pcc@: since you are working on the perf FYI testing phase, I am assigning the bug to you. 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/tools/build/+/7075ba2599232e8478e67fc99b080e34fc0fb8bb

commit 7075ba2599232e8478e67fc99b080e34fc0fb8bb
Author: Peter Collingbourne <pcc@google.com>
Date: Thu Mar 15 16:00:36 2018

Add Android CFI builders on chromium.perf.fyi (build).

Bug: 821860
Change-Id: Ie754a12137bfa8e62ef46d0232800bdcd67c8845
Reviewed-on: https://chromium-review.googlesource.com/963295
Reviewed-by: Emily Hanley <eyaich@chromium.org>
Reviewed-by: John Budorick <jbudorick@chromium.org>
Commit-Queue: Peter Collingbourne <pcc@chromium.org>

[modify] https://crrev.com/7075ba2599232e8478e67fc99b080e34fc0fb8bb/scripts/slave/recipe_modules/chromium_tests/chromium_perf_fyi.py
[modify] https://crrev.com/7075ba2599232e8478e67fc99b080e34fc0fb8bb/masters/master.chromium.perf.fyi/master.cfg
[modify] https://crrev.com/7075ba2599232e8478e67fc99b080e34fc0fb8bb/masters/master.chromium.perf.fyi/slaves.cfg

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f74f4fd03b8b201d9102591a781bb25c4b0c6e64

commit f74f4fd03b8b201d9102591a781bb25c4b0c6e64
Author: Peter Collingbourne <pcc@chromium.org>
Date: Wed May 30 22:16:04 2018

mb: Change Android CFI perf builders to build with ThinLTO instead of CFI.

For the moment, this is the configuration that we plan to deploy in
official builds.

Bug: 821860
Change-Id: I70ff3df79988a741676b52a18a0d6c3fe4f2438e
Reviewed-on: https://chromium-review.googlesource.com/1053230
Reviewed-by: John Budorick <jbudorick@chromium.org>
Commit-Queue: Peter Collingbourne <pcc@chromium.org>
Cr-Commit-Position: refs/heads/master@{#563020}
[modify] https://crrev.com/f74f4fd03b8b201d9102591a781bb25c4b0c6e64/tools/mb/mb_config.pyl

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/tools/build/+/5903a1e038f51f827cbfcc2ce99c3c7d5ed83faf

commit 5903a1e038f51f827cbfcc2ce99c3c7d5ed83faf
Author: Peter Collingbourne <pcc@google.com>
Date: Fri Jun 15 20:48:22 2018

Assign slave{244..247}-c1 to "Android CFI Builder Perf FYI".

Each of these machines has a different configuration:
244: n1-standard-8
245: n1-highmem-8
246: n1-standard-16
247: n1-highmem-16

This is part of an experiment to discover the minimum capacity
required to run Android ThinLTO perf bots. Once we have found the
minimum, this change will be reverted and we will update the
configuration of slave270-c1.

Bug: 821860,  853050 
Change-Id: I99ad99c396a0cad6c731a7972dc22bef5d0b80aa
Reviewed-on: https://chromium-review.googlesource.com/1102178
Commit-Queue: Peter Collingbourne <pcc@chromium.org>
Reviewed-by: Dirk Pranke <dpranke@chromium.org>
Reviewed-by: John Budorick <jbudorick@chromium.org>

[modify] https://crrev.com/5903a1e038f51f827cbfcc2ce99c3c7d5ed83faf/masters/master.chromium.perf.fyi/slaves.cfg

I let the perf bots run in this configuration for two weeks and there were no compile failures (there were some runhooks failures but they seem unrelated).

I also collected build durations for each of the configurations and calculated the median over the two weeks for each of the bots. They are as follows:

244 (n1-standard-8): 1311s
245 (n1-highmem-8): 1159s
246 (n1-standard-16): 1100s
247 (n1-highmem-16): 1157s

Next I will look at the median duration on the existing perf bots. That should let us know how much additional capacity we'll need.

Looking at the 2000 most recent builds on "Android Builder Perf" on chromium.perf the median duration is 1106s. So assuming that we stick with n1-standard-8 machines the duration will increase by around 200s on average, or around 20%. I think that means that we'll need 1 or 2 more machines on "Android Builder Perf", probably 2 to be safe.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/tools/build/+/c38d4ba2777a93d09121183258d229cfadd962fa

commit c38d4ba2777a93d09121183258d229cfadd962fa
Author: Peter Collingbourne <pcc@chromium.org>
Date: Fri Jul 20 19:24:03 2018

Revert "Assign slave{244..247}-c1 to "Android CFI Builder Perf FYI"."

This reverts commit 5903a1e038f51f827cbfcc2ce99c3c7d5ed83faf.

Reason for revert:
The experiment has finished and we have downgraded slave270-c1
to n1-standard-8.

Original change's description:
> Assign slave{244..247}-c1 to "Android CFI Builder Perf FYI".
> 
> Each of these machines has a different configuration:
> 244: n1-standard-8
> 245: n1-highmem-8
> 246: n1-standard-16
> 247: n1-highmem-16
> 
> This is part of an experiment to discover the minimum capacity
> required to run Android ThinLTO perf bots. Once we have found the
> minimum, this change will be reverted and we will update the
> configuration of slave270-c1.
> 
> Bug: 821860,  853050 
> Change-Id: I99ad99c396a0cad6c731a7972dc22bef5d0b80aa
> Reviewed-on: https://chromium-review.googlesource.com/1102178
> Commit-Queue: Peter Collingbourne <pcc@chromium.org>
> Reviewed-by: Dirk Pranke <dpranke@chromium.org>
> Reviewed-by: John Budorick <jbudorick@chromium.org>

TBR=dpranke@chromium.org,pcc@chromium.org,jbudorick@chromium.org

# Not skipping CQ checks because original CL landed > 1 day ago.

Bug: 821860,  853050 
Change-Id: I5fa0180fc70b0f673abea6d58b0024d219d28580
Reviewed-on: https://chromium-review.googlesource.com/1142903
Reviewed-by: John Budorick <jbudorick@chromium.org>
Commit-Queue: Peter Collingbourne <pcc@chromium.org>

[modify] https://crrev.com/c38d4ba2777a93d09121183258d229cfadd962fa/masters/master.chromium.perf.fyi/slaves.cfg

FYI, the name of the FYI Android CFI builders now are:
https://ci.chromium.org/p/chrome/builders/luci.chrome.ci/android-cfi-builder-perf-fyi
https://ci.chromium.org/p/chrome/builders/luci.chrome.ci/android_arm64-cfi-builder-perf-fyi