8/26    Amazon medium instances (3.75GB RAM): 0.160/hour = $1382/year
        Generic standard Linux VPS (4GB RAM): $480/year
        Amazon costs more (but does offer superior scaling options).
        \_ Amazon is $670 if you buy a year's usage up front (heavy util).
           Why is heavy util less expensive than light util?

6/27    You can't put 16GB RAM in the new 13" Macbook Pro.
        Fuck you Apple!
        \_ Really?  I know they say you can't put more than 8GB in the 13"
           MBP, but I have 16GB in my 2011 13" MBP and it works perfectly.
           \_ I thought the new ones have soldier on HD and RAM just like
              the Macbook Air?
              \_ AFAIK, only the new 15" MBP has the RAM soldered on its
                 motherboard.  It looks like the 13" MBP's RAM can still
                 be upgraded, since OWC is selling a 16 GB RAM upgrade for
                 it:
         http://eshop.macsales.com/item/Other%20World%20Computing/1600DDR3S16P
                 \_ That's true but the upcoming 13" retina coming out in
                    Sep/Oct (reliable source from an Asian supplier) will
                    be ALL soldier on component. FUCK YOU APPLE!!!

3/29    A friend wants a PC (no mac). She doesn't want Dell. Is there a
        good place that can custom build for you (SSD, large RAM, cheap video
        card--no game)?
        \_ As a side note: back in my Cal days more than two decades ago when
           having a 387SX made me the only person with floating-point hardware,
           most machines were custom built.
           most machines were custom built.  -- yuen
           \_ did that make you a better programmer/gamer/etc?
              \_ No.  I was writing a pre-emptive multi-threading library as
                 part of an effort to port some floating-point-heavy Unix code
                 to DOS for CS199.  I couldn't figure out how to save the
                 "context" of Borland C++'s floating-point emulation library.
                 The options were to either mask context switching around all
                 floating point operation or get a co-processor.  I ended up
                 getting the co-procoessor.  -- yuen
                 to DOS for CS199 (http://ssli.ee.washington.edu/~bilmes
                 mypubs/bilmes1992-icmcmood.pdf).  I couldn't figure out how
                 to save the "context" of Borland C++'s floating-point
                 emulation library.  The options were to either mask context
                 switching around all floating point operations or get a
                 co-processor.  I ended up getting the co-procoessor.  -- yuen
        \_ Why not Dell? What is it about Dell she doesn't like?
        \_ Fry's? HP? Lenovo? The only place I have done this is Lenovo.

6/5     In an effort to stabilize our services, we'll be rebuilding parts of
        the CSUA infrastructure over the course of this summer.  To give us
        some wiggle room, I've temporarily decreased soda's allocated RAM from
        8GB to 2GB.  If you need to run something that requires large amounts
        of memory, please send mail to root@csua.org and we'll try to
        accommodate your request.  --jordan
--------------------------------------------------------------------------------

3/19    When you're explaining the stack to people do you draw it with the
        highest addresses at the top or at the bottom?
        \_ When I explain any memory layout including stacks, I draw with the
           highest addresses at the bottom.  But I've seen people doing the
           other way.  -- yuen
           \_ do you by any chance have seen or have a jpg of the full memory
              layout of the program, eg: .text, .bss, .initdata, etc etc?
              If not, can one get that info from the ELF spec?  Thanks.

5/22    Duke turns of its Usenet server:
        http://news.duke.edu/2010/05/usenet.html
        \_ Imminent death of the net predicted.
        \_ :( Sucks. I wish Berkeley hadn't turned off Usenet, either.

1/26    What's a good motherboard I can get that will fit in a 1U
        case, with sata connectors, gig-e ethernet, and lots of
        slots for RAM?  I also don't want to have to use expensive
        double buffered RAM.  Can you tell I havne't bought a PC in
        years?  thanks.
        \_ Buy a mac. Haven't you heard? if you don't buy a Mac people
           will put the electronic Black Spot on you and never talk to you
           again.  Or put you down for being poor.

12/7    How many TCP retransmits are too many? Here is what I get:
            3594143433 segments received
            3760174421 segments send out
            3801829561 segments retransmited
        \_ rephrase. you can never have too much money. or too little.
           what is, is.
           \_ You always get a few, but I have a bunch of squid servers that
              I just noticed have 500/second, which seems really high to me.
              They do about 20MB/sec at peak so this is a retransmit every
              40K? This is kind of hard to believe, because they are working
              fine.
              More details have been requested so here they are:
              Each server has apache and squid installed and serves about 10M
              requests/day out of memory in 5ms or less. They are configured
              as memory only squid servers. They are middle tier caching
              servers, sitting between an application layer and an API.
              They have 1Gb/sec uplinks to shared switches, which each have
              1 Gb uplinks to the core routers. No other servers on the
              switches are showing this behaviour and the switches themselves
              are not overloaded. Further investigation shows that 99% of
              the connections on the servers are sitting in TIME_WAIT, which
              I actually think is the cause of these retrans. I am still trying
              to figure out who is rudely dropping all their connections to
              these servers, but it is hard, since all traffic is through a load
              balancer (NetScaler).

4/20    Heil....HeilHitlerHe.. #In Memory of Adolf Hitler# ..Heil....HeilHitlerHe
        Heil....ilHitlerHeil.... We will always remember ....Heil....ilHitlerHeil
        Heil....Heil............. and cherish you. Your .....Heil....Heil........
        Heil....Heil............. acts of selflessness ......Heil....Heil........
        Heil....Heil........... will be passed down from ....Heil....Heil........
        HeilHitlerHeilHitler... generation to generation. ...HeilHitlerHeilHitler
        HeilHitlerHeilHitler... The lies that dishonor your .HeilHitlerHeilHitler
        ........Heil....Heil..... name will be vanquished. ..........Heil....Heil
        ........Heil....Heil.... You were a true patriot ............Heil....Heil
        ........Heil....Heil.... and a lover of all men, ............Heil....Heil
        HeilHitlerHe....Heil... all races, all religions. ...HeilHitlerHe....Heil
        ilHitlerHeil....Heil.. #In Memory of Adolf Hitler# ..ilHitlerHeil....Heil

4/9     Which is better designed/engineered for reliability, the SDCard
        or the Compact Flash card? I'm IEEE and EE dumb but I do understand
        a lot of RAM terms and design architectures.
        \_ Mechanically, I think CF cards are more durable because there is
           more plastic and the contacts are hidden.  For example, I'd be
           comfortable to let my toddler kid play with a CF card but not an
           SD card.  Electrically, I have no idea.

4/1     http://www.google.com/mobile/m/brainsearch/intro_android.html

3/30    CF/SD Disk database:
        http://robgalbraith.com/bins/multi_page.asp?cid=6007

3/19    Any comments on IE8?  How's its memory usage?
        \_ Who cares.  Microsoft will eventually push it onto everyone.
           Accept your fate.
           \_ Well, if IE8 uses less memory than Firefox 3, I'll get it now
              instead of wait for it to be pushed.

3/16    Does anyone know what type of flash is used in the iPhone and how many
5A      writes the flash memory could handle before failing? I'm wondering why
        MobileSafari doesn't support paging to flash "disk". As is, it reloads
        pages all the time, which sucks.

2/20    Why flash ram will get you into trouble, in the long run:
        http://www.pcper.com/article.php?aid=669
        This is worse than over-clocking your computer. There's data
        involved.
        \_ What's your point? That flash supports a fairly limited number of
           write cycles? We knew that already.
           \- I am not the OP but dont think you're inquiry is very
              thoughtful. This may be another case of "physical difference"
              [between rotating mag storage and memory cell storage] cant
              be abstracted away by software emulation" [in this case
              focusing on the write/erase asymmetry]. An older example of
              this with another hyped technology was ATM emulation of
              Ethernet and the problem of doing broadcasts.

2/13    Question about memory relocation:
        These days most h/w has a relocation register. Could the relocation
        address be stored on disk or in kernel memory vs. in a register? Yes,
        that would be slow but is it possible? Do you *need* a relocation
        register or does it exist purely for performance reasons? I was
        reading some paper written by IBM in the 1960s that seemed to
        imply the former and I don't understand why that would be.
        \_ What "relocation register" is this?  If you're talking about the
           one that (say) the IBM 7090 used for time-sharing, that's been
           obsolete for 30 years; modern machines use virtual memory.  (And
           the tables defining the virtual-memory mapping are in fact stored
           in kernel memory, since they're too big to fit in a register.)
           Or did you have something else in mind?
           \_ Yes, but modern machines still have a base register. Do
              they need it?
              \_ What do you mean by a "base register"?  It would help if you
                 could say exactly which register you're talking about on some
                 real architecture (x86, PowerPC, MIPS, etc.), or at least if
                 you could describe what this register does.
        \_ Because comparing 1960s computer hardware to today's is like asking
           why drivers need a whip to make their cars drive faster because
           you saw a picture of a horse drawn carriage.
           \_ Well, no. Modern hardware still does it that way. The
              question is: Does it *have* to?

1/21    If I have a linked list of structs and many of those structs have
        members that are structs then what is the best way to free() the
        memory when I am done with them? I thought I would walk the list
        and do a free() on each member of each struct, but that generates
        errors like free(): invalid pointer, presumably because I don't
        always allocate memory in each struct. No, I never took a class in
        memory management (obviously). In a Java World I don't worry
        about all this! ;)
        \_ Why don't you just check the pointer before calling free? i.e.
           if (p) { free(p); }
           \_ This was my first inclination and it doesn't change anything.
           \_ This should have no effect in this case.  The if only checks if
              p is null, and if p is null, most versions of free will ignore
              p is NULL, and if p is NULL, most versions of free will ignore
              it anyway.
                   \_ All versions.  It's a requirement in the standard.
                      \_ Hah, you assume all libc implementations are
                         compliant. Well, okay, things are better these days.
              That said, if you aren't allocating the memory for the pointer,
              you NEED to set the pointer to NULL.  Preferably at the point
              that the enclosing struct is allocated.  Otherwise that pointer
              may just be nonsense, which the if won't catch.
        \_ Well, do the structs contain other structs, or pointers to other
           structs?  If struct A contains struct B, struct B is allocated
           and free'd as part of struct A.
           If struct A contains a pointer to struct B, you need to make sure
           you're allowed to free the pointer before you do, perhaps there
           are multiple pointers to B, and B shouldn't be free'd until all
           the pointers are done.
           If you are supposed to free pointer to B, but you think you might
           be accidentally freeing it twice, you're going to have to be more
           careful and figure out where exactly you should free it.  There's
           no easy way out of that.
        \_ You need to define the difference between and owning an object and
           referring to one.  It's a logical difference -- the owners are
           responsible for lifetime, referring pointers just are assigned.
        \_ You're not doing something like the following, right???
                while (p != NULL) {
                    free(p);
                    p = p->next;
                }

12/18   I'm loving the Privacy Browsing feature on Safari. I turn it
        on, browse porn, then turn it off, and no one using my
        computer would ever know that I browsed porn (no history,
        no cache, nada). Loving it man.
        \_ The guy with the backdoor knows. Also, the shadow knows.
           \_ That may be the case and I don't really care about the
              guy with the backdoor. The most important thing is that
              my kids and my wife don't know about it.
              \_ But it's advertised as being useful for gift shopping!
                 You horrible, horrible man.
                 \_ And tax cuts are good for THE PEOPLE!

12/5    What would you guys think of this?
        TYAN Tank barebone
        http://tyan.com/product_barebones_detail.aspx?pid=353
        2x Intel Quad Xeon E5420
        http://www.newegg.com/Product/Product.aspx?Item=N82E16819117147
        16GB ram
        http://www.newegg.com/Product/Product.aspx?Item=N82E16820134633
        SAS hard drives
        http://www.newegg.com/Product/Product.aspx?Item=N82E16822116058

        Total cost <$3000, and it seems like this would be an ESX supported
        configuration if we ended up going that way (and there's no reason
        that it would preclude any other virtualization option either)

        The goal here would be to eliminate the other servers in various states
        of 'broken' (scotch, screwdriver, lifesaver) and replace it with this
        one.  Once it was up and tested, we'd migrate soda there and then reuse
        the hardware from Soda for something else.  Thoughts?
        \_ Why do you believe soda would not run on a $1-1.5k machine:
           3gz proc + 1tb disk + ~4gb memory. I suppose something on the
           higher end of this range with two slower processors might be
           better than a 1proc. I'm guessing you'ld be better off with
           one two lower end machine. Especially if you want one to play
           around on.
           \_ He's talking about replacing 4 servers with 1 server, not
              just replacing soda.  Soda is already way overpowered for
              what it's doing right now.  It doesn't need to be a dedicated
              machine anymore.
              After that I really can't tell what you're trying to say.

9/20    I am running NAT on an Extreme Networks switch connected to a
        Cisco switch at GigE speeds. Copying a file across the same link
        w/o using NAT gets about 25 MB/sec, but with NAT turned on I get
        2.5 MB/sec. The CPU is 98% idle and there is plenty of RAM. I know
        that NAT in s/w has an overhead but could it realistically be a
        factor of 10 slower?
        \_ what kind of file copy?
           \_ NFS or SFTP. Same symptoms with either.
        \_ Are you sure it is not your test platform that is slower? This
           seems really crappy. Are you perhaps having a duplex mismatch
           problem on the switch the NAT is connected to?
           \_ What test platform? You mean the one I am copying with? I
              said that I get 25 MB/sec with NAT off and 2.5 MB/sec with
              NAT on - using the exact same hardware. Therefore, I doubt
              it is a duplex problem or it would be present both times.

9/4     Someone asked yesterday about the Obama/Ayers link.  It took me a while
        to find a link I was happy with, this one is ok.  This guy goes into
        excruciating detail about the Annenberg Challenge and CAC over a
        number of blog posts.  The full Annenberg documents were only recently
        released, and this post predates the release.  The documents are still
        being digested, so this is far from the final word.
        Anyway, it is pretty obvious that Obama and Ayers had a close working
        relationship for at least 6 years (the real relationship almost
        cetainly predates this), in which they funneled millions to Ayers'
        causes and students.  It's also quite obvious that Obama has been
        highly disingenous about this relationship, and many of his statements
        about it have been blatant falsehoods. ("Our kids went to school
        together", etc)
        http://csua.org/u/m9b
        \_ Yes, Obama supports terrorism and wants to KILL WHITEY.
        \_ It looks like Obama has really been trying to improve the schools,
           which is great. I think I am going to give him another $2300.
           \_ Might want to look into what he was spending that money on
              first.  The challenge itself was deemed an utter flop. $100mil
              down the hole.
              \_ Deemed an utter flop by who? This guy is not a reliable
                 source.
                 http://preview.tinyurl.com/554pov
                 The Annenberg Challange is hardly the radical far-left
                 proposal this guy makes it out to be.
                 \_ Sorry, it was a flop _in Chicago_. That is, it had no
                    effect on student performance in Chicago. Your link is an
                    overall evaluation, so not really interesting for this
                    discussion.
                    Also, it was deemed a flop by the Consortium of Chicago
                    School Research (CCSR) in 2003.
                    \_ Yes, that is interesting. I will have to do some more
                       research. But I think it is amusing that an effort to
                       try and weaken the hold of the teacher's union is called
                       a communist tactic by this guy.
                       \_ I with you agree there.  -op
                       \_ I agree with you there.  -op

        here are some new numbers (VM Size in MB reported by Task Manager):
                                        2.0.0.16        3.0.1
        - Startup, empty the cache.     13              23
        - Go to http://www.yahoo.com          17              40
        - Open 6 news pages in          43              69
          6 new tabs.
        - Close the 6 tabs above.       35              54
        - Open the same 6 news pages    44              71
          in 6 new tabs again.
        - Close the 6 tags again.       38              53
        So, after opening and closing the same tabs a second time, Firefox 3
        still uses more memory than Firefox 2.  Any idea?  Thanks.
        \_ The delta between 2 "close the 6 tabs" lines is 3 megs for FF2
           and -1 megs for FF3.  The general problem with FF2 was it tended
           to leak memory over time.  For a desktop app, an extra 10 megs
           is minimal, but since people like to have long lived browser
           sessions, FF2's habit of slowing growing up to 2 gigs of memory
           is a real problem.
        \_ Dunno, I notice better memory usage with lots of tabs open.
           Firefox 2 used to go into the 500 MB range after a while,
           Firefox 3 typically doesn't go above 400 MB.
        \_ I used to have to restart my browser every day. I don't have to
           with FF3.
           \_ FF3 does seem more crashy though.
        \_ Anyway, you made your point. He has extensive political connections
           to an ex-WU guy. A big yawner to me, but perhaps middle America
           cares. I kind of doubt it.

8/8     Just got my 16GB Rally2 USB flash drive, for $50.  I can't believe the
        pace of technology.
        \_ http://woot.com has an 8GB Kingston DataTraveler 100 for $20+s/h.

6/9     Video RAM question:
        Let's say I need to display a shape model that was ~3 million
        plates that are each 72 bytes (216MB). Do I have to have a video
        card that has more than 216 MB of RAM or can the system dip into
        main RAM in order to display it? As I understand it, the VRAM is
        a buffer but what happens when the buffer fills? In this instance
        my video card has 64 MB of RAM and the image displays fine. Is
        the rest of the image resident in system RAM? Do people get video
        cards with more VRAM for performance reasons (VRAM is faster and
        dual-ported) or because they can display images they never could
        otherwise? Compare a system with 16GB RAM and 64 MB VRAM to one
        with 1 GB RAM and 2 GB VRAM.
        \_ That wildly depends on how the data is presented to your card.
           \_ Please elaborate.
              \_ If you present the data to your card the right way, it can
                 store the vectors directly.  If you don't, it can't.
                 \_ So what if it can't if I have 16 GB of main RAM? Can I
                    use that or not? Why would I not want to? That is,
                    what does storing directly buy me? Can I display my
                    model in either case but with a performance penalty?
                    Excuse my ignorance, but I am not a gamer or anime
                    freak and never had a graphics class. I realize you
                    touched a little bit below, but I've been reading
                    about rendering pipelines all day and there's not been
                    much that talks about the role of the h/w in this or
                    what happens when you run out of video RAM.
                    \_ If you do "immediate" mode then you're limited by the
                       bus speed to the video card.  If you can cram it all
                       onto the card, you're theoretically maximizing the
                       hardware performance. -emarkp
                       \_ Are there any other penalties? Just the penalty
                          of the speed of the VRAM vs. RAM and the
                          offloading of the main CPU in favor of the GPU?
                          I don't see the bus speed as a limiting factor.
                          You have to load it all into VRAM over some bus
                          anyway, right? I realize say AGP might be
                          faster than PCI but I am more concerned by how
                          the size of VRAM impacts the problem than by
                          issues of throughput, which are clear-cut. If I
                          have 256 MB of VRAM and a 1 GB image then what
                          happens? Does the video card swap to main RAM?
                          \_ At this point, you leave generalities and it
                             depends on the card, and the quality of the
                             driver.  If the card runs low on memory, it can
                             dump whatever it needs to render, and then pull it
                             back from main memory when necessary. -emarkp
                 Typically cards do 16 bits of floating point, but that's
                 moving up to 32.  The rendering pipeline today is basically
                 transforming triangles modelspace->worldspace, lighting,
                 worldspace->cameraspace.  This can be done in "immediate"
                 mode, where each triangle is provided from the program running
                 in the CPU to the video system (D3D, OpenGL, etc.=> video
                 card) or in buffered mode (historically "display lists" on
                 OpenGL though that included much more than just geometry)
                 commonly called "vertex buffer objects".  If there's room on
                 the card, you can push a bunch of geometry into the card, then
                 just tweak the input values and do everything very
                 efficiently.  If there isn't room on the card, then the video
                 memory is mostly 1) frame buffer (what you see) and 2) texture
                 memory.
                 http://en.wikipedia.org/wiki/Rendering_pipeline
                 John Carmack (Id software, etc.) believes he's solved the
                 megatexture problem (imagine a whole world or large
                 environment that can be textured down to tiny detail),
                 with little or no performance issues (implemented in their
                 upcoming game, Rage).  He says his next project is to do the
                 same with geometry.  So things may be changing in the next few
                 years. -emarkp

4/2     Anyone try vmware player and notice that network performance
        becomes terrible? I mean things like webpages timing out, etc.
        This is in NAT mode on a windows XP host running a windows 2K guest.
        \_ I use VM Server on XP Pro running XP Pro guests in NAT mode.  No
           such problem.
        \_ I use VM Server on XP Pro running two XP Pro guests in NAT mode.  No
           such problem.  However, if I run three or more guests, everything
           becomes slow (not just network), and Task Manager on the host shows
           that the CPU rarely goes to idle even though I'm not doing anything
           It's not a problem with RAM since Task Manager shows that Commit
           Charge Total is still less than the physical RAM.
           on the host or the guests.  It's not a problem with RAM since Task
           Manager on the host shows that Commit Charge Total is still less
           than the physical RAM.
        \_ It's a little bit worse, but not much worse. Do you have enough
           RAM? Everything is slow without it - and it should be allocated.
           \_ I actually meant the network performance inside the guest.
              It seems fine on the host. The host has 2 GB, and the guest
              is allocated 512 MB, so I don't think that should be a problem.
              I'm not doing anything particularly resource-intensive.
              \_ 512 MB might not be enough. Increase that. In my
                 experience (Win XP guest on MacOS host) I started with
                 512 MB allocated and was unhappy until I doubled that.
                 \_ Hmm, Win2K should require substantially less memory than
                    XP, but it's worth a try I guess. My suspicion is
                    actually that a large number of TCP connections is
                    handled badly somewhere (maybe in vmware's nat service),
                    since webpages nowadays tend to retrieve random images
                    from a whole bunch of different places, etc.
                    \_ Maybe, except that my network performance (also
                       using NAT) is not anywhere near as bad as yours.

3/31    Dear Chinese people who are into feng-shui and szhuan4-ming4.
        There are a few main ways to szhuan4-ming4. If memory serves
        right, one of them is called zji3-wai1 and the other one is
        called zji3-ping2. Is that right? Where can I get more informatino
        on this? Thanks for any pointer.
        \_ http://en.wikipedia.org/wiki/Fortune-telling#Asian_fortune_telling
           It lists four major methods and some other methods.
        \_ http://chinese-astrology.blogspot.com/2005/08/zi-wei-dou-shu-what-is-it.html
           The first one is called Zi3 Wei1 but I don't know other methods.

2/21    Cold Boot Attacks Against Disk Encryption:
        http://www.schneier.com/blog/archives/2008/02/cold_boot_attac.html
        http://citp.princeton.edu/memory

2/20    In python, say I have a function:
        def hello(a,b,c): ...
        and I have a list l that contains 3 entries (l[0], l[1], l[2]),
        how can I pass it easily in the following manner?
        hello(l)
        \_ hello(*l)
           \_ Oh cool, thanks!!!

12/14   What the heck are SDHC flash memory cards?
        what's the SD?
        can i use SDHC cards with my old stuff?
        \_ SD - secure digital, HC - High Capacity. You probably cannot
           use SDHC in old stuff b/c HC is a different memory addressing
           scheme. Some old stuff like Treo 650s can be hacked to work
           w/ SDHC cards, but most can't.

3/15    YEEEHAAAA!!!!
        'ATLANTA - A panel of Georgia lawmakers signed off Thursday on a plan
        to create a Confederate heritage month, even as legislative leaders
        reacted coolly to a push to apologize for the state's role in slavery.
        Sen. Jeff Mullis' bill would dub April as Confederate History and
        Heritage Month to honor the memory of the Confederacy and "all those
        millions of its citizens of various races and ethnic groups and
        religions who contributed in sundry and myriad ways to the cause of
        Southern Independence."'
        http://news.yahoo.com/s/ap/20070315/ap_on_re_us/confederate_month
        \_ WTF?  Is that a joke?
           \_ Nope.  YEEEEHAAAAA!!!

3/4     trn crashes on me upon startup:
        "*** glibc detected *** malloc(): memory corruption: 0x08091618 ***
        Abort"
        Any idea?  Thanks.

10/25   Anyone know if ACP-EP memory is a problem?  Looks like it's the
        least expensive where I am, but I don't have any idea of its quality.

8/30    I just configured a Dell XPS M1710 with the 2.0-GHz 4MB-cache
        Core 2 Duo with WinXP Pro and 3 years warranty.  $3,106 after tax!
        Dang, some rich kids will be spending bux on their computer for
        undergrad WoW.
        \_ Excuse my lameness, but it this faster than a 3GHz Pentium D ?
           I know MHz doesn't matter anymore, but it doesn't feel right.
           \_ I believe all Pentium D's up to 3.6 Ghz are blown out of the
              water by a 2.0-GHz C2D with 4MB-cache.
              http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=12
              The above has desktop Core 2 Duo vs. Pentium D, not mobile C2D,
              but you also need to consider the 1.86 C2D desktop has only
              2MB of cache.
           \_ In my (limited) experience, Core (1) Duo 1.86GHz >> Dual Xeon 3.0

8/25    USB teddy bear holds data, scares children:
        http://tinyurl.com/m9ak4  -John

8/4     What happens if you boot a computer with no RAM installed?  Will it
        POST?  Can you get into the BIOS config?  Or does it just beep at you?
        \_ I think you get a blank screen and the infamous 3 beeps.
        \_ Depends alot on your PC BIOS, but most just simply refuse to POST.
        \_ Absolutely.  No way.  Sometimes.  -proud American

8/1     I'm looking for a 1GB and a 2GB USB flash drive.
        All the reviews I've seen on amazon seem to be very mixed.
        I'd like something rugged, simple, fast, and reliable.
        Any suggestions? Thanks
        \_ I like the Kingston DataTraveler Elite 2GB.  I backup my pr0n to the
           the encrypted partition, it's fast, and it's reliable.  Other flash
           drives have the cool slideout (so you don't lose the cap), but it's
           hard to find one with real hardware-based encryption.  FYI, the
           "Privacy Edition" is the same thing except you don't have a public
           partition (basically so ppl don't have to think), and I don't care
           for the "Migo Edition" and similar solutions where you load your
           user prefs from your flash drive (too heavyweight IMO).
           http://www.newegg.com/Product/Product.asp?Item=N82E16820189017
           If you don't want hw encryption, you can go much cheaper, just go
           to http://newegg.com, pick 1GB or 2GB, and sort by Rating.
        \_ Get a U3 Smart flash drive.  This week Office depot 1GB $25.
           http://www.newegg.com/Product/Product.asp?Item=N82E16820189017
        \_ One of my recent new toys has been a Corsair FlashVoyager, which
           I've been very happy with. It supports partitioning and all that
           good stuff and is fairly snappy for a USB thumbdrive. Plus it's
           built like a brick and hard to lose (rubbery surface stays in my
           pocket much, much better). IMO, teeny-tiny ones (Cruzer Micro or
           smaller) are just too loseable
           http://www.newegg.com/Product/Product.asp?Item=N82E16820233009
           I think it's also slightly cheaper than a few months ago.
           --michener
           \\\_ Thanks y'all! -op
        \_ I've got a Memorex 2GB Traveldrive.  Never had any problems
           it.  I carry all kinds of porn wherever I go. -proud American

7/14    So as I understand it, Intel introduced Hyperthreading because the
        penalty for a cache miss or branch misprediction on Netburst was huge
        so to make up for it they could work on a different thread while
        waiting for memory latency.  Given that the Core2 architecture has
        such a wide execution path, couldn't they use HT to try and keep all
        those execution units full?
        \_ I attended one of Intel's talks on campus and the "huge penalty"
           myth is a myth. As the number of pipelines increases, the depth
           of the miss and a pipe flush is longer, yes. However, keep in mind
           that since each stage in the pipe is shorter the clock cycle is
           also faster. Thus, in terms of absolute time (time=number of
           pipes that need to be flushed * cycle time), the increase in P4
           from the old architecture is only increased by ~20% time, which
           is insignificant in computer science speak.
             As for keeping execution units full, it's very much application
           dependent. Even with the huge instruction reorder mechanisms some
           applications still can't utilize them all.

6/21    When they say Conroe chips will run on a 1066MHz FSB, what kind of
        RAM does that imply?  DDR2-522 or DDR2-1066?

4/4     To whoever recommended crucial, thanks for the tip, but
        I just cancelled my order with them.  Their credit card
        clearing operation is pathetic, and customer service was
        rude and unhelpful.  They were unwilling and/or unable to
        try to figure out how to process an out-of-US order.  Fuck
        them.  Thanks anyway.  Any other suggestions?  -John
        \_ Wow, sorry to hear about that.  I've also used http://memoryx.net
           \_ Don't apologize, I appreciate the tips.  I'll try them. -John
        \_ I've bought ram from 1-800-4-Memory (http://www.18004memory.com
           They were cheap and decent, but I have no idea how the service
           is outside the US. Another option could be to try amazon.co.uk.
        \_ Crucial is just a memory maker.  They're resold by a zillion other
           companuies around the world.  http://pricegrabber.com, http://pricewatch.com,
           http://shopping.com, froogle, http://pricescan.com, etc to find a reseller.
        \_ John, here is a hint:  American Express.
           No, I am not kidding.  Most of online store in USA requires
           billing address in USA.  If you are using some non-USA card,
           you are going to get into trouble.   American Express appearently
           only check the zip code. So, as long as the zip code is correct,
           it will go through.  In my part of town, my zip code is 3 digit,
           and i managed to get it through by padding it with zeros at the
           front of digits :p           kngharv
           \_ I don't think it is true--I work with Versign and AVS is
              available for AmEx as well.  It's up to the vendor whether
              to use it or not.
           \_ The thing is, I already have 2 gold cards (one Swiss and
              one Chilean.)  AmEx is fine, but I have never ever ever
              had trouble with either of my Visas except with US vendors.
              I'd prefer not to have another card floating around just to
              deal with people who're too lazy to implement proper ID
              checking (which this is a case of.)  And as for zip codes,
              many of the shops I've had trouble with don't understand that
              a lot of the world doesn't work that way (my zip code right
              now is "Las Condes", go figure.)  Thanks though.  -John
        \_ Amazon seems to sell some Crucial memory.  I've bought from
           Amazon abroad and shipped to the US, and that went smoothly.
           Perhaps the reverse is also smooth.

4/4     Anyone have a recommendation for where I can get a cheap 1GB
        memory dimm for a 12" Powerbook?  -John
        \_ http://www.crucial.com (I don't know if it is cheap)
           \_ Seconded. And the price is worth not having to ship back
              faulty or non-compatible units.
                \_ Thanks much, I just ordered there.  Price-wise it's
                   perfect--even with import tax & shipping it's much
                   cheaper than Apple.  -John
           \_ Thirded, always had good experience with http://crucial.com, the prices
              are not rockbottom but they are good, and I've never had a
              problem, plus they have fast shipping.
              \_ I just cancelled my order with them.  Their credit card
                 clearing operation is pathetic, and customer service was
                 rude and unhelpful.  They were unwilling and/or unable to
                 try to figure out how to process an out-of-US order.  Fuck
                 them.  Thanks anyway.  Any other suggestions?  -John

4/1     April is the cruelest month.
        \_ breeding lilacs out of the dead land,
           mixing memory and desire, stirring dull
           roots with spring rain.

3/16    Mr. Phelps, this thumbdrive will self-destruct in 5 seconds:
        http://tinyurl.com/pqkd4 (everythingusb.com)

2/17    What's the basic technological difference between EEPROM and
        flash? Both program and erase using electricity...
        \_ http://en.wikipedia.org/wiki/Flash_memory
        \_ i thought eeprom erases with uv light
           \_ ee = eletrically eraseable.
              you are thinking of eprom or just plain shitty prom.

2/7     I have a little multi-threaded server I'm writing, and I log at
        the start of each call and at the end of each call.  I log by
        having a global lock file, lock, write, flush, unlock.  This
        seems like a bottleneck, is there a better way to log from a
        multi-threaded server? Perhaps something like syslogd where I
        could send messages to another process that would log for me?
        (This would avoid the flush because it could keep running even if
        the server crashes)  Order in the log is not terribly important,
        since a quick run through sort on the date will order that for me.
        \_ What about a separate thread to manage the log? Put the msgs
           in a shared queue, and have the logging thread write out the
           messages later on.
           \_ Not a bad idea, but if the server crashes I won't get the
              messages just before the crash.
              \_ There isn't really a way to avoid this problem in a
                 threaded app, except to move logging into a separate
                 process, such as syslog. If you are going to use
                 syslog(3), then you probably should openlog(3) prior
                 to starting your threads.
        \_ How about each thread keeps its own log file, as well as writing
           into a shared buffer which is flushed periodically to the common
           log.  That way in normal operation you have just one log to look
           at, but when the server does down, you can examine the per-thread
           logs.  In normal operation, other threads will not have to wait
           while one thread flushes its own log.
        \_ I don't think there is a problem with just using write(2), it is
           atomic and writes to buffer cache so a crash of the process won't
           be a problem.  With some applications, mmap(2) is better, but
           depends on what you are doing.  --jwm
           \_ Note that write(2) is only atomic if nbytes is less than
              PIPE_BUF (which is at least 512 according to POSIX). That
              said, if you know your log messages will be reasonably short,
              this is the way I'd go. -gm
              \_ I was assuming he wanted to write a file, and in that case
                 I suspect that the atomicity extends at least to the page
                 size, though I may be wrong.  --jwm
              PIPE_BUF (which is at least 512 according to POSIX). You
              can also get short writes, even if nbytes is less than 512,
              if you're writing to a pipe or other space-limited fd. -gm
           \_ Really?  I didn't know that.  Related, why is it called
              write(2) rather than just write()?
              \_ That represents that it's in section 2 of the man pages which
                 is the system call section. "man write" shows you the page
                 for the write utility to send messages to someone's tty.
                 "man 2 write" gets you the write system call.
                 \_ On Slowlaris: man -s 2 write
                    \_ Unix needs it's own diversity day.
              \_ Stop geeking and find a hot gf during undergrad before it's
                 too late!
           \_ I'll add to my previous comment some speculation. If you're
              using stdio or iostreams, I suspect your flush is not syncing to
              disk, but merely calling write(2) to flush it's internal buffer.
              To test this you could use ktrace to see the calls it is making.
              Of course you will still may need a lock to protect the library
              you are using as I suspect these are not threads safe.  And if
              you don't like locks, you could use sprintf() and write(2) with
              no locks. --jwm
              \_ Tried using sprintf and write(2) with no locks, and it
                 doesn't quite work.  The log file is a bit messed up.  It
                 appears that concurrent calls to write can screw things
                 up.  (Not threadsafe)  But, it seems like in that case
                 the lock shouldn't be costing as much as I had supposed
                 anyway, since the write is just being buffered somewhere.
         \_ Having every thread in your application serially accessing
            a piece of code that does I/O is a really bad thing, you
            really don't want to do this.  Grabbing a lock and
            sticking data on some list that another thread comes and
            consumes will be a lot faster (just makes sure that other
            thread doesn't hold the lock while writing the data, then
            you've lost everything you gained in the first place.)  As
            for needing to get everything logged in case your
            application crashes, you are never going to get that
            anyway.  If you have 10 threads waiting to grab the lock
            and the server crashes those 10 logging statements will be
            lost no matter what.
            How often you log should influence your choices here, are
            we talking tons of debugging logging or just 1 or two
            lines a second?  If a lot of logging keep in mind that
            things like gettimeofday are syscalls and those are more
            expensive. Is it really that important that you get the
            date of the log statement exactly right?  Can you get the
            time before or after you get the lock?  If you really need
            to order your logs, maybe just use a long that you
            increment per statement?
            Getting a lock is done entirely in user space (and
            normally in just a few asm instructions AS LONG AS THERE
            IS NO QUEUE FOR THE LOCK.  Locks get a lot more expensive
            to use (by orders of magnitude) when any blocking has to
            occur.  If you are worried that a lot of threads are going
            to be blocking at a time keep the critical section as
            small as possible, it really helps.
            Finally, while logging can be more complex than it looks
            at first, it is also a pretty solved problem.  There are
            tons of free logging libraries out there that do all this,
            do it well, and do it fast.  It might be worth your while
            to just use existing code.
            \_ Interesting. Thanks.  What could I search for to find some
               of these logging libraries?  I didn't have much luck last
               time I tried.  (The critical section was already tiny, just
               the write and the flush.  The string is all built outside
               the critical section.) -op
               \_ You do understand that I/O is THE most expensive thing
                  you can do don't you?  Just because it is only 2 lines
                  of C code doesn't mean the critical section is fast.

1/25    I'm working on a program in windows which is having memory problems.
        When I allocate memory with new or malloc, it seems to reserve more
        memory than I request (using VM Valiator to monitor this).  Even if I
        delete the memory there is still a "reserved" chunk that doesn't seem
        to ever get used by other code.  I'm using VC++ 6 btw.  Anyone have an
        idea of what might be going on?
        \_ When you malloc memory it gets a page at a time.  Also freeing
           memory will rarely actually give that memory back.  You do
           understand memory pages right?
           \_ Yes.  The memory involved here is in multiples of megabytes
              however.

12/11   NEC: products of the future
        http://www.nec-design.co.jp/showcase

11/16   Your friend's kid's gadget is cooler than yours
        http://www.nytimes.com/2005/11/17/technology/circuits/17pogue.html
        \_ That's awesome!
        \_ Wait how does it know what you clicked on? Don't you need
           to store the picture somewhere (in memory) and an input
           method (like a tablet)?
        \_ I got to play with one of these a couple of months ago before it
           came out.  It is pretty cool, although the one problem it has it the
           only feedback it can give is audio. Thus, it wouldn't be very easy
           to use in a classroom w/o disturbing everybody.  Also, in order to
           use it, you need to write on special paper that costs about $1 a
           sheet.  Its covered by very tiny dots so the pen knows where it is
           located on the paper at all times.  But I will say that their future
           plans for the Fly pen seem really cool (wireless sync and such).
           \_ It comes with headphones and paper is about 8 cents/sheet.
        \_ I think reading its manual alone will make me dizzy.

11/07   Does 168pin DDR (PC3200) memory needs to be used in pairs? Thx!
        \_ Ideally, in pairs.  Dual-channel mode will be activated if the
           memory is of the same size and in the right slots.
           If not added in pairs or of memory of different size or in the
           wrong slots, then the memory will operate in single-channel mode.
           Your mobo manual should tell you the same thing.
           Your mobo manual will also tell you which slots to use for a pair
           (e.g., for my Asus, it was every other slot that was a pair).
           Same deal with DDR2.
           \_ Dual-channel mode is much faster?
              \_ From what I've read, for your typical program you will see
                 1-5% difference.
                 \_ Dual channel nearly doubles your memory bandwidth.
                    If "typical applications" don't use memory bandwidth,
                    then this guy is right.  But stuff I care about is
                    often limited by memory bandwidth....  --PeterM
                    \_ Yeah, an easy test is to run Doom 3 or whatever and
                       measure fps, swap the memory to the wrong slots, and
                       see if there's a difference.
                       Repeat for real program that you're using.

11/4    I'm putting 1.5GB of RAM in a computer that takes PC2100. Is there
        any reason that a mix of PC3200 and PC2700 RAM should not work?
        \- kids today are lazy and stupid. STFW, the answer is obvious.
           \_ Yes, the answers I found seems to say that it should work fine,
              but with this RAM config, the computer now experiences kernel
              oops, so I'm trying to figure out where to start with debugging.
              \_ Have you tried loading your computer with this many sticks
                 of memory before?  I'd be more suspicious signal integrity
                 or load problems first.  That or even if your motherboard
                 is designed to support that memory configuration (assuming
                 yours is an older design).
        \_ I have never had success mixing ram. Now a days I always order
           from the same vendor and if possible the same part number. If
           I can't I replace all the ram so it all matches. Linux/FreeBSD
           seem to be far more sensitive to this than windows.

11/3    In Windows XP, how can I trace all the files called by an executable?
        I'm trying to run Quicken off of my USB thumbdrive by doing this:
        1. Installed a copy (I bought at Fry's) on XPProAtHome machine.
        2. Copied executables to USB thumbdrive.
        3. Used SysInternal's procexp.exe to see what files qw.exe used when
           it was running on XPProAtHome.
        4. Copied some additional dll's and other files that procexp.exe
           highlighted to USB thumbdrive.
        5. Copied additional dll's and other files to appropriate directories
           on XPProAtWork machine.
        6. Ran qw.exe off the thumbdrive using XPProAtWork.
        7. Quicken hangs at splash screen.
        Apparently, just copying the executables and additional dll's and files
        is not sufficient.  Procexpe.exe also highlighted some registry keys in
        use on the XPProAtHome machine, but I haven't tried copying those to
        the registry in XPProAtWork yet.  Before I do that, is this even
        feasible?  I googled on 'tracing files called by executables' but the
        results mostly relate to networking issues.  Any help is appreciated.
        \_ try filemon?
        \_ Why not mount the drive then install onto it?  Doesn't Quicken ask
           where you want to install it?
           \_ But it will still modify the registry of the target machine,
              right?  So will it run if I try to take it to another machine?
              -op
              \_ Oh, probably... why not just put your data files on the thumb
                 drive?  Why do you actually need the bins there?
                 \_ Yes, I plan on putting my data files on the thumbdrive too.
                    I want to put the binaries on there so that I can run it
                    on any PC without having to install it on that PC.

10/13   Remember rapidly climbing memory prices?  All the memory makers were
        in on it.
        http://www.cnn.com/2005/TECH/10/13/samsung.price.fixing.ap/index.html

10/8    In memory of the Rush, buried by Classes of '07 and '08,
        March 23, 1905. Requiescat in Pace.

9/30    I've got a Via C3 (using kernel-image-2.4.27-2-586tsc) and 1504 MB
        of RAM. Only 904 MB shows up with the "free" command. Is there a way
        to get this kernel to see/use all the available ram? or do I need
        a newer (or custom compiled) kernel?
        \_ you need a kernel compiled with bigmem support, or something like
           that. or highmem. i forget the exact term.
           \_ Great! Thanks. I think I found it: CONFIG_HIGHMEM4G   -op
           \_ Great! Thanks. I think I found it: CONFIG_HIGHMEM   -op

9/23    Anyone ever built their own TiVo/PVR? any suggestions/advice
        before I build one? I'm reading my way through http://byopvr.com
        trying  to figure out how to build one.
        \- isnt the hardware heavily subsidized? why would you want
           to build one from scratch? or is this a geek learning exercise?
                \_ definitely the latter. granted I won't have to pay a
                   monthly service fee, I'll probably pay more up front
                   if I build one.
           \_ I built mine out of mythtv because for some reason, I've
              generally been against paying for a subscription of any kind.
              You also get more features from building your own.  Another
              reason these days with TiVo adding various restrictions is
              that you're not under direct control of some corporation's
              whim.  I admit, TiVo probably is more stable, but mythtv is
              what works for me.
              \_ How much did you spend on the hardware?
                 \_ I've used parts that are just laying about few times.
                    The latest incarnation I'm working on cost me about $400
                    for the shuttle-like case, cpu, and memory, I believe.
           \_ I like to be able to watch recorded programs remotely :)
        \_ You mean like mythtv?
        \_ Decide on size of machine and noise level you find acceptable.
           \_ how about those VIA C3 chip.  I know that you can run
              800Mhz without any fans.  The newer 1GHz chip from
              xxx nm fab reportly can run without fan neither.  tried it?

9/21    What's the benefit of ECC RAM?
        \_ When the RAM has a bit error it alerts you about it and
           attempts to correct it. There is an extra 'parity bit' that
           works like such bits do in RAIDs. Without ECC RAM you have no
           idea if your data is corrupt or not. Always buy ECC RAM for
           data you care about. I recommend always buying it if your
           motherboard supports it.
           http://www.anandtech.com/guides/viewfaq.html?i=3
           \_ Is there EC RAM which you still have an idea if your data is
              corrupt or not, although not being able to correct it?
              \_ That would be called parity.
              \_ There is. It's just called 'parity RAM' but I never see
                 it (or support for it). It seems ECC has rendered plain
                 old parity RAM obsolete.
                 \_ See, it works like this:  if you have only 8 bits
                    in your RAM line, (RAM stored by bytes), you can
                    add a chip and get 9 bits: 8 data, and parity.
                    If you have 32 bits in your RAM line, you can add 4
                    bits and do ECC on those 32 bits.  So ECC RAM is
                    party RAM, it's just stupid not to do ECC since you
                    have 4 bits per 32 instead of just 1 per 8.  All RAM
                    these days is at least 32 bits "wide", so everyone
                    just does ECC.... --PM
                    \_ Yes, but you can still just 'check' the data and
                       not correct it, but why do it when ECC is so much more
                       useful and can check 2 bits (and correct 1) instead of
                       just checking 1?
                       \_ the idea is that the more parity bits you have,
                          the higher you probability of catching the error.
                          with only 1 bit, you'll not catch anything. with
                          4 bits, you catch more. but with 4 bits, you
                          might as well correct as well.
                          \_ Why can't I catch 1-bit errors with 1-bit parity?

8/9     I had a motherboard die after like 90% of the capacitors leaked.
        Appareantly this is a widespread problem (badcaps.net). I'm
        wondering, (a) why I hadn't heard of this earlier, and (b) why
        there hasn't been a big recall of defective components.
        \_ (a) because you don't read /., http://theinquirer.net, ....  Do you
               read anything?  Try to keep up a little, will ya?  This is
               over a year old stuff.
               \_ I read /. almost everyday. When was it on /.?
                  \_ Years ago.
           (b) Abit sort of recalled.  Problem is too widespread to recall.
               Trying to do so would probably take out a lot of companies.
        \_ Not only that, you apparently don't read the MOTD or archives,
           this was mentioned more than once here. The answer is to get
           replacement caps (assuming your other components are intact)
           and solder them on to the board. Also, MBs are under $100 US,
           the cheapies can be found for sometimes less than fifty, so
           it's such a cheapie item that a general recall would've killed
           some of the MB companies because the margins are razor thin.
           And yes, there is a sort-off coverup going on.
        \_ I just had my thinkpad t41 crap out under warranty... does anyone
           know if they had this problem too?  It died slowly, e.g. showing
           signs of going out of spec in the video memory first.
           \_ Just about EVERY company was affected, including IBM.