Compiling Phrot

[mdettweiler]

Originally Posted by rogue

I have attached the latest. Don't forget to apply the patch described here.
<snip>

Hmm...I can't quite figure out where to insert that code into the yeainit.c file. (I have absolutely no knowledge of C or any of its derivatives.) Can you possibly attach at patched version of yeainit.c?

[rogue]

Originally Posted by Anonymous

Hmm...I can't quite figure out where to insert that code into the yeainit.c file. (I have absolutely no knowledge of C or any of its derivatives.) Can you possibly attach at patched version of yeainit.c?

Modify this:

y_size_t y_init(y_size_t n)
{
y_size_t i = (1L << (sizeof(y_size_t) * 8 - 2)), j,
k = sizeof(y_size_t) * 8 - 1, old = Y_LENGTH;

to look like this:

y_size_t y_init(y_size_t n)
{
n+=(n&1); /* round up to even, as we must be even sized */

y_size_t i = (1L << (sizeof(y_size_t) * 8 - 2)), j,
k = sizeof(y_size_t) * 8 - 1, old = Y_LENGTH;

[mdettweiler]

Originally Posted by rogue

Modify this:

y_size_t y_init(y_size_t n)
{
y_size_t i = (1L << (sizeof(y_size_t) * 8 - 2)), j,
k = sizeof(y_size_t) * 8 - 1, old = Y_LENGTH;

to look like this:

y_size_t y_init(y_size_t n)
{
n+=(n&1); /* round up to even, as we must be even sized */

y_size_t i = (1L << (sizeof(y_size_t) * 8 - 2)), j,
k = sizeof(y_size_t) * 8 - 1, old = Y_LENGTH;

Okay, I was able to modify the source as per your instructions, and build Glucas, but I can't quite figure out how to build phrot itself. I tried executing the command "make" from within the glprov directory (I put the glprov and glucas directories within the same directory), but I get the following error messages:

Makefile:1: ../Make.inc: No such file or directory
make: *** No rule to make target `../Make.inc'.  Stop.

Where did I go wrong?

[rogue]

Originally Posted by Anonymous

Okay, I was able to modify the source as per your instructions, and build Glucas, but I can't quite figure out how to build phrot itself. I tried executing the command "make" from within the glprov directory (I put the glprov and glucas directories within the same directory), but I get the following error messages:

Makefile:1: ../Make.inc: No such file or directory
make: *** No rule to make target `../Make.inc'.  Stop.

Where did I go wrong?

Make.inc is just a file that is in the parent directory of the folder that has the glprov source. In mine I have:

CC_OPT=-O3 -mtune-nocona -msse2 -msse2 -m64
HOSTTYPE=c2d
PHILMATH=/Distributed/phil

Change PHILMATH to be the folder that contains the glprov folder.

I have attached a zip file that you will need to build an additional library needed by the phrot makefile. Put it at the same level as the glprov folder. You will need to use "make" in the libs folder, then you should be able to make phrot.

I assume you changed YEALIB in the phrot Makefile to point to your glucas installation. You should also change this line:

YEACFLAGS:=${CC_OPT32} -fnested-functions

to this:

YEACFLAGS:=${CC_OPT}

-fnested-functions should have no affect on the build. You might want to change this line:

OPT:=-O2 -ffast-math

to

OPT:=-O3

and compare the results.

[mdettweiler]

Originally Posted by rogue

Make.inc is just a file that is in the parent directory of the folder that has the glprov source. In mine I have:

CC_OPT=-O3 -mtune-nocona -msse2 -msse2 -m64
HOSTTYPE=c2d
PHILMATH=/Distributed/phil

Change PHILMATH to be the folder that contains the glprov folder.

I have attached a zip file that you will need to build an additional library needed by the phrot makefile. Put it at the same level as the glprov folder. You will need to use "make" in the libs folder, then you should be able to make phrot.

When I run "make" from the libs folder, I get the following error:

max@max:~/Documents/prime-searching/phrot/libs$ make
gcc -O3 -mtune-nocona -msse2 -msse2 -m64 -O3 -o math52.c2d.o -c math52.c
cc1: error: unrecognized command line option "-mtune-nocona"
make: *** [math52.c2d.o] Error 1

Am I using a wrong version of gcc or something? (As far as I know, I have the latest version--I'm running Ubuntu 8.04 beta with everything up to date.)

I assume you changed YEALIB in the phrot Makefile to point to your glucas installation. You should also change this line:

YEACFLAGS:=${CC_OPT32} -fnested-functions

to this:

YEACFLAGS:=${CC_OPT}

-fnested-functions should have no affect on the build.

Actually, in the makefile I found in the phrot folder, I don't see any "-fnewsted-functions" thing at all--it instead said "YEACFLAGS:=${CC_OPT32_DEBUG}". So, I just left it as it was.

You might want to change this line:

OPT:=-O2 -ffast-math

to

OPT:=-O3

and compare the results.

I didn't see that in the glprov makefile either. Am I looking in the wrong place?

I must confess that I know essentially nil about C/C++, and am not very experienced in compiling things from source--I've compiled software from source before, but have never had to mess with editing things in makefiles and whatnot, so this is all new to me.

[rogue]

I guess I zipped up a different Makefile than I thought. Use CC_OPT instead of CC_OPT32_DEBUG.

It should be -mtune=nocona, not -mtune-nocona. Sorry about that.

[rogue]

Originally Posted by em99010pepe

A few results from my AMD 64 3000+:

So for your P4 you will get half the time.

Based upon the timings I see, I highly recommend that you run phrot instead of LLR. Read this link for more info.

[rogue]

One last note for those of you using phrot. You need to specify the -b3 and -o options when you run it. -b3 will write output residues that are compatible with LLR. -o will write those residues to a file in addition to what is written to the terminal. This file can then be sent when you complete your range.

Finally, if enough people start to use phrot, someone should consider supporting phrot with LLRNet.

[em99010pepe]

Originally Posted by rogue

Based upon the timings I see, I highly recommend that you run phrot instead of LLR. Read this link for more info.

I am running LLRnet....is there a compiled version?

[Siemelink]

Originally Posted by rogue

One last note for those of you using phrot. You need to specify the -b3 and -o options when you run it. -b3 will write output residues that are compatible with LLR. -o will write those residues to a file in addition to what is written to the terminal. This file can then be sent when you complete your range.

Finally, if enough people start to use phrot, someone should consider supporting phrot with LLRNet.

Interesting stuff! Given that I do millions of tests I want to try as well. But I can not compile at the moment. Does someone have a windows compile available?

Willem.

[tnerual]

Originally Posted by Siemelink

Interesting stuff! Given that I do millions of tests I want to try as well. But I can not compile at the moment. Does someone have a windows compile available?

Willem.

here is how to do the stuff under windows ...

1. download phrot here:
http://www.mersenneforum.org/attachm...8&d=1165306798

2. unzip files
3. change name of "phrot.0.5.cygp3" to "phrot.exe"
4. download cygwin1.dll here:
http://www.dll-files.com/dllindex/dl....shtml?cygwin1
5. uncompress cygwin1.dll in same directory as phrot
6. create a "totest.txt" file containing:

ABC $a*$b^$c+1
1 2 3
4 5 6
7 8 9

7. run phrot in a "dos windows":

phrot -b3 -o totest.txt

that's all
this will test 1*2^3+1 then 4*5^6+1 and 7*8^9+1 (edit totest.txt to match your needs)

can a moderator move this technical discussion out of this thread and create a new one just for phrot ?