From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Mosberger Date: Thu, 10 May 2001 04:14:08 +0000 Subject: Re: [Linux-ia64] kernel update (relative to 2.4.4) Message-Id: List-Id: References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-ia64@vger.kernel.org >>>>> On Wed, 9 May 2001 09:21:49 -0500 (CDT), Jack Steiner said: Jack> BTW, The SN platform does have a hi-res synchronized clock that can be Jack> mmapped & read from user space. The clock is located in the "chipset" Jack> and is synchronized across all nodes/FSBs. Good. This was indeed the working assumption: on machines where ITCs aren't synchronous, there needs to be some other mechanism to obtain fine-grained time stamps. So, I think what you'll need is a device driver for your chipset which will allow mapping the hi-res clock into user space where libc can use it to implement the *clock* routines. In the kernel, we can make gettimeoffset() a platform-specific function, so SN1 and other platforms with non-synchronous clocks can use an external hi-res timer to implement it. The only challenge is to decide when to use ITCs and when to use other hardware. Unfortunately, I don't believe SAL provides info on whether the ITC clocks are synchronized (I think it should!). For now, we could implement it such that an external hi-res timer is used in favor of ITCs whenever such hardware is present. --david