@Andy, I'm unable to PM for some reason :(

[Jason G]

Hi Andy:  Contiuing our PM discussion,
             I meant a Linux port ..... Have you access to the pre-release area ? Last post in that section seems to be middle of last year, It'd be good to get some action going.

Jason.

[Jason G]

Starting next week I should have a lot of things out of the way, and two weeks holiday.

[Jason G]

Are you there "ahaveland" ?  I cant' see mto PM you still  .  Oh well, Here's what I tried to write and I hope you find this thread:

Hi Again,
  My PM seemed to stop functioning for a while, not sure why.  Anyway I wanted to find out if you had access to the pre-release tester's area where we've been working. The Linux section hasn't been posted to in there since mid 2007 so it would be really nice to get something happening in there. I'm very rusty with Linux (Actively used Slackware and Debian distributions a little over 10 years ago) but do have Ubuntu installed in a VM already. 

Reorientation and some guidance setting up developer tools might be in order for me first. My holidays start next week so I'll have more time then.

I would like to gradually migrate my current, non-operational, systems over to FreeBSD, so assisting with the Linux port might be a good intermediate step for me.

Jason

[Hmm, the issues with PM's seem to only apply when I'm trying to PM yourself ... oh well, maybe Gecko_R7 will know what's up]

[ahaveland]

Hi Jason, I'm here... just noticed!

Well, I'm pretty intimately familiar with system aspects of linux, but not so much with compiling, but I have spent about 48 hours non-stop trying to work out how it all works.

All these macros and preprocessors and n levels of abstraction, it's a wonder that it anyone could get it to compile. I still haven't succeeded in compiling v1.3, or Eric's code, and forget Alex's v8 for now - I don't have Accelerate.h or cblas.h to help.

No, I don't have access to the latest sources - it may be better to adapt alex's code to those, or forget about portability and cross platform considerations and go for a system specific rewrite. At least it would speed up development without having to take into account every other combination of platforms and options.

The code I have seen has grown into an asteroid sized rat's nest!

After hundreds of attempts, I finally got latest ICC/IPP working and compiling, and 1.3 version almost there - everything compiles now, but I have a linking problem that I have tried to solve for the last day without any success, and no help whatsoever with google. Seems that no one has done this before.

There have been some changes in ipp since v9.  I am sure I have the right libraries in ldflags, but they can't be found.

Here is the current sticking point, which I hope someone could shed some light on!

icpc -o seti_boinc -pthread -L /home/haveland/andy/intel/cce/10.1.015/bin   -L /home/haveland/andy/intel/cce/10.1.015/lib -L /home/haveland/andy/intel/ipp/5.3.2.068/em64t/lib   -limf   -lsvml     -lippsemergedem64t   -lippsmergedem64t   -lippcoreem64t   -i-static  -L/usr/lib -nodefaultlibs -L/home/haveland/andy/intel/ipp/5.3.2.068/em64t/lib   seti_boinc-main.o seti_boinc-analyzeFuncs.o seti_boinc-analyzeReport.o seti_boinc-analyzePoT.o seti_boinc-pulsefind.o seti_boinc-gaussfit.o seti_boinc-lcgamm.o seti_boinc-malloc_a.o seti_boinc-seti.o seti_boinc-seti_header.o seti_boinc-timecvt.o seti_boinc-s_util.o seti_boinc-version.o seti_boinc-worker.o seti_boinc-chirpfft.o seti_boinc-spike.o seti_boinc-progress.o seti_boinc-fft8g.o seti_boinc-gdata.o seti_boinc-schema_master.o seti_boinc-sqlrow.o seti_boinc-sqlblob.o seti_boinc-xml_util.o   -L/home/haveland/andy/seti_dev/boinc/api -lboinc_api -L/home/haveland/andy/seti_dev/boinc/lib -lboinc    -lssl -lcrypto /usr/lib                                           /libcrypto.a /usr/lib/libssl.a -Wl,-Bdynamic -ldl -Wl,-Bdynamic -lm -Wl,-Bstatic -lnsl -Wl,-Bdynamic -lrt -Wl,-Bstatic -lz -Wl,-Bstatic -lstdc++ -Wl,-Bstatic -lgcc_eh -Wl,-Bdynamic -lpthread -Wl,-Bdynamic -lc
ipo: remark #11000: performing multi-file optimizations
ipo: remark #11005: generating object file /tmp/ipo_icpc5Mniuk.o
lcgamm.cpp(115): (col. 5) remark: LOOP WAS VECTORIZED.
lcgamm.cpp(131): (col. 7) remark: LOOP WAS VECTORIZED.
lcgamm.cpp(93): (col. 11) remark: LOOP WAS VECTORIZED.
lcgamm.cpp(93): (col. 26) remark: LOOP WAS VECTORIZED.
analyzeFuncs.cpp(762): (col. 5) remark: LOOP WAS VECTORIZED.
analyzeFuncs.cpp(776): (col. 5) remark: LOOP WAS VECTORIZED.
analyzeFuncs.cpp(828): (col. 5) remark: BLOCK WAS VECTORIZED.
ld: skipping incompatible /usr/lib/libssl.so when searching for -lssl
ld: skipping incompatible /usr/lib/libssl.a when searching for -lssl
ld: skipping incompatible /usr/lib/libcrypto.so when searching for -lcrypto
ld: skipping incompatible /usr/lib/libcrypto.a when searching for -lcrypto
/tmp/ipo_icpc5Mniuk.o: In function `seti_analyze(ANALYSIS_STATE&)':
/tmp/ipo_icpc5Mniuk.c:(.text+0x16c1f): undefined reference to `ippStaticInit'
/tmp/ipo_icpc5Mniuk.c:(.text+0x16fdb): undefined reference to `ippGetCpuFreqMhz'
/tmp/ipo_icpc5Mniuk.c:(.text+0x17b90): undefined reference to `ippsFFTInv_CToC_32fc'
/tmp/ipo_icpc5Mniuk.c:(.text+0x181e8): undefined reference to `ippsFFTFree_C_32fc'
/tmp/ipo_icpc5Mniuk.c:(.text+0x18df4): undefined reference to `ippsFFTInitAlloc_C_32fc'
/tmp/ipo_icpc5Mniuk.o: In function `v_BaseLineSmooth(float (*) [2], int, int, int)':
/tmp/ipo_icpc5Mniuk.c:(.text+0x27c2c): undefined reference to `ippsFFTFree_C_32fc'
/tmp/ipo_icpc5Mniuk.c:(.text+0x27ce1): undefined reference to `ippsFFTInitAlloc_C_32fc'
/tmp/ipo_icpc5Mniuk.c:(.text+0x27e40): undefined reference to `ippsFFTInv_CToC_32fc'
/tmp/ipo_icpc5Mniuk.c:(.text+0x28152): undefined reference to `ippsFFTFwd_CToC_32fc'
make[2]: *** [seti_boinc] Error 1
make[2]: Leaving directory `/home/haveland/andy/seti_dev/seti_boinc/client'
make[1]: *** [all-recursive] Error 1

ippStaticInit and ippGetCpuFreqMhz is in lib ippcoreem64t, but it doens't want to play.

ippsFFTInv_CToC_32fc and ippsFFT* are in ippsmergedem64t
They are specified but still can't be found, and it's driving me crazy!

Once I get something that works, then I stand a better change of being incrementally trying out stuff and replacing functions and understanding wtf is going on!

Perhaps the latest 2.4 source would help.

Cheers,
Andy.

[Jason G]

Well the way I got started (Way back with R2.0 With the Windows Builds) was to follow the How Tos , to the left of the forum,  The Linux one there would probably make sure your tools/ libs & everything is set up properly, and familiarise you with the structure a bit better. I think I'll start there myself to get me reoriented with Linux on the Holidays.

  The IPP function calls 'look like' a symptom of those tools, paths & directives for IPP not set right, or something missing, , might be worth checking how these need to be set up.

There is a wealth of stuff related to AK build Windows port in a thread in the development area.  I think you'll need access to that info, Let's see if either of us can get hold of Gecko_R7 at some point. 

I'm not sure whether it would be easiest to start from our windiows port codebase, or directly from Alex's with a couple of minor alterations.  I will give that some thought over the next few days. It would be nice if they could be integrated    and be multi-platform, as the facility is there...

Jason

[ahaveland]

Well, I just got it to compile by switching to dynamic instead of static, and the benchmark completed in half the time, but that is by 1.3 standards...
After installing the libraries, it seems to be running linux ssse3 now.

Obviously static would be better, and Alex's even better as it wouldn't need to use restricted IPP libraries... Meanwhile I haven't a clue why it wouldn't link statically.

Now while it's running, to see how it's doing, I'll start learning the basics of sse primitives and porting...

Porting could evolve from both ends and end up with convergence, and then integrated into the main codebase and tidied up a lot... though this is extremely time consuming.

[Jason G]

Well, I just got it to compile by switching to dynamic instead of static, and the benchmark completed in half the time, but that is by 1.3 standards...
After installing the libraries, it seems to be running linux ssse3 now.

Obviously static would be better, and Alex's even better as it wouldn't need to use restricted IPP libraries... Meanwhile I haven't a clue why it wouldn't link statically.

Now while it's running, to see how it's doing, I'll start learning the basics of sse primitives and porting...

Porting could evolve from both ends and end up with convergence, and then integrated into the main codebase and tidied up a lot... though this is extremely time consuming.

Well have fun doing it!  Yeah, on win32, using the static ipp libs requires some include files differences and linking to the staticlib folder content libraries instead.  Dunno how that would function on Linux.

Great that you got a build running so quickly.  If you choose to start on tha AK code I'll need to get you some information.. I'll see if Gecko_R7 can help out here.[PM'd him]