Processes' memory
|
Prev: GPS resolutio (was Re: SYSENTER/SYSEXIT_vs._SYSCALL/SYSRET)
Next: F r e e C I S C O C e r t i f i c a t i o n s
From: EricP on 29 Jan 2010 00:31 (groups trimmed) Anne & Lynn Wheeler wrote: > On 01/28/10 20:16, Andy "Krazy" Glew wrote: >> Lynn: can you expand on this? "Shared Variable" sounds like it is >> relevant to parallelism or concurrency; system services sounds like >> syscalls. Abominations sounds bad, but makes me curious. >> >> I assume "ape" = A Programming Language. Iverson? Vectors? +/array? > > note ... i see the above in email and in google ... but it hasn't shown > up at my usenet server ... there seems to be some connectivity problem > with usenet "gateways" .... ibm-main mailing list usenet gateway > recently has had several things go totally missing; ... in any case. > > re: > http://www.garlic.com/~lynn/2010c.html#28 Processes' memory > > say little like peek/poke with shared memory ... but process could > either be an asynchronous operation or a synchronous system services > (apl masking kernel call behind shared variable). > http://en.wikipedia.org/wiki/APL_Shared_Variables > > a discussion of shared variable here ... avoiding mentioning the > cms\apl abomination > http://www.jsoftware.com/papers/APLDesign.htm > > there was cambridge science center that did virtual machines, cms, > cms\apl, gml, etc. > > the phili science center had falkoff, iverson, and apl\360 > > the palo alto science center does 5100, apl\cms and the 370/145 apl > microcode assist > > ..... > > cms\apl then shows up in large scale applications in the (initially > cp67-based) internal HONE system. > > a lot of apl was used for lots of things that are done in spreadsheets > today As many timeshare companies discovered :-( > the cambridge science center did a little external time-sharing service > .... for other internal locations ... as well as some number of > non-employees ... students & others from various institutions of higher > learning in the boston area. > > a big early use of cms\apl on the cambridge science center cp67 system > .... was the corporate hdqtrs business planning people who loaded the > most sensitive and most valuable corporate resource on the cambridge > system (all the customer information) and did cms\apl business > modeling. Big innovation from cms\apl was large enuf workspace size and > the ability to do file i/o involving large amounts of data ... for real > world applications. the cambridge cp67 had to had to have sufficient > security that the most valuable of corporate assets didn't leak out to > unauthorized & non-employee users. > Bob Bernekey did some work on parallel APL The Role of APL and J in High-performance Computation (1993) http://www.snakeisland.com/aplhiper.htm Eric (who does not knows apl or HPC)
From: Terje Mathisen "terje.mathisen at on 29 Jan 2010 01:30 Anne & Lynn Wheeler wrote: > prior terminal emulation was via real 327x controller, real 327x coax > cable, PC card that emulated 327x terminal ... file upload/download was > emulated screen input/output ... and ran into various bottlenecks. Bottlenecks indeed! :-) I had written (from scratch) an implementation of "Super Kermit" (up to 64K of window size, up to 9K packets and selective retransmit) in Pascal a year or two previously, then decided I wanted to see if I could make it run on our mainframe, using the PC in VT100/VT220 mode via a protocol emulator. I modified the Pascal-VS (?) mainframe Kermit to work with large packets so it could fill an entire screen buffer, on the PC side the receiver had to implement a very simplified VT terminal rule set so that all cursor and font escape sequences could be ignored. It was an ugly hack, but it turned out to run at least as fast as a 3270PC or 3270AT with a hardware coax connection. :-) Terje -- - <Terje.Mathisen at tmsw.no> "almost all programming can be viewed as an exercise in caching"
From: Anne & Lynn Wheeler on 29 Jan 2010 01:48 Anne & Lynn Wheeler <lynn(a)garlic.com> writes: > later there was the a74 (7437) ... which was much more like a real 370 > in small form factor (with several press/news articles from 1988): > http://www.garlic.com/~lynn/2002d.html#4 IBM Mainframe at home > > some a74 changes I tried to get released (including paged mapped > filesystem): > http://www.garlic.com/~lynn/2003f.html#56 ECPS:VM DISPx instructions re: http://www.garlic.com/~lynn/2010c.html#15 Processes' memory wiki page http://en.wikipedia.org/wiki/PC-based_IBM-compatible_mainframes mentions both 5100 and xt/370 ... and then jumps to p/370 ... skips over a74 all together (which was in expansion box with big thick cable to back of PC). the xt/370 cards were offered for at/370 ... big difference was PC/AT hard disks were much faster ... which was big boost. for version 2 of xt/370 rel.2 (9/84), cp88 "dos bypass" (i.e. pc/370 support software running on the 8088 side ... getting some pathlength boost by doing some stuff directly rather than everything going thru dos) and my cms paged-mapped filesystem for cms (on the 370 side) http://www.garlic.com/~lynn/2010c.html#14 Processes' memory i shipped cms PAM (paged-mapped) filesystem internally and on xt/370 ... but it never got out in standard cms release ... misc. past posts mentioning paged mapped filesystem http://www.garlic.com/~lynn/subtopic.html#mmap minor example of PAM versus non-PAM for constrained real storage ... was traditional filesystem had to load executable image from standard file with simulated i/o ... several executable images were larger than available real storage ... so it involved bringing in pieces from emulated cms area on xt dos disk into virtual pages in 370 memory .... and then paging those pages out to emulated paging area on (same) xt dos disk ... so additional parts of the executable images could be read (from xt disk) into additional virtual pages (also going back out to the xt disk). eventually when the whole process was complete ... it would start execution ... which tended to be in virtual pages now out in paging area (on xt disk) ... replacing virtual pages (having to be written out to paging area (on xt disk). PAM just sets up page mapped tables to point directly at the executable and starts execution (w/o having to copy the executable image thru 370 memory from one area on the xt disk to another area on the same xt disk). It can actually do a little bit better than that ... rather than having to page fault each individual 4k page from the paging area on the xt disk ... it can concurrently start fetch of several pages from the filesystem while allowing execution (appropriately serializing execution by fiddling the page invalid bits). In traditional CMS environment with much less real storage constraint .... whole images could still be page mapped at one time ... and a much higher level of multiple block asyncronous concurrent page transfers could go on (dynamically tuned to match the amount of available real storage). -- 42yrs virtualization experience (since Jan68), online at home since Mar1970
From: Morten Reistad on 29 Jan 2010 03:45 In article <m3ljfhzfjv.fsf(a)garlic.com>, Anne & Lynn Wheeler <lynn(a)garlic.com> wrote: > >On 01/28/10 20:16, Andy "Krazy" Glew wrote: >> Lynn: can you expand on this? "Shared Variable" sounds like it is >> relevant to parallelism or concurrency; system services sounds like >> syscalls. Abominations sounds bad, but makes me curious. >> >> I assume "ape" = A Programming Language. Iverson? Vectors? +/array? > >note ... i see the above in email and in google ... but it hasn't shown >up at my usenet server ... there seems to be some connectivity problem >with usenet "gateways" .... ibm-main mailing list usenet gateway >recently has had several things go totally missing; ... in any case. I have a slightly different problem; I see the replies to posts by you, and several others, a day or so before I see the post. And I get my own posts from around 3 weeks back reinjected into the news stream somewhere. -- mrr
From: Peter Flass on 29 Jan 2010 19:07
Morten Reistad wrote: > In article <m3ljfhzfjv.fsf(a)garlic.com>, > Anne & Lynn Wheeler <lynn(a)garlic.com> wrote: >> On 01/28/10 20:16, Andy "Krazy" Glew wrote: >>> Lynn: can you expand on this? "Shared Variable" sounds like it is >>> relevant to parallelism or concurrency; system services sounds like >>> syscalls. Abominations sounds bad, but makes me curious. >>> >>> I assume "ape" = A Programming Language. Iverson? Vectors? +/array? >> note ... i see the above in email and in google ... but it hasn't shown >> up at my usenet server ... there seems to be some connectivity problem >> with usenet "gateways" .... ibm-main mailing list usenet gateway >> recently has had several things go totally missing; ... in any case. > > I have a slightly different problem; I see the replies to posts > by you, and several others, a day or so before I see the post. > > And I get my own posts from around 3 weeks back reinjected into > the news stream somewhere. > Things must be frozen up in Norway this time of year;-) |