which bus numbers do these correspond to again? I should know this but don't.
thanks
ron
* ron minnich rminnich@lanl.gov [040322 18:21]:
which bus numbers do these correspond to again? I should know this but don't.
It refers to the Hypertransport Links, Down (LDT0), Across (LDT1), UP (LDT2). Is that what you mean?
Stefan
On Mon, 22 Mar 2004, Stefan Reinauer wrote:
It refers to the Hypertransport Links, Down (LDT0), Across (LDT1), UP (LDT2). Is that what you mean?
yes, thanks.
I think I picked these names, and if I did I will take the blame: they are hard to remember.
Should we bite the bullet and change the names? ldt0, ldt1, ldt2?
ron
* ron minnich rminnich@lanl.gov [040322 18:30]:
It refers to the Hypertransport Links, Down (LDT0), Across (LDT1), UP (LDT2).
I think I picked these names, and if I did I will take the blame: they are hard to remember.
These names are also spread across several pieces of AMD64 and Hypertransport documentation, since it illustrates hypertransport graphs of different wiring quite nice. Still, it's unclear from a technical point of view.
Should we bite the bullet and change the names? ldt0, ldt1, ldt2?
Should not be too hard, since we are not really using them heavily. I still have some defines used in coherent_ht.c which should go away and replaced by something more sophisticated.
Stefan
On Mon, 22 Mar 2004, Stefan Reinauer wrote:
- ron minnich rminnich@lanl.gov [040322 18:30]:
It refers to the Hypertransport Links, Down (LDT0), Across (LDT1), UP (LDT2).
I think I picked these names, and if I did I will take the blame: they are hard to remember.
These names are also spread across several pieces of AMD64 and Hypertransport documentation, since it illustrates hypertransport graphs of different wiring quite nice. Still, it's unclear from a technical point of view.
arg. So we can pick either one and it won't sync with different bits of documentation.
how about: ldt0_down, ldt1_across, ldt2_up?
ron
* ron minnich rminnich@lanl.gov [040322 18:48]:
arg. So we can pick either one and it won't sync with different bits of documentation.
how about: ldt0_down, ldt1_across, ldt2_up?
I think going with LDT[012] is safe. A comment in one or two algorithms explaining down, up, across should be enough... I vote for ldt[012] for the names, as it will allow machines with possibly more links without changing the names
Stefan
On Mon, 22 Mar 2004, Stefan Reinauer wrote:
- ron minnich rminnich@lanl.gov [040322 18:48]:
arg. So we can pick either one and it won't sync with different bits of documentation.
how about: ldt0_down, ldt1_across, ldt2_up?
I think going with LDT[012] is safe. A comment in one or two algorithms explaining down, up, across should be enough... I vote for ldt[012] for the names, as it will allow machines with possibly more links without changing the names
OK we'll try to cut this in soon. Or if you want to go for it that's fine too.
ron
* ron minnich rminnich@lanl.gov [040322 18:55]:
OK we'll try to cut this in soon. Or if you want to go for it that's fine too.
Looking through the code I only find 2 occurences of the UP/DOWN/ACROSS names:
Config.lb: register "up" = "{.chip = &amd8131, .ht_width=16, .ht_speed=600}"
Is this information ever being used? Otherwise I suggest we get rid of it completely.
The other is auto.c and coherent_ht.c: #define CONNECTION_0_2 UP #define CONNECTION_1_3 UP
Any others?
Stefan
Stefan Reinauer stepan@suse.de writes:
- ron minnich rminnich@lanl.gov [040322 18:55]:
OK we'll try to cut this in soon. Or if you want to go for it that's fine too.
Looking through the code I only find 2 occurences of the UP/DOWN/ACROSS names:
Config.lb: register "up" = "{.chip = &amd8131, .ht_width=16, .ht_speed=600}"
Is this information ever being used? Otherwise I suggest we get rid of it completely.
It is currently unused. I don't think it ever has been.
The other is auto.c and coherent_ht.c: #define CONNECTION_0_2 UP #define CONNECTION_1_3 UP
Any others?
All I know about.
Eric
On Mon, 2004-03-22 at 10:36, Stefan Reinauer wrote:
- ron minnich rminnich@lanl.gov [040322 18:30]:
It refers to the Hypertransport Links, Down (LDT0), Across (LDT1), UP (LDT2).
I think I picked these names, and if I did I will take the blame: they are hard to remember.
These names are also spread across several pieces of AMD64 and Hypertransport documentation, since it illustrates hypertransport graphs of different wiring quite nice. Still, it's unclear from a technical point of view.
Should we bite the bullet and change the names? ldt0, ldt1, ldt2?
Should not be too hard, since we are not really using them heavily. I still have some defines used in coherent_ht.c which should go away and replaced by something more sophisticated.
From my understanding of the AMD docs, the mapping between LDT0-2 to
up, down, across is not that simple and static. "Up" can be either LTD0 or LDT1 or LDT2 so as for "down" and "across".
The terminology 'link' is heavily and badly overloaded in LinuxBIOS, we have link for HT links and link for how many branches a device has (bridges have two 'links'), and there are some other link I have no idea about.
The same happens for the field name "enable" of some structures. Some of them are verbs (enable() operation), some others are actually adjectives (been 'enabled'). I though that I am the ONLY who can't speak English in LinuxBIOS project.
Ollie
Li-Ta Lo ollie@lanl.gov writes:
On Mon, 2004-03-22 at 10:36, Stefan Reinauer wrote:
- ron minnich rminnich@lanl.gov [040322 18:30]:
It refers to the Hypertransport Links, Down (LDT0), Across (LDT1), UP (LDT2).
I think I picked these names, and if I did I will take the blame: they are hard to remember.
These names are also spread across several pieces of AMD64 and Hypertransport documentation, since it illustrates hypertransport graphs of different wiring quite nice. Still, it's unclear from a technical point of view.
Should we bite the bullet and change the names? ldt0, ldt1, ldt2?
Should not be too hard, since we are not really using them heavily. I still have some defines used in coherent_ht.c which should go away and replaced by something more sophisticated.
From my understanding of the AMD docs, the mapping between LDT0-2 to
up, down, across is not that simple and static. "Up" can be either LTD0 or LDT1 or LDT2 so as for "down" and "across".
The terminology 'link' is heavily and badly overloaded in LinuxBIOS, we have link for HT links and link for how many branches a device has (bridges have two 'links'), and there are some other link I have no idea about.
The same happens for the field name "enable" of some structures. Some of them are verbs (enable() operation), some others are actually adjectives (been 'enabled'). I though that I am the ONLY who can't speak English in LinuxBIOS project.
The challenge is the lack of good alternative names. Suggestions for better alternatives are welcome before we freeze the freebios2 tree.
Eric
-
ebiederman＠lnxi.com -
Li-Ta Lo -
ron minnich -
Stefan Reinauer