This is the mail archive of the
gdb@sources.redhat.com
mailing list for the GDB project.
Re: Always cache memory and registers
- From: "John S. Yates, Jr." <jyates at netezza dot com>
- To: <gdb at sources dot redhat dot com>
- Date: Mon, 23 Jun 2003 15:59:55 -0400
- Subject: Re: Always cache memory and registers
- Organization: Netezza Corporation
- References: <1056381193.18735.ezmlm@sources.redhat.com>
- Reply-to: "John S. Yates, Jr." <jyates at netezza dot com>
----- Original Message -----
From: Andrew Cagney <ac131313@redhat.com>
Sent: Sun, 22 Jun 2003 18:26:13 -0400
> The only proviso being that the current cache and
> target vector would need to be modified so that the
> cache only ever requested the data needed, leaving it
> to the target to supply more if available (much like
> registers do today). The current dcache doesn't do
> this, it instead pads out small reads :-(
How much freedom would the target have? If a random
byte in the middle of a cache-line was requested could
a target back up the starting address to the beginning
of the cache-line? What happens if host requests byte
at address X and target returns data from range Y to Z
where X does not fall within that range? (Suggestion:
host places returned data in cache and reattempts the
access; this will lead to a request of X from the target;
presumably the target has its own rationale for why it
is returning unrequested data and will eventually return
the request byte; at that point the host cache access
will succeed returning the request byte to its caller.)
/john
--
John S. Yates, Jr. 508 665-6897 (voice)
Netezza Inc 508 665-6811 (fax)
200 Crossing Blvd. Framingham, MA 01701