On 2017-09-06 21:54, R0b0t1 wrote:
On Wed, Sep 6, 2017 at 3:05 PM, Brian Inglis wrote:
On 2017-09-06 09:24, Alan Ludwig wrote:
I'm trying to get an estimate of the order of magnitude of the total number
of hours in service for newlib. Is it thousands of hours? Millions of hours?
Tens of millions of hours? For which versions?
Usually, I'd work this backwards from estimates of the number of downloads,
installed seats, size of the user base, etc. However, newlib doesn't seem to
have any of this sort of information available at the download or support
websites.
Do you have any information that I might use to estimate the hours in
service for newlib? The numbers can be very rough. I'm only trying to choose
an order of magnitude. Is it 10^4 or say 10^7 total hours? Any information
that I might use to make these gross estimates would be appreciated.
Even estimating downloads you'd need IP address logs from sourceware for git
source, for those who rebuild everything from scratch, and from all Cygwin
mirrors, for those who install packages.
Cygwin usage would be difficult to estimate as it could range from tens to
thousands of hours per year on server and desktop platforms.
From earlier replies, embedded platforms would probably swamp those uses.
I was going to point out that newlib ships with the arm-none-eabi
toolchain distributed by ARM Holdings and leave it at that, but Joel
Sherril's answer is very insightful.
That's ~1500 licensees buying 15G chips/year - aggregate ~100G chips to date,
and the earlier replies from Joel, Dave McGuire, Andrew Pinski, and Jon Beniston
make it clear that it must be used in thousands of products with volumes from
the teens (for the cool space deployments) to many millions, running for many
years in widely deployed commercial products including automotive, network, and
phones, so Jon's estimate of (perhaps many) trillions of hours seems closest.
Between the automotive and space products they will also have travelled many Gm.