syscall with AT_FDCWD and zero flags instead to achieve
the same behavior.

Fixes #24403

Change-Id: I36fc9ec9bc938cd8e9de30f66c0eb9d2e24debf6
Reviewed-on: https://go-review.googlesource.com/100878
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

runtime: use Android O friendly faccessat syscall on linux/amd64

The Android O seccomp policy disallows the access syscall on amd64, see
https://android.googlesource.com/platform/bionic/+/android-4.2.2_r1.2/libc/SYSCALLS.TXT

Use the faccessat syscall with AT_FDCWD instead to achieve the same
behavior.

Updates #24403

Change-Id: I9db847c1c0f33987a3479b3f96e721fb9588cde2
Reviewed-on: https://go-review.googlesource.com/100877
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

sync: make WaitGroup more space-efficient

The struct stores its 64-bit state field in a 12-byte array to
ensure that it can be 64-bit-aligned. This leaves 4 spare bytes,
which we can reuse to store the sema field.

(32-bit alignment is still guaranteed because the array type was
changed to [3]uint32.)

Fixes #19149.

Change-Id: I9bc20e69e45e0e07fbf496080f3650e8be0d6e8d
Reviewed-on: https://go-review.googlesource.com/100515
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

net: use golang.org/x/net/dns/dnsmessage for DNS resolution

Vendors golang.org/x/net/dns/dnsmessage from x/net git rev
892bf7b0c6e2f93b51166bf3882e50277fa5afc6

Updates #16218
Updates #21160

Change-Id: Ic4e8f3c3d83c2936354ec14c5be93b0d2b42dd91
Reviewed-on: https://go-review.googlesource.com/37879
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

runtime: fix another typo in runtime-gdb.py

tuple, touple,
gdb, gdv,
let's call the whole thing off.

Change-Id: I72d12f6c75061777474e7dec2c90d2a8a3715da6
Reviewed-on: https://go-review.googlesource.com/100836
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/compile: extract common noding code from func{Decl,Lit}

Passes toolstash-check.

Change-Id: I8290221d6169e077dfa4ea737d685c7fcecf6841
Reviewed-on: https://go-review.googlesource.com/100835
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

cmd/compile: fix duplicate code generation in swt.go

When combining adjacent type switch cases with the same type hash, we
failed to actually remove the combined cases, so we would generate
code for them twice.

We use MD5 for type hashes, so collisions are rare, but they do
currently appear in test/fixedbugs/bug248.dir/bug2.go, which is how I
noticed this failure.

Passes toolstash-check.

Change-Id: I66729b3366b96cb8ddc8fa6f3ebea11ef6d74012
Reviewed-on: https://go-review.googlesource.com/100461
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

runtime: print goid when throwing for split stack overflow

Change-Id: I66515156c2fc6886312c0eccb86d7ceaf7947042
Reviewed-on: https://go-review.googlesource.com/100465
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

runtime: refactor gdb PC parsing

Change-Id: I91607edaf9c256e6723eb3d6e18c8210eb86b704
Reviewed-on: https://go-review.googlesource.com/100464
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

cmd/compile: cleanup closure.go

The main thing is we now eagerly create the ODCLFUNC node for
closures, immediately cross-link them, and assign fields (e.g., Nbody,
Dcl, Parents, Marks) directly on the ODCLFUNC (previously they were
assigned on the OCLOSURE and later moved to the ODCLFUNC).

This allows us to set Curfn to the ODCLFUNC instead of the OCLOSURE,
which makes things more consistent with normal function declarations.
(Notably, this means Cvars now hang off the ODCLFUNC instead of the
OCLOSURE.)

Assignment of xfunc symbol names also now happens before typechecking
their body, which means debugging output now provides a more helpful
name than "<S>".

In golang.org/cl/66810, we changed "x := y" statements to avoid
creating false closure variables for x, but we still create them for
struct literals like "s{f: x}". Update comment in capturevars
accordingly.

More opportunity for cleanups still, but this makes some substantial
progress, IMO.

Passes toolstash-check.

Change-Id: I65a4efc91886e3dcd1000561348af88297775cd7
Reviewed-on: https://go-review.googlesource.com/100197
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

Revert "cmd/compile: implement CMOV on amd64"

This reverts commit 080187f4f72bd6594e3c2efc35cf51bf61378552.

It broke build of golang.org/x/exp/shiny/iconvg
See issue 24395 for details

Change-Id: Ifd6134f6214e6cee40bd3c63c32941d5fc96ae8b
Reviewed-on: https://go-review.googlesource.com/100755
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

cmd/compile/internal/ssa: track stack-only vars

User variables that cannot be SSA'd, either because their addresses are
taken or because they are too large for the decomposition heuristic, do
not explicitly appear as operands of SSA values. Instead they are written
to directly via the stack pointer.

This hid them from the location list generation, which is only
interested in the named value table. Fortunately, the lifetime of
stack-only variables is delineated by VarDef/VarKill ops, and it's easy
enough to turn those into location list bounds.

One wrinkle: stack frame information is not explicitly available in the
SSA phases, because it's owned by the frontend in AllocFrame. It would
be easier if the set of live LocalSlots were returned by that, but this
is the minimal change to fix missing variables. Or VarDef/VarKills
could appear in NamedValues, which would make this change even easier.

Change-Id: Ice6654dad6f9babb0286e95c7ec28594561dc91f
Reviewed-on: https://go-review.googlesource.com/100458
Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

cmd/compile: improve PPC64.rules to reduce size of rewritePPC64.go

Some rules in PPC64.rules cause an extremely large rewritePPC64.go
file to be generated, due to rules with commutative operations and
many operands. This happens with the existing
rules for combining byte loads in little endian order, and
also happens with the pending change to do the same for bytes
in big endian order.

The change improves the existing rules and reduces the size of
the rewrite file by more than 60%. Once this change is merged,
then the pending change for big endian ordered rules will be
updated to use rules that avoid generating an excessively large
rewrite file.

This also includes a fix to a performance regression for
littleEndian.PutUint16 on ppc64le.

Change-Id: I8d2ea42885fa2b84b30c63aa124b0a9b130564ff
Reviewed-on: https://go-review.googlesource.com/100675
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

math/big: add comment about internal assumptions on nat values

Change-Id: I7ed40507a019c0bf521ba748fc22c03d74bb17b7
Reviewed-on: https://go-review.googlesource.com/100719
Reviewed-by: Ian Lance Taylor <iant@golang.org>

runtime: improve arm64 memclr implementation

Improve runtime memclr_arm64.s using ZVA feature to zero out memory when n
is at least 64 bytes.

Also add DCZID_EL0 system register to use in MRS instruction.

    Benchmark results of runtime/Memclr on Amberwing:
name          old time/op    new time/op    delta
Memclr/5        12.7ns ± 0%    12.7ns ± 0%      ~     (all equal)
Memclr/16       12.7ns ± 0%    12.2ns ± 1%    -4.13%  (p=0.000 n=7+8)
Memclr/64       14.0ns ± 0%    14.6ns ± 1%    +4.29%  (p=0.000 n=7+8)
Memclr/256      23.7ns ± 0%    25.7ns ± 0%    +8.44%  (p=0.000 n=8+7)
Memclr/4096      204ns ± 0%      74ns ± 0%   -63.71%  (p=0.000 n=8+8)
Memclr/65536    2.89µs ± 0%    0.84µs ± 0%   -70.91%  (p=0.000 n=8+8)
Memclr/1M       45.9µs ± 0%    17.0µs ± 0%   -62.88%  (p=0.000 n=8+8)
Memclr/4M        184µs ± 0%      77µs ± 4%   -57.94%  (p=0.001 n=6+8)
Memclr/8M        367µs ± 0%     144µs ± 1%   -60.72%  (p=0.000 n=7+8)
Memclr/16M       734µs ± 0%     293µs ± 1%   -60.09%  (p=0.000 n=8+8)
Memclr/64M      2.94ms ± 0%    1.23ms ± 0%   -58.06%  (p=0.000 n=7+8)
GoMemclr/5      8.00ns ± 0%    8.79ns ± 0%    +9.83%  (p=0.000 n=8+8)
GoMemclr/16     8.00ns ± 0%    7.60ns ± 0%    -5.00%  (p=0.000 n=8+8)
GoMemclr/64     10.8ns ± 0%    10.4ns ± 0%    -3.70%  (p=0.000 n=8+8)
GoMemclr/256    20.4ns ± 0%    21.2ns ± 0%    +3.92%  (p=0.000 n=8+8)

name          old speed      new speed      delta
Memclr/5       394MB/s ± 0%   393MB/s ± 0%    -0.28%  (p=0.006 n=8+8)
Memclr/16     1.26GB/s ± 0%  1.31GB/s ± 1%    +4.07%  (p=0.000 n=7+8)
Memclr/64     4.57GB/s ± 0%  4.39GB/s ± 2%    -3.91%  (p=0.000 n=7+8)
Memclr/256    10.8GB/s ± 0%  10.0GB/s ± 0%    -7.95%  (p=0.001 n=7+6)
Memclr/4096   20.1GB/s ± 0%  55.3GB/s ± 0%  +175.46%  (p=0.000 n=8+8)
Memclr/65536  22.6GB/s ± 0%  77.8GB/s ± 0%  +243.63%  (p=0.000 n=7+8)
Memclr/1M     22.8GB/s ± 0%  61.5GB/s ± 0%  +169.38%  (p=0.000 n=8+8)
Memclr/4M     22.8GB/s ± 0%  54.3GB/s ± 4%  +137.85%  (p=0.001 n=6+8)
Memclr/8M     22.8GB/s ± 0%  58.1GB/s ± 1%  +154.56%  (p=0.000 n=7+8)
Memclr/16M    22.8GB/s ± 0%  57.2GB/s ± 1%  +150.54%  (p=0.000 n=8+8)
Memclr/64M    22.8GB/s ± 0%  54.4GB/s ± 0%  +138.42%  (p=0.000 n=7+8)
GoMemclr/5     625MB/s ± 0%   569MB/s ± 0%    -8.90%  (p=0.000 n=7+8)
GoMemclr/16   2.00GB/s ± 0%  2.10GB/s ± 0%    +5.26%  (p=0.000 n=8+8)
GoMemclr/64   5.92GB/s ± 0%  6.15GB/s ± 0%    +3.83%  (p=0.000 n=7+8)
GoMemclr/256  12.5GB/s ± 0%  12.1GB/s ± 0%    -3.77%  (p=0.000 n=8+7)

    Benchmark results of runtime/Memclr on Amberwing without ZVA:
name          old time/op    new time/op    delta
Memclr/5        12.7ns ± 0%    12.8ns ± 0%   +0.79%  (p=0.008 n=5+5)
Memclr/16       12.7ns ± 0%    12.7ns ± 0%     ~     (p=0.444 n=5+5)
Memclr/64       14.0ns ± 0%    14.4ns ± 0%   +2.86%  (p=0.008 n=5+5)
Memclr/256      23.7ns ± 1%    19.2ns ± 0%  -19.06%  (p=0.008 n=5+5)
Memclr/4096      203ns ± 0%     119ns ± 0%  -41.38%  (p=0.008 n=5+5)
Memclr/65536    2.89µs ± 0%    1.66µs ± 0%  -42.76%  (p=0.008 n=5+5)
Memclr/1M       45.9µs ± 0%    26.2µs ± 0%  -42.82%  (p=0.008 n=5+5)
Memclr/4M        184µs ± 0%     105µs ± 0%  -42.81%  (p=0.008 n=5+5)
Memclr/8M        367µs ± 0%     210µs ± 0%  -42.76%  (p=0.008 n=5+5)
Memclr/16M       734µs ± 0%     420µs ± 0%  -42.74%  (p=0.008 n=5+5)
Memclr/64M      2.94ms ± 0%    1.69ms ± 0%  -42.46%  (p=0.008 n=5+5)
GoMemclr/5      8.00ns ± 0%    8.40ns ± 0%   +5.00%  (p=0.008 n=5+5)
GoMemclr/16     8.00ns ± 0%    8.40ns ± 0%   +5.00%  (p=0.008 n=5+5)
GoMemclr/64     10.8ns ± 0%     9.6ns ± 0%  -11.02%  (p=0.008 n=5+5)
GoMemclr/256    20.4ns ± 0%    17.2ns ± 0%  -15.69%  (p=0.008 n=5+5)

name          old speed      new speed      delta
Memclr/5       393MB/s ± 0%   391MB/s ± 0%   -0.64%  (p=0.008 n=5+5)
Memclr/16     1.26GB/s ± 0%  1.26GB/s ± 0%   -0.55%  (p=0.008 n=5+5)
Memclr/64     4.57GB/s ± 0%  4.44GB/s ± 0%   -2.79%  (p=0.008 n=5+5)
Memclr/256    10.8GB/s ± 0%  13.3GB/s ± 0%  +23.07%  (p=0.016 n=4+5)
Memclr/4096   20.1GB/s ± 0%  34.3GB/s ± 0%  +70.91%  (p=0.008 n=5+5)
Memclr/65536  22.7GB/s ± 0%  39.6GB/s ± 0%  +74.65%  (p=0.008 n=5+5)
Memclr/1M     22.8GB/s ± 0%  40.0GB/s ± 0%  +74.88%  (p=0.008 n=5+5)
Memclr/4M     22.8GB/s ± 0%  39.9GB/s ± 0%  +74.84%  (p=0.008 n=5+5)
Memclr/8M     22.9GB/s ± 0%  39.9GB/s ± 0%  +74.71%  (p=0.008 n=5+5)
Memclr/16M    22.9GB/s ± 0%  39.9GB/s ± 0%  +74.64%  (p=0.008 n=5+5)
Memclr/64M    22.8GB/s ± 0%  39.7GB/s ± 0%  +73.79%  (p=0.008 n=5+5)
GoMemclr/5     625MB/s ± 0%   595MB/s ± 0%   -4.77%  (p=0.000 n=4+5)
GoMemclr/16   2.00GB/s ± 0%  1.90GB/s ± 0%   -4.77%  (p=0.008 n=5+5)
GoMemclr/64   5.92GB/s ± 0%  6.66GB/s ± 0%  +12.48%  (p=0.016 n=4+5)
GoMemclr/256  12.5GB/s ± 0%  14.9GB/s ± 0%  +18.95%  (p=0.008 n=5+5)

Fixes #22948

Change-Id: Iaae4e22391e25b54d299821bb7f8a81ac3986b93
Reviewed-on: https://go-review.googlesource.com/82055
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

cmd/compile: document new line directives

Fixes #24183.

Change-Id: I5ef31c4a3aad7e05568b7de1227745d686d4aff8
Reviewed-on: https://go-review.googlesource.com/100462
Reviewed-by: Ian Lance Taylor <iant@golang.org>

runtime, syscall: add RawSyscall6 on Solaris and make it panic

The syscall package currently declares RawSyscall6 for every GOOS, but
does not define it on Solaris. This leads to code using said function
to compile but it will not link. Fix it by adding RawSyscall6 and make
it panic.

Also remove the obsolete comment above runtime.syscall_syscall as
pointed out by Aram.

Updates #24357

Change-Id: I1b1423121d1c99de2ecc61cd9a935dba9b39e3a4
Reviewed-on: https://go-review.googlesource.com/100655
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>

test/codegen: port all small memmove tests to codegen

This change ports all the remaining tests checking that small memmoves
are replaced with MOVs to the new codegen test harness, and deletes
them from the asm_test file.

Change-Id: I01c94b441e27a5d61518035af62d62779dafeb56
Reviewed-on: https://go-review.googlesource.com/100476
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

test/codegen: add codegen tests for div

Change-Id: I6ce8981e85fd55ade6078b0946e54a9215d9deca
Reviewed-on: https://go-review.googlesource.com/100575
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

cmd/asm: move manual tests out of generated file

Thanks to Iskander Sharipov for spotting this in an earlier CL of mine.

Change-Id: Idf45ad266205ff83985367cb38f585badfbed151
Reviewed-on: https://go-review.googlesource.com/100535
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Iskander Sharipov <iskander.sharipov@intel.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

runtime: don't use floating point in findnull on Plan 9

In CL 98015, findnull was rewritten so it uses bytes.IndexByte.

This broke the build on plan9/amd64 because the implementation
of bytes.IndexByte on AMD64 relies on SSE instructions while
floating point instructions are not allowed in the note handler.

This change fixes findnull by using the former implementation
on Plan 9, so it doesn't use bytes.IndexByte.

Fixes #24387.

Change-Id: I084d1a44d38d9f77a6c1ad492773f0a98226be16
Reviewed-on: https://go-review.googlesource.com/100577
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>

test: check that size argument errors are emitted at call site

Add tests for the "negative size argument in make.*" and "size argument
too large in make.*" error messages to appear at call sites in case the
size is a const defined on another line.

As suggested by Matthew in a comment on CL 69910.

Change-Id: I5c33d4bec4e3d20bb21fe8019df27999997ddff3
Reviewed-on: https://go-review.googlesource.com/100395
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

runtime: fix typo in gdb script

Change-Id: I9d4b3e25b00724f0e4870c6082671b4f14cc18fc
Reviewed-on: https://go-review.googlesource.com/100463
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/compile: remove BTQconst rule

This rule is meant for code optimization, but it makes other rules
potentially more complex, as they need to cope with the fact that
a 32-bit op (BTLconst) can appear everywhere a 64-bit rule maches.

Move the optimization to opcode expansion instead. Tests will be
added in following CL.

Change-Id: Ica5ef291e7963c4af17c124d4a2869e6c8f7b0c7
Reviewed-on: https://go-review.googlesource.com/99995
Reviewed-by: Keith Randall <khr@golang.org>

cmd/compile: reject type switch with guarded declaration and no cases

Fixes #23116.

Change-Id: I5db5c5c39bbb50148ffa18c9393b045f255f80a3
Reviewed-on: https://go-review.googlesource.com/100459
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

cmd/compile: use key position for key:val elements in composite literals

Fixes #24339.

Change-Id: Ie47764fed27f76b480834b1fdbed0512c94831d9
Reviewed-on: https://go-review.googlesource.com/100457
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

cmd/compile: deduplicate racewalk switch cases

Only the contiguous ones, to keep the patch simple. Remove some
unnecessary newlines, while at it.

Change-Id: Ia588f80538b49a169fbf49835979ebff5a0a7b6d
Reviewed-on: https://go-review.googlesource.com/94756
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

cmd/asm: VPERMQ's imm8 arg is an uint8

The imm8 argument consists of 4 2-bit indices, so it can take values up
to $255. However, the assembler was treating it as Yi8, which reads
"fits in int8". Add a Yu8 variant, to also keep backwards compatibility
with negative values possible with Yi8.

Fixes #24378.

Change-Id: I24ddb19c219b54d039a6c1bcdb903717d1c7c3b8
Reviewed-on: https://go-review.googlesource.com/100475
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

compress/gzip, compress/zlib: fix Writer documentation inconsistencies

Fixes #24379.

Change-Id: Ibdc763a0c2b56e26f4269f8be429880e34a2558f
Reviewed-on: https://go-review.googlesource.com/100495
Reviewed-by: Joe Tsai <joetsai@google.com>

cmd/vendor: fix JSON format

Change-Id: I9c5a4a4031c12d67c7e75e9a276a766927abf83d
Reviewed-on: https://go-review.googlesource.com/100415
Reviewed-by: Cherry Zhang <cherryyz@google.com>

cmd/compile: enable inlining variadic functions

As a side effect of working on mid-stack inlining, we've fixed support
for inlining variadic functions. Might as well enable it.

Change-Id: I7f555f8b941969791db7eb598c0b49f6dc0820aa
Reviewed-on: https://go-review.googlesource.com/100456
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/compile: eliminate mkinlcall's isddd parameter

These are always set to n.Isddd(), which is readily available within
mkinlcall.

Change-Id: I3d7fbc9dc19a40d6b905691c666eee9bcd031a00
Reviewed-on: https://go-review.googlesource.com/100455
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/internal/obj/ppc64: implement full operand support for l*arx instructions

The current implementation of l*arx instructions does not accept non-zero
offsets in RA nor the EH field. This change adds full functionality to those
instructions.

Updates #23845

Change-Id: If113f70d11de5f35f8389520b049390dbc40e863
Reviewed-on: https://go-review.googlesource.com/99635
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>

cmd/compile: remove some unused parameters

As reported by unparam.

Passes toolstash -cmp on std cmd.

Change-Id: I55473e1eed096ed1c3e431aed2cbf0b6b5444b91
Reviewed-on: https://go-review.googlesource.com/97895
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

cmd/internal/obj/arm64: support logical instructions targeting RSP

Logical instructions can have RSP as its destination. Support it.

Note that the two-operand form, like "AND $1, RSP", which is
equivalent to the three-operand form "AND $1, RSP, RSP", is
invalid, because the source register is not allowed to be RSP.

Also note that instructions that set the conditional flags, like
ANDS, cannot target RSP. Because of this, we split out the optab
entries of AND et al. and ANDS et al.

Merge the optab entries of BIC et al. to AND et al., because they
are same.

Fixes #24332.

Change-Id: I3584d6f2e7cea98a659a1ed9fdf67c353e090637
Reviewed-on: https://go-review.googlesource.com/100217
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

Revert "runtime: convert g.waitreason from string to uint8"

This reverts commit 4eea887fd477368653f6fcf8ad766030167936e5.

Reason for revert: broke s390x build

Change-Id: Id6c2b6a7319273c4d21f613d4cdd38b00d49f847
Reviewed-on: https://go-review.googlesource.com/100375
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>

cmd/internal/obj/arm64: fix branch-too-far with TBZ like instructions

The compiler now emits TBZ like instructions, but the assembler's
too-far-branch patch code didn't include that case. Add it.

Fixes #23889.

Change-Id: Ib75f9250c660b9fb652835fbc83263a5d5073dc5
Reviewed-on: https://go-review.googlesource.com/94902
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>

cmd/compile: fix failure to reset reused bit of storage

This is the "3rd bug" that caused compilations to sometimes
produce different results when dwarf location lists were
enabled.

A loop had not been properly rewritten in an earlier
optimization CL, and it accessed uninitialized data,
which was deterministically perhaps wrong when single
threaded, but variably wrong when multithreaded.

Change-Id: Ib3da538762fdf7d5e4407106f2434f3b14a1d7ea
Reviewed-on: https://go-review.googlesource.com/99935
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>

go/token: add example for retrieving Position from Pos

There are few uses for the majority of the API in go/token for the
average user. The exception to this is getting the filename, line, and
column information from a token.Pos (reported and absolute. This is
straightforward but figuring out how to do it requires combing through
a lot of documentation. This example makes it more easily discoverable.

Updates #24352.

Change-Id: I0a45da6173b3dabebf42484bbbed30d9e5e20e01
Reviewed-on: https://go-review.googlesource.com/100058
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

C: add Filippo Valsorda's @golang.org email

Change-Id: I3758a4350b600af304b2cff7ad59c7368a01ab5c
Reviewed-on: https://go-review.googlesource.com/100215
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

runtime: convert g.waitreason from string to uint8

Every time I poke at #14921, the g.waitreason string
pointer writes show up.

They're not particularly important performance-wise,
but it'd be nice to clear the noise away.

And it does open up a few extra bytes in the g struct
for some future use.

Change-Id: I7ffbd52fbc2a286931a2218038fda52ed6473cc9
Reviewed-on: https://go-review.googlesource.com/99078
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

runtime: simplify range expressions in tests

Generated by running gofmt -s on the files in question.

Change-Id: If6578b150e1bfced8657196d2af01f5d36879f93
Reviewed-on: https://go-review.googlesource.com/100135
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/compile: avoid extra mapaccess in "m[k] op= r"

Currently, order desugars map assignment operations like

    m[k] op= r

into

    m[k] = m[k] op r

which in turn is transformed during walk into:

    tmp := *mapaccess(m, k)
    tmp = tmp op r
    *mapassign(m, k) = tmp

However, this is suboptimal, as we could instead produce just:

    *mapassign(m, k) op= r

One complication though is if "r == 0", then "m[k] /= r" and "m[k] %=
r" will panic, and they need to do so *before* calling mapassign,
otherwise we may insert a new zero-value element into the map.

It would be spec compliant to just emit the "r != 0" check before
calling mapassign (see #23735), but currently these checks aren't
generated until SSA construction. For now, it's simpler to continue
desugaring /= and %= into two map indexing operations.

Fixes #23661.

Change-Id: I46e3739d9adef10e92b46fdd78b88d5aabe68952
Reviewed-on: https://go-review.googlesource.com/91557
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

cmd/compile/internal/ssa: emit IMUL3{L/Q} for MUL{L/Q}const on x86

cmd/asm now supports three-operand form of IMUL,
so instead of using IMUL with resultInArg0, emit IMUL3 instruction.

This results in less redundant MOVs where SSA assigns
different registers to input[0] and dst arguments.

Note: these have exactly the same encoding when reg0=reg1:
      IMUL3x $const, reg0, reg1
      IMULx $const, reg
Two-operand IMULx is like a crippled IMUL3x, with dst fixed to input[0].
This is why we don't bother to generate IMULx for the case where
dst is the same as input[0].

Change-Id: I4becda475b3dffdd07b6fdf1c75bacc82af654e4
Reviewed-on: https://go-review.googlesource.com/99656
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Keith Randall <khr@golang.org>

cmd/compile: implement CMOV on amd64

This builds upon the branchelim pass, activating it for amd64 and
lowering CondSelect. Special care is made to FPU instructions for
NaN handling.

Benchmark results on Xeon E5630 (Westmere EP):

name                      old time/op    new time/op    delta
BinaryTree17-16              4.99s ± 9%     4.66s ± 2%     ~     (p=0.095 n=5+5)
Fannkuch11-16                4.93s ± 3%     5.04s ± 2%     ~     (p=0.548 n=5+5)
FmtFprintfEmpty-16          58.8ns ± 7%    61.4ns ±14%     ~     (p=0.579 n=5+5)
FmtFprintfString-16          114ns ± 2%     114ns ± 4%     ~     (p=0.603 n=5+5)
FmtFprintfInt-16             181ns ± 4%     125ns ± 3%  -30.90%  (p=0.008 n=5+5)
FmtFprintfIntInt-16          263ns ± 2%     217ns ± 2%  -17.34%  (p=0.008 n=5+5)
FmtFprintfPrefixedInt-16     230ns ± 1%     212ns ± 1%   -7.99%  (p=0.008 n=5+5)
FmtFprintfFloat-16           411ns ± 3%     344ns ± 5%  -16.43%  (p=0.008 n=5+5)
FmtManyArgs-16               828ns ± 4%     790ns ± 2%   -4.59%  (p=0.032 n=5+5)
GobDecode-16                10.9ms ± 4%    10.8ms ± 5%     ~     (p=0.548 n=5+5)
GobEncode-16                9.52ms ± 5%    9.46ms ± 2%     ~     (p=1.000 n=5+5)
Gzip-16                      334ms ± 2%     337ms ± 2%     ~     (p=0.548 n=5+5)
Gunzip-16                   64.4ms ± 1%    65.0ms ± 1%   +1.00%  (p=0.008 n=5+5)
HTTPClientServer-16          156µs ± 3%     155µs ± 3%     ~     (p=0.690 n=5+5)
JSONEncode-16               21.0ms ± 1%    21.8ms ± 0%   +3.76%  (p=0.016 n=5+4)
JSONDecode-16               95.1ms ± 0%    95.7ms ± 1%     ~     (p=0.151 n=5+5)
Mandelbrot200-16            6.38ms ± 1%    6.42ms ± 1%     ~     (p=0.095 n=5+5)
GoParse-16                  5.47ms ± 2%    5.36ms ± 1%   -1.95%  (p=0.016 n=5+5)
RegexpMatchEasy0_32-16       111ns ± 1%     111ns ± 1%     ~     (p=0.635 n=5+4)
RegexpMatchEasy0_1K-16       408ns ± 1%     411ns ± 2%     ~     (p=0.087 n=5+5)
RegexpMatchEasy1_32-16       103ns ± 1%     104ns ± 1%     ~     (p=0.484 n=5+5)
RegexpMatchEasy1_1K-16       659ns ± 2%     652ns ± 1%     ~     (p=0.571 n=5+5)
RegexpMatchMedium_32-16      176ns ± 2%     174ns ± 1%     ~     (p=0.476 n=5+5)
RegexpMatchMedium_1K-16     58.6µs ± 4%    57.7µs ± 4%     ~     (p=0.548 n=5+5)
RegexpMatchHard_32-16       3.07µs ± 3%    3.04µs ± 4%     ~     (p=0.421 n=5+5)
RegexpMatchHard_1K-16       89.2µs ± 1%    87.9µs ± 2%   -1.52%  (p=0.032 n=5+5)
Revcomp-16                   575ms ± 0%     587ms ± 2%   +2.12%  (p=0.032 n=4+5)
Template-16                  110ms ± 1%     107ms ± 3%   -3.00%  (p=0.032 n=5+5)
TimeParse-16                 463ns ± 0%     462ns ± 0%     ~     (p=0.810 n=5+4)
TimeFormat-16                538ns ± 0%     535ns ± 0%   -0.63%  (p=0.024 n=5+5)

name                      old speed      new speed      delta
GobDecode-16              70.7MB/s ± 4%  71.4MB/s ± 5%     ~     (p=0.452 n=5+5)
GobEncode-16              80.7MB/s ± 5%  81.2MB/s ± 2%     ~     (p=1.000 n=5+5)
Gzip-16                   58.2MB/s ± 2%  57.7MB/s ± 2%     ~     (p=0.452 n=5+5)
Gunzip-16                  302MB/s ± 1%   299MB/s ± 1%   -0.99%  (p=0.008 n=5+5)
JSONEncode-16             92.4MB/s ± 1%  89.1MB/s ± 0%   -3.63%  (p=0.016 n=5+4)
JSONDecode-16             20.4MB/s ± 0%  20.3MB/s ± 1%     ~     (p=0.135 n=5+5)
GoParse-16                10.6MB/s ± 2%  10.8MB/s ± 1%   +2.00%  (p=0.016 n=5+5)
RegexpMatchEasy0_32-16     286MB/s ± 1%   285MB/s ± 3%     ~     (p=1.000 n=5+5)
RegexpMatchEasy0_1K-16    2.51GB/s ± 1%  2.49GB/s ± 2%     ~     (p=0.095 n=5+5)
RegexpMatchEasy1_32-16     309MB/s ± 1%   307MB/s ± 1%     ~     (p=0.548 n=5+5)
RegexpMatchEasy1_1K-16    1.55GB/s ± 2%  1.57GB/s ± 1%     ~     (p=0.690 n=5+5)
RegexpMatchMedium_32-16   5.68MB/s ± 2%  5.73MB/s ± 1%     ~     (p=0.579 n=5+5)
RegexpMatchMedium_1K-16   17.5MB/s ± 4%  17.8MB/s ± 4%     ~     (p=0.500 n=5+5)
RegexpMatchHard_32-16     10.4MB/s ± 3%  10.5MB/s ± 4%     ~     (p=0.460 n=5+5)
RegexpMatchHard_1K-16     11.5MB/s ± 1%  11.7MB/s ± 2%   +1.57%  (p=0.032 n=5+5)
Revcomp-16                 442MB/s ± 0%   433MB/s ± 2%   -2.05%  (p=0.032 n=4+5)
Template-16               17.7MB/s ± 1%  18.2MB/s ± 3%   +3.12%  (p=0.032 n=5+5)

Change-Id: Ic7cb7374d07da031e771bdcbfdd832fd1b17159c
Reviewed-on: https://go-review.googlesource.com/98695
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>

cmd/asm: fix ARM64 vector register arrangement encoding bug

The current code assigns vector register arrangement a wrong value
when the arrangement specifier is S2, which causes the incorrect
assembly.

The patch fixes the issue and adds the test cases.

Fixes #24249

Change-Id: I9736df1279494003d0b178da1af9cee9cd85ce21
Reviewed-on: https://go-review.googlesource.com/98555
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

math/big: optimize shlVU and shrVU on arm64

This CL implements shlVU and shrVU with arm64 HW instructions "LDP" and "STP" to reduce load cost,
it also removes unnecessary checks on the number of shifts for better performance.

Benchmarks:

name                              old time/op    new time/op    delta
AddVV/1-8                           21.6ns ± 1%    21.6ns ± 1%      ~     (p=0.683 n=5+5)
AddVV/2-8                           13.5ns ± 0%    13.5ns ± 0%      ~     (all equal)
AddVV/3-8                           15.5ns ± 0%    15.5ns ± 0%      ~     (all equal)
AddVV/4-8                           17.5ns ± 0%    17.5ns ± 0%      ~     (all equal)
AddVV/5-8                           19.5ns ± 0%    19.5ns ± 0%      ~     (all equal)
AddVV/10-8                          29.5ns ± 0%    29.5ns ± 0%      ~     (all equal)
AddVV/100-8                          217ns ± 0%     217ns ± 0%      ~     (all equal)
AddVV/1000-8                        2.02µs ± 0%    2.03µs ± 0%    +0.73%  (p=0.008 n=5+5)
AddVV/10000-8                       20.5µs ± 0%    20.5µs ± 0%    -0.01%  (p=0.008 n=5+5)
AddVV/100000-8                       246µs ± 5%     250µs ± 4%      ~     (p=0.548 n=5+5)
AddVW/1-8                           9.26ns ± 0%    9.32ns ± 0%    +0.65%  (p=0.016 n=4+5)
AddVW/2-8                           19.8ns ± 3%    19.8ns ± 0%      ~     (p=0.143 n=5+5)
AddVW/3-8                           11.5ns ± 0%    11.5ns ± 0%      ~     (all equal)
AddVW/4-8                           13.0ns ± 0%    13.0ns ± 0%      ~     (all equal)
AddVW/5-8                           14.5ns ± 0%    14.5ns ± 0%      ~     (all equal)
AddVW/10-8                          22.0ns ± 0%    22.0ns ± 0%      ~     (all equal)
AddVW/100-8                          167ns ± 0%     166ns ± 0%    -0.60%  (p=0.000 n=5+4)
AddVW/1000-8                        1.52µs ± 0%    1.52µs ± 0%      ~     (all equal)
AddVW/10000-8                       15.1µs ± 0%    15.1µs ± 0%    +0.01%  (p=0.008 n=5+5)
AddVW/100000-8                       163µs ± 4%     153µs ± 3%    -5.97%  (p=0.016 n=5+5)
AddMulVVW/1-8                       32.4ns ± 1%    33.0ns ± 1%    +1.73%  (p=0.040 n=5+5)
AddMulVVW/2-8                       56.4ns ± 2%    55.9ns ± 1%      ~     (p=0.135 n=5+5)
AddMulVVW/3-8                       85.4ns ± 1%    85.1ns ± 0%      ~     (p=0.079 n=5+5)
AddMulVVW/4-8                        129ns ± 1%     129ns ± 0%      ~     (p=0.397 n=5+5)
AddMulVVW/5-8                        148ns ± 0%     148ns ± 0%      ~     (all equal)
AddMulVVW/10-8                       270ns ± 0%     268ns ± 0%    -0.74%  (p=0.029 n=4+4)
AddMulVVW/100-8                     2.75µs ± 0%    2.75µs ± 0%    -0.09%  (p=0.008 n=5+5)
AddMulVVW/1000-8                    26.0µs ± 0%    26.0µs ± 0%    -0.06%  (p=0.024 n=5+5)
AddMulVVW/10000-8                    312µs ± 0%     312µs ± 0%    -0.09%  (p=0.008 n=5+5)
AddMulVVW/100000-8                  2.89ms ± 0%    2.89ms ± 0%    +0.14%  (p=0.016 n=5+5)
DecimalConversion-8                  315µs ± 1%     312µs ± 0%      ~     (p=0.095 n=5+5)
FloatString/100-8                   2.56µs ± 1%    2.52µs ± 1%    -1.31%  (p=0.016 n=5+5)
FloatString/1000-8                  58.6µs ± 0%    58.2µs ± 0%    -0.75%  (p=0.008 n=5+5)
FloatString/10000-8                 4.59ms ± 0%    4.59ms ± 0%      ~     (p=0.056 n=5+5)
FloatString/100000-8                 446ms ± 0%     446ms ± 0%    -0.04%  (p=0.008 n=5+5)
FloatAdd/10-8                        184ns ± 0%     178ns ± 0%    -3.48%  (p=0.008 n=5+5)
FloatAdd/100-8                       189ns ± 3%     178ns ± 2%    -6.02%  (p=0.008 n=5+5)
FloatAdd/1000-8                      371ns ± 0%     267ns ± 0%   -27.99%  (p=0.000 n=5+4)
FloatAdd/10000-8                    1.87µs ± 0%    1.03µs ± 0%   -44.74%  (p=0.008 n=5+5)
FloatAdd/100000-8                   17.1µs ± 0%     8.8µs ± 0%   -48.71%  (p=0.016 n=5+4)
FloatSub/10-8                        148ns ± 0%     146ns ± 0%    -1.35%  (p=0.000 n=5+4)
FloatSub/100-8                       148ns ± 0%     140ns ± 0%    -5.41%  (p=0.008 n=5+5)
FloatSub/1000-8                      242ns ± 0%     191ns ± 0%   -21.24%  (p=0.008 n=5+5)
FloatSub/10000-8                    1.07µs ± 0%    0.64µs ± 1%   -39.89%  (p=0.016 n=4+5)
FloatSub/100000-8                   9.48µs ± 0%    5.32µs ± 0%   -43.87%  (p=0.008 n=5+5)
ParseFloatSmallExp-8                29.3µs ± 3%    28.6µs ± 1%      ~     (p=0.310 n=5+5)
ParseFloatLargeExp-8                 125µs ± 1%     123µs ± 0%    -1.99%  (p=0.008 n=5+5)
GCD10x10/WithoutXY-8                 278ns ± 4%     289ns ± 5%    +3.96%  (p=0.040 n=5+5)
GCD10x10/WithXY-8                   2.12µs ± 2%    2.15µs ± 2%      ~     (p=0.095 n=5+5)
GCD10x100/WithoutXY-8                615ns ± 1%     629ns ± 5%      ~     (p=0.135 n=5+5)
GCD10x100/WithXY-8                  3.42µs ± 1%    3.53µs ± 2%    +3.38%  (p=0.008 n=5+5)
GCD10x1000/WithoutXY-8              1.39µs ± 1%    1.38µs ± 1%      ~     (p=0.460 n=5+5)
GCD10x1000/WithXY-8                 7.47µs ± 2%    7.49µs ± 3%      ~     (p=1.000 n=5+5)
GCD10x10000/WithoutXY-8             8.71µs ± 1%    8.71µs ± 0%      ~     (p=0.841 n=5+5)
GCD10x10000/WithXY-8                28.4µs ± 2%    27.2µs ± 2%    -4.24%  (p=0.008 n=5+5)
GCD10x100000/WithoutXY-8            78.9µs ± 1%    79.1µs ± 0%      ~     (p=0.222 n=5+5)
GCD10x100000/WithXY-8                240µs ± 1%     228µs ± 1%    -4.98%  (p=0.008 n=5+5)
GCD100x100/WithoutXY-8              1.87µs ± 2%    1.89µs ± 1%      ~     (p=0.095 n=5+5)
GCD100x100/WithXY-8                 26.6µs ± 1%    26.3µs ± 0%    -1.14%  (p=0.032 n=5+5)
GCD100x1000/WithoutXY-8             4.44µs ± 2%    4.47µs ± 2%      ~     (p=0.444 n=5+5)
GCD100x1000/WithXY-8                36.7µs ± 1%    36.0µs ± 1%    -1.96%  (p=0.008 n=5+5)
GCD100x10000/WithoutXY-8            22.8µs ± 1%    22.3µs ± 1%    -2.52%  (p=0.008 n=5+5)
GCD100x10000/WithXY-8                145µs ± 1%     142µs ± 0%    -1.88%  (p=0.008 n=5+5)
GCD100x100000/WithoutXY-8            198µs ± 1%     190µs ± 0%    -4.06%  (p=0.008 n=5+5)
GCD100x100000/WithXY-8              1.11ms ± 0%    1.09ms ± 0%    -1.87%  (p=0.008 n=5+5)
GCD1000x1000/WithoutXY-8            25.4µs ± 1%    25.0µs ± 1%    -1.34%  (p=0.008 n=5+5)
GCD1000x1000/WithXY-8                515µs ± 1%     485µs ± 0%    -5.85%  (p=0.008 n=5+5)
GCD1000x10000/WithoutXY-8           57.3µs ± 1%    56.2µs ± 1%    -1.95%  (p=0.008 n=5+5)
GCD1000x10000/WithXY-8              1.21ms ± 0%    1.18ms ± 0%    -2.65%  (p=0.008 n=5+5)
GCD1000x100000/WithoutXY-8           358µs ± 0%     352µs ± 1%    -1.71%  (p=0.008 n=5+5)
GCD1000x100000/WithXY-8             8.72ms ± 0%    8.66ms ± 0%    -0.71%  (p=0.008 n=5+5)
GCD10000x10000/WithoutXY-8           690µs ± 0%     687µs ± 1%      ~     (p=0.095 n=5+5)
GCD10000x10000/WithXY-8             16.0ms ± 0%    12.5ms ± 0%   -22.01%  (p=0.008 n=5+5)
GCD10000x100000/WithoutXY-8         2.09ms ± 0%    2.07ms ± 0%    -0.58%  (p=0.008 n=5+5)
GCD10000x100000/WithXY-8            86.8ms ± 0%    83.4ms ± 0%    -3.95%  (p=0.008 n=5+5)
GCD100000x100000/WithoutXY-8        51.2ms ± 0%    51.2ms ± 0%      ~     (p=0.548 n=5+5)
GCD100000x100000/WithXY-8            1.25s ± 0%     0.89s ± 0%   -28.98%  (p=0.008 n=5+5)
Hilbert-8                           2.46ms ± 2%    2.53ms ± 1%    +2.89%  (p=0.032 n=5+5)
Binomial-8                          5.15µs ± 4%    4.92µs ± 1%    -4.43%  (p=0.032 n=5+5)
QuoRem-8                            7.10µs ± 0%    7.05µs ± 0%    -0.59%  (p=0.008 n=5+5)
Exp-8                                161ms ± 0%     161ms ± 0%    -0.24%  (p=0.008 n=5+5)
Exp2-8                               161ms ± 0%     161ms ± 0%    -0.30%  (p=0.016 n=4+5)
Bitset-8                            40.4ns ± 0%    40.3ns ± 0%      ~     (p=0.159 n=5+5)
BitsetNeg-8                          158ns ± 4%     155ns ± 2%      ~     (p=0.183 n=5+5)
BitsetOrig-8                         374ns ± 0%     383ns ± 1%    +2.35%  (p=0.008 n=5+5)
BitsetNegOrig-8                      620ns ± 1%     663ns ± 2%    +7.00%  (p=0.008 n=5+5)
ModSqrt225_Tonelli-8                7.26ms ± 0%    7.27ms ± 0%      ~     (p=0.841 n=5+5)
ModSqrt224_3Mod4-8                  2.24ms ± 0%    2.24ms ± 0%      ~     (p=0.690 n=5+5)
ModSqrt5430_Tonelli-8                62.3s ± 0%     62.4s ± 0%    +0.15%  (p=0.008 n=5+5)
ModSqrt5430_3Mod4-8                  20.8s ± 0%     20.8s ± 0%      ~     (p=0.151 n=5+5)
Sqrt-8                               101µs ± 0%      97µs ± 0%    -3.99%  (p=0.008 n=5+5)
IntSqr/1-8                          32.7ns ± 1%    32.5ns ± 1%      ~     (p=0.325 n=5+5)
IntSqr/2-8                           161ns ± 4%     160ns ± 4%      ~     (p=0.659 n=5+5)
IntSqr/3-8                           296ns ± 7%     297ns ± 6%      ~     (p=0.841 n=5+5)
IntSqr/5-8                           752ns ± 7%     755ns ± 6%      ~     (p=0.889 n=5+5)
IntSqr/8-8                          1.91µs ± 3%    1.90µs ± 3%      ~     (p=0.746 n=5+5)
IntSqr/10-8                         2.99µs ± 4%    3.00µs ± 4%      ~     (p=0.516 n=5+5)
IntSqr/20-8                         6.29µs ± 2%    6.19µs ± 2%      ~     (p=0.151 n=5+5)
IntSqr/30-8                         14.0µs ± 1%    13.8µs ± 2%      ~     (p=0.056 n=5+5)
IntSqr/50-8                         38.1µs ± 3%    37.9µs ± 3%      ~     (p=0.548 n=5+5)
IntSqr/80-8                         95.1µs ± 1%    94.7µs ± 1%      ~     (p=0.310 n=5+5)
IntSqr/100-8                         148µs ± 1%     148µs ± 1%      ~     (p=0.548 n=5+5)
IntSqr/200-8                         587µs ± 1%     587µs ± 1%      ~     (p=1.000 n=5+5)
IntSqr/300-8                        1.31ms ± 1%    1.32ms ± 1%      ~     (p=0.151 n=5+5)
IntSqr/500-8                        2.48ms ± 0%    2.49ms ± 0%      ~     (p=0.310 n=5+5)
IntSqr/800-8                        4.68ms ± 0%    4.67ms ± 0%      ~     (p=0.548 n=5+5)
IntSqr/1000-8                       7.57ms ± 0%    7.56ms ± 0%      ~     (p=0.421 n=5+5)
Mul-8                                311ms ± 0%     311ms ± 0%      ~     (p=0.151 n=5+5)
Exp3Power/0x10-8                     584ns ± 2%     573ns ± 1%      ~     (p=0.190 n=5+5)
Exp3Power/0x40-8                     646ns ± 2%     649ns ± 1%      ~     (p=0.690 n=5+5)
Exp3Power/0x100-8                   1.42µs ± 2%    1.45µs ± 1%    +2.03%  (p=0.032 n=5+5)
Exp3Power/0x400-8                   8.28µs ± 1%    8.39µs ± 0%    +1.33%  (p=0.008 n=5+5)
Exp3Power/0x1000-8                  60.1µs ± 0%    59.8µs ± 0%    -0.44%  (p=0.008 n=5+5)
Exp3Power/0x4000-8                   818µs ± 0%     816µs ± 0%    -0.23%  (p=0.008 n=5+5)
Exp3Power/0x10000-8                 7.79ms ± 0%    7.78ms ± 0%      ~     (p=0.690 n=5+5)
Exp3Power/0x40000-8                 73.4ms ± 0%    73.3ms ± 0%      ~     (p=0.151 n=5+5)
Exp3Power/0x100000-8                 665ms ± 0%     664ms ± 0%    -0.16%  (p=0.016 n=4+5)
Exp3Power/0x400000-8                 5.99s ± 0%     5.97s ± 0%    -0.24%  (p=0.008 n=5+5)
Fibo-8                               116ms ± 0%     117ms ± 0%    +0.42%  (p=0.008 n=5+5)
NatSqr/1-8                           113ns ± 2%     112ns ± 1%      ~     (p=0.190 n=5+5)
NatSqr/2-8                           249ns ± 2%     250ns ± 2%      ~     (p=0.365 n=5+5)
NatSqr/3-8                           379ns ± 1%     381ns ± 2%      ~     (p=0.127 n=5+5)
NatSqr/5-8                           838ns ± 3%     841ns ± 5%      ~     (p=0.754 n=5+5)
NatSqr/8-8                          1.97µs ± 3%    1.97µs ± 4%      ~     (p=1.000 n=5+5)
NatSqr/10-8                         3.04µs ± 4%    3.04µs ± 4%      ~     (p=1.000 n=5+5)
NatSqr/20-8                         6.49µs ± 3%    6.50µs ± 2%      ~     (p=0.841 n=5+5)
NatSqr/30-8                         14.3µs ± 2%    14.2µs ± 2%      ~     (p=0.548 n=5+5)
NatSqr/50-8                         38.5µs ± 3%    38.3µs ± 3%      ~     (p=0.421 n=5+5)
NatSqr/80-8                         96.3µs ± 1%    96.1µs ± 1%      ~     (p=0.421 n=5+5)
NatSqr/100-8                         149µs ± 1%     148µs ± 1%      ~     (p=0.310 n=5+5)
NatSqr/200-8                         591µs ± 1%     592µs ± 1%      ~     (p=0.690 n=5+5)
NatSqr/300-8                        1.31ms ± 1%    1.32ms ± 0%      ~     (p=0.190 n=5+4)
NatSqr/500-8                        2.49ms ± 0%    2.49ms ± 0%      ~     (p=0.095 n=5+5)
NatSqr/800-8                        4.70ms ± 0%    4.69ms ± 0%      ~     (p=0.222 n=5+5)
NatSqr/1000-8                       7.60ms ± 0%    7.58ms ± 0%      ~     (p=0.222 n=5+5)
ScanPi-8                             326µs ± 0%     327µs ± 1%      ~     (p=0.222 n=5+5)
StringPiParallel-8                  71.4µs ± 5%    67.7µs ± 4%      ~     (p=0.095 n=5+5)
Scan/10/Base2-8                     1.09µs ± 0%    1.10µs ± 1%      ~     (p=0.810 n=5+5)
Scan/100/Base2-8                    7.79µs ± 0%    7.83µs ± 0%    +0.53%  (p=0.008 n=5+5)
Scan/1000/Base2-8                   78.9µs ± 0%    79.0µs ± 0%      ~     (p=0.151 n=5+5)
Scan/10000/Base2-8                  1.22ms ± 0%    1.23ms ± 1%      ~     (p=0.690 n=5+5)
Scan/100000/Base2-8                 55.1ms ± 0%    55.1ms ± 0%    +0.10%  (p=0.008 n=5+5)
Scan/10/Base8-8                      512ns ± 1%     534ns ± 1%    +4.34%  (p=0.008 n=5+5)
Scan/100/Base8-8                    2.90µs ± 1%    2.92µs ± 0%    +0.67%  (p=0.024 n=5+5)
Scan/1000/Base8-8                   31.0µs ± 0%    31.1µs ± 0%    +0.27%  (p=0.008 n=5+5)
Scan/10000/Base8-8                   741µs ± 0%     744µs ± 1%      ~     (p=0.310 n=5+5)
Scan/100000/Base8-8                 50.5ms ± 0%    50.7ms ± 0%    +0.23%  (p=0.016 n=5+4)
Scan/10/Base10-8                     485ns ± 0%     510ns ± 1%    +5.15%  (p=0.008 n=5+5)
Scan/100/Base10-8                   2.68µs ± 0%    2.70µs ± 0%    +0.84%  (p=0.008 n=5+5)
Scan/1000/Base10-8                  28.7µs ± 0%    28.8µs ± 0%    +0.34%  (p=0.008 n=5+5)
Scan/10000/Base10-8                  717µs ± 0%     720µs ± 1%      ~     (p=0.238 n=5+5)
Scan/100000/Base10-8                50.3ms ± 0%    50.3ms ± 0%    +0.02%  (p=0.016 n=4+5)
Scan/10/Base16-8                     439ns ± 0%     461ns ± 1%    +5.06%  (p=0.008 n=5+5)
Scan/100/Base16-8                   2.48µs ± 0%    2.49µs ± 0%    +0.59%  (p=0.024 n=5+5)
Scan/1000/Base16-8                  27.2µs ± 0%    27.3µs ± 0%      ~     (p=0.063 n=5+5)
Scan/10000/Base16-8                  722µs ± 0%     725µs ± 1%      ~     (p=0.421 n=5+5)
Scan/100000/Base16-8                52.7ms ± 0%    52.7ms ± 0%      ~     (p=0.686 n=4+4)
String/10/Base2-8                    248ns ± 1%     248ns ± 1%      ~     (p=0.802 n=5+5)
String/100/Base2-8                  1.51µs ± 0%    1.51µs ± 0%    -0.54%  (p=0.024 n=5+5)
String/1000/Base2-8                 13.6µs ± 0%    13.6µs ± 0%      ~     (p=0.548 n=5+5)
String/10000/Base2-8                 135µs ± 1%     135µs ± 2%      ~     (p=0.421 n=5+5)
String/100000/Base2-8               1.32ms ± 1%    1.33ms ± 1%      ~     (p=0.310 n=5+5)
String/10/Base8-8                    169ns ± 0%     170ns ± 0%      ~     (p=0.079 n=5+5)
String/100/Base8-8                   635ns ± 1%     633ns ± 1%      ~     (p=0.595 n=5+5)
String/1000/Base8-8                 5.33µs ± 0%    5.30µs ± 0%      ~     (p=0.063 n=5+5)
String/10000/Base8-8                50.7µs ± 1%    50.7µs ± 1%      ~     (p=1.000 n=5+5)
String/100000/Base8-8                499µs ± 1%     500µs ± 1%      ~     (p=1.000 n=5+5)
String/10/Base10-8                   517ns ± 1%     512ns ± 1%    -1.01%  (p=0.032 n=5+5)
String/100/Base10-8                 1.97µs ± 0%    2.01µs ± 1%    +2.13%  (p=0.008 n=5+5)
String/1000/Base10-8                12.6µs ± 1%    12.1µs ± 1%    -4.16%  (p=0.008 n=5+5)
String/10000/Base10-8               57.9µs ± 1%    54.8µs ± 1%    -5.46%  (p=0.008 n=5+5)
String/100000/Base10-8              25.6ms ± 0%    25.6ms ± 0%    -0.12%  (p=0.008 n=5+5)
String/10/Base16-8                   149ns ± 0%     149ns ± 1%      ~     (p=1.000 n=5+5)
String/100/Base16-8                  514ns ± 0%     514ns ± 1%      ~     (p=0.825 n=5+5)
String/1000/Base16-8                4.01µs ± 0%    4.01µs ± 0%      ~     (p=0.595 n=5+5)
String/10000/Base16-8               37.7µs ± 0%    37.8µs ± 1%      ~     (p=0.222 n=5+5)
String/100000/Base16-8               373µs ± 1%     372µs ± 0%      ~     (p=1.000 n=5+5)
LeafSize/0-8                        6.64ms ± 0%    6.66ms ± 0%    +0.32%  (p=0.008 n=5+5)
LeafSize/1-8                        74.0µs ± 1%    71.2µs ± 1%    -3.75%  (p=0.008 n=5+5)
LeafSize/2-8                        74.1µs ± 0%    70.7µs ± 1%    -4.53%  (p=0.008 n=5+5)
LeafSize/3-8                         379µs ± 0%     374µs ± 0%    -1.25%  (p=0.008 n=5+5)
LeafSize/4-8                        72.7µs ± 0%    69.2µs ± 0%    -4.79%  (p=0.008 n=5+5)
LeafSize/5-8                         471µs ± 0%     466µs ± 0%    -1.05%  (p=0.008 n=5+5)
LeafSize/6-8                         377µs ± 0%     373µs ± 0%    -1.16%  (p=0.008 n=5+5)
LeafSize/7-8                         245µs ± 0%     241µs ± 0%    -1.65%  (p=0.008 n=5+5)
LeafSize/8-8                        73.1µs ± 0%    69.4µs ± 0%    -5.10%  (p=0.008 n=5+5)
LeafSize/9-8                         538µs ± 0%     532µs ± 0%    -1.01%  (p=0.008 n=5+5)
LeafSize/10-8                        472µs ± 0%     467µs ± 0%    -1.07%  (p=0.008 n=5+5)
LeafSize/11-8                        460µs ± 0%     454µs ± 0%    -1.22%  (p=0.008 n=5+5)
LeafSize/12-8                        378µs ± 0%     373µs ± 0%    -1.34%  (p=0.008 n=5+5)
LeafSize/13-8                        344µs ± 0%     338µs ± 0%    -1.61%  (p=0.008 n=5+5)
LeafSize/14-8                        247µs ± 0%     243µs ± 0%    -1.62%  (p=0.008 n=5+5)
LeafSize/15-8                        169µs ± 0%     165µs ± 0%    -2.71%  (p=0.008 n=5+5)
LeafSize/16-8                       73.3µs ± 1%    69.5µs ± 0%    -5.11%  (p=0.008 n=5+5)
LeafSize/32-8                       82.7µs ± 0%    79.2µs ± 0%    -4.24%  (p=0.008 n=5+5)
LeafSize/64-8                        135µs ± 0%     132µs ± 0%    -2.20%  (p=0.008 n=5+5)
ProbablyPrime/n=0-8                 44.2ms ± 0%    43.9ms ± 0%    -0.69%  (p=0.008 n=5+5)
ProbablyPrime/n=1-8                 64.8ms ± 0%    64.4ms ± 0%    -0.60%  (p=0.008 n=5+5)
ProbablyPrime/n=5-8                  147ms ± 0%     147ms ± 0%    -0.34%  (p=0.008 n=5+5)
ProbablyPrime/n=10-8                 250ms ± 0%     249ms ± 0%    -0.29%  (p=0.008 n=5+5)
ProbablyPrime/n=20-8                 456ms ± 0%     455ms ± 0%    -0.29%  (p=0.008 n=5+5)
ProbablyPrime/Lucas-8               23.6ms ± 0%    23.2ms ± 0%    -1.44%  (p=0.008 n=5+5)
ProbablyPrime/MillerRabinBase2-8    20.6ms ± 0%    20.6ms ± 0%    -0.31%  (p=0.008 n=5+5)
FloatSqrt/64-8                      2.27µs ± 1%    2.11µs ± 1%    -7.02%  (p=0.008 n=5+5)
FloatSqrt/128-8                     4.93µs ± 1%    4.40µs ± 1%   -10.73%  (p=0.008 n=5+5)
FloatSqrt/256-8                     13.6µs ± 0%     6.6µs ± 1%   -51.40%  (p=0.008 n=5+5)
FloatSqrt/1000-8                    69.8µs ± 0%    31.2µs ± 0%   -55.27%  (p=0.008 n=5+5)
FloatSqrt/10000-8                   1.91ms ± 0%    0.59ms ± 0%   -69.17%  (p=0.008 n=5+5)
FloatSqrt/100000-8                  55.4ms ± 0%    17.8ms ± 0%   -67.79%  (p=0.008 n=5+5)
FloatSqrt/1000000-8                  4.56s ± 0%     1.52s ± 0%   -66.59%  (p=0.008 n=5+5)

Change-Id: Icce52c69668f564490c69b908338b21a2288e116
Reviewed-on: https://go-review.googlesource.com/79355
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

test: move rotate tests to top-level testsuite.

Remove old tests from asm_test.

Change-Id: Ib408ec7faa60068bddecf709b93ce308e0ef665a
Reviewed-on: https://go-review.googlesource.com/100075
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>

cmd: re-generate all stringer files

The tool has gotten better over time, so re-generating the files brings
some advantages like fewer objects, dropping the use of fmt, and
dropping unnecessary bounds checks.

While at it, add the missing go:generate line for obj.AddrType.

Change-Id: I120c9795ee8faddf5961ff0384b9dcaf58d831ff
Reviewed-on: https://go-review.googlesource.com/100015
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>