Sign in to follow this  
  • entries
    455
  • comments
    639
  • views
    422037

Testing again...

Sign in to follow this  
_the_phantom_

64 views

Friday night I decided to make effective use of my time; I went out, had a few drinks and danced around like a loon (a fact my lower back has been pointing out to me all day, heh). Of course, with only a couple of days left to hand in a dissertation and with it still short of the 10,000 word mark this was of course a great plan.

Well, it was for two reasons;
1 - I needed to relax a bit before the big push to the end this weekend
2 - I needed to run the program to get some results and a full run is estimated to take somewhere in the region of 2h to complete...

However, I came home around 3am to find the program had compeleted and I had results; huzzah!

Today was spent looking at them... well, firstly looking at them and then deciding that, while the format was good for output and for others to look at from the point of view of analysis and graph drawing in Excel it kinda sucked slightly [grin]

So, a bit of time was spent today writing a program to parse the data into slight more useful formats;
- 4 text files which contained the min, max, average and totals for all the tests (56 in total)
- one text file per test which is pure CSV data for import into Excel

Having pulled in this data I noticed a couple of things;
- Massive spikes at the start of some of the tests *grumbles* This was mucking with my max and average results a bit
- The GPU only 'drive' method was taking a VERY long time to complete

The former is fixed by discarding the first X iterations of data and collecting from then onwards, the latter I might have helped via a change from glTexImage() to glSubTexImage() for the 'driving' value upload. I might even change the internal format of the driving texture from a 4 component to a single component value as it only has to be replicated across all four values in the TLM update function anyways. This would save memory and allow for faster uploads of the data.

The net result of all of this is that I'm going to have to run all my tests again [sad]

On the upside I get a chance to 'fix' the bad data and get a new metric; frames per second.

I wasn't collecting it initally, didn't see the point, however it might be worth doing so because if the GPU only results stay bad on the timing scale I can at least point to the fps and say 'but look, here it was better, I just don't know why' heh [grin]

So, off to make a quick change to the GPU only driving code and then to bed; long day tomorrow, I only have until Monday 8:15am BST to get this done and be there before 8:30am to get it all printed off.

Oh, current word count is now 10,551 (I was happy when I crossed that 10,000 barrier), so I'm just under 1,500 off the min word count, however the critical evaluation is recommended to be around 4 pages which takes most of that anyways and I still need to;
- Talk about results
- Talk about "normals"
- Clean up previous text so it makes sense (which has already gained me a fair few words in just one section)

After all that it's just the fiddly stuff to do and diagrams to get done and in.

No worries [grin]
Sign in to follow this  


0 Comments


Recommended Comments

There are no comments to display.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now