Tweak RNG.

Tweak RNG.

R=bmeurer@chromium.org

Committed: https://crrev.com/623cbdc5432713badc9fe1d605c585aabb25876c
Cr-Commit-Position: refs/heads/master@{#32132}

[Benedikt Meurer, 2015-11-20 07:04:30]

LGTM.

[Yang, 2015-11-20 07:10:16]

The CQ bit was checked by yangguo@chromium.org

[commit-bot: I haz the power, 2015-11-20 07:10:22]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1462293002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1462293002/1

[commit-bot: I haz the power, 2015-11-20 07:21:23]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-11-20 07:21:23]

Try jobs failed on following builders:
  v8_linux64_asan_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux64_asan_rel/builds/...)

[Yang, 2015-11-20 07:26:41]

The CQ bit was checked by yangguo@chromium.org

[Yang, 2015-11-20 07:26:41]

The patchset sent to the CQ was uploaded after l-g-t-m from bmeurer@chromium.org
Link to the patchset: https://codereview.chromium.org/1462293002/#ps20001
(title: "fix bug")

[commit-bot: I haz the power, 2015-11-20 07:26:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1462293002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1462293002/20001

[commit-bot: I haz the power, 2015-11-20 07:55:31]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-11-20 07:55:58]

Patchset 2 (id:??) landed as
https://crrev.com/623cbdc5432713badc9fe1d605c585aabb25876c
Cr-Commit-Position: refs/heads/master@{#32132}

[mjmalone, 2015-11-20 23:37:26]

On 2015/11/20 07:55:58, commit-bot: I haz the power wrote:
> Patchset 2 (id:??) landed as
> https://crrev.com/623cbdc5432713badc9fe1d605c585aabb25876c
> Cr-Commit-Position: refs/heads/master@{#32132}

Hey guys, I think this patch is still no good. The upper bits of an MWC are low
entropy, so the upper bits of your xor'd result I believe will be pretty low
entropy too. I could be reading the code wrong, but I'm guessing the top bit of
the result will be 0 most of the time. The right fix for MWC1616 to perturb the
upper bits would be to << one generator and then add the full output from the
second generator, mod 2^32. That way you use the (low entropy) upper bits from
one generator to perturb the (high entropy) lower bits of the other. The
algorithm isn't great in general though, as is being discussed elsewhere.