Use global rand.Rand instance in mathrand.

Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want. Update all of the global
functons to use our global interface instead of the rand.Rand package
one.

Additionally, as an optimization, remove "defer" calls from the Locking
wrapper. This speeds up operations by a factor of 3x.

BUG=chromium:675813
TEST=None

Review-Url: https://codereview.chromium.org/2588213002
Committed: https://github.com/luci/luci-go/commit/f469f2540cad71872c674950f530bdbc5de13e02

[Vadim Sh., 2016-12-20 01:40:38]

https://codereview.chromium.org/2588213002/diff/1/common/data/rand/mathrand/m...
File common/data/rand/mathrand/mathrand.go (right):

https://codereview.chromium.org/2588213002/diff/1/common/data/rand/mathrand/m...
common/data/rand/mathrand/mathrand.go:58: // Generate a new Rand instance and
return it. Our callers expect this to be
this comment is stale now

https://codereview.chromium.org/2588213002/diff/1/common/data/rand/mathrand/m...
common/data/rand/mathrand/mathrand.go:248: // The callback's r must not be
retained or used outside of hte scope of the
typo: hte scope

[dnj, 2016-12-20 02:12:48]

Description was changed from

==========
Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want.

BUG=None
TEST=None
R=vadimsh@chromium.org
==========

to

==========
Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want. Update all of the global
functons to use our global interface instead of the rand.Rand package
one.

Additionally, as an optimization, remove "defer" calls from the Locking
wrapper. This speeds up operations by a factor of 3x.

BUG=None
TEST=None
==========

[dnj, 2016-12-20 02:14:06]

Description was changed from

==========
Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want. Update all of the global
functons to use our global interface instead of the rand.Rand package
one.

Additionally, as an optimization, remove "defer" calls from the Locking
wrapper. This speeds up operations by a factor of 3x.

BUG=None
TEST=None
==========

to

==========
Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want. Update all of the global
functons to use our global interface instead of the rand.Rand package
one.

Additionally, as an optimization, remove "defer" calls from the Locking
wrapper. This speeds up operations by a factor of 3x.

BUG=chromium:675813
TEST=None
==========

[dnj, 2016-12-20 02:16:12]

https://codereview.chromium.org/2588213002/diff/1/common/data/rand/mathrand/m...
File common/data/rand/mathrand/mathrand.go (right):

https://codereview.chromium.org/2588213002/diff/1/common/data/rand/mathrand/m...
common/data/rand/mathrand/mathrand.go:58: // Generate a new Rand instance and
return it. Our callers expect this to be
On 2016/12/20 01:40:38, Vadim Sh. wrote:
> this comment is stale now

Done.

https://codereview.chromium.org/2588213002/diff/1/common/data/rand/mathrand/m...
common/data/rand/mathrand/mathrand.go:248: // The callback's r must not be
retained or used outside of hte scope of the
On 2016/12/20 01:40:38, Vadim Sh. wrote:
> typo: hte scope

Done.

[Vadim Sh., 2016-12-20 02:31:56]

lgtm

https://codereview.chromium.org/2588213002/diff/60001/common/data/rand/mathra...
File common/data/rand/mathrand/mathrand_test.go (right):

https://codereview.chromium.org/2588213002/diff/60001/common/data/rand/mathra...
common/data/rand/mathrand/mathrand_test.go:174: //
BenchmarkOurDefaultSourceViaCtx-32            20000000        77.8 ns/op
awesome!

[dnj, 2016-12-20 02:36:07]

https://codereview.chromium.org/2588213002/diff/60001/common/data/rand/mathra...
File common/data/rand/mathrand/mathrand_test.go (right):

https://codereview.chromium.org/2588213002/diff/60001/common/data/rand/mathra...
common/data/rand/mathrand/mathrand_test.go:174: //
BenchmarkOurDefaultSourceViaCtx-32            20000000        77.8 ns/op
On 2016/12/20 02:31:56, Vadim Sh. wrote:
> awesome!

To be fair it's b/c I'm not doing an allocation per round, but yeah!

[dnj, 2016-12-20 02:36:09]

The CQ bit was checked by dnj@chromium.org

[commit-bot: I haz the power, 2016-12-20 02:36:18]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-12-20 02:47:13]

CQ is committing da patch.

Bot data: {"patchset_id": 60001, "attempt_start_ts": 1482201369279150,
"parent_rev": "fd0a102c24af05830dccbce7aeb7cf8d3295e90c", "commit_rev":
"f469f2540cad71872c674950f530bdbc5de13e02"}

[commit-bot: I haz the power, 2016-12-20 02:47:16]

Description was changed from

==========
Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want. Update all of the global
functons to use our global interface instead of the rand.Rand package
one.

Additionally, as an optimization, remove "defer" calls from the Locking
wrapper. This speeds up operations by a factor of 3x.

BUG=chromium:675813
TEST=None
==========

to

==========
Use global rand.Rand instance in mathrand.

By default, mathrand will use a new *rand.Rand if none is installed in
the Context. This leads to a very bad performance case if the user
forgets to install an instance in the Context!

Now that we have locking, use a single global rand.Rand instance. The
user can always change this if they want. Update all of the global
functons to use our global interface instead of the rand.Rand package
one.

Additionally, as an optimization, remove "defer" calls from the Locking
wrapper. This speeds up operations by a factor of 3x.

BUG=chromium:675813
TEST=None

Review-Url: https://codereview.chromium.org/2588213002
Committed:
https://github.com/luci/luci-go/commit/f469f2540cad71872c674950f530bdbc5de13e02
==========

[commit-bot: I haz the power, 2016-12-20 02:47:17]

Committed patchset #4 (id:60001) as
https://github.com/luci/luci-go/commit/f469f2540cad71872c674950f530bdbc5de13e02