|Re: [eigen] 3.0.4 coming soon, testing appreciated|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] 3.0.4 coming soon, testing appreciated
- From: Gael Guennebaud <gael.guennebaud@xxxxxxxxx>
- Date: Wed, 23 Nov 2011 23:16:45 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:from:date:message-id:subject:to :content-type:content-transfer-encoding; bh=TMp7duuvRXYO0YWwnxFliMit2nJcdG2dcCkEF/Ya9KA=; b=k6Tz/KIT9ht9yog6RR4u4pf4wHcmCEEyJ4Dmj0HoJ7+yrmHa7QBXIJfyn+SXNcToMe OtDMm9LU6Py0cUoucPJ8xBkgcBLSdmXQ0YIH06BKuGoI+RZAuSICycLu23buylVSqRCt Cem71qJyFm/MUzd5FWdHkJgc69sXoJQ1Zpw6k=
here it takes about 3s with ICC 12.1 and 100 iterations.
On Wed, Nov 23, 2011 at 11:09 PM, Rhys Ulerich <rhys.ulerich@xxxxxxxxx> wrote:
>> it seems that you compiled the tests without optimizations as they
>> should not take more than 1s.
> There's an EIGEN_REPEAT=100 in effect in those results. Each
> iteration is definitely under 1s. Please advise if you think the
> timings still look suspect.
>> Also comparing the running times of different runs does not make sense
>> because the involved matrix sizes are different for each run (unless
>> you specify a unique seed, e.g.: ./product_trmm_1 s0).
> That is good to know. I'll do that while gathering the details with
> perf as Benoit suggested.
> - Rhys