freenode/lisp - IRC Chatlog
Search
10:56:16
p_l
shka: yes, but the latencies make it very far from "let's just think of memory instead of multiple completely disparate systems"
10:57:00
p_l
and in my experience, companies balk when I suggest "ok, let me hit ebay, I'll have the 56GBit low-latency fabric ordered by tomorrow on the cheap" ;)
10:59:19
shka
anyway, those would be different beast, i agree, but this approach has some advantage over ultra expensive huge IO mainframes
10:59:56
p_l
most companies that lease mainframes have very specific reasons for doing so (combination of RAS requirements, power, software base, etc.)
11:01:05
flip214
you can get 56gbit IB, 2 dual-port adapters and 2 cables (so two machines can talk 112gbps) for < €1000
11:03:05
flip214
so all of the data needs to have been seen before the first data byte can be transmitted
11:04:25
p_l
shka: there are some low-level details that impact how fast you can go with Ethernet without changing it into non-Ethernet, including how it's designed for Store-and-Forward data
11:05:25
p_l
shka: IB doesn't have store-and-forward inside subnet (most sites anyone sees are single subnet)
11:08:17
p_l
Netherlands used Myrinet on mixed Myrinet/10GbE links, in poland I think we did something shitty with plain IP... ;)
11:08:45
p_l
flip214: IB uses IPv6 *addressing*, and intra-subnet packets have 16bit address for wormhole routing
11:09:32
p_l
the 16bit addresses are internal affair of the subnet that software stack is not supposed to bother itself with (outside of management software for subnet itself)
11:11:54
p_l
it's much harder to do wormhole switching when you have to match complex 128bit addressing to ports ;)
19:27:33
rpg
I suspect that CL-DOT cannot handle subgraphs, but I'm not 100% sure. Can anyone confirm?
21:29:51
fiddlerwoaroof
When I quickload it in Lispworks, I get an error about defining (METHOD DESCRIBE-OBJECT :AFTER (SYMBOL T)) visible from package Common-Lisp
21:31:13
fiddlerwoaroof
I guess this settles it: "Users can write methods for describe-object for _their own_ classes if they do not wish to inherit an implementation-supplied method."
21:32:52
aeth
I would guess that quite a lot of things do not run in LispWorks. e.g. I try to test in everything but clisp (too many missing non-standard features) and the proprietary Lisps. And things frequently have broken in subtle ways by the next time I test in CCL, ECL, etc.
21:33:43
aeth
e.g. Once there was an issue with the way I coerced some numbers between numeric types that only showed up in CCL iirc.
21:35:31
aeth
Anything large and complicated enough has a good chance to not run in a random CL without testing finding some minor things that the other CLs tolerate.
21:39:10
fiddlerwoaroof
I've found that usually it works just fine and, when it breaks, the fix is fairly simple
21:39:44
aeth
There are a lot of things that aren't in the standard but are de facto standard, and an implementation not complying can cause pain. e.g. CLISP has a 64-bit fixnum that's far too small, and doesn't have double-float or single-float specialized arrays.
21:40:05
aeth
So a library that assumes single-float specialized arrays and assumes that the Lisp will have them and type check and error if the type is wrong will not work as intended on CLISP.
21:41:45
fiddlerwoaroof
Yeah, but this isn't really a question of unspecified behavior: the standard doesn't allow users to defmethod DESCRIBE-OBJECT on standard classes.
21:44:35
aeth
My example was bad... Looks like ECL doesn't type error like SBCL and CCL or just put the double there (because it's a T array) like in CLISP.
21:51:14
aeth
fiddlerwoaroof: Well, a similar issue pops up, then. Sometimes only one implementation follows the standard correctly, and the bug is only spotted because the code tries to be portable.
22:35:39
rpg
I was just looking at some ways to draw, e.g., DAGs with sink nodes, and the techniques rely on putting certain graph nodes is a subgraph, and then setting the rank on the subgraph.
22:37:40
scymtym
rpg: https://github.com/scymtym/cl-dot/tree/wip-clusters i'm not sure clusters do what you describe, though. the way i use them is as nodes containing nodes
22:38:19
rpg
Can you draw links from nodes inside a cluster to nodes outside the cluster, or only to/from the cluster nodes themselves?
22:38:32
scymtym
also, the branch has some unrelated stuff because i'm waiting for https://github.com/michaelw/cl-dot/pull/21 to be merged
22:40:14
scymtym
for edges to connect to clusters, the graph may need the compound=true (or similar, i don't remember exactly) attribute
22:49:28
rpg
I think I'll have to tweak it a little, because I'm not going to add clusters and then add their components, but rather add certain nodes to subgraphs in order that I can control their rank.
22:50:02
rpg
I'm not sure why, but it seems like things like [rank=sink] only work as subgraph attributes, not node attributes.