|Re: [eigen] [patch] Aligned new/delete|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] [patch] Aligned new/delete
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Thu, 7 May 2009 20:42:20 +0200
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=/XADACSNUtMhohzFiLy3oFnSyhS9Fd5mN23Ruq5V2s8=; b=stBqJX7oScsxIYbVXaSa4BjhOVsyVMS9EL/QOq2MLrdW+8N9CVJuGAW9TPB0HKm/2h NFNDrAIU9mFUwLeR12atSDLvoq9tvY9qrx4JZY+xHnV5R2rrx0Xp27Xp0FI18bkuI7FA SK3wibHMEYJ5ujStz33jps6gWSclZX4ZbnEAw=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type:content-transfer-encoding; b=AZtDnFPpmS3pSYtBMjDu+EqEFM0pN27x15m5sM3X+Nvna3K/d7cMjedGD/wqgft1f2 NyFge82KyalxMaJFbZzq0L/EijPLgkXWDKiFwLeIggEJ71Daz/mbH/zyK+O+SlRch28L sIqHxnCJdmvXWGWI3ijiiJLstnDmy42kPQfYw=
can someone with experience with exceptions and overloading new/delete
2009/5/6 Hauke Heibel <hauke.heibel@xxxxxxxxxxxxxx>:
> I made the parameter std::nothrow_t unnamed and ... added the (still
> *shame*) missing no-throw delete.
> On Wed, May 6, 2009 at 3:40 PM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
>> This needs review by someone more competent than me in these
>> matters... I just learnt the meaning of the throw() declaration.
>> One remark: the parameter nt is unused, so perhaps make it unnamed to
>> avoid warnings?
>> 2009/5/6 Hauke Heibel <hauke.heibel@xxxxxxxxxxxxxx>:
>>> Two things are covered in this patch.
>>> 1. Each operator new needs its corresponding operator delete. This was
>>> not the case for in-place new/delete.
>>> 2. Operators new were declared with throw() which is not correct since
>>> in the current standard implementation they may throw std::bad_alloc.
>>> The throw() declaration was moved from new to delete which must never
>>> 3. Added no-throw new/delete. In order to allow user using no-throw
>>> new and delete I just added them. The difference is that instead of
>>> throwing an exception they return 0 in case of an allocation failure.
>>> For the in-place new and delete I piped the calls to the global
>>> in-place new and delete functions since actually no allocation is
>>> performed. It is the job of the user to take care that when using
>>> in-place new and delete the memory is properly aligned.
>>> I ran the unit tests and all besides the new matrix exponential passed
>>> - could not submit to the dashboard since the daily limit is reached.