Our Dear Leader Sam Hocevar has previously blogged about
PIC
and inline ASM. Today I came across a sort of extension to this
problem.
Consider the following code, which implements a double word compare
and swap using the x86
cmpxchg8b instruction (for a bonus you
can lock it to make it atomic).
#include <stdio.h>
typedef struct double_word_t
int a;
int b;
double_word;
/* atomically compare old and mem, if they are the same then copy new
back to mem */
int compare_and_swap(double_word *mem,
double_word old,
double_word new)
char result;
__asm__ __volatile__("lock; cmpxchg8b %0; setz %1;"
: "=m"(*mem), "=q"(result)
: "m"(*mem), "d" (old.b), "a" (old.a),
"c" (new.b), "b" (new.a)
: "memory");
return (int)result;
int main(void)
double_word w = .a = 0, .b = 0 ;
double_word old = .a = 17, .b = 42 ;
double_word new = .a = 12, .b = 13 ;
/* old != w, therefore nothing happens */
compare_and_swap(&w, old, new);
printf("Should fail -> (%d,%d)\n", w.a, w.b);
/* old == w, therefore w = new */
old.a = 0; old.b = 0;
compare_and_swap(&w, old, new);
printf("Should work -> (%d,%d)\n", w.a, w.b);
return 0;
This type of CAS can be used to implement lock-free algorithms
(I've
previously
blogged about that sort of thing).
The problem is that the
cmpxchg8b uses the
ebx
register, i.e. pseudo code looks like:
if(EDX:EAX == Destination)
ZF = 1;
Destination = ECX:EBX;
else
ZF = 0;
EDX:EAX = Destination;
PIC code reserves
ebx for internal use, so if you try to
compile that with
-fPIC you will get an error about not being
able to allocate
ebx.
A first attempt to create a PIC friendly version would simply save
and restore
ebx and not
gcc anything about it,
something like:
__asm__ __volatile__("pushl %%ebx;" /* save ebx used for PIC GOT ptr */
"movl %6,%%ebx;" /* move new_val2 to %ebx */
"lock; cmpxchg8b %0; setz %1;"
"pop %%ebx;" /* restore %ebx */
: "=m"(*mem), "=q"(result)
: "m"(*mem), "d" (old.b), "a" (old.a),
"c" (new.b), "m" (new.a) : "memory");
Unfortunately, this isn't a generic solution. It works fine with
the PIC case, because
gcc will not allocate
ebx for
anything else. But in the non-PIC case, there is a chance that
ebx will be used for
addr. This would cause a
probably fairly tricky bug to track down!
The solution is to use the
#if __PIC__ directive to either
tell
gcc you're clobbering
ebx in the non-PIC case,
or just keep two versions around; one that saves and restores
ebx for PIC and one that doesn't.
New Jersey is one of the states holding presidential primaries on this Super Tuesday and since you asked, I voted for John McCain.
The chief reason for my vote is of course the war against Islamic terrorism. Now that we are finally gaining the upper hand in Iraq, we can't afford to elect someone who will fail to see victory through. Mind you this isn't necessarily a reason not to vote Democrat. The anti-war wing of that party has been thoroughly marginalized. Hillary atleast has a commendable reputation for viciousness. Even Obama has made some encouraging remarks (ok that was probably a red herring but I have the audacity to hope you see.) But while the Democrats do not take the anti-war crowd seriously, they do depend on them for a great deal of their funding. A Democratic president might feel the need to throw them some bones inadvertantly screwing up things in the process. On the Republican side, Huckabee might leave tactical planning to Jesus. Romney would toe the line but there is no sign that he has any original thoughts on the matter. While I continue to support the President, it must be admitted that this whole war is stretching out as long as it is because mistakes were made. We need options beyond "staying the course" and "cutting and running" and McCain a military hero and early supporter of the surge could be the one to find them.
The other big issue is supreme court appointments. One, possibly two justices may need to be replaced in the next four years. We need to ensure that the court does not veer leftward again as it will inevitably do during a Democratic administration. Here there is a big split in the Conservative establishment. Ideologues consider McCain to be insufficiently pure and worry he will end up appointing another Souter or O'Connor (both appointed by Republicans it should be noted.) Pragmatists point out that McCain is viewed more favorably than Romney or Huckabee amongst the independent and moderate voters who will be crucial to victory and that even the slight chance of a bad choice is better than the certainty of a bad choice if Clinton or Obama get to make it.
Social issues are important to many Republican voters and in this area too there is some grumbling about McCain being too liberal. Some have even suggested that if McCain beats Huckabee or Romney, a Democratic victory and four years out of power would be preferable in order to bring the GOP back to the straight and narrow. Although this contingent is loud I frankly don't think they have the influence people think (As evidenced by the actual performance of Huckabee or previous champion Fred Thompson.) When push comes to shove, they will rally behind the party candidate especially if McCain makes a bona fide social conservative (Huckabee himself?) his vice-president.
On economic issues, the president's job is to not interfere and let business run itself. I don't see any of the Republican candidates being much different in this regard. Or the Democrats for that matter.
[Careful readers will have noticed I have avoided mentioning Ron Paul. Paulestinians, consider this carte blanche to send me your long, rambling, incoherent screeds about how Dr. Paul has the answers to everything. Because God knows I haven't read enough of them lately.]
Update: Kevin asks what I think about the candidates in terms of "geek" issues such as net neutrality, DMCA etc. IMO there is not a whole lot of difference between the major candidates on these issues. Republicans one would think would be more beholden to corporations in general. Democrats on the other hand have a closer association with the media who are behind many of the sillier IP laws. Overall, I don't think it ought to be much of a factor in choosing the president even for Free Software types. Rather we should focus on the legislative branch (in Congress as well as the states) which holds the real power here and in the court of public opinion.
No Debian content here. Just like last year I decided to watch one of the largest sport events in the USA - the final game of the National Football League Super Bowl XLII (roman numerals are different for the sake of it 
). I came to like the Japanese take on American football, so I watched the Superbowl last year and was appalled at the lack of any entertainment value in the game. I was told that it was one the worst Super bowls ever, so I gave it another try.
Just like last year the American football looks to be very lawyer-friendly: bundles of complex rules and exceptions. The game could have been much simpler. Like for example, a lot of rues come from the fact that the game timer almost never stops. It is good for TV companies as they can plan their broadcast times better, but horrible for the stop-go game like the American football is. Just stopping the timer on every point the ball is stopped like they do in final seconds would cut out a lot of rules. Playclock and all. Rules on use of timeouts and stuff. Of course the game time would need to be adjusted down a bit. Now you have 4×15 minutes of ‘game’ where at most 20 minutes of that is in action. That game time is a lie - most of the time people are just standing around.
Terminology irks me quite a lot. Why is it a “1st down” when there have been two “1st down”s just a few minutes ago? Something along the lines of “down counter reset” would be much clearer. And what is cryptic “1st & 10″ doing in the middle of the screen? It would have been much clearer if that info would be split: “Downs: 1/4″ and a measuring arrow between the ball and the yellow line with “10yd” written on it. And then there is an obscured “two in two” or “two for two” or “two in O” phrases which apparently means the same as “two 1st downs out of two tries” in the context. Can’t the commentators be a bit more elaborate and actually say the units of measurement? “Giants strength first tonight”? How about “in the lead”, English anyone?
The camera placing editing is just horrible - the camera direction is switching all the time and that is confusing as hell. In normal football when professional camera teams are working, you always get all the shots from one side and even one general angle. If for clarity reasons a camera shot from other side is needed (usually in a replay) they always say that the image is from the reverse side in large letters on the screen. Here a lot of close-up shots were all over the place and even rotated around inside one shot! That might look fancy, but is just crappy camera work and editing.
Then commentators start talking walls of numbers and walls of names. Unless you know by hearth who plays at what position, there is no way to understand what those people are saying. Than is just sloppy commentary. Is FOX incompetent in all possible ways?
Then again the funky rules came into play just at the end of the first quarter - I have no idea why the ball was moved to the end line. Apparently the defender did or didn’t something. Did not see anything worth penalizing the team like that at all. They dropped the term ‘pass interference’ later on. Isn’t interfering with the pass the whole point of the defense play in this game??? Lawyer game, I’m telling you. And later on, pulling on guys arm while he is catching and being tackled apparently is just fine. And holding is illegal???
On the big penalty on the Giants the judge could have just said that the ball is out of bounds. Giving a 10yd penalty for such a minor thing is just ridiculous.
The halftime show was kind of meh. Staged to extreme, but no soul nor sense. And what were all those people doing on the field? Isn’t that bad for the grass? And the commercials .. Average at best. Horrible at worst. For example a government commercial saying parents to protect their prescription medicine to .. help out your neighborhood drug dealer??? Oh and does anyone still doubt that media is controlled by republicans after that Coke commercial where the democrat is portrayed as bald and ugly and the republican is portrayed as the nice guy going for the compromise. And then they both Segway into the sunset? Brainwashing and prejudice embedded in a Coke commercial !?! WTF is going on in USA !?
The 12-man-on-the-field penalty in the 3rd quarter is another legalese bullshit in the game. It is clear that the extra player had no effect on the game so it is just ridiculous for the judge to penalize the team for that. A couple small penalties were given for Patriots as well, judge bias was not as prevalent as last year.
The camerawork stayed the same through the game - confusing, flashy and sloppy. Other cameras are in the shot line all the time. The game however did compensate for that all. The 4th quarter was very nice and exiting. I had no idea of the teams before the game, but I did like the Giants more - they just seemed to have more will and soul in their game. So I rooted for them since the 2nd quarter and it was very satisfying to see them win the game in the final seconds.
Summary: good game, great final quarter, horrible coverage and crappy sideshows (adverts and halftime show).
As Internet-connected people, we tend to interact with a lot of
people who aren't necessarily geographically co-located. Hence
today's cool little program I found: gworldclock. You pick
an arbitrary number of timezones to have in your list, and the
current time in all of them is displayed in a neat list, quietly
ticking away regularly. Small? Yes. Trivial? Getting close to
it. Really Useful? You betcha. It also has a very tidy
'rendezvous' mode, where you choose a time, a timezone (from your
list of selected timezones) and it shows the corresponding time in
all of your other chosen timezones. Again, something that 20
minutes of coding will probably produce, but the whole point of this
open source thing is not having to re-invent the wheel...
The 'g' does indicate a GTK application, but there is a
kworldclock for people who wouldn't be seen dead with a GTK
application on their desktop, and a console-only version,
tzclock, for anyone who lives the text-mode life.
Yes 
In the past I was not blogging small things if they were not worth a full blog entry on their own (or if I did not have the time to make up a full entry of it :). Inspired by the example of 
My. That's a mouthful of a title. I went to this event on Friday, which
was run by 
Hello! I am in San Francisco. Offers of beer, entertainment and marriage will be given appropriate consideration. Interestingly, moving from Melbourne to here seems to have left me back on GMT, which I'm sure isn't quite what I want.
Well I've had a busy weekend, but I'm sober now.
I made a new release of xen-tools, which has a couple of minor bugfixes and not much else. I also released a new update for Debian of the xen-shell which fixes a couple of bashisms.
Finally I've managed to sign up two new users to my 
As the Debian release timetable for Lenny is now clear and the soft freeze is only a few weeks hence, I've decided that I won't look for the libqof1 to libqof2 transition before Lenny. There just isn't time to complete the GDA backend in time.
Given three groups of groups of people A, B, C, responsible for 3 queues Qa, Qb, Qc ; and an elements e which have to go through Qa, then Qb, then Qc.
Not being able to sleep sucks. It is now 6 AM, and if I wanted to get up as planned, I have less than two hours.
Since I can't concentrate at all with that little sleep, I guess I will make it at least four hours. Still, tomorrow will suck, especially as I will be dealing with crypto code.