Request from Phil - Page 8

fatphil · 2006-12-13, 07:43

Quote:

Originally Posted by Ken_g6

Hm - I tried to compile the latest source here, but I seem to be missing some dependencies. Can anyone who's built it help me?

For one, I got and compiled GLucas, but compiling it didn't produce libyeafft.*.a (where I think * should be "athlon-pc-cygwin"? Or is it athlonxp or something?)

For another, I don't have /usr/local/include/Make.inc. What goes there? I assume ${HOSTTYPE}, and maybe ${PHILMATH}?

Here's my linux machine's one:

HOSTTYPE=athlon
PHILMATH=/home/phil/projects/maths
GCC_MARCH=-march=athlon-xp
GCC=gcc-4.1
NASM=nasm -f elf

You do need to rename the libyeafft.a once you've built it to have the right name for the Makefile to pick up. (This lets you have multiple versions of the library hanging around for multiple targets, but take care of the autogenerated header file, that needs backing up too.)

Ken_g6 · 2006-12-13, 15:09

Quote:

Originally Posted by fatphil

Here's my linux machine's one:

HOSTTYPE=athlon
PHILMATH=/home/phil/projects/maths
GCC_MARCH=-march=athlon-xp
GCC=gcc-4.1
NASM=nasm -f elf

You do need to rename the libyeafft.a once you've built it to have the right name for the Makefile to pick up. (This lets you have multiple versions of the library hanging around for multiple targets, but take care of the autogenerated header file, that needs backing up too.)

Quote:

bash-3.2$ gcc --version
gcc (GCC) 3.4.4 (cygming special) (gdc 0.12, using dmd 0.125)

Do I need to compile and install a newer gcc? It didn't come with Cygwin.

I didn't get *.a out of makeing Glucas-2.9.0.tar.gz; just Glucas.exe and some .o files. It doesn't even look like the Makefile contains an ar command to make a .a file.

fatphil · 2006-12-13, 16:46

Quote:

Originally Posted by Ken_g6

Do I need to compile and install a newer gcc? It didn't come with Cygwin.

I didn't get *.a out of makeing Glucas-2.9.0.tar.gz; just Glucas.exe and some .o files. It doesn't even look like the Makefile contains an ar command to make a .a file.

Any gcc >= 3.0 should be fine.

I didn't realise that you were on cygwin, you need to jump through some hoops if you want to use cygwin. Can you grab the latest SubVersion files from the YEAFFT/GLucas site (oxixares.com)? They've got my cygwin patches. If not, then I need to give you the patched versions.

I have to ask what you expect to achieve from compiling your own binary? It's hard work on windows. You'll probably not get anything different from what I've built. I'm not trying to be dismissive, it's just that I know how much hard work it is compared to linux or BSD. (the linker has incompatibilities, basically, you need to munge external symbol names.)

Ken_g6 · 2006-12-13, 17:06

I'm just looking for a binary to try out, and I didn't see one in this thread!

Are you distributing your build?

Edit: I just found it! ".cygp3" wasn't an extension I recognized as executable.

fatphil · 2006-12-13, 19:53

Quote:

Originally Posted by Ken_g6

I'm just looking for a binary to try out, and I didn't see one in this thread!

Are you distributing your build?

There's no part of the system that's "secret", but there may be quirks that make self-building a bit of an uphill struggle. Cygwin's the quirkiest. The next full release of YEAFFT should make it easier.

Quote:

Edit: I just found it! ".cygp3" wasn't an extension I recognized as executable.

'extensions' are a feature of filesystems I refuse to recognise! ;-p

Ken_g6 · 2006-12-14, 05:53

Well, here's a benchmark on my Athlon XP 2200+ (1.8GHz). This may be the wrong forum for it, but I don't see a thread for phrot anywhere else. So:

Actually testing 635481*1048576^20480+1 (20482/49152 limbs)
635481*2^409600+1 [-244748,-353119,372855,204230] is composite LLR64=F4D8A6DFADB
1882C. (e=0.10944 (0.143278~6.95512e-16@1.053) t=2326.28s)

By comparison, using the Seventeen or Bust client which uses George's code, in the past I've tested similar numbers in 1122 seconds:
601467*2^409600+1 is not prime. Res64: C89279AB140D4564

Phil, does phrot not use the Discrete Weighted Transform that George uses?

2006-12-14, 07:26

Quote:

Originally Posted by Ken_g6

Phil, does phrot not use the Discrete Weighted Transform that George uses?

No. This is a programm especially for non base 2 prime testing. As Phil wrote earlier in this thread about llr being faster on base 2:

Quote:

Originally Posted by Phil

That's basically LLR/PRP/PFGW's sweet spot - absolutely no point in running Phrot for those forms on an x86 machine. That and GW's SSE2 convoluions work, and YEAFFT's don't, so I have to stick with 8 plain old FPU registers.

fatphil · 2006-12-14, 07:32

Quote:

Originally Posted by Ken_g6

Well, here's a benchmark on my Athlon XP 2200+ (1.8GHz). This may be the wrong forum for it, but I don't see a thread for phrot anywhere else. So:

Actually testing 635481*1048576^20480+1 (20482/49152 limbs)
635481*2^409600+1 [-244748,-353119,372855,204230] is composite LLR64=F4D8A6DFADB
1882C. (e=0.10944 (0.143278~6.95512e-16@1.053) t=2326.28s)

By comparison, using the Seventeen or Bust client which uses George's code, in the past I've tested similar numbers in 1122 seconds:
601467*2^409600+1 is not prime. Res64: C89279AB140D4564

Phil, does phrot not use the Discrete Weighted Transform that George uses?

Nope it does not. And I see no need to split information about phrot into a second thread as you obviously haven't been bothered to read more than a tiny fraction of this thread. Quite why are you posting non-riesel/sierpinski base 5 examples here - in what way do you think they're relevant? Did you not consider perhaps that there's some reason I'm only mentioning the program in this particular sub-forum (and to GC/W testers by mail)?

Do not use phrot on x86 machines for numbers of the form k*2^n+/-1.
Or for Mersennes.
Or for GFNs.
Or for GEFs.
Or for numbers of arbitrary form.

Ken_g6 · 2006-12-14, 15:18

Quote:

Originally Posted by fatphil

Quite why are you posting non-riesel/sierpinski base 5 examples here - in what way do you think they're relevant?

There's no need to be short with me.

The reason is, the leader of the Seventeen or Bust project was curious about using your code for non-x86 machines.

fatphil · 2006-12-15, 05:45

Quote:

Originally Posted by Ken_g6

There's no need to be short with me.

The reason is, the leader of the Seventeen or Bust project was curious about using your code for non-x86 machines.

Does Louie not have the ability to benchmark it himself? Or even to read this thread? Or is he incapable of contacting me directly? Or even simply unwilling to?

If the answer to even one of those is yes, then it really doesn't reflect very well on him.

However, this is off-topic for this S/R5 subforum.

2006-12-13, 17:06	#81
Ken_g6 Jan 2005 Caught in a sieve 613₈ Posts	I'm just looking for a binary to try out, and I didn't see one in this thread! Are you distributing your build? Edit: I just found it! ".cygp3" wasn't an extension I recognized as executable. Last fiddled with by Ken_g6 on 2006-12-13 at 17:13

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Request	pinhodecarlos	Lounge	3	2017-10-26 18:58
Bug/request	Dubslow	YAFU	4	2012-03-31 03:07
Odd request?	Xyzzy	Lounge	23	2011-03-08 17:50
GMP-ECM Request	rogue	GMP-ECM	4	2009-11-23 15:07
A little request	10metreh	Aliquot Sequences	8	2009-10-07 18:13

2006-12-14, 05:53	#83
Ken_g6 Jan 2005 Caught in a sieve 5×79 Posts	Well, here's a benchmark on my Athlon XP 2200+ (1.8GHz). This may be the wrong forum for it, but I don't see a thread for phrot anywhere else. So: Actually testing 6354811048576^20480+1 (20482/49152 limbs) 6354812^409600+1 [-244748,-353119,372855,204230] is composite LLR64=F4D8A6DFADB 1882C. (e=0.10944 (0.143278~6.95512e-16@1.053) t=2326.28s) By comparison, using the Seventeen or Bust client which uses George's code, in the past I've tested similar numbers in 1122 seconds: 601467*2^409600+1 is not prime. Res64: C89279AB140D4564 Phil, does phrot not use the Discrete Weighted Transform that George uses?