FACM wrote:Sometimes, waiting hours for big packages to recompile with a new USE flag on is a little irritating, but sometimes I need the excuse to walk away from the computer.
You, sir or madam, are everything that is wrong with the typical Gentoo user.
Use flags do not make your computer faster. They do not reduce size or memory usage. Anyone telling you otherwise is lying to you. Compiling your own binaries won't improve performance either. That last one can be tested.
What I propose is this: I've currently got a stock install of Ubuntu running in Parallels. I'll do a stage 1 install of Gentoo in a secondary VM. We'll then run some code that I've already written to perform some benchmarks on my Mac, but is entirely compatible with Linux as well. I'll post the code and the results here. We will then run some math on the results (we're gonna run the code multiple times: it uses a pseudorandom number generator to add two large matrices) and check for an Honestly Significant Difference between the results and it turns out that there is one in Gentoo's favor, I'll consent that you're not a kool-aid drinking troll.
Hi there, let's clear some things up here.
First, I made no statement as to why I changed USE flags. I made no claims that USE flags accelerated anything, reduced binary size [though someone else did mention this earlier], or anything of the sort. My exact statement only indicates that changing USE flags requires recompiling the source. If you're curious, I added xinerama support to my system so I could get dual monitors working the way I wanted them to. Feel free to point me to a laundry list of configs and packages I could change manually that's easier than adding xinerama to my USE flags and "emerge system". I doubt one exists.
Secondly, compiling your own binaries can improve performance. The cases where it will be an Honestly Significant Difference are few, but they are notable. My simple, far-from-scientific test was comparing GIMP on my Macbook (1.8 GHz Core Duo, standard build) to GIMP on my tower (3.0 Ghz P4, -O3 -march=prescott). The plasma plugin ran about 3 seconds faster on the tower for a 2000x2000 image using the same seed value. Fairly impressive, considering the weaker CPU. Not sure how much threading the plasma plugin does, though. For something like OpenOffice, Firefox, etc where the code wouldn't benefit from SSE/2/3 or your preferred architecture's floating-point instructions, there won't be a great difference in speed.
On a note related to your benchmark proposal: Macs, despite the UNIX base, will not always run UNIX code as fast as an actual UNIX machine. I'm not sure if it would affect VMs or not. Some things are moved around to other functions for reasons I'm not clear on. A benchmarking tool discovered this, and had to rewrite the mac client to make it match the UNIX scores on the same hardware. I'll try and find a link to it since I'm being ambiguous and vague.
That covers your issues with me. Now, can you explain why the sudden outburst of hatred at me? [Beyond "This the Religious Wars forum"?]Instead of explaining sudden outbursts of hatred directed at other individuals, let's instead not have any of that. Direct your outbursts of hatred at Gentoo, if you wish, but we'll be skipping the personal attacks. Thank you.