Memory consumption on sieving implementations

README.md

Memory consumption on sieving implementations

Metadata of this repository is here.

Future Plan

This repository will move to GitHub from Gist. Be careful.

fit-log_1.txt

*******************************************************************************
Sat Oct 19 10:45:32 2019


FIT:    data read from plot_data_file using (column(2)):(log(column(3)))
        format = x:z
        x range restricted to [90.0000 : 260.000]
        #datapoints = 20
        residuals are weighted equally (unit weight)

function used for fitting: fit_f1(x)
	fit_f1(x) = a_1 * x + b_1
fitted parameters initialized with current variable values

iter      chisq       delta/lim  lambda   a_1           b_1          
   0 3.1393630766e+05   0.00e+00  9.02e+01    1.150000e+00   1.000000e+00
   6 6.6493607765e-01  -8.54e-03  9.02e-05    1.384034e-01  -1.216164e+01

After 6 iterations the fit converged.
final sum of squares of residuals : 0.664936
rel. change during last iteration : -8.54453e-08

degrees of freedom    (FIT_NDF)                        : 18
rms of residuals      (FIT_STDFIT) = sqrt(WSSR/ndf)    : 0.1922
variance of residuals (reduced chisquare) = WSSR/ndf   : 0.0369409

Final set of parameters            Asymptotic Standard Error
=======================            ==========================
a_1             = 0.138403         +/- 0.005579     (4.031%)
b_1             = -12.1616         +/- 0.6185       (5.086%)

correlation matrix of the fit parameters:
                a_1    b_1    
a_1             1.000 
b_1            -0.998  1.000 

Makefile

.DEFAULT_GOAL := do
.PHONY: do
do: sieving-mem.pdf
.PHONY: clean-log
clean-log:
	$(RM) fit-log_*
sieving-mem.pdf: sieving-mem.plot sieving-mem.gpdata
	$(MAKE) clean-log
	gnuplot $<

sieving-mem.gpdata

# SVP-dim    Sieve-dim    Used-memory
129     98      4.2
129     106     12.0
131     100     5.3
131     108     16.1
133     103     8.0
133     103     12.1
135     108     16.1
135     108     16.2
136     108     16.2
136     112     28.5
137     112     14.1
139     108     16.2
141     105     10.6
143     110     21.3
145     114     37.7
147     118     67
149     117     59
151     124     160
153     123     139
155     127     246

sieving-mem.pdf

sieving-mem.plot

# vim: set ft=gnuplot :

plot_data_file = 'sieving-mem.gpdata'

set term pdfcairo size 6in,3.5in

set size ratio 0.5
set key left
set grid
set xlabel 'Dimension'
set xrange [90:260]

stats plot_data_file using 3 nooutput
yrange_max = (2 ** 30) * STATS_max
set ylabel 'Memory consumption (GiB)'
set yrange [0:yrange_max]
set logscale y

print 'Plot Max. sieving dim. -- Used memory'
stats plot_data_file using (column(2)):(log(column(3))/log(2))
set output 'sieving-mem.pdf'
plot plot_data_file using (column(2)):(column(3)) title 'Maximum sieving dimension'

print 'Plot Solved dim. -- Used memory'
stats plot_data_file using (column(1)):(log(column(3))/log(2))
set output 'sieving-mem.pdf'
replot plot_data_file using (column(1)):(column(3)) title 'SVP Challenge solved dimension'

####

print 'Fitting 1'
a_1 = 1.15
b_1 = 1
fit_f1(x) = a_1 * x + b_1
set fit logfile 'fit-log_1.txt'
fit fit_f1(x) plot_data_file using (column(2)):(log(column(3))) via a_1, b_1
set output 'sieving-mem.pdf'
replot exp(1) ** fit_f1(x) title 'Fitting of max. sieving dimension'

####
print 'Additional data'
set output 'sieving-mem.pdf'
replot (384) w lines lc 'blue' dt 2 title '384 GiB (An ABCI node)'
set output 'sieving-mem.pdf'
replot (476 * (2 ** 10)) w lines lc 'red' dt 2 title '476 TiB (All the ABCI nodes)'
set output 'sieving-mem.pdf'
replot (512) w lines lc 'green' dt 1 title '512 GiB ([ADH+] used)'