KatieMember Since 15 Jun 2006
Offline Last Active Jan 07 2016 08:43 AM
- Group Members
- Active Posts 881
- Profile Views 8,603
- Submitted Links 0
- Member Title Member
- Age Age Unknown
- Birthday Birthday Unknown
Posted by Katie on 01 October 2011 - 02:18 PM
Don't bother doing general studies, it's just wasting time. Steer clear of what are regarded as "soft" subjects. "Media studies" for example being one. Likewise English Lit is useful if you're about to embark on a degree in English Lit, but CS depts will bin your application.
Good news time; compsci courses aren't very oversubscribed!! In other words; good A level grades and you're sorted. No need to be captain of the rugby team and play the lute and also walk on water to differentiate yourself like if you were applying for a "afternoon lectures only" subject like (say) English Lit.
Here is an example of ICL's entry policy; http://www3.imperial.ac.uk/computing/teaching/undergraduate/computing/admissions/advice/
Notice they do not require A-level computing, however TBH, it's probably a useful building block.
Similarly, here's Cambridge's equivalent -- http://www.cam.ac.uk/admissions/undergraduate/courses/compsci/requirements.html
Lots of science, lots of maths. Sorry.
Right, now you know that you need to be doing lots of maths, lots of science, you can go look for a suitable college to go to. Generally this will be a "sixth form" college, not a "further education" college. Colleges in the UK tend to be split into two types; sixth forms are more academic the others are "vocational"; they do courses in (for example), hairdressing, beauty therapy, office administration and so on. If the website does a lot of talking about vocational studies, mentions things like RSAs or diplomas or NVQs, they've maybe not got the focus you'll need.
Your school should really have routes into suitable colleges -- So if you pick a set of four or five A-levels that you'd be interested in studying that fit those sorts of entry requirements, they'll be able to point you in the right direction. I know you're in London, but you can't be the FIRST person at the school to aspire to university :-)
You do not really need a masters degree to do well in the IT industry. I know people who have done the BSc/MSc/PhD routes and while they generally have a useful toolbox of exotic maths and reading papers about FTL neutrinos doesn't make them frown as much as it does me, they aren't intrinsically better software engineers. If you're going to do a masters, do an MBA later on in your career -- being both technical AND business savvy is pretty valuable. But you shouldn't be doing that until you've got some time served at the coalface. Also, if you're sneaky about it, you can locate a company to work which will pick up the bill for doing the MBA (which is NOT CHEAP by the way!)
It's not that an MSc/PhD isn't something to do, but do bear in mind that a) slots for people doing them are restricted, b) you'll be a student for longer. Compsci masters and doctorates are not a key to instant success. On the other hand, it's not unknown for master/doctoral projects and groups to suddenly become startup companies and also I've found that people who have done them seem to have larger social networks within the IT industry. I'm unusual, for example, for being a mere grad still being in fairly close contact with a lot of my graduating year. Whereas masters and doctorates people seem to be more connected. That DOES help with employability, but if you know about this upfront, you can do it without spending five years longer eating pot noodle.
Posted by Katie on 19 September 2011 - 01:36 AM
Posted by Katie on 17 September 2011 - 01:19 AM
Posted by Katie on 14 September 2011 - 01:40 AM
If you put a unit cube through the transform, then the determinant will tell you what the volume of the output cube will be; in other words it tells you the volume scaling that the matrix will do.
Hence, well behaved 3D transforms will all have positive non-zero dets; and all rotation/translations which you expect to do no scaling will have a det of 1.
This also helps explain why zero det matricies can't be inverted. In order for it to have det zero, it must shrink one of your unit cube dimensions to zero. Hence there can't be a matrix which goes the other way, because there's no way to reinflate that deleted dimension back to one. Hence there's no inverse for the matrix.
Obviously, since you HAVE to squash one of the dimensions away to go from the 3D world to the 2D render plane, this means all projection matricies must have det = 0. And hence, this is why there is no general case system for taking a pixel in the window and finding it's location in 3D space, because that would amount to being able to invert your projection matrix.
Posted by Katie on 13 September 2011 - 03:43 PM
I remember reading adverts in the 1990s for the languages which would replace C and C++ with languages so easy that managers would be able to write the code themselves... I'll bet no-one's ever heard of any of them these days.
Easily 20 years this conversation's been going.
It's actually getting very very hard to hire good C/C++ people. And that's a problem because Google runs on hardcore C++ stuff; KVM and Xen need good C hackers, performance libraries for Android need C and C++ coders... Just this week Google is talking about tech to run C++ code natively inside sandboxes because they know the world isn't replacing that stuff any time soon. Good. Means my pay will go up for my rarity value...
We've all been waiting 50 years for Cobol to die and it still won't.
Posted by Katie on 03 September 2011 - 08:02 AM
No, they're not. They're bad if you're an academic computer scientist. Academic computer scientists never actually have to ship code. Globals work, use what works, ship code.
"really messes up my goal of keeping things modular."
Don't have a goal of being modular. Have a goal of "shipping code". What value to you is "being modular"? None. It's an **ideal**. It's not a goal. Your goal is to ship code. That means you get to sacrifice ideals sometimes. This could be one of those times.
You're over-thinking this problem. Everyone always does. Everyone thinks there's a problem with "logging" because it's always either too verbose or not verbose enough and it's never quite right and everyone thinks the way to fix it is to somehow make the logging production more complication.
The problem isn't with the logging. It's with the tool on the other end. Make your program be either "verbose" or "succinct" based on a flag. That's all you need on that front.
Then just output strings but have smarter tools looking at the output.
The UNIX world manages to handle tons and tons of logging using "syslog" which is hardly sophisticated. "syslogd", the program on the other end, that has versions which'll do everything up to handling global networks with millions of servers...
Don't try and solve your logging analysis problems during your log generation.
Posted by Katie on 31 August 2011 - 10:07 AM
Posted by Katie on 17 August 2011 - 05:39 AM
Lay them all out in a big block of ram, save that to disk, gzip it. When you come time to load it, you're only loading one file -- you hoof it into one giant block of ram (only 1 allocation, only 1 directory walk). Ungzipping a file is pretty easy on loading it. In addition your compression should be even better; larger files generally compress better than a set of smaller files because the compression can find more similarities to reduce.
Instead of generating the mipmaps on load every time, generate them on first run and save them out to disk. Again, do this one big block if possible. On load, pull it in, rattle down it transmitting the mipmap levels individually.
If you're on an OS which supports mapping the files, you don't need to load them or do an actual memory allocation. Map the file into your address space, provide suitable "linear access" hints to your buffer caching system so it will do read-aheads, start accessing the memory and away you go. The OS will try and load pages ahead of your accesses in the background; you don't have to wait for the "read()" call to complete. The effect of this is that you can already be generating and uploading mipmaps for texture 1 while the disk IO channels are still loading texture 20.
Instead of asking the GLU library to generate your mipmaps, you could build them yourself by using rendering of the fullsize image to pixel buffer objects and copies internal to the card to make them. That way the mipmap data doesn't have to transit the host-GPU bus.
Posted by Katie on 16 August 2011 - 02:15 PM
Posted by Katie on 14 August 2011 - 08:18 AM
Posted by Katie on 12 August 2011 - 05:41 AM
Go to job fairs, talk to people there.
Find the developers, talk to them in the pub.
Get recommended by friend who already works there.
Join to do something other than development -- retaining good testers is a nightmare and most companies are always looking for more.
Posted by Katie on 10 August 2011 - 02:57 PM
You are james bond and I claim my five pounds?
Posted by Katie on 10 August 2011 - 05:03 AM
Well, one reason is historic; power-of-two numbers are easy to manipulate quickly. Games on 8bit computers and consoles often used 16x16 tiles because it was a compromise between detail, storage space for the tiles, and being able to access the pixel data quickly.
In addition screen sizes were also often powers-of-two. For example on the CPC, 16x16 tiles means you can have a 20x10 screen with 40 pixel lines left over for a status display at the bottom.
When hardware assist started to appear for the graphics, it was often power-of-two oriented, again for hardware simplicity reasons. Early OpenGL and DirectX (and other APIs such as Glide) similarly limited textures to 256x256 for hardware reasons, meaning that 16x16 or 32x32 tiles would fit better and not waste precious texture memory.
Posted by Katie on 10 August 2011 - 04:49 AM
Except in anything but a very badly run meeting, ideas aren't just "shot down" without there being justifications -- particularly when the idea/explanation comes from (say) the highly experienced developer who has historically shipped good product. It's not that the intern in the room doesn't get to have an voice an opinion, it's just that don't get to say "no" without a reason. They're quite allowed to say "Ah, could I just point out this reason why this may not work...", but they don't get to just say "nope". Well, not and keep their job for very long.
Popularity is, historically in a number of fields, a poor way of choosing correctness. No-one likes an answer which is difficult and would prefer an easy one and popularity contents will select the latter; which is fine as long as you can guarantee that there will only ever be easy answers to things.
Posted by Katie on 09 August 2011 - 12:47 PM
I can't see where the magic technology is. I can see that you've implemented a "hard and soft layers" architecture involving scripting to connect component parts together. Could you be more specific about what it is I'm missing?