Just hooked up my first Raspberry Pi--the $35 computer.

http://www.raspberrypi.org/

Used 4GB sd card with Debian 6 "Wheezy" from elementI4.

apt-get update -- to bring things up to date

apt-get open-cobol -- got version 1.1

FTP'd some source code over and compiled.

Worked without a single hitch!

yes OC runs really nicely ont he raspberry pi. i've been using it on one for the last few months (got my raspi in april) with no major problems.

CE version fails one test LINE SEQUENTIAL write but from what i've read (iirc) the correct output for that is still up for debate, 1.1 builds and runs well whith the tests taking 16minutes +- a few depending on gcc flags and options used (this is on raspbian armhf distro).

I ran the telco benchmark yesterday and the million number file takes 105seconds +-1sec

all in all it's pretty good.

[note: elastic cobol is also available for the raspi now as well, it has some nice bits such as GUI but it uses JVM which with my irrational dislike of java is a minus for me]

As I guess you don't need to run the binary files generated by the raspi on other systems I suggest to run the compiler/runtime and the programs optimized (only works on GCC 4.2+ [run gcc --version to check]).
Build OpenCOBOL again

and to enable this for compilation with CE (for enabling this with 1.1 you need to lookup COB_CFLAGS in defaults.h, export this together with -march=native)

Please tell us the new benchmark results, just for the record.

Simon

the defaults settings for gcc in the raspbian os for the raspberry pi are already set to be -march=armv6 -mfpu=vfp -mfloat-abi=hard so set at their optimum settings already & -O3 doesn't make a difference really at all, maybe a second or two but well within my margin of error for timing so might even already be set as the default setting for gcc too. i could start adding other optimisations but then you start getting to the 1 optimisation helps one program but slows down another so should only be used sparingly and tweaks individually.

i could use -mcpu=arm1176jzf-s but so far on other things it's not made a measurable difference to speed of execution over just using -march=armv6

Thank you for this information. I guess it's done by system default for CFLAGS?

Simon