CCRL Discussion Board

Slashdot has a lame roundup of existing Z68 mainboards. Among the comments was an interesting tidbit Re: I want more RAM Slots : Once you get above about 8GB of RAM, you really shouldn't trust it for anything serious without ECC. Even the best listed bit-error rates (which are all pretty much a gues...

Kirill, I think what you're saying is that the quad channel SB-E appeals to you because it permits 32 or 64GB memory configurations without the expense of registered ECC and a server mainboard with 16 or 18 DIMM sockets on two or three channels. You're not really looking at saturating 50GB/s memory ...

The only way to verify a WDL50 table will be by verifying a DTZ50 table and re-converting it to WDL50 twice and comparing the two copies. You can generate once and then exhaustively probe each WDL50 position against any verified DTZ >= 50 table. Can't you? It's almost the same thing, with a little ...

Kirill, I didn't mention SandyBridge-E because I haven't decided which performance attribute is most important. The four channel memory controller rocks if your computation is memory bound in the right way. Sequential access (reading from queues) will have trouble saturating this, even with many cor...

AMD's Bulldozer was released today. Underwhelming. Almost inexplicable given the increase in die size (2 billion transistors). You really have to hope for AMD that there are some performance gremlins to address to unleash the full potential. This part desperately needs some remedial gremicide to bec...

Kirill: Amassing some preliminary dominant move statistics is much on my mind. I was planning to use KR-KN as my first test case. 462 * 64 * 64 < 2^9 * (2^3)^4 = 2^21 = 2 million. Not out of the range for loading into memory in an R workspace. I could slap some columns onto the data frame concerning...

Kirill: My default is to view this as a big computation that's performed in full glory twice or thrice on highly specialized platforms. I can't get my head around terabytes for every man. Perhaps you think bigger than I do. I see an unconquered mountain, you see a tourist attraction. But I appreciat...

Guy: One chapter of one book by Dijkstra was the biggest influence on my programming career taking everything together. Don't recall the title of that particular book, but I do know it was on a reading list promulgated in one of P.J. Plaugher's old columns--if memory serves half as well as it used t...

Kirill, a quick comment on distributed file systems. HDF5 lives in the HPC space and has some application level support for this (mounds of documentation to wade through). I have no idea if what it provides overlaps with your ambitions. On another front, distributed file systems is the raison d'être...

I'm stealing ten minutes to skim the Schaeffer paper kindly posted by Guy. From page 8: To reduce the cost, the program only iterates over all positions until a "small" number of changes occurs in an iteration. The positions that change value are saved in a queue. This is in line with a mu...

KPPP-KPP pawn slices I wish I had more time for this discussion. I found an hour over the weekend to compute the pawn slices of KPPP-KPP. R has a "partitions" package to compute the seven partitions of 5, so I don't have to (five pawn stacked in one file, down to all five pawns in distinc...

Kirill, your points on independence and accuracy are good ones, though I would have guessed DTC were similarly independent. Perhaps I don't fully grasp the metrics yet. The chess glossary defines DTC as "any conversion of material" which may be "either a promotion or capture". Do...

Guy, I downloaded the spreadsheet set you referenced. The definitions on the first sheet will be useful to me starting out. There's a lot of data there, but it doesn't look like an easy data set to visualize. On my own map, it could serve as an annotation that certain kinds of specialized test cases...

Guy, That was a quick response. I've seen many of your previous posts here in passing. It's good to hear from you. I presume that generating WDL tables is expedited over full DTx tables due to less memory pressure in space. I'm not sure it buys you much in time at the memory manipulation layer, sinc...

By Gaviota engine I meant tbtools for EGTB probing. UPDATE: Found the "edit" button. I'm not used to that. Previous post amended. I spend a fair amount of time on sites where the rules of engagement are fire and forget.

Hello everyone! This is my first post on CCRL after exchanging a couple of private emails with Kirill, who encouraged me to post here. Long ago I came up with an unorthodox approach to data compression. My approach--which I have yet to prove viable for objects bigger than 1kB--requires extraordinari...