[Cbe-oss-dev] installation-report: debian 7.1 wheezy installation success on PS3

[Declan Malone]

On 20 August 2013 21:23, Geoff Levand <geoff at infradead.org> wrote:

> On Tue, 2013-08-20 at 20:30 +0100, Declan Malone wrote:
> > $ ./hello_ppe.elf
> > Hello world!
>
> OK, thanks for testing.
>
> Please let us know the output from the command 'uname -a'.
>
> Also, could you try this one?
>
>
> https://www.kernel.org/pub/linux/kernel/people/geoff/cell/ps3-linux-docs/CellProgrammingTutorial/src/example3_2
>
> -Geoff
>
>
>
Here it is:

$ uname -a
Linux cluster 3.5.0 #1 SMP Mon Jul 23 23:15:07 CEST 2012 ppc64 GNU/Linux
$ make
gcc -Wall  -lspe2 abs_ppe.c -o abs_ppe.elf
/usr/bin/ld: skipping incompatible
/usr/lib/gcc/powerpc-linux-gnu/4.6/../../../../lib/libspe2.so when
searching for -lspe2
/usr/bin/ld: skipping incompatible
/usr/lib/gcc/powerpc-linux-gnu/4.6/../../../../lib/libspe2.a when searching
for -lspe2
/usr/bin/ld: skipping incompatible /usr/lib/../lib/libspe2.so when
searching for -lspe2
/usr/bin/ld: skipping incompatible /usr/lib/../lib/libspe2.a when searching
for -lspe2
/usr/bin/ld: skipping incompatible
/usr/lib/gcc/powerpc-linux-gnu/4.6/../../../libspe2.so when searching for
-lspe2
/usr/bin/ld: skipping incompatible
/usr/lib/gcc/powerpc-linux-gnu/4.6/../../../libspe2.a when searching for
-lspe2
spu-gcc -Wall   abs_spe.c -o abs_spe.elf
$ ./abs_ppe.elf
out[00]=1
out[01]=2
out[02]=3
out[03]=4
out[04]=5
out[05]=6
out[06]=7
out[07]=8
out[08]=9
out[09]=10
out[10]=11
out[11]=12
out[12]=13
out[13]=14
out[14]=15
out[15]=16
out[16]=17
out[17]=18
out[18]=19
out[19]=20
out[20]=21
out[21]=22
out[22]=23
out[23]=24
out[24]=25
out[25]=26
out[26]=27
out[27]=28
out[28]=29
out[29]=30
out[30]=31
out[31]=32
out[32]=33
out[33]=34
out[34]=35
out[35]=36
out[36]=37
out[37]=38
out[38]=39
out[39]=40
out[40]=41
out[41]=42
out[42]=43
out[43]=44
out[44]=45
out[45]=46
out[46]=47
out[47]=48
out[48]=49
out[49]=50
out[50]=51
out[51]=52
out[52]=53
out[53]=54
out[54]=55
out[55]=56
out[56]=57
out[57]=58
out[58]=59
out[59]=60
out[60]=61
out[61]=62
out[62]=63
out[63]=64
$ ./abs_ppe.elf | perl -nle '$i=$_; /out\[(\d+)\]=(\d+)/; print $i unless
$1 +1 == $2'
(no output... all lines in form out[x]=x+1)
$ ./abs_ppe.elf | perl -nle '$i=$_; /out\[(\d+)\]=(\d+)/; print $i unless
$2 == $.'
(no output,...no gaps in above sequence)

I also tried a much more complex test program that I wrote a while back,
and it worked fine. It uses 6 SPEs to calculate a version of Reed-Solomon
codes. It uses double-buffered DMA and mailboxes so I figured it might be a
useful test:

$ svn checkout svn://svn.code.sf.net/p/gnetraid/code/ gnetraid-code
$ cd gnetraid-code/PS3-IDA/07-shebang
$ make
$ dd if=/dev/urandom of=400Mb bs=1M count=400
$ ./rabin-ida-ps3.pl split -n 5 -k 3 400Mb # splits file into 5 parts using
Rabin's IDA
$ rm 400Mb-0.sf 400Mb-4.sf # delete some "shares"
$ ./rabin-ida-ps3.pl combine -o output 400Mb-*.sf # reconstruct file using
remaining 3 shares
$ cmp 400Mb output && echo files match
files match

That's maybe of no use, but I thought I'd test it anyway, fwiw.
Thanks,
dec
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/cbe-oss-dev/attachments/20130820/d5027ff1/attachment.html>