View Single Post
Old 2019-12-02, 23:04   #415
bchaffin
 
Sep 2010
Portland, OR

7·53 Posts
Default

Quote:
Originally Posted by rogue View Post
You can always find the latest pfgw executable at https://sourceforge.net/projects/openpfgw/. The thread title is not correct.

Can you post the output from the various -a options when using special modular reduction? That is probably specific to your CPU as I do not see any roundoff errors with that candidate using special modular reduction on my laptop.
Sure, see below. I'm running Ubuntu 18.04 on an i9-9980XE, which is a Skylake core. The pfgw64 binary is from Sourceforge, I didn't build it myself.

Here's the special reduction output (this was run with -r, but it makes no difference):

Code:
pfgw64 -r -V '-q(10^210731-1)/9'
PFGW Version 4.0.0.64BIT.20190528.x86_Dev [GWNUM 29.8]

Special modular reduction using AVX-512 FFT length 40K, Pass1=128, Pass2=320, clm=1 on 10^210731-1                 
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 700001/700030 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a1
Special modular reduction using AVX-512 FFT length 48K, Pass1=128, Pass2=384, clm=1 on 10^210731-1                 
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 700001/700030 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a2
Special modular reduction using AVX-512 FFT length 56K, Pass1=128, Pass2=448, clm=1 on 10^210731-1                 
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 700001/700030 ERROR: ROUND OFF 0.5>0.45
PFGW will automatically rerun the test with -a3
Special modular reduction using AVX-512 FFT length 60K, Pass1=192, Pass2=320, clm=1 on 10^210731-1                 
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 700001/700030 ERROR: ROUND OFF 0.49999>0.45
PFGW will automatically rerun the test with -a4
Special modular reduction using AVX-512 FFT length 64K, Pass1=128, Pass2=512, clm=1 on 10^210731-1                 
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 700001/700030 ERROR: ROUND OFF 0.49999>0.45
PFGW will automatically rerun the test with -a5
Special modular reduction using AVX-512 FFT length 72K, Pass1=192, Pass2=384, clm=1 on 10^210731-1                 
Detected in MAXERR>0.45 (round off check) in prp_using_gwnum
Iteration: 700001/700030 ERROR: ROUND OFF 0.49999>0.45
PFGW will automatically rerun the test with -a6
(10^210731-1)/9 ERROR DURING PROCESSING! (878.9319s+0.0013s)
And here's the output when running the generic routine:

Code:
pfgw64 -r -V '-qR(210731)'
PFGW Version 4.0.0.64BIT.20190528.x86_Dev [GWNUM 29.8]

Generic modular reduction using generic reduction AVX-512 FFT length 72K, Pass1=192, Pass2=384, clm=1 on A 700034-bit number
R(210731) is composite: RES64: [780BD8CCB26C7E56] (661.1415s+0.0012s)
Thanks!
bchaffin is offline   Reply With Quote