/prog/ - sse asm question

Name: Anonymous 2011-04-20 22:59

I was playing around with some asm and was wondering. Why is it that when i use sse to manipulate data in parallel that it seems to accumulate errors that i otherwise dont get with my output?

Name: Anonymous 2011-04-20 23:14

>>1
Ask tech-support.

Name: PROFESSIONAL AUTIST 2011-04-20 23:15

UM, IT'S FUCKING CALLED SATURATION ARITHMETIC, YOU FUCKING LARDCUNTFUCK.

Name: Anonymous 2011-04-20 23:22

>>3
Well... no need to get angry... I was just asking.

Name: Anonymous 2011-04-20 23:29

>>3
Umm... also that does not sound like the problem i am having. Let me explain. When doing operation on small sets lets say of 1000 numbers than the errors are small. Only a few decimal places will be off. But if i do operations on large sets like a set of 10000000 numbers than the errors accumulate. The errors also dont present them selfs when I only use the scaler aspects of sse but if i use the parallel aspects then the errors arise.

Name: nambla_dot_org_rules_you 2011-04-21 0:24

>>5
Read the following paper before you look like a bigger dumbass than what you already are.

http://download.oracle.com/docs/cd/E19957-01/806 3568/ncg_goldberg.html

Name: Anonymous 2011-04-21 0:24

http://download.oracle.com/docs/cd/E19957-01/806-3568/ncg_goldberg.html

Name: nambla_dot_org_rules_you 2011-04-21 0:25

>>3
Pull my finger bitch.

Name: Anonymous 2011-04-21 0:39

OP, try Common Lisp. It uses error-free rational arithmetics, instead of relaying on a good-for-nothing buggy FPU.

Name: Anonymous 2011-04-21 1:11

ok im sry but all the responses i have gotten are just terrible...
>>9
wtf are u talking about ... have i not said several time i am using sse. I did not say anything about the fpu. And why suggest a higher up language... i am talking asm(Assembly).

>>6
>>7
yes , i know about rounding errors... that is clearly not the source of the problem. If it was then both the scaler and the parallel would have the same output.

It clearly seems that i over estimated u guys... u dont seem to have any idea what i am talking about.

Name: Anonymous 2011-04-21 1:55

And nothing of value was lost.

Name: Anonymous 2011-04-22 5:52

>>10
u dont seem to have any idea what i am talking about.

That's because we are pathetic. And not telepathetic. Here's the deal, young boy: you wanna help, you gonna post some code. No code - no help. Got it? Now move along.

Name: Anonymous 2011-04-22 11:08

>>12

I don't even want to see his code, code written by retards is bad enough as it is, assembly code written by retards is even worse.

Name: Anonymous 2011-04-22 11:13

>>1
Because SSE always operates in round-to-nearest, fast denormal mode. The amount of error is negligible for most multi-media applications, for which SSE is intended.

Name: nambla_dot_org_rules_you 2011-04-22 11:41

>>13
And how do you feel about code written by an autistic jew?

Name: Anonymous 2011-04-22 11:42

>>15
Go fuck an autistic jew.

Name: nambla_dot_org_rules_you 2011-04-22 11:44

>>16
I will the next time your mom has the day off.

Name: Anonymous 2011-04-22 11:52

>>15

I haven't seen some but it would probably be very well structured.

Name: Anonymous 2011-04-22 13:14

>>15
U MENA RMS

sse asm question

1 Name: Anonymous 2011-04-20 22:59

2 Name: Anonymous 2011-04-20 23:14

3 Name: *PROFESSIONAL AUTIST* 2011-04-20 23:15

4 Name: Anonymous 2011-04-20 23:22

5 Name: Anonymous 2011-04-20 23:29

6 Name: nambla_dot_org_rules_you 2011-04-21 0:24

7 Name: Anonymous 2011-04-21 0:24

8 Name: nambla_dot_org_rules_you 2011-04-21 0:25

9 Name: Anonymous 2011-04-21 0:39

10 Name: Anonymous 2011-04-21 1:11

11 Name: Anonymous 2011-04-21 1:55

12 Name: Anonymous 2011-04-22 5:52

13 Name: Anonymous 2011-04-22 11:08

14 Name: Anonymous 2011-04-22 11:13

15 Name: nambla_dot_org_rules_you 2011-04-22 11:41

16 Name: Anonymous 2011-04-22 11:42

17 Name: nambla_dot_org_rules_you 2011-04-22 11:44

18 Name: Anonymous 2011-04-22 11:52

19 Name: Anonymous 2011-04-22 13:14