Long term build times

[Garbanzo]

As time goes by computers get faster and software gets bigger. Over the long run, will build times get longer or shorter? How does the time to update world for a general purpose desktop compare to 10 years ago? What might it look like 10 years from now?

[Muso]

[toralf]

IMO the compile times of llvm, clang, gcc, rust,chromium et al exploded in the past 2 years.

[fedeliallalinea]

[Hu]

Does a general purpose desktop include the inner-operating-system-masquerading-as-a-browser (modern Chrome, modern Firefox) as an integral component?

[Muso]

[Ant P.]

[eccerr0r]

What circumstances?

Same hardware, same software: the baseline, won't change.
Faster hardware, same software: time goes down
Same hardware, newer software: time goes up
Faster hardware, newer software: usually same as before.

Firefox itself also is increasing in compile time, and that's not counting the rust dependency (plus there's no distrust (yeah, right) whereas old firefox could distcc).

Has anyone gotten back down to a 1-minute kernel compile with modern hardware? My "fast" machine is old and slow now compared to modern machines, but even when it was new, I couldn't get kernel compile times back to a minute and a half or so when I had my dual overclocked celerons, though it got close - maybe 2 and a half minutes.
On the other hand, that dual celeron machine will take much longer to build that kernel.

Now my i7 takes probably 5 mins or so for a recent kernel, but I haven't timed it in a while.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[Goverp]

[C5ace]

My former employer has a industrial application written in C++. The applications had to be customized and recompiled for each customer. Compile time was about 6 hours. They rewrote the application in C. Compile time was cut to 1.5 hours.

[eccerr0r]

When people moan about compile times, the bean counters then end up using interpreted languages... an even worse solution...
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[Anon-E-moose]

[John R. Graham]

[eccerr0r]

Well, in 2005, the P90 is already almost a decade old, so that's not a proper comparison - else I could say my old K6-233 takes a 2 months to build current software and conclude software compile times are O(n!) as time goes on...

Need to at least compare mediocre P3 (the P4 was introduced in 2005) compilation times, say a P3-600, and that 1 week compilation time drops to a day - 24 hours - and we're back to where we started...

Software compilation times are disgusting.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[John R. Graham]

Not really:

[eccerr0r]

I also wonder if it really took 24 hours to build gcc on a P3-600 or 1 week on a P90 too, unless you were running on poorly configured hardware.

I had a personal record of a 5 hour Linux 0.99pl12 compile time on a 386DX40 - but this is also not a proper compile time comparison as I know it was swapping to a slow hard drive the whole time, a "then modern" disk upgrade on the same machine dropped the compile time to an hour; but even this was not a proper comparison as 486DX2's were already out by then.

Also it took me almost 6+ hours to compile gcc-2.something on a 33MHz SPARC clone with only 8MB RAM, but once again that machine was suboptimal as it was badly swapping all along the way.

All in all, this is a hard question to answer unless there are some baselines available (like no/minimal swapping, using "modern" hardware available at the time and not already outdated, etc.) I do have to say that SSDs are probably the largest bump that happened to computer speeds in recent times as CPU CPS and IPC have been slow to increase, but compile times should not be disk limited -- unless it was swapping.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[John R. Graham]

I have to eat a little crow here.