I might be alone, but I love statistics and numbers.. so I pulled some from the tester db:

Trunk auto tester:
  oldest build: 2010-09-05 03:53:13 (469 days)
  number of build runs: 18673
  average builds per day: 39.8145

Pull auto tester:
  oldest build: 2011-12-14 01:09:38 (4 days)
  number of build runs: 2702
  average builds per day: 675.5 .. but that's wrong since day 1 and 2 had much much fewer runs:

+------------------+----------+
| date(start_time) | count(*) |
+------------------+----------+
| 2011-12-14       |        6 |
| 2011-12-15       |       32 |
| 2011-12-16       |      596 |
| 2011-12-17       |     1331 |
| 2011-12-18       |      738 |
+------------------+----------+

There would be even more pull test runs except that I haven't (and likely won't) start the pull tester on Sean's osx box, so no osx testing for pulls right now.  That's hardware he shares with other stuff and the pull tester would keep it busy 24/7 and that's too much to ask.

Hardware involved (will go up on the website at some point):

1 freebsd/32 - dual core p4 3ghz
1 freebsd/64 - vm on a quad core core2 box 2.66ghz
1 linux/32   - dual core pD 2.8ghz
2 linux/64   - 1) the above quad core core2 box, shares with the vm, obviously
             - 2) dual core athlon 64 X2
3 windows/32 - 1) dual core atom N280 1.66ghz
               2) dual core pD 3ghz
               3) ec2 c1.medium (2 core Xeon E5410 2.33ghz)

Any other slices of data anyone is interested in seeing?

Later,
Brad

I noticed that on 15 Dec, the time required for to run DMD tests
dropped by 20-30% on all platforms.
I reduced CTFE memory consumption at that time, but I wouldn't expect
it to impact the compiler tests much.
It's about the same time that the pull tester started. Did you improve
something?

On 19 December 2011 06:03, Brad Roberts <braddr at puremagic.com> wrote:
> I might be alone, but I love statistics and numbers.. so I pulled some from the tester db:
>
> Trunk auto tester:
> ?oldest build: 2010-09-05 03:53:13 (469 days)
> ?number of build runs: 18673
> ?average builds per day: 39.8145
>
> Pull auto tester:
> ?oldest build: 2011-12-14 01:09:38 (4 days)
> ?number of build runs: 2702
> ?average builds per day: 675.5 .. but that's wrong since day 1 and 2 had much much fewer runs:
>
> +------------------+----------+
> | date(start_time) | count(*) |
> +------------------+----------+
> | 2011-12-14 ? ? ? | ? ? ? ?6 |
> | 2011-12-15 ? ? ? | ? ? ? 32 |
> | 2011-12-16 ? ? ? | ? ? ?596 |
> | 2011-12-17 ? ? ? | ? ? 1331 |
> | 2011-12-18 ? ? ? | ? ? ?738 |
> +------------------+----------+
>
> There would be even more pull test runs except that I haven't (and likely won't) start the pull tester on Sean's osx box, so no osx testing for pulls right now. ?That's hardware he shares with other stuff and the pull tester would keep it busy 24/7 and that's too much to ask.
>
> Hardware involved (will go up on the website at some point):
>
> 1 freebsd/32 - dual core p4 3ghz
> 1 freebsd/64 - vm on a quad core core2 box 2.66ghz
> 1 linux/32 ? - dual core pD 2.8ghz
> 2 linux/64 ? - 1) the above quad core core2 box, shares with the vm, obviously
> ? ? ? ? ? ? - 2) dual core athlon 64 X2
> 3 windows/32 - 1) dual core atom N280 1.66ghz
> ? ? ? ? ? ? ? 2) dual core pD 3ghz
> ? ? ? ? ? ? ? 3) ec2 c1.medium (2 core Xeon E5410 2.33ghz)
>
> Any other slices of data anyone is interested in seeing?
>
> Later,
> Brad
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos

Hrm, I hadn't added any of the additional machines to the mix at that point, so it's not that.  Those entered mostly on the 16 and 17th, if I remember right.  I can't think of anything I did around that point that would have sped anything up.

With multiple machines eligible to run the tests now, timing from one run to another can vary quite a bit just due to differences in hardware generations.  That's a negative side effect, but I need the extra throughput.  Today I'm going to add hostname tracking to each build.  The current ip address isn't sufficient since everything inside my house appears as just the one external address and there's 2 machines handling some of the platforms.

On 12/19/2011 4:25 AM, Don Clugston wrote:
> I noticed that on 15 Dec, the time required for to run DMD tests
> dropped by 20-30% on all platforms.
> I reduced CTFE memory consumption at that time, but I wouldn't expect
> it to impact the compiler tests much.
> It's about the same time that the pull tester started. Did you improve
> something?
> 
> On 19 December 2011 06:03, Brad Roberts <braddr at puremagic.com> wrote:
>> I might be alone, but I love statistics and numbers.. so I pulled some from the tester db:
>>
>> Trunk auto tester:
>>  oldest build: 2010-09-05 03:53:13 (469 days)
>>  number of build runs: 18673
>>  average builds per day: 39.8145
>>
>> Pull auto tester:
>>  oldest build: 2011-12-14 01:09:38 (4 days)
>>  number of build runs: 2702
>>  average builds per day: 675.5 .. but that's wrong since day 1 and 2 had much much fewer runs:
>>
>> +------------------+----------+
>> | date(start_time) | count(*) |
>> +------------------+----------+
>> | 2011-12-14       |        6 |
>> | 2011-12-15       |       32 |
>> | 2011-12-16       |      596 |
>> | 2011-12-17       |     1331 |
>> | 2011-12-18       |      738 |
>> +------------------+----------+
>>
>> There would be even more pull test runs except that I haven't (and likely won't) start the pull tester on Sean's osx box, so no osx testing for pulls right now.  That's hardware he shares with other stuff and the pull tester would keep it busy 24/7 and that's too much to ask.
>>
>> Hardware involved (will go up on the website at some point):
>>
>> 1 freebsd/32 - dual core p4 3ghz
>> 1 freebsd/64 - vm on a quad core core2 box 2.66ghz
>> 1 linux/32   - dual core pD 2.8ghz
>> 2 linux/64   - 1) the above quad core core2 box, shares with the vm, obviously
>>             - 2) dual core athlon 64 X2
>> 3 windows/32 - 1) dual core atom N280 1.66ghz
>>               2) dual core pD 3ghz
>>               3) ec2 c1.medium (2 core Xeon E5410 2.33ghz)
>>
>> Any other slices of data anyone is interested in seeing?
>>
>> Later,
>> Brad
>> _______________________________________________
>> phobos mailing list
>> phobos at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/phobos
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos