From f52703a2c13a4e69525be6ce9883eead2f96eb6f Mon Sep 17 00:00:00 2001 From: afanfakh Date: Fri, 16 May 2014 11:28:29 +0200 Subject: [PATCH 1/1] New version. --- fig/bt.eps | 144 +++++++++------- fig/bt.pdf | Bin 7604 -> 7587 bytes fig/cg.eps | 226 ++++++++++++++++++------- fig/cg.pdf | Bin 7537 -> 7967 bytes fig/cg_per.eps | 134 +++++++++------ fig/cg_per.pdf | Bin 7145 -> 7135 bytes fig/compare_class_C.pdf | Bin 27237 -> 26672 bytes fig/file.eps | 154 ++++++++++------- fig/file.pdf | Bin 7288 -> 7276 bytes fig/file3.eps | 132 ++++++++------- fig/file3.pdf | Bin 6520 -> 6512 bytes fig/lu_pre.eps | 136 +++++++++------ fig/lu_pre.pdf | Bin 7106 -> 7092 bytes paper.tex | 361 +++++++++++----------------------------- 14 files changed, 684 insertions(+), 603 deletions(-) diff --git a/fig/bt.eps b/fig/bt.eps index d30c63c..2f44b28 100644 --- a/fig/bt.eps +++ b/fig/bt.eps @@ -1,7 +1,7 @@ %!PS-Adobe-2.0 EPSF-2.0 %%Title: bt.eps %%Creator: gnuplot 4.6 patchlevel 0 -%%CreationDate: Mon Mar 24 19:20:05 2014 +%%CreationDate: Thu May 15 12:08:40 2014 %%DocumentFonts: (atend) %%BoundingBox: 50 50 320 239 %%EndComments @@ -432,7 +432,7 @@ SDict begin [ /Author (afanfakh) % /Producer (gnuplot) % /Keywords () - /CreationDate (Mon Mar 24 19:20:05 2014) + /CreationDate (Thu May 15 12:08:40 2014) /DOCINFO pdfmark end } ifelse @@ -454,128 +454,150 @@ LTb 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont 602 448 M ( 0.3) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 799 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont 602 799 M ( 0.4) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1151 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.5) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1502 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.6) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1854 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.7) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2205 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.8) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2557 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.9) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2908 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3260 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 1.1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3611 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 1.2) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 686 308 M ( 1) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -1578 448 M +1700 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 1.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -2470 448 M +2714 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 2) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -3363 448 M +3728 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 2.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -4255 448 M +4741 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 3) Cshow -1.000 UL -LTb -5147 448 M -0 63 V -0 3100 R -0 -63 V -0 -3240 R -( 3.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 1.000 UL @@ -587,30 +609,36 @@ LTb -4461 0 V Z stroke LCb setrgbcolor +/Helvetica findfont 140 scalefont setfont 112 2029 M currentpoint gsave translate -270 rotate 0 0 M (Normalized energy and performance inv.) Cshow grestore +/Helvetica findfont 140 scalefont setfont LTb LCb setrgbcolor +/Helvetica findfont 140 scalefont setfont 2916 98 M (Frequency scaling factors) Cshow +/Helvetica findfont 140 scalefont setfont LTb 1.000 UP -1328 659 M +/Helvetica findfont 140 scalefont setfont +1416 659 M (Optimal scaling factor=1.315) Lshow -864 3470 M +/Helvetica findfont 140 scalefont setfont +889 3470 M (BT Class C) Lshow 1.000 UL LT1 -0.00 0.00 0.00 C 1216 587 M +0.00 0.00 0.00 C 1293 587 M 32 -121 V 32 121 V -32 1513 R 0 -1634 V stroke gsave [] 0 setdash -1216 587 M +1293 587 M 32 -121 V 32 121 V stroke @@ -621,57 +649,61 @@ LTb 1.000 UL LT0 0.00 0.00 1.00 C LCb setrgbcolor -4496 3478 M +/Helvetica findfont 140 scalefont setfont +4496 3464 M (Normalized performance inv.) Rshow +/Helvetica findfont 140 scalefont setfont LT0 -0.00 0.00 1.00 C 4580 3478 M +0.00 0.00 1.00 C 4580 3464 M 399 0 V 686 2908 M -74 -157 V -81 -129 V -88 -128 V -97 -130 V -106 -131 V -117 -132 V -131 -133 V -146 -134 V -164 -135 V -186 -136 V -212 -137 V -245 -138 V -286 -139 V -338 -140 V -3363 868 L -3858 726 L -4478 582 L +84 -157 V +92 -129 V +963 2494 L +109 -130 V +121 -131 V +133 -132 V +149 -133 V +165 -134 V +187 -135 V +211 -136 V +241 -137 V +279 -138 V +325 -139 V +384 -140 V +3728 868 L +4291 726 L +4995 582 L % End plot #1 % Begin plot #2 stroke LT1 1.00 0.00 0.00 C LCb setrgbcolor -4496 3338 M +/Helvetica findfont 140 scalefont setfont +4496 3296 M (Normalized energy) Rshow +/Helvetica findfont 140 scalefont setfont LT1 -1.00 0.00 0.00 C 4580 3338 M +1.00 0.00 0.00 C 4580 3296 M 399 0 V 686 2908 M -74 -250 V -81 -188 V -88 -177 V -97 -165 V -106 -152 V -117 -140 V -131 -126 V -146 -111 V -164 -97 V -186 -80 V -212 -62 V -245 -41 V -286 -19 V -338 8 V -406 39 V -495 79 V -620 130 V +84 -250 V +92 -188 V +963 2293 L +109 -165 V +121 -152 V +133 -140 V +149 -126 V +165 -111 V +187 -97 V +211 -80 V +241 -62 V +279 -41 V +325 -19 V +384 8 V +461 39 V +563 79 V +704 130 V % End plot #2 stroke LTb diff --git a/fig/bt.pdf b/fig/bt.pdf index 80a301799d226cb7eebd82123432f003f2a30378..626ffffd2990f3b1bba200fafdde36b408fad1ac 100644 GIT binary patch delta 1619 zcmdmDz1Vt!N4<5ih^4^s;##g&0R@$<^B6frp1*j*&(;GVB-MrdMBBZ!WRo#Ts+-dM_vIw_Tia)0JnO zb+mQXs@V5)$-TTQ)@lt=hB4^^*#RZ$#U=6+OO8MF;ya;slut-Z!SJ|bM+EZ{Ysn?Q zIhI&oYdVtghoO^ug3R3m3qCdouq&UiVLb8V)`y$Pl3QM63hw=7v*>b#i#WUc^vW*1 z$=Qp;uk3R_-~HlTr2T0pWA^&P1;)od2%Rig+gfNg+bSq{%s!){%{dt{MA_M-NI7lq>$_+FkPXeT`;9; zD<5a}V!;{gJV(#@nGc+y^3|m3Qc_urtdxFct&?0k2HgHvxAP~gpXEi z?Dw8LO5>UQYu%CyoE0}XPt4%Bk;P&Y=NYiUS>0ZJ#>xkZw_ondaeJKcEM&r2v1xg0 zZ59gjH1CxAFhA>yT*}j?pO==)#^3x>{_Ac1USGwy0G*sJsSPg_8QXup{`|6F>;B*A zc+zT5ME?KwKN zZ~L!jS--!2erfX~e2;d(u0@5@-XFyr`zL zoBe-T=j+p_{9|9ejGWEqZNk8+&9lC*J+oRT+_&tP*!LH;OUoGNWtmPnwYBS7UEyz$ zERMIFB0Gd+ZQYo}XMb+I$0jIr??kIzuj$9LZ-XwgN64BP?K@uo!7^Zn{lXttFCM>o z@t^mE=kqqNdv&IJ@xz9+&B|Lp^sTjGwY#opQMZis=XRDqS(_7Iu`~X-@KEnX;>-+2 z`K+|PvaA1wObj+Md-%_yuzz0fe@)Hn-JUIe3(vHuy%tc{6Ll*U zLJUo<6b!A5tPCuytPCbMiYw!m+#}vZoE1qD?rOw1T3vr~pqMgIW^NXf^kXsux!X6f zQo+zv!4T|Z69WY!149!{E`8tp6qm%3R1Fs^kbr@yp{b#fp@E^Xwt>;)SrodnY z0fjsTE-=Hu%+lBrUChwX(i}s~z{G5Fx{O-A5vG)}ks*e16C)!GF$-f;3^8LfOm)VV z7-B|-77#IhB)6Lyo0&|OmbHmBHgmBwurzkIFf?~Gv~aR?F>`e^a&f6h1g$mp1v>VKB zu3`&$`19H1s@;L_I~2CCJ+xwuZeC>Fx%l^lCDwZ3MJ-81uM`RyCwa|k++gakYqG;G ze#c$Ox*lG~_$O>(&^*b@Bg4;8Cct!FnZbV@Z@kTvMwj#5FV0REls&F&WvyX#KDlIG zjP%~dZ?;vDF6HNbMT#Af;i*6DD{;76gp-dAPH2NmBL^&)9^m_BT{(p-ZwAZ>iOq#-#Vcm)_=d* zH9E#6m}FeDI*`U=8L(i#-F&|A-@EI}u2+@cnVVPtYW3r-|2A3QUGR9v-Iu!`mp5xD zcYnBBkykl6v(?^j_wW4OkH0Oq%THYTcZ8{XTcj!;I$jQ=*UEkKcEX z@m}-P_=C&p%T&&D%k!A!c-#JuvaQ~GG5=fm^UZa}i|QMElCSg&JcxRz_}w|Bx_;rN zwYRoAbe_>D1GnPPS~_2hQ6gjVjoum9nT3U8~kkz$C3rn=PmO;!8;Q!G~)#((G0%$JGa|(1`tNaYnLDuRQdF z;)PbHHvC3;h@4Bvu^4%rO z9%ko$2l9N$_#ALla(aU;~J{O-?mSwn#NGNCt8yw~G}K zr%+bhTEaI!PeIo=QMX7T#K^=-!O+Um%E-XV%5-vpxH4|Z)#6RWSrI7Vu11Wb)%7Pk zh$$0g=4Mt&KPEGfyL}Uj6pTz148cw|FjX)zFf`HR()Z0zaY-ym)o=m&#L&dR*vQn- z+`!1dQrp0Aa;J2^ow>0BD9{WP@)Wqh3EwFlToWT!IUyK zw!lzsVr+>aW?^cCA!cljsm{azL(Is~5+cTr>*K);&0 zS{k{SJGvOUn7A0b89F-~Svc9*5L6M%WoO4#T#{H+Qc;we#$|43Zpfvo>gw;t1pvAX Be!2hv diff --git a/fig/cg.eps b/fig/cg.eps index aa625f0..b6a1752 100644 --- a/fig/cg.eps +++ b/fig/cg.eps @@ -1,7 +1,7 @@ %!PS-Adobe-2.0 EPSF-2.0 %%Title: cg.eps %%Creator: gnuplot 4.6 patchlevel 0 -%%CreationDate: Mon Mar 24 16:08:11 2014 +%%CreationDate: Thu May 15 12:07:41 2014 %%DocumentFonts: (atend) %%BoundingBox: 50 50 320 239 %%EndComments @@ -432,7 +432,7 @@ SDict begin [ /Author (afanfakh) % /Producer (gnuplot) % /Keywords () - /CreationDate (Mon Mar 24 16:08:11 2014) + /CreationDate (Thu May 15 12:07:41 2014) /DOCINFO pdfmark end } ifelse @@ -454,128 +454,150 @@ LTb 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont 602 448 M ( 0.3) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 799 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont 602 799 M ( 0.4) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1151 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.5) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1502 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.6) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1854 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.7) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2205 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.8) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2557 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 0.9) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2908 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3260 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 1.1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3611 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 140 scalefont setfont -4482 0 R ( 1.2) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 686 308 M ( 1) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -1578 448 M +1700 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 1.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -2470 448 M +2714 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 2) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -3363 448 M +3728 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 2.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -4255 448 M +4741 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 140 scalefont setfont 0 -3240 R ( 3) Cshow -1.000 UL -LTb -5147 448 M -0 63 V -0 3100 R -0 -63 V -0 -3240 R -( 3.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 1.000 UL @@ -587,30 +609,36 @@ LTb -4461 0 V Z stroke LCb setrgbcolor +/Helvetica findfont 140 scalefont setfont 112 2029 M currentpoint gsave translate -270 rotate 0 0 M (Normalized energy and performance inv.) Cshow grestore +/Helvetica findfont 140 scalefont setfont LTb LCb setrgbcolor +/Helvetica findfont 140 scalefont setfont 2916 98 M (Frequency scaling factors) Cshow +/Helvetica findfont 140 scalefont setfont LTb 1.000 UP -1757 589 M +/Helvetica findfont 140 scalefont setfont +1903 589 M (Optimal scaling factor=1.56) Lshow -864 3470 M +/Helvetica findfont 140 scalefont setfont +889 3470 M (CG Class C) Lshow 1.000 UL LT1 -1653 587 M +0.00 0.00 0.00 C 1790 587 M 32 -121 V 32 121 V -32 1713 R 0 -1834 V stroke gsave [] 0 setdash -1653 587 M +1790 587 M 32 -121 V 32 121 V stroke @@ -621,57 +649,143 @@ LTb 1.000 UL LT0 0.00 0.00 1.00 C LCb setrgbcolor -4496 3478 M +/Helvetica findfont 140 scalefont setfont +4496 3464 M (Normalized performance inv.) Rshow +/Helvetica findfont 140 scalefont setfont LT0 -0.00 0.00 1.00 C 4580 3478 M +0.00 0.00 1.00 C 4580 3464 M 399 0 V 686 2908 M -74 -53 V -81 -56 V -88 -60 V -97 -62 V -106 -66 V -117 -71 V -131 -74 V -146 -79 V -164 -84 V -186 -90 V -212 -96 V -245 -104 V -286 -111 V -338 -119 V -406 -129 V -495 -140 V -620 -153 V +84 -53 V +92 -56 V +101 -60 V +109 -62 V +121 -66 V +133 -71 V +149 -74 V +165 -79 V +187 -84 V +211 -90 V +241 -96 V +279 -104 V +325 -111 V +384 -119 V +461 -129 V +563 -140 V +704 -153 V % End plot #1 % Begin plot #2 stroke LT1 1.00 0.00 0.00 C LCb setrgbcolor -4496 3338 M +/Helvetica findfont 140 scalefont setfont +4496 3296 M (Normalized energy) Rshow +/Helvetica findfont 140 scalefont setfont LT1 -1.00 0.00 0.00 C 4580 3338 M +1.00 0.00 0.00 C 4580 3296 M 399 0 V 686 2908 M -74 -214 V -81 -192 V -88 -181 V -97 -169 V -106 -156 V -117 -143 V -131 -130 V -146 -115 V -164 -100 V -186 -83 V -212 -65 V -245 -45 V -286 -22 V -338 4 V -406 36 V -495 75 V -620 125 V +15 -36 V +14 -36 V +15 -36 V +16 -34 V +15 -35 V +16 -34 V +16 -33 V +16 -33 V +16 -33 V +17 -32 V +17 -32 V +17 -32 V +18 -31 V +17 -31 V +18 -30 V +18 -30 V +19 -30 V +19 -29 V +19 -29 V +19 -29 V +20 -28 V +20 -28 V +21 -27 V +20 -27 V +22 -27 V +21 -27 V +22 -26 V +22 -25 V +23 -26 V +23 -25 V +23 -24 V +24 -24 V +24 -24 V +25 -24 V +25 -23 V +26 -22 V +26 -22 V +26 -22 V +27 -22 V +28 -21 V +28 -20 V +28 -21 V +30 -19 V +29 -20 V +31 -19 V +31 -18 V +31 -18 V +32 -18 V +33 -17 V +34 -17 V +34 -16 V +35 -16 V +36 -15 V +36 -15 V +37 -14 V +38 -14 V +39 -13 V +40 -13 V +41 -12 V +41 -12 V +43 -11 V +43 -11 V +44 -10 V +46 -10 V +47 -9 V +47 -8 V +49 -8 V +50 -7 V +51 -7 V +53 -6 V +54 -5 V +55 -5 V +56 -4 V +58 -4 V +60 -2 V +61 -2 V +63 -1 V +64 -1 V +66 0 V +68 1 V +70 2 V +72 3 V +73 4 V +76 4 V +78 5 V +80 6 V +83 8 V +85 8 V +87 9 V +90 10 V +93 11 V +96 12 V +98 13 V +102 15 V +105 16 V +108 17 V +112 18 V +115 19 V +119 21 V % End plot #2 stroke LTb diff --git a/fig/cg.pdf b/fig/cg.pdf index 816c5a6085e9b383b8cb528474e6be5216749633..ce68c72c9b5168918e6acff6a4b2747ab0538f3e 100644 GIT binary patch delta 2063 zcma)&do%MPO0BqG8X=}}pN#x|xd&-!TH=27Tr%RmbW@@u3NxrD;yf9c!^5a+$*&U zKWLF_Dr<5SoAMp|cXjV(o4eA=^S8j-&O4!&??&rN#05ZCmkp=-;7?OF2EK?_>I1i% ztjsp&Pp54~ZsJspINP&}Ra>he8hmE|uOHLF@!nj@&(>2m6^gDmxy_!lRXf0T3bhB4 z^Hfp}H_5FDbZeo91+L)i~?Fnw!`G%e!oM{DVw2ZS1y)!ja>B9`B#j3Mn zogau$mr16l5^f)IS-qw<^RTgPww=hkD0=qNI@yz{<4KV%SUc%Uer0C{jOvAtPno;= zjD*1b>5rD{CK=0+`RLm6B3`EE&_JVT;M{Q(GU(vf{J(xDzhJMOUf-tTdv!E#Az+;s zXcCRSt6~>sCQ0Rjs7R)~z*sa*lV2f3W!AH7iv{1qCk&nqefJV|IO@KMs^}VMek-20 z)9Gs^aT*LdPXB4}$%n!c=;4at)e?E$f_@w#^yGt}SGIE1vqMU-qI}aoeJjN?*4g)+ zQ6@cV{0V6JmiH6o?3KvEyA|0^l4CTFV7eBU*DEAv2cg`4xjAaN^kZd^@{7=aoiPk! zP%2`**B-quy5&sgc7!(&%f}&wj8)1Y+lj3{U~L{LiN++Jz~sz4Q5~5MSNJm`hH|fW zzCI3n^s`b;hTv4QqW;K>f4*8!2)Zi=)tpRK`P`uMuSr#BgQY}|sL)}Pue?NuWYsg5 z-?k_hMMQcndBsrd&JP4SKFQuJvba2=Yio}5Sx2~qJ&Vam_PZ}V9aibm5*R^~8jLe) zLPpL& zc3Y{RPCoNv|AFfyMo(}3Iyeeh1_mwM(2!8>F`;#k`VR}$LLELAr3O3swEwM7P7`{g zPJ8C*rUS4Pj*Fps=KC*lm^7v69S&Uqo{n%I4v$se@7=fJYqGVReO}OUue4~Bwr%+^ zqB%hfCwm=GQZ@YB(S4Uo(=Q3KDf~cWZ``)u^>Z2fjoz>d>1R8etK(K0SY&4z%a+zc zm#!@~GPMa3#6Ko|ANQ-ga5uDX$7lRG6LLIgBlS|u$87y`SMLZv$PQ3V;IAcrIq^I~ zGcOh35Xxvh9vzR;m6I<#duew|QN2Cus7T*b7tZs4HL zh~^^EJ}h(^k^$f;NPq%TFhsF~q2>R0Nj1#kpx3?SgJdl-%^K5Syi#}W5L07L?DkDWvy{gHtnaqlP+W{(dbVfQeQ z0OEG@VE^46PsE9zn@~dlR}9JJ^TsqR!3~2UlW|1cF$|Ujx{zJmTnRX;Bl!OuXP{In ZG&zNpl)_75M?#5sEDma6;d;Ui`fpp(Nmu{? delta 1588 zcmbPl_t9#Cd;Rp&dD7-0ZQpCS^e($b+|%#Q$ufys>YDVnONS@YEXKo2XDRQca{izZ#T=UyAI z&}H(9z2SCK6%H~sy;y1=p+49A!Tgd2RWS{)UiA`xAoJ}!$tyM_XOnc<0;MIFZG*X^?k8b z`MS{j2esk)+@;Ug%s<=I(p;aogfls1V#<kCK{6FQl*O3Kne;W?~a&S}xIO+;ZUV z>e>?<%vuh_7=&2gUnS%xV|!M`)?3GR>nygk&F{n0|1i0q3&>HO80C-@;Ps~d?`t&` z3*lAG$LG!L51QQ(?0saT?Gep{!mM?|?AE$#+dON^!Yt*h<~~@)5MaPzk;Ia>KDi*k zDyY=B?PTrr)zVv6vCnNf-07U8I59)j`B~-41s=8Zu)n1y!`yOqh7Yjc4DoJL(Sbg2UPaHMk z&ifm`*Dv4n+sbXv1dAV=J{*`^Ra&}s{kQGUH}8}BRDUpJ@f-Jm9l;S#_pN$m?3c-H z(rTKto}W=|!!j$Q>XR1J-yA)kmk=TMVX5=pnoVt&6=w=)>FBD*olkqcQ&VXB$D91u zPyJ?FDjPeR~Fj>DhlJfD~RW$$yvX(tQ) z_0{W=U&__po^wO1QPo&X^_jw*+?Ve^x(Xp5?ZCtN%0RKHe5Tc|EHnv!$uU<}<9qLMo=I zsfnrP28Oz+$)?7-hK8xBx&}a?o04Q=WMG(WHW)>Ef=wgOOhQ=6T z24<5l%Bs~Hm|#j78=GS&H!-%r5VJ7Bw942VQ=N$chM19|r3uJ9eq^_s8XHb-m9vX+ zF*P=@Fg9{Ca5Qo>Ha0UhGInz{wJ7yp^O==z(nCssmb3qwjdoAJ&S zkJmQN*eGpn#<5Gj_QkpLb83^{`34BymVZ6Bc6Z7`rM{A5{y z9$x*lx^lb8bT5J8l%0(WBbAQ|g!Cz@sW?CQY+uY$!qRBGxcP>o%$x-wCzLyS+Z9m$gcH zb!&JO|7Ummq{lDvv}Muv6+QJy?->%c_nG%eSYCe6Un0g(6tDDMz1imFd7;|y>w)3Z zKdxBk7JU8T%vY8>r}$lsmHEni><;6t4ZcLw$CT26{)1}I^{fVf5eUMx33B|h7Wt=?yre~7> zlIOB}mcA-=)^2?<;&}GI@H|HXsfVv z{bD_!#$F$y+Af|^{Qugme(iHvH^blEdG*2b==a%IziqjZcq6z)r<#A?3FfU?d%v&e z{q)as!Vj@}p(P9M$8=bTKaui|7q{P>)AH&68pA%KA=zr8#X?AM2`cBPbGh3Ne&Yqp|tccTOrP&rQyS`-$LPVe6Qx^%E z^Fa7ai(lzxSrs*00fv%Ccg_)^kqGgJuMe^jMB1Ob0j25+)2+1f_ z(DhB!tyBmxG__JNv@)_XGP1I=oV-p{8Mow1(I(=o=ofQWBgWC{`jhKKl!-EPv$J>* zlNrd}zKN9zhNcRJU?&?}Di|3UnrL$A`{t*(B$lLVxLAP%3``A84ULRU49v9+j3%F! zoNQ}sqyP#u1BE;VE-=Hu%*4VNUChwH97D{&(rEH*DfM~-3>ia1OLGhpj4UlM#LO%$ zF~p3FG1VEFVTc(S7#V=|@*z3h#L{50zqDX`*I?3YE#U$uUF zZ~gw2{@JC<42rCut|^ABZGX|zbn&Ld65~62MGvPw?o?uNuGemB>JPCNP{Zv$zPT&HQ>`|t2KBeqP|ErAwn`>)yk%j z?u083j|yCzb;a-rr?a>3G^hIAm1?KAsq$!R{>-!aq5AHkf9^tWa<+Q*K)*6LV%=KI7a= zfgz<2GOCm_Z#hllV3X~Mbocc`s)1?5KznlM7`eWg?cHt>37X zcxf}+v}NYgmg%N&ik$PD^+>}gb}GYyse4s_ET3DF@o(ye z>(4KL{OrSz>hksR^8MD|ZpMqtPWUZ9LnGLJewn^~ng8whUwf9S`tw)qzq!wD{yEzh zP2UeKez-5FXZ_ljRZG|N&;A@4-&up z=HdR-yAR5iGH`iva>uVZ@2WR#%i$fNsov*fC#8IdiIWoAB(t--XtIsQJ|C9^O}E~S z%lug8J=8tl_vA$NOC7o5*RFn=k(t%Q^C+grWLnJ{u`NC`H!@l3 zpLh9K7j&#Xj3bsYE6KE5Beaq4SQeMuNFMcr+d%xOnF+T6H>o&YSGq+Z7zRcp&Qx&hMZnj%IHFx8gwLP90 zmSxlJCX3IXcj-e>YTKb>4aJTc3`su4Zt0F~%3l|ruAR$W6y4QQn5Qek8C#Zk?$t`e z$y*=XDb%=oWZ~lNIp<1yBGPDC8+{ff)v7CYHwNVul757-9yN#*-&Xsn;7~$QT+Km}8h= zY+!*QW^Q1KA!cNPsm{n8L(Is)*Z{0|vbi)bqv2#%Y1>$1XJ=y<3pWcxLuX?{b7KQ% xLvv>%H%kjA17{aAH&+)s8-glgx$Nw?ic1oUN-By{)40sc4UM=|RbBnvxBw(bGgtrs diff --git a/fig/compare_class_C.pdf b/fig/compare_class_C.pdf index 805353c39818cc9c43b1d83409e9411174676cf1..0fb5371d8b867b2a70192e0cbd14c2c64d7b962c 100644 GIT binary patch delta 12024 zcmajFRaBiz(=8kzK(Igv!QCAeuy6_P?jGFT^$G6o?ry=|-7UB-g1dX*?Dzd|zW>{2 zj8ivVJx9&1>e+qKJ-YXPL;ae8ij@L1WbM}3(K=7mFY$k-3vEU3B^K*n#QR~WF{XTfZ&(;4^m`2k%F0T=-s@29~k8H`H3hbDDEnw=O*^4&%A#D;#`xD z;|_;3$oQLIPW1RE;9uV)w&R3-XPh1b-(vvpLV@(B+!*2m0r{NZ;rUBmUZmpyrDcNS z0rEuTg6XUtH3%v@wm%N|dSLv!vqNzSn-7x@lOf^_jMTkGZyQ*JR`hC3U<7=&Wt3IWSbW z-e+k+;V`H7MoQ_GZK)k&RL+n(pLS4BaV|y2tM|5VpIwc%1lKy?7(ik^(Lpx~fUGdm zgjd~ZP4;vbRr7R;5GW}2lN=0Pz>U)j2&z@|ma0OlH}sUIQH%|k(6-v7AF5R`l7HjF zHT?OaQ757oMOw>brLqYJ{IXxGv!^OVapMD7Ip_7Zw4Kt@ml#q>kHCw?Y4{=4ndVzE z8}#eJoz&{%=fDn2LsO~}*T$1(LG6U#tGOc*mPdKm8rp>OC{IS)g)>Km)Bn5Eb|q?{qky2MD0ygUj`=1sZ)#txpxidtvB@RF^3 z`2wls!WgQU31Uw<7c!ximqmbWe6Ef>Y1ApJE5N_>h3umLQ^I&a(%B+aT|T0oW$jKr z{=9nz_j#oHHAiFuWTbfNCWedVW&%?3`7xPH1h{b#U=EN~1BWz3o;zfG2cIx5zVH^jT#dpl4J7~O??Vc@Z^2bC2sX0VVC~+HSa^iD{?%*c>U7oU-CRN-Y0zzDL2i1E2ci(AsrJVqRZ!1m3{4vg_ML(li78K z@uS8N)nG0EWUWuh@BS(3_d@@GQuvmXO<_9DdiJ7UbP(d9bFqm8=$s0qcC~BN?6A9Q z_R{YCERYN!YAoQG_=I|0iOef(Mnf~_rOD!V7P!vE=s#AcOlopkuudh)4q;tn+fQbkS<_mGn%U;3 zOlx3d)i~I1__WM_dj{~#hTTpg1dVc3pHa9pr9I9R*eLV6O88d%wn}-3;_wv4Th((7 zPW~LW20FP_l3Xfv308x^dh(o65>aIjzQ1c~wIndX;lr?knG@vT_y9H#2Ma$xqLZ_u ziGdBGd*+^}kB-i8+z<8cC&C|_-q1HcJ^C-Pf55?EUlmb>QFWQ3p%X))6O+JQ*+~EV zFdBcRL2f0@9&JN$QG$t{1b2Z}5&d|sDeRsPDuzKm@mXKhMnI*#JWO0$; znKtl2T^tPagXw=~cn`-qY5r=Mnx3AS=99L|@U`w7(JCjMFnbla?;;VRRZ{4(359tz zLR}ObK2DrBckv9EHI|cbwMCgu`<`lB;ztU zR-4k=DO>MAH%*YBPXJ2CF3&FCCm4$6fUMhi5*A5Ysv9&tX%rUg!6y0G9HW~7r!TKxGsl`ae1bv+v>JH%#!)4s5gVK80FW(E1n=0Z2* zI{ej?LN2pj=A;6DMjl*icH?0Gq-Q7Z@QW*Z-%24vqH~O;x+3O8HF<~gYt}klHVWee z@XRZJKL|}D+?D6Yx8Y#>;bqDx>gI-OXYqT-%n}+?$6XY1=94||p_O?gJoZvD!O}2b zp-F(gqZRfLb_0-e5^U`x;elWzBjtA9C9szRkF#>@uES(+YM=aW#+_cgNiQ-4GU8bT z7D90;&o-7%UayxfzU|27oQK`if;4h6t>>eYTYO&5&pDIQ2Wp`Pwt=c?Z3g!$K03Cu zcQsZ)(OhnY5!ZaEM#$nhzww0!ag@bEta%)oZ$%YYs zgQ>LmgSQSD81)HParoi$z`;Glvn?#GZ+Liq^GipUM}}%8!d+JPz&Wr}fgO98)}VZf-~m4Xoro~CtX z447QxM~C-XyFM{wZ`^DttgqgnF8Y>@*-)LYy5e={={zs4g^Ch6uOupT!=I!2HrLO~ zd!wJvIxjC6k?D`Up34Ad)dg^Bdvw`oeQ{PThZt}E59Y==%W3=|N zo;zZam`}9`Q2HBi-otnCPy)QPI)imZTFru|aDJ5v1=v?YwGJ&LlEcRw%nNmUDnz=F zVk6aEWuk8#e47Kx_=U^#l<-4@aK*n5rM^-bkblP)Wv2~4VhO17s9gg%PEW3)!W@b^ zGbW?T#1|ZuaLT6=nvOg2n#43qDR&0_)ob>=koqXYL<0$=&Eyz!@ zJ~mVrhiQIPN%z^1xdPxh321N5QQba|-aN;0m&rRnKlKfUu8LW>%be_QI>!=Or&FOf z(FivF7N?3`LZAUEz3oFnByqB=5ECuE>%IA|l4$@7Z+yusyZ!hLUo?}_Pgw=uYeovk zsMIKOD{2R=e)^2h=%$viX($W{?Ej7izeV{w+HPJmw=J<{lmpPg;&?g6?n;JOeRU`3 zKXEYmi4-2tJarS*C18qCkwIqhgQu_wiB*q}h6{L1{m-h zo*lV%XYts6C;*^N*_RDw*6xD!WQ1&717vH?$Ocx<1EuhmBYiBi439Vz@XTaRnopFi z3)%7Y=+*W2(6s0;PU;r|G5cSe<;XW-uuYn1Ko{J^~3)1^K2gHDuGjULUfC;{&A|%Wy7ERKXt;xMa6^GYIK)cQ52#zM;uTY zY;ss;QGk=^Cm|pBDo89tCnam*7et7P{3TSj_emfo6{^o`SP(2-U=<`1O~|CECy@jM z`@^=B*8j$_#{)7kw~VIaujB5syQ&;Bq1*0?=Y5TyX=Tdo{wILG7-_ll)+Fj>b-tVd zm0o|=*rFLdjW-X&Tqqh%R~;Eb%;}$t;=cwb2H@RZb6h3f>{8Pto6E_qY$o1}BS{Td zN1^{lE25#n%gT#t#9L!`Or^j)rI53~FN`di3HD#{=Rw+JoGHXGK70{?Wk#C$qyYON z$yNWI5!IcAtu&-d^mV)5yrN9-YsfQZe;$&Lnu-jd|D%={qvhw~#f9=gWDH+sc7M<~ z2QVA%LeUoPk_5TNPq)2$lI}^#5X*|j&zg@D^J%5{4dk{jmo;2`Zj@tlXa78%m5R6d z__CFZGTe4Ei%_zL2ghz=gZsQ2CjZ%{yz6f8YsP`;{azp0d!i!ss&|HK)!K{}0rkdj z;9+|O>z1a&;7I#mkTkC$uau>&y0o(C9XR*%(NA5E-LWs(X>%+f6y*)LEDU{Z+!2U} z9~NzyhdCnb4zc93$2z9OifsMjPfnWF7f`w=mP$GHH-<*k71=;(E67@G{2 z^ux=Z)}0$0mGR@9oY8>NNA-y02=gx|Cg5Vx2&L86YGfNPb~~HKuS)G2?by}2fq+8p z1u^&R?Pga%X((rj+EU8 z(+FMIByFK+t6{@S$*RRhP~9NM+YjKE>GOrD5zv9;E8r*}GHuzZX3DG*{SJOex}B8P zj~kM*(UfzVw0<4A-P)>O$cbvR0JQXW+gej3*=dO36Bo5v>gcThYN`|ivAT%5FHNBX7nsO3=vh2h~)R#&HvzE&A=3+q1@cA z!(~${lX>5WME~Pgkn)s}_?Dr*heQ_3c4b{}gq45rI9)}i1->?Bi|O2o6o62`vaQb@ zeSyNXGHKG!cdf$NS!N`EHOmS=2Hc{0Q1~C2X?a0CIQI6WM4xY;*qzigr8O6sa49G8 zOc8k;p!SPuQyrgBHDQ7T%1K4$M}6RBU+%C%4W9DjZFUPv=CWnHNTQ^_DnFX14hS@m9M% zOTJd-BrjYmq}Fr=#>#Q)k-{``KB2nD%Am>!&GgDswHS&l4ztF#EDO2e&-r_riDgx) zFtS1d*g7wt2{X@ghW2VreF9-hYd(-K39>X)k3Z}kF9oic@oKv|16WG~xHk`s-`vEG zn`}2)RXWU2@wYvnyl2yt-IT{0YfOJHn9L{~Y_3~ns9tVFau*Xd!FehQ_$n2+qnht^I6u;z+^4_ryyB^$}#p7E+;G3@Yp_|d;(axwY*l{ncq>c#a+-cSLC zo$7O5hvIO}R_B0s$MA(phVz$Gg@e~FME;~vT+=2I{p3^zWp!%{VD8O7tDaSR70iqM z6gqTduhi@`m@HQ!Fpb*&H}{I?)m-{L>rMT&<_Ub}+fW4PU(LG$6GRXC_uYF-p0CV% z0#S{y-ITi_E?g*FY$3*?p(ziH?o6KG&2_afV=fQ7ieU>k)AO11H#eKGsmk}8dtUfN z|B|Bm19|K4x2aAe8e~kZH`#tm-|%47rOxw@tBt`owAvM8WP1(e^h#pGf5rbctP2d|O_C^9Bm8K0M2d$IV41|Jw>r^Jkv7z4IcCmPfoXo2?p;)y}tB>(Jy*tX}Uwmr+9%1LNt1? zeMdFhvg!=OK8vm~2wyB9)xApg%l`m~z7VB57xj7>8 zwQq5y?G-I9?Bb`(1W1dgRntbqN(f8Cc4cHPPnf+#xtocILH@cvZXUoaGRF9m3&O*|rgt zK^<6=eeTf-lgd(il{i{+i<-U`AXlYPeXxALoKgJ<2*X>Z%9&8~B!a zLJRq@-rq7f+xerMxOJ=pa$DE<8{IZGi-vXY?Kh&Y2@+7aP*_V`H&@ZhfK}Z@ef7E8 z`ge0P1q4ox#ANo+_2zZFkn|cK{sP8yTGN#XEin${iA8Q|H~C@4n41}3p(NE=`Uu}8 zu&MgO(ychon6L-M{{1`4$iX+SVkBh7k%oZY&sv*dk`0j8Yi+4=5QSxn@_Tyr@trNl z8vpg!qNtH=Zilu*b9-wl;5FtmN8rZP!qnz6fKKAxlKYw5o1KSaOp2m^Kb$+W3I{Wy zU?EtT39q;G@W8oL-~sj_h<_igT>c|8Jq`CV!2A>UOn4CXS$#1dxg}udzSXr>*WPZ8 zPfJmg!ArYYY#nepgKyqMe3;{?*Upei$?gci>UeFusASC8)#C~Ryi6$n7SH7m2jntr z>^D1%%pW&9{wdq$_&K`GuHaPAB`x2Lj5d4G7gnO_{LGd`A-jdl_g!#0q#<|yxS<2| zO{qnNEMvE-g#a5iguaYEFaKbd zd@&N`Txh(m=aqweuu%Q?ujtJ84dU2XatN994a{Rge%a47_l3h>Pt9Pldq3X2US~QE zzxpUeG9#rv1>^??+dIi!C$4_-1P!VGgY7o{FnW+?383#UU(S1;YJq5rXh?Wc;Ik#)EBIH=wH=#!iS-bgHs~VA5bL}WLQ^s{|E2_!i++E_)oZib?5Q&_won4Fa- zNrZnYea~9Hgor7TU353qs5!Tb7J1r@e3vDSSKh@n-wdcM2Rwc0K&F~MIo46H{_Dr} zk*&GgjgLX7d%3|o=+xVRWv3Tm2|Hh|hqy+jwf_DQK%9RYW7=0VE^=+^o8thmKGJ^InFHpx4O zvTL|Qr16ulFJZrtGc-SYNnbm=}13Cmbxn?Citdn6P58ETt1MJGH67kgTdrNun#FFUf9iH>mP) z?*NM|d<<3T??HM9s zA%9Ap?n%>EEv{i1z;fOFgM%!hB11oiodc-($ibs+LRcT=GfD7?2W>$pRr*iMOoH_L zpVywAn;+iSegMJDo#}$F!NKDnRy=l@zh{TNRMLYRZ9RGI=WeG@tMK}q@WI_D^v+D3 zk7wV_|C9iKAI;vM-bRA^a+tDe_Y0}+If#Z{M9;)_C|IRB7ix#Aw+ib2=zC5P#?%1w z94B5g!!-Q4`8B`A8IC%CG`1OHhV&cG77lk1qTYJhSb$Jw1N7uy!1x)ar7>peSR~nJ z&z01uWlKM?;U4vWncxZG8;(m^`K^fXJ;L@JS0R!eHr8bzv#Zx!OyEpHHu77%?984; z4r|6kK40Sk1xJ9UME$gK&Y!=J! z%7hVz;Py`veM_72=IdBi_nk?OMp?^ z4}TNrxHJ}pOIF@z{vq{;$2>7mJHxK{!+}&E1&Q&endLp;t58{iknX#eaqnJk_x5Vv z`&k~2b0GKSj!8h-b+;W=UVvMRGSPObISH2U_NgsOnh7fy|HKn_t+@FynbD?fZGBFo z9l^Evwes&_G-=l6h%6l8WVglq{Tg?~?P^tWSBVkIqn{k)NAzT|{}*a$8>LQnoE;@v$O?Y=#zvh zG)H*oKHs77OJ$)g9yIjaeN50`YBv#=0sqwl1^xkXowrfVk-mzo&2a9gWaLg6+CXRu`k z`LM=zQ0h03e)8)CjwL@l)O7lUHRJ&lc#nsYa&;PV12H)gfcvyis$7TRLMU)ljn8 zj*1ae(5fE`bG*u3t^aS<5faiud*g&jRTpAp^wh6;GgfOUo_zn`? zFg-FoH|pO8?1WT?8d`GbvN;srvLjvC7>$k_o{epW0Pv|Fw3PXvC@<>&InXB=yXLFf zHLKru%O-jipBC5PXL~z$bi5WbO}}1u(|2Z*s$s#?)yo#2j{R_GOogr?;sS|=NJKs-#Z;J9#z?WYwNzBNVQApbET2I%s^$>JU~>xzgvO9!p%US% z86gRM37xnK%hj&jLcIBWAvLRLQTs;IPkyfa3lxnttydD+C%<31W4iyS*Wjo@yMInW zsaTm9yA#lcw;xkgY;W96O6lX^A zUxb0v4UWd;iewollCYQa6wUW{uDJ~z&?I*k_SDXAmZbM|$U+}cB7Nf>>A?)MBo7e* zfJ?q3m@t0Itb|e=o;eP3g#z6YwgHV=w5RY}T%@?LY;0Vx#5LcM%aI%UL+&khL~?u4 z+OHZh6ZdZg7#2bYe(-YJ*rXLKkJy{?S>FO@6dxsF0xAU>(dA&XBSsrTC-?Eu@rqS7!drC zS(-KR7v!&Ic?$`%y$DXY!ZBhbqH78Uhv?*~`Of)yhXvsPdfjO;_Cu2ZdO-)o7{Q@APW z{+TBEcR5khGXNoqW2b9q%L2uv$EScokF(x=BP7{}13FD@FC za`3uRxjNc$|L4IX+gR9j!0^K$Fk{uhlm}&3jm45nds*sUU&PU=A4X91Hb0gCyFmh zK}Fd5{+`LBl=Wk-!^y7ml3zQ`p%e5H{LRE$eU|&PyM*r$@2Wp~wXJhK zQh)m=GT&0Ml5t{$+So)<47eEEKowcakBcnr7Kc5#(Bmm;d1k=FfHnfd%er8}5?f#!(-S3kPiLzD}cs z5d=_?%t?Q0u59(^q_o96?clFjzT<*c8j5G8z-37azm=;|gjyqaPwzro$M>22v|dP99>k(nN_!VFWKGe?R z49$ti#{iH?$0k7i0Aqov;N23|R@MhAKCE~)(kz2fkQkQr+*eF$X20HP7UWYvs zJXylt+!JLI99X1=W{5m&B`dQ0#1ITjAN{D5MBI>eIxOIT z5`xpn4n*!TCxmmy<<}*svanv_v(Ky6O}WDH!##B@0UcAI2>lS$8N$M{)=N zy8Eq?=zguCIA3UO9+e5s`zqlH-_v#+aU7A;z%c;VB3w1>&D-wE^ZngrEQ{85pIa73 zYP3Pa|M``Zv4S`JJ!|&OC|zpVSY?l;;#y7th_2{pkIfe@vQ+nCk)13rkS$foB4Rbf zi{|wIo2I8u1FyPSuOV-rE#gjJc9Ste;!@2jG5TCJVFM@rri@hE$Q4!LbWQebI7gs1 zyk5Lxw$M1VI`n%LLoOQNL@;Y4FA63SY&sX4&=$? zQE7>(PvC;K^S7_XxzcO=KDxcGz=2LSFiIr&9#j1m+VuC6Wj9K;-TU=gBE>$?{zJ;f z5}>)C`z^g55l4#3g_?Kpk^^i%mE^qd-K>tH_T0QJJRI%53~!cWBn6y+rPza` z7s$K~A!;j~D;A#zR^}_xBZk_O@D2$#p7#C3Et6NV{j21wj7TzH8YGg@C-r2`VVsci zU|yO6QXJ%C@A1L#?+NeU<)s^g!+s{;5x&19v7Nc6;9_=ywVR%7rgMnv0~~Lj9&u-_ zo@d=dmEQ3Ytk?2lk*KP9Y)KGsW8bKQ$1%#+bA(Kw-I48T%(3EGdFQ8Qt37bAg8qds znVMR$zNicVwoZiiOZvBOZ8`>KDc|UIy_e@i{hS3;orCDdF)eFvwWBqc_b~IE7-#xC zI6a!`$}7HXP;_+S^~~d;apdZMYGcl#SblrMNi{ z8?Ep$T#heLp1;r5$21 z?%i5q71l08nn@cbU?dDN@u`*-2&u7EL7B)uiE?F*ax3|+16j=YJ|y&DsYm8nk=H1A z*f_~D$^5$*?}qo&mbXp%1}VEZpB{wdlYS=LQbxwIcVbjbMth?^^G_g66ayn@X6A`Y zzRJfI{*z>5RU^~GP2F95cksiYe<)3|wP+MBYpyQBe} zNnWS7KyL`|WErQvXR^$fmw^tQzXkB4Z}Z4$f6d^H?{m)L)Zf3cyi8wq?;Zit!I zKYsrn(wdD+NO%6%Ojl?l>!u=qT#%>l>=w@VB8GsW{O}2`c(SMhhDv}lW=^oBEJ(OFW<})shorVq}VNDn}@%#V0 zdPP$iC}c1Q%*+g8XJ)5kW=SYB1H-Vfuq8muC;&D#W@6_57_lue;(r~Mf9m4+Ul=op zjUDvAFc65D_5Xr_S^gEr0%rRc#t!~(4&Z+z27$O({zYa7asG?U&cghUQ2)Q8LF{bI z|K@Ul!2gQl9NlA^sn#J`(Q$ delta 12621 zcmaiaRZv~svURXvA-F?u_YE6|-~@Nq;O=f4ci*_XyE_DTcXv&2*FWdn`*7}e9{yFe zs=7vv?$NX7!(6l0sF{X1n1_g!1Zl{`;B$QGoT^T7{M4S;1%m@+#E}l!TBwl|lGDu2 z;J*PmK>_L}FHDUS)AsDJq_>LLTSJh<+4-~K_x zAm8$SfAnneB=|IrbM-XzeMpbLH*|Kdxta<_l8*HRTnFZ(;UHX>c!J-81R;|lp(Agh zpUQ5?!M?r<^gW}>{ zYr~IbAz02T8;lb%GmwFm`m|^p=IR_tdrrL*S%U&r^F!UTMjK|@$iD!(TS+N+FBdXA z;nw`E5dXxpSbowEg!my20mu?#?A^;hEnFtp4m}IY1y|o#3^iKdRrd&c=yyza*p~&; zZ8TKTHY1UAeDrb=sJOZ>hgotgAT3J(OZ1}mAh9MivLj(o@k~XhmJ7xCM~z7m@1%!` zGM#Zo38puoTs5Cg3}d97-q?sP^9tL)_;)ey1ze~f@6RyVW%(zS*vf>sI+J`lv;d$A zC4Icwc&EQrAuE5rmXAVY@>R>RvEbS8*<_6vC$@p*pupHAFXyv)?_P1-(2e2Ze4jsc z+wtWTQ=S?~l)4~!CNV;^ccduMs&ZbobEX3g!Q8joT=gh?|3p!FST$xXwV9iaJ6BaZ z4X7x(yX8c;fOitv93;>ZK~oib$f8q3YtYD6o)!02SsJLBS;;OOcljV#Vz&6h^tG}j zF#M`cyLA9f1{W)9p_i*sIu9;`z}_&8*9Xs?7vBt&(!^gIC{SE7-ttCXr*fOux2&vd zDgpIIo?5yjiQlc)pa@2ih*v!DNNRwb#KT>YY_C+_7p9w3#if#J+b@c=D(qBMy`K3a zw~$UPHB(7mIrN9pLjxo#u^;EU5IlQ z`ov6Fkqb%jXC~(MYj`(g3L54)vL<@&pbU|ER56nRcc*JRvg>`}^4 zZ;d-*ooZ8LU?}D~gNe^tN|v^qV(+;cI-j!Q0_-1@-Xc#ZLdhjdUrt$BNK)8PMO_0l zA%-j`ogC5pEfJizp2U*LUdjNqQDw0T_<&+kl_NngwlQ*aaxngB{jbOtoB@Lb1!Mt$ zD`5EQ*@e^yvDk^3iERxm5IER4IR8!g8zXRl<-b#b$Ms^t2Xy51{`yuB;4lUz@Hd0} zdOOBJbPyXmCo?}k!v7X_%Q*1x)+Shr+b&{I zrF#3>4O44qBG}oh$@tg88E)XdTfJxV1L$RgXo+=Br&?E~(?jKAq5Q$ar059MG}e z_qz#Pp>WpP4hy*tzY*`%DSP$v-a5|BZqwuLBb_V#r%xooBO_rasmT)Oq~Y(t^6U(} zv~};(UxGrkNF`gJ`yqUm6C-n?6KRTPPH#UO#?x?7ER1Vau($W6CS~oQ4S!%K-wW;% zqpw;R$gbPAwxE*-WKbGiAm@-Np(61kcGG zQ94rbgr*~%vq@|X-Zoyq_q+Kg45?v^O{TwTO>^_o4fQ zp+-HRb%RZCL~x}YST~!z)2Q3fZISmbK5ah8|In}AjeLTwi|YEZMJYbgj1GD(jHiYsCN=h*mz?G^b3{A*)!P0)3MEr!Fi1LwB(r$I8uv$dtcsDE$ zb{CRY{mglW!9x|d00hhHc_9QW-f#OTXLI6b{Wu~6&+2{eGp)~NgrguHV@XY_N_Qf&8UW#7Z4YHramk{}@d-v0CZFZH#-s%P$Bdnb(79D|C5+>PNC zN}_{KX9gZB1)PJ*RLG&luSZAxu1Qopd0qR0<~1NKwa7E^O|B{*oAV#zyAhZ_oNN8p zwq%F0kOy8Bb-c%i3fMu`Z8NfS4q%kI(GK#-H$Q#Xf+xCWgJ*P{%<&ID<{F3z9*!mA zFeJyB*7y2|dLHt)-AO4 ze1`ze)fk-D#UELT{zJ2!B6Ih)D(Gy`X;9XuS2fY!?Nmv4YEVGw8_SqX5k?_)Y19yU zu90FI6HkSB;#B3SI%TcDz9mkl%ly%CWQkhmq+ZC6B07$lIPbY?0JjS58kSxlelwr6 zKrM&dpsftm1b&SUW3m302GeR|(i7LKe(2U1`|R_|!lV!7j#aKsk&8oFtXmdC+4Ukm zr{Jr;A0z5zzvK#X8c5)VPIwQDIh9L7;-Q3xXoF&?v^JsOo~X{TkhJo^lf zAi6cbmw2+_XRXv_e}avZUGqF}e&REpX2~{y{YXJ8a>f}%s3Y@7ac(fHDiL0E7gIVz ztm-2Wr&KC5;ui@v>+t8=2n1~owChk5e3V!VwLQL~WTeNvnQhu6q=nT518g{>Fua9O z%+NbwXLxSaF42-Cq+;hd3Uh8%5w0^*|Kc_D?(kQP@&S>);h*|qQS>VI05QmZ1&sW> z6o$E#6zq0T0ZFMA9;TtB7WNrFYmCJ~Ad7N3mDhs3)Cr|wZfYoOc{0LyO6})tyLx-W zpN(GmI8P$Zdy0*mrm<c=#wn{{|67QuFeWdG#b+rZVEntQOU$Hb`w9+~Yy+GR$>&+ThUkVv40#;~ zdI>yGSyXC)W^nb-$23mep-T?}-0eWpk3WuHowp&kq|Wx_&Q?M5E(U|;R0GAn(5l&s zmcw2vG@17k3&mXF#9RX`4N19jXU>;br*l1*f)PWcp7dGVw2b?`hKTvAlQCJ9b2S+d zq1bAF&{GUoh7;;+`VoZLr`Kk!bn#u6KZ!wsmMdyTa+*(&_)-Ejtv{FRu1SUFPe5R* z*5G+6HTJKY=BQ4lteVbIX-UX1bLtgsjuJELEv{KsB8se;`BGl#9C&$#uDkL}Z5O zz7Id$V8LnnpFjjkO8Mdnn|$tBmj^;0?fBmGgayqAwm@@T4qo1w5@mP#ynStXLCY{V zVt0Gdq^yUr(s^tfbbc#19pj5oV@q8%K_iu;drI@=6HvntoJMx}ZuYia=Rbo0;(L|N zayYl5SRojhWZcN|a=Ge2W7i;sW)LxVaA+ix*HPNgocNiP(w)iKGn2`iBz<;mdVMhh zV$uU=WLI6h(D!^1;1i(uq`X|v#b9IicW7TbZyx(BH#Jp)7{A_5!a`wtFqoxu>27&x z^teKFZY`DH<1%S6Go&LXRpuUm=)#>3=xF_oSJ);<-6I4?B?`ZKHckGND^!$axjR3h zb)dG^f@0WnK0nP}A*&Z&=zeiNk#C?wrrJ@~()mam2L`wjwR+l3jIQbN*15QNh=1jM ze>j23Q_-w99ym=(?1~3^KOkjr7t0+YA#U-|VIfdB2Vurx1{MCr?=C=(*ZrqP?Q}3h ziHLuwcfSQv1?@9oUifKq()aS&{BZQ#P(BOxoMGqEABRULM;#%z5ouKI{i2bUfJV|U z_nP8#?K3kKwEBwu*3e)5xb_Nu-+NE_=+L`h4xR|IYqYD4#R4!!umBZP)=!;_EZ6GX zq#~kX)1K0v^e1?sy*fdj_GmKJ7aI|jx7HV#Tbo*%Xw8Hb)M8`Dw@Fs|SVrIxgQb%t zyYW!_`UCe9pqhrEy{=r+a;jb}zkat5HG->UJP1>1+i3)F5@ ztIY(&p&fjWe-T?<4YD21r&yE@DYc+<(Lc34x4k#XwZ_9bY6UT#8^?=Yk9;dUXo@7m zb?QE;2}^U_XiYeZITctUK1|xDZ5?tnY&~*xZfyrLzJ>J94^4x&QTqxGu1w8&MM z4PincKGt+}45Kj4pG(~cAMCf$>+XU4jL&II?{qu)kOVW&2z5`3$jeD-R#TBT8#h|} z)6bAHDk^~(4+`{N_;z`}?LVQ+Jx`eSr3lzyOULbIAh#m-!QQO?ti7x z{n6R=wz46Oy}*Q9F`<3(o7=~OaL~lw*l(M9x-4{Bw<(^lO(yu|t-ER5^&%$yKHvFpiH2WK^UD6lN z@;0kwJ4uZpuPRN_TC;8KS=p^G?XYD8rHCJUXCqkPrjiH5076He(I<=|%h)Hhqq=dg z)Vh7CHP$9a_Q^pn^SMj?{XV>q4MCk~3O#OuO>HKtJO4q;hdidv2RFT|uJeTk_nBnf z0V&X0hl1_6=gLc!+Sz??!2@T~q$d!9sbaWltZ0RtRhqZssN*P4d~d`&PNFE_p`nG4 zU3`$zdiJA=$sTa_(eQohQ{lGF>G3;R_O92_CI|K-;`m3AA+TFusBBNy75Ytcz&+t` zw9XEx^BnFk>tW0RfH>D z8pm=ZNaYMLWBopUam1BDlAI`NC}~vhqu#9-Lsuc|AURJRbaOK{*Qdfhy^_Ix+7NUt z6pIR}1jpZoHX85V9SO@+BmP7Mg!B8#NE~l8(}W>W@ewP}i3$E}Q_L zE0bGEqB*e2?a1yQpQ_zP2i-|xrnvBoWO;{XdAESV7aUj*a^}sG^!n21O>yfI%8%zk zd+Qb2P6hpvYB7!Kzlu@;C#d^F0~#a)HyNl4-}!C#o{VrKckw8QRtSmY>&vUlG2h~f zn(%jI`=JyQbv~RlczYy^n_<3ZPcfKCfBxP!XY~d35SRSW4=ymWS(V#M0Es~Tez0m_ ziBjX?L-MHx=t3<@727{NP(w8@up1E~1Nj#?mzCsM04^`GlX1SbVpDWcArus}y@DbQ z$UK!7J-@CP(`mX0wtF5G)ThLs=mb@=@(sYfrg6m{t z11PwB$M=i!UZ6i6Oy~)lUK=tA@{lQ2^YS6D>g1?$J8)QFSayrD#l1*>CVvBj*uqbB zhdsC>R*9~8j!8-}dn$aFN(SNSW}1U|X`G*}S{h%&O4+Pt_xhG?X;8Ckpntd+yLNc8zcKF&3WLR-tZ_UL{=TH+A3BYX=Z<^HD)*?ldQP6T zB%?7Xs;n_La{iEx%5->WwzcQEf2-%`l4y<{RPfAkS20Z|OC-PkRF(ar6))RWunfr1 zw-X9R`ocba{Z%>-$rO7_MKN*&wU6(4=}x)YwRQPAPOFU+%^3u7pc7s`F(9e~uF9`U zIqNvflM)~Q>7R~hA>Mmb=Ro?qv_LBpwLPC=!>XxSz4k*>i7MDH`h@|AMEY<3xI@hN^TKK1dFFP4*LW)eq{03g}GD6@kw87f~NTY##k*Awzp z+f6*j6xvP5*E3KYx=jNI>mE@%^OJ0@P);?%ckE4UPIAz2pFx0~x|`&57h?xQ2bG;1 z{1jp}!a5n`Z*ksAbnZYwzd4qsYy$KwR39=gYBL;kAObEVruq$CwkN+L1seie6$H9m zdl1V`N>5NArI-fsv|+b^tOqBfkC^XIY56I}Nm|zlKAahNSwA1@p8Uw09L=R)34N-Y zf-)y(kX|6YPxHyKY++lTO1zr(o!#XavRzgncqQ8iV8| z<)!Ihv;+562$-j%V46jJ5~;Mnqzn}9?ehyQVL+Ig_<^!5MZndD*QX8Rg4wh}M+ zh!KzgTc)EVkM9qrCK*P%Whe1+2u&NpvtoTOU*{G4#6Enp`2gF&v7LTr(7U znPFP#!lbF(tF)H4kjIQE9q+`~X=?;&WF&EyNVisn&mgQ43kY`gu8}=Gk@wDy9s%m+ zi|b~O`)_A;tgC_r?nW#$D+?~;jKHPz1}>n!z?azcwlRWLfe1zB^f4{1gq8jL5jOWE zirFfnhql87ZM@1USzJ5EDBN9LENSo4bMBI?N^+N_ng-{le%BSumXV?qQznXyXt&u) zLoN=q=`E}~{&~H+{Cvtn_Nm|1>U%4*sR}b)W#p?E)LSm5EjzzfHCZUS8%W6%vsFNi zwH*%Ph9nqOTiqksR<&k+>xPq(a@Lv@SV!TijtWLc!EG}&R=_tKUWyG11Q@=7)|Czl zl+E11G%xLv3>?eI0=c+o1}Li!NWmNmD=*SvEh~co2dw&!s^Wf@I|COlkPF!_#l>Bh$jCuz*d48Qz44h? zdRyd?4>zk8yCid?+!5aCGTk5<(>^)Pd-mpel8GNNKQc%mFbmPb(&jvR4D7`wT-sME zBqBW4ZH{l$mGbfOU^}z!%{dn*1em6x86kPK?pE^YbG|_t;B12if~`NUw;$Vd!aZTB7!6_t6az(?{p$E=*o`&uLdxc#fb2)wRZcuI=9X z?&)Jr^_uT>%#X`=mIPS*>M3LdrRwSvIx}!a6z1faM!HYI)sSD4*7bziU*)%}uQ{Gu z$-Bqo9*zhn^SRsSvP_R6)4s^~1=D@9gGl+b+NZ2;F6)AdM4`GVS-FGMs-n!YMv83f zJNy}iBf%=iT1BAjk+v2To=?Z2tYWP$0o3YekhfM{M^GXCwKpLDoAl5qioep?yT%Au zai6l(z687IXzQLcfqplAmv#5e+gnS}M5~-Xi*+D?K=TVWh|LHu{p{D%W?knZTSZS9 zO~vgM_T!Uoh0#>{AnN2x$K#hF4V~1%*_nQgSbqGQK)}# zyuLlD91^0e@0?0|HyiPy>J#`~AI=220u9XBr@U#BVZrrr7~ zHeL(eIy|XDOnm0g;9HDgsj?q=?t=(q8+k-CN|l?r#m<`32I2R-CT7lf(=}GsguTn9 za1d1Z4crFA9)wo2)&jXW-jSuarEDrO>da{C-36C4Y&4=E_BMbv3?pNzqp{JkuV~(Y zkt4K|L)JPmQIcL_qb%k+J+k2yjYtO>zH{hf>-d{f{VTO6-T`E)4|SidF(=iBe9_dX zoqU&GQM-{@Lc8%P_!qV5Fvb~(`a`HFW^8Iyxp0~GBHr9jn}TCBqo-s$Y-^Zk^cD#> zD=wA(+#VMceDJo+Mn$LR#byw{gAE|tVI2N6HM*5}(g)BNt^~-}6T~iPmGhE(r_WB% z16Pj*(==<7J&c6xR}V9d*p56K+bJxLj7}Q`4tG=xYadj65y47iQxy=19<9H&f9mkk z@KIwwHZ~zJRrP_6N=e^1tTQgcahf)eoV7E#Hz5YuOY_?bqIj(>WPYmBTl3sA7v552 zzZ}WTNJC(D(B82d>ErbUVKy5l(MrSl3%Yci@9FKt;OwRfT*&YMtzQWPP(T7+hC2*H zFbaHZm4Vsx>uQJ}obD&EdngrJZt@iyoSNAh#AA~SUCK21Ic?*~SpzcL{yS~&5Hh{R zZ+oCHnQVNGAVcYwmv-rqpq(xU0`RwS)ULADbU333Md0W!S{uHlK6^^pZ7%}xjOjz@ zV%Ah-d{g_l& zMamOy@b5$NI~_&Db7VO~enS#Ks~$~Pb_m%>4YYA*{OF@Gw>jlW% zcx23LVsEhK2gLaPeradd9fRqz@yW4ga1O6g+oWT_H=jB{Jz$U zKe9VY3KCUSR8$S>HY-f|VoSj3xukhelW zF|(ZZ@TJOqq~$bI^?a^%Clv7qB@T$y6E4nRRy3@w&{vu4P9lW-kHiT>*fpBgC#u5y z_GDta;*eTf5nt9QM%cy zhO8n=iu7$Iol_%)tIy1FDqw8q;_)OYm(RnKq)p`^{z|HM7(#r|;^o^GrGuc8gUZcW z_wO9>8mzPPtDC$HgN>K_T1)s{4*pI2Z-X0xTm2Dw+XmKqXLZ$}vG$)HJ1apGQns^Y zu7uWGx69H{(LY)7jz#~Vi}<6D8gRn)ew%Lll_RyJKo+84e3kaabYIPgb{Kg$W-LP^ zxcT70$RlJt4|9n`_($|mx;PGKnOrH#D)~gthg=VQ0=~vgO^c1|VhF(7;NBOc3=is8 z+h5Z75$R|^Jx@Fmv(K@=HM&lA4W{R|NxTryjiSAz6{Emtq>H}(DI7!g2dHtD1hGKi z$+Y>biiMmX5a;`)qDD$Zla7v1Ck_vY*2c0gqhG*7b5>S0N=$bb6&C>lbo2|YJBk~Z ziPdZB+3y1vx9xGij`Zwqv{#(j%;2-13?d13qho(|U;eGm8BXN+( zf#o%%xS0Y&|D!~BUkVw-*O2$|%f>W^nn`5@&^DJ*H}*O!HB_bd?eO%pQy2=wFF$C0 zmp0}ZX_DL4z4JbECYwZ3{}kW;SnKlX2vP2OnB90H?h9w^*Qa|Z7mw-B4@JF%1pAJ5 zJTOP99DdEq-*H!XSfXRSypd%;(=^LrPwW+R?U#f^IYdHo_7*RL+)=(>rC%rY4@Y++ z1p+1%S9@ztsh0|Zin6}xWuMbdjIc?z?nDnIA08GD%wL%$u#7Kym{v2f=o$LQ4#OWiNtjt3QfNm=3bk1#pNbeHdTph=gIL2DMti4b5!J$XNBwVaA@Av>FEqHz{e~yZ$PK+6 z7}zXDB2R37Z=g#GdSDf_*HGQFc`QTYy`G-WDD>){goF))BJwzCCc`z$hlAY|6eUG4 zpvXV{mX`Zsm)q1c@QDGQ+OMg{Xxoh#i<4*?8^!do%!9yQSAw-jEO0!@s(#x_0^7^l zD_qH@{e+yuUbnDo>G9Sz`Sj{U(I9QY_?HZ&M8~4h89sqy(h?hX_qC$k5jr|%KAj&_ zO1qD<8#F%@C?J&rsAv~qYN$44axPo1FXeL-PnPLW7f7Ma^$4nZ!5MetKP&)6LmHiW zf6f!@H2Gs7^^xl4nqr!+K=ws13>)M!!M4(;es!|@jU4H*LYgViKJ1z%E*{8 zq9Rv?oapIuV|rMiGv;IcoGltMzZTm^* zXa_2!nd5Tann}jhWsAn^&QwmhO*yf<9!Tg&pi&weFQ1UR;{8Y{B+vDxTp908=2mn= zBTm7#i%6pjH7qmjJ;e8p+&pF?HfDoR$|C{8r}W8lf_;Jgg5M;1G~*?+Bzm~VTXcVb zdORF>o)jS2x04F^A@C(=ov8HDPnjL|i1r39P2o9%q?9}>^U~)0D0q0`+ z(OL6Lo+iDO4Cew2p(+|rcJT={Iaq*PanqJ2W*@OFxgZl0GYt@Uc zs!-V2CO3Tpxox?H34#m)Jc2aPH+|PHJibIYWq$G&&d~n&+MN4mc z@%|_p#=#}Y=Ohx1m;3w!f`9A>ybmqvsz~rKF+W)W5zQ+cFy6j^R?Lh6v=T*hMF~G= zVGfrq{J36y4Hz}qFQd4lwq`gHXi01!c9AZ;P&536!4DhfgUnw%%-AmzyYqd;hxPhG ze(fjF@W=P)#y_+sU)(KG>+#)}4CtUj3DpkdNMz#5X0i{lY%f2CXgJDSPVm_ZODO6M zSXOW53h2kEpItRyrZi~JL10vs(MVbzS+1hdbCGbD#4?!!qogziEqt|7ZvMgnRV#a( zr&OI}8d&8OR`qD9ECDyRpEsFC*{e(D1x8*{-syu$AIrj4R>MSJ zxXu-|Z3>NS5~@3+7{tw`*#-rzD%S!+Hnj+U1V=xVx^OeNz-5FXAmx}CV(aynfwFPF zV3uy$>7(i8Z~AQ7T;gHt$4CRVnvbNQ@utD6YxGx99my0w+kor zV77kKx6Zqm-z)x(2OXl>W!gH(KJ?4wUKFc=B(*>Yz%AH^hplIRWB5NylT?Ll4V@U? z=GPSG19rLT89_aW-b}o&y(8G09)djfNb5v8Z=teYPRC?rUeM?tFyp4{rpNubcHftt z*LNM)o8{))Cgg}b(z3kpD6@mpiv2*+UKRT+HrFhV5#p>HAiwX`Pma$|p0FEKME+Ba zZ5#WW$DOywop!fV+x6=E4`DCI(oQpc8_maA{r6Ot9N%=EOrhs^?gXoph8WsXv#7Ac z1CWkfdk%a;?gWj6rsMd{v70?EWg=#fdDtj4{oSKQC493%XrB6Z+-Bz29?hwMWqtB|bTBLBybI^#_sXg>0%$w;+>^*FcHRUyFuK}qFjH|} zv`h?%0A+%@T`$WeVf}D>8!6YRzj~B@h5Gu`Bv*cLlIpFVmYPMvzjfOn8StHc{8JJt z&!2QLI8!Wi0MZ=mmlVo5b-AJy+t2ndI2?vDThV(bAc%JU=(>M6=>BDo14rq>LvwL* zu!7U{0AOXhIIy-pB_taQ7?GX?G+`6K{KfZzS2TJf*I(X86`rKA;RnHezO(12w3M&r zfz;Dbw%yAn=?zfF5r*k8oIe*b@aCp=%JCOqOlI!Hcv1%2m58hC&SZ!Ub3K98bROlo zje^IPOk-?)(0qL8d!Wh%A zIdx4bS!S<62$ij}ui8G`)ZI~l;4dOzQ0Fx3!tsF8O)A@(jsZELd5Rd$I-m7#yE{3L z;w_8$wT_eu<)ZTTwLPL!DXybAH%O@Z?E^E=Oj#TTGf0od=a1r2Egr^WhG7rUuj5j? zR8WuRZ~Cvuu2c8six1rnlr+SChuOD2=LZ;l(2JXVc*TsryZd)MFAe|B-UW>;d>1zP z2)tGUo%X$R9aO)=6JC3yxvjeq`3zFeZ*YAVYGY&L$8Qt~fT0JoW$p diff --git a/fig/file.eps b/fig/file.eps index 15d5963..5161419 100644 --- a/fig/file.eps +++ b/fig/file.eps @@ -1,7 +1,7 @@ %!PS-Adobe-2.0 EPSF-2.0 %%Title: file.eps %%Creator: gnuplot 4.6 patchlevel 0 -%%CreationDate: Mon Mar 24 19:12:58 2014 +%%CreationDate: Thu May 15 12:29:17 2014 %%DocumentFonts: (atend) %%BoundingBox: 50 50 320 239 %%EndComments @@ -432,7 +432,7 @@ SDict begin [ /Author (afanfakh) % /Producer (gnuplot) % /Keywords () - /CreationDate (Mon Mar 24 19:12:58 2014) + /CreationDate (Thu May 15 12:29:17 2014) /DOCINFO pdfmark end } ifelse @@ -454,120 +454,140 @@ LTb 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont 602 448 M ( 0.4) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 843 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont 602 843 M ( 0.5) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1239 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 0.6) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1634 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 0.7) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2029 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 0.8) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2425 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 0.9) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2820 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3216 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 1.1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3611 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 1.2) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 686 308 M ( 1) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -1578 448 M +1700 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 1.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -2470 448 M +2714 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 2) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -3363 448 M +3728 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 2.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -4255 448 M +4741 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 3) Cshow -1.000 UL -LTb -5147 448 M -0 63 V -0 3100 R -0 -63 V -0 -3240 R -( 3.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 1.000 UL @@ -579,103 +599,113 @@ LTb -4461 0 V Z stroke LCb setrgbcolor +/Helvetica findfont 200 scalefont setfont 112 2029 M currentpoint gsave translate -270 rotate 0 0 M (Normalized energy and performance inv.) Cshow grestore +/Helvetica findfont 140 scalefont setfont LTb LCb setrgbcolor +/Helvetica findfont 200 scalefont setfont 2916 98 M (Frequency scaling factors) Cshow +/Helvetica findfont 140 scalefont setfont LTb 1.000 UP -1578 646 M +/Helvetica findfont 200 scalefont setfont +1700 646 M (Optimal scaling factor) Lshow -864 3453 M +/Helvetica findfont 140 scalefont setfont +889 3453 M ( ) Lshow 1.000 UL LTb -1525 1342 M +1639 1342 M 0 585 V 0 4 R 0 -593 V stroke LT1 -1493 609 M +0.00 0.00 0.00 C 1607 609 M 32 -121 V 32 121 V -32 733 R 0 -854 V stroke gsave [] 0 setdash -1493 609 M +1607 609 M 32 -121 V 32 121 V stroke grestore 1.000 UL LT0 -1435 1931 M -188 0 V -1435 1342 M -188 0 V +1538 1931 M +213 0 V +1538 1342 M +213 0 V stroke LTb % Begin plot #1 1.000 UL LT0 0.00 0.00 1.00 C LCb setrgbcolor -4496 3478 M +/Helvetica findfont 190 scalefont setfont +4496 3443 M (Normalized performance inv.) Rshow +/Helvetica findfont 140 scalefont setfont LT0 -0.00 0.00 1.00 C 4580 3478 M +0.00 0.00 1.00 C 4580 3443 M 399 0 V 686 2820 M -74 -99 V -81 -102 V -88 -106 V -97 -109 V -106 -112 V -117 -116 V -131 -121 V -146 -124 V -164 -129 V -186 -134 V -212 -139 V -245 -143 V -286 -150 V -338 -155 V -3363 920 L -3858 752 L -4478 577 L +84 -99 V +92 -102 V +963 2513 L +109 -109 V +121 -112 V +133 -116 V +149 -121 V +165 -124 V +187 -129 V +211 -134 V +241 -139 V +279 -143 V +325 -150 V +384 -155 V +3728 920 L +4291 752 L +4995 577 L % End plot #1 % Begin plot #2 stroke LT1 1.00 0.00 0.00 C LCb setrgbcolor -4496 3338 M +/Helvetica findfont 190 scalefont setfont +4496 3233 M (Normalized energy) Rshow +/Helvetica findfont 140 scalefont setfont LT1 -1.00 0.00 0.00 C 4580 3338 M +1.00 0.00 0.00 C 4580 3233 M 399 0 V 686 2820 M -74 -293 V -81 -211 V -88 -198 V -97 -185 V -106 -170 V -117 -156 V -131 -141 V -146 -125 V -164 -107 V -186 -90 V -212 -68 V -245 -47 V -286 -20 V -338 9 V -406 45 V -495 90 V -620 146 V +84 -293 V +92 -211 V +963 2118 L +109 -185 V +121 -170 V +133 -156 V +149 -141 V +165 -125 V +187 -107 V +211 -90 V +241 -68 V +279 -47 V +325 -20 V +384 9 V +461 45 V +563 90 V +704 146 V % End plot #2 stroke LTb @@ -692,3 +722,5 @@ stroke grestore end showpage +%%Trailer +%%DocumentFonts: Helvetica diff --git a/fig/file.pdf b/fig/file.pdf index a1ea9235b8b82ed7f9fd6038d18ed997c775aa5d..e87756bb71d69dc64c3d93900ab979b1cdeff308 100644 GIT binary patch delta 1628 zcmexi@y23;Tm9T=zWLJT0&VYqYggp(CwlVPuV;;#e@kIwo+DqIja@!&FW0}X znYde;eXaR5M`M|ZpMR*f&vAde#=Ot&*Is+y{d;tu{mb-vcA90|EPJ)|b!r_6P94uo z96QZ6*3NO#T5yhYu0)BFl*DftR>ce{XXT%wONmNU zm8n?Imfk!)r#LP>J|;;nD3HN$YQx+{g?R=Ydj$_MtF^KoUbBC_Wf_y1HKV05W3|_T zZ(&S#>Zbn{N>p@ic=XKON0v3P+-~YCZlCXFk61XiU1R07v7VsVuqNi^!(B2zWS5?Q z`d>cS>tk1u=dNIyRrjCfRla`N z`i&*Wd)rC%cm2Mw)C3;AALo1ZBGZFMNl6_sQzp1Q`7!s(C+(;>vAUOD*YA{wwn*<# z<^7@TZo=O`xwht};=(DXZ#gKfh`3r=V%1&zY472MzkMbL&K8i04cz;=YinWcmuJ}@ zzlH8<-}Lw+M@X;Vtwkr=@3c?3={wL4FDF~l#mNtvJk-p%i=B}S|T{Gjebu}S9a=bud5B%`#Rf9d<6pV9w2VIJ3m9GcPYMQn!5(TX22X9nLMw z#SaDxnoPAhzHHL>Nt3+x{#w8K^Vu)@fl{+%Q}a@aONvqxbGd-5{G=={GX(<$5N&B< zxml4#SWv~%)X*~3EG1Fb(8Sn6*U&IEP1gVjbdypI%*`y5Ee#9}QYS}<77?fLov5`$ zNJgoGu5Y4lr9z0Isg;7Em64T^rInTWWJfV&+>$k7O~hHDB<`+8jHA`{Cku)y6J_S+ zYvKV+W*~R_CRQpKnkpEAoor~XU}RuuqRFN2o1fy6SdyyYVg(W~Ff}waG%_|cG}JaQ znj9lF(ay|R0TgHk3V8}#V1|L2nGrCkQN;|63`{V@3`~tDYfG!u8)Hft8(Cl|H!-rr z5VOFh*9=pgu_cC>ks(waKa$%`4NNR2-iC*#Gdh&6< zv}Ga}U&^WVr9Vo)w{h+lx9G=Pmd=ma&i?gN@U`&mTCb{Em%S)I^=dm8Q^xydJB>pEm1VuF2i4(~dpd$~R}%5zZY)8WfYA62+T0EKmIK zWwK(8ft?_$$YjYwx`LB*g_$IUnRZJw+VVCo4}UGtEhsss@<2e&oO)3+^}cs$liua( z+}%1$P5d!?U;G0P;l+Vd0=PYliuh{n+Dtzll+@spZ@za>O}A*J(4Jlui(dB+Y4rsk z%woi4o^F}LtNde4wk7}Gz|R2;vkW# zGK4*E25N3EtoJ<-syA(l<^AbeRp+jx`|uw6)8r!FW%FIDy(N4@V%dKFwfawtY<{~d zeTv;k1^L16L z_pG-#6|hpsY2{j>1>f>?t@936E8J|8tMN|SP{YIB+0SNZx;eDIar?YB&4*=jLVS~- zU-nrTYZ&Nq+iM<6yP_AT#spPKt$hKF*H-Fe%@)pn6|+cVC6lDj1evujorLY#tl!ru zUUWPV+HX0tlUZA!yesGlx9!gNs&_a(qzXLN@7(#SZNup&k44+gOj$Vj%$)W7soy3& zR5CW}_Dqpp?GZEiL~q2>4@oEIxav)tuRhhd!DG%QS)t`+RjQrI-(nweqzYM_+8LzM zH|bJrO0;^Xa={0+qBA~MQ~VZoo!MKmLg4xTUpANc?N`}*^{$CZ%}XgRDN0SuFSlK^4=~)WlSC14G@^WK&~ZL&MZmT>~J{H8D4}NVZJ2ut>2m zoa`f7M4ZA$qSg|=`FRSuzKOa;3L!=&RtknzmR5#FR#q01O~sUPOBRSV5od*%xI1wQ z8^u+LQn>k)cmR_y$ko1yMG8hH3Wi_@n_5iPl@jMMHZnCdH!w0Z&^9of93VB(&fHW1 z6hH8kiYQzASAMYhvW)Vq|LKX5nmL=w@i-Xk=*VX71={LAe*590(#*4u15@1J&{y7^|3!Rogk z`UDUEzIy)Q=PZl$7AJSJ_EpKIbA`>(RI>P`S)RNl zU}=V6W5((H^chbyjygG81RY+iZ`@*G+|nlEq%7(*-Cz8W^Tt5U2NM*T&hb1t!y^^M z)*HeWYZuEQAFFI}iLrsjv3`xfm$lrHWscphbEafO@s|WuJ+X6~$gnfM&r6a0`lj&J zsp7lS*Y~BZf7W?3c#EN2TKt@Q$sW88XWatk_AQuaw4zptWqL2~j)WQ62cCNfJDRF9 z6{a=4GHuG5?=@jl=|z1(F(U@Gmjy>>%wfB9r_bmomvN=~BZj84m5$BlxrFK&N}r#( zHS=#o`@%op*ZaG8esnBSjpniFmJ2@n`EA|)y}Ne*HIJKJX#URJ|5;XfZe8zhY0fv- z7AN>ju)N)4vu{D^yXk*_d`Q2ucK(^iv&?Vio;iE<*Z05wHg4W;v-WZjqms8x(dOL8 zA0KSL@!_hl_|3$Ri+=97DVp~!so28q%SF%nuTP_Y=DfTtp3dwrv5k3s{_LWa`MZCW zx62FOvu~{1%CYmRhnvXN?^BgF$QpbSTpjvk-T~=k707J=JIb*4>-^U~!;-x} zYhvHOSg@}9(ZaH2T5=lZdzk~STW{Iznt$rX*JTpvdo>@dF&Fen5^jskcABwN?MP*r z;LfF2L^8r=O?tL7B`}0f_u%H~yZ+bw;GePWm1b&QN^wb1s%T;^7m%5sl*MJHV4wh^ z&CLuqCo*5;Q?WEnGBQswP1H3sF}Bb(G)zs?H2?zL#5B_+Q*-0wM6)!*$zei8#3_6u zWGxYrQL3Qpo2Xl<5MpR*rC?}fWMyn%Wo0qhURW8oWR-9eaaJgZxT_K4Xm$O`{6fk^ znYsCjNFb9L$lbn)l?sNY3Wi`O8(JtB85o*qa_Rf#r?@1Rq-wZWfdmXp4NVP=j13Kq zv<-|VM~hFgGdEQL1)70Eo&pz`VPI-uW{56kXkZ8oepGb^7G|cCwI$T*O)#Vkfi|Hl zH#9QAG}g@62t%Er1*SS9LrgJCV{@=xek8Y>7+DxkzAs@DYvJbNXldbW=4fu>Y+>wd z;pl2$?Bwj`YV7LdY-wicXlFxEMJ$({9anKlVo^y&QED2OnSr?_m#V6(zZ(|-+PoGs delta 1403 zcmexh^uuU^SG};IK->F&qIQQEA4~Szf9qngZMYK3_Bnk1?Y))N6W_i&=<@qCOZXf2$8TqHis<}&vgq#Y2`lYf z?@IMpmh4Jld~yBs%cN$Tc!sOCOJINbUI|9ykUl*iYmvX zOpZ@gYq=-vxmKUVq-4n8Zg{G`$~gRE;5N?gIhj@RnZ6$x^h6StXV12|$+D9tYL9Ew zoX2r{I5qDbDm}Wod*4iZ@%S^c8_GJaOb=k3^@3#)i{qnGo`$XYTsjiZzP5*GT{yBJ zX_0_s@rp2^Q0+*eT?{)?B7ZbC?7Zyblo?Vs%ai-;l4}?QIR2QQ#!Ano1824?~-cA!R;Go|2)2N?wLQf|Grl|Q_9~FRo1Lpr0J&l<>k8_ z`*;2QRlR51<3yWxa&NCbyj3UPZ$8H&$9>rg?ug|!3F}U)rJrnnT=BoyCP)75;ca|H zH*U^e{kDJiV`umH!s37x$5ZtD_+L*xpR{NC+4}9m!G#klOC4s43)dNWM^+xczWmS6 zM`slHOiwMGdo-kN+l9mnhgZa>E|uyoDi$)IbN0w_)um-eRo58m=j-WuofPwbd!6%V z(6p&Zq3&IWUuh!80Q5MQ-4OU zNE5wnu~A>^1H7G2%vpWAyWisc$F6yPUnh8f^*qlzP4?>EzcJN)?Z1QmSFTkq=?{*o zI`eAham}Nx$2HH$srTz1);oT!<4w_$OSXL;uL`f+-*h~rsme#Qv|qFBWU|s6-em&m z^_mZI^jkOBn05Zi(RZD7XQ_^>-fHnjO>eesuK)UD=TGLQ`HYN8lWm!$87((QGGF9V zF-@~DHczoI)J;t`HP$sWOik4_00Lc;l(aO%MDsK=1EZwLenLgWDSRSiE#aG=r=aVb zs9U5EVq{{aU}$AxWnye)Wj5JdSQ)ouk#G}nR!E4rs}bX9b^XZ~g_VgibMqOIKqg_3 zyL}Uj6pTz148cw|Hk)i9F3w|YWNK(`U}R{hZD2S#SbUP5g|PxChzu0+6u7_)15*n# zOLQ?q1E4`@>I^K*jVCKhsMVWdNEsTMSfDF6G%~?7*385JL!F@oraB`-OfgFnQ?Oos zB)6LwSsG5hE@2aE>1=FZXyW2xU~Xh;>SSi*Y-C|*;O6S=V(RMXTSTmayq66yc| diff --git a/fig/lu_pre.eps b/fig/lu_pre.eps index 1368d03..fcd8b42 100644 --- a/fig/lu_pre.eps +++ b/fig/lu_pre.eps @@ -1,7 +1,7 @@ %!PS-Adobe-2.0 EPSF-2.0 %%Title: lu_pre.eps %%Creator: gnuplot 4.6 patchlevel 0 -%%CreationDate: Mon Mar 24 15:37:21 2014 +%%CreationDate: Thu May 15 12:23:20 2014 %%DocumentFonts: (atend) %%BoundingBox: 50 50 320 239 %%EndComments @@ -432,7 +432,7 @@ SDict begin [ /Author (afanfakh) % /Producer (gnuplot) % /Keywords () - /CreationDate (Mon Mar 24 15:37:21 2014) + /CreationDate (Thu May 15 12:23:20 2014) /DOCINFO pdfmark end } ifelse @@ -454,96 +454,110 @@ LTb 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont 602 781 M ( 1.2) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1280 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 1.5) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 1780 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 1.8) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2279 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 2.1) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 2779 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 2.4) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 3278 M 63 0 V 4398 0 R -63 0 V +/Helvetica findfont 200 scalefont setfont -4482 0 R ( 2.7) Rshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 686 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 686 308 M ( 1) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -1578 448 M +1700 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 1.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -2470 448 M +2714 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 2) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -3363 448 M +3728 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 2.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb -4255 448 M +4741 448 M 0 63 V 0 3100 R 0 -63 V +/Helvetica findfont 200 scalefont setfont 0 -3240 R ( 3) Cshow -1.000 UL -LTb -5147 448 M -0 63 V -0 3100 R -0 -63 V -0 -3240 R -( 3.5) Cshow +/Helvetica findfont 140 scalefont setfont 1.000 UL LTb 1.000 UL @@ -555,75 +569,87 @@ LTb -4461 0 V Z stroke LCb setrgbcolor +/Helvetica findfont 200 scalefont setfont 112 2029 M currentpoint gsave translate -270 rotate 0 0 M (Normalized time) Cshow grestore +/Helvetica findfont 140 scalefont setfont LTb LCb setrgbcolor +/Helvetica findfont 200 scalefont setfont 2916 98 M (Frequency scaling factors) Cshow +/Helvetica findfont 140 scalefont setfont LTb 1.000 UP -864 3361 M +/Helvetica findfont 200 scalefont setfont +889 3361 M (LU Class B) Lshow +/Helvetica findfont 140 scalefont setfont +889 714 M +( ) Lshow 1.000 UL LTb % Begin plot #1 1.000 UL LT0 0.00 0.00 1.00 C LCb setrgbcolor -4496 3478 M -(Normalized predicted perf.) Rshow +/Helvetica findfont 190 scalefont setfont +4496 3443 M +(Normalized predicted time) Rshow +/Helvetica findfont 140 scalefont setfont LT0 -0.00 0.00 1.00 C 4580 3478 M +0.00 0.00 1.00 C 4580 3443 M 399 0 V 686 448 M -74 52 V -81 56 V -88 61 V -97 67 V -106 73 V -117 82 V -131 90 V -146 101 V -164 114 V -186 129 V -212 147 V -245 170 V -286 198 V -338 234 V -406 281 V -495 344 V -620 430 V +84 52 V +92 56 V +101 61 V +109 67 V +121 73 V +133 82 V +149 90 V +165 101 V +187 114 V +211 129 V +241 147 V +279 170 V +325 198 V +384 234 V +461 281 V +563 344 V +704 430 V % End plot #1 % Begin plot #2 1.000 UP stroke LT1 1.00 0.00 0.00 C LCb setrgbcolor -4496 3338 M -(Normalized real perf.) Rshow +/Helvetica findfont 190 scalefont setfont +4496 3233 M +(Normalized real time) Rshow +/Helvetica findfont 140 scalefont setfont LT1 1.00 0.00 0.00 C 686 448 BoxF -760 491 BoxF -841 610 BoxF -929 613 BoxF -1026 729 BoxF -1132 765 BoxF -1249 922 BoxF -1380 1020 BoxF -1526 1060 BoxF -1690 1127 BoxF -1876 1340 BoxF -2088 1474 BoxF -2333 1673 BoxF -2619 1795 BoxF -2957 1993 BoxF -3363 2225 BoxF -3858 2508 BoxF -4478 2924 BoxF -4779 3338 BoxF +770 491 BoxF +862 610 BoxF +963 613 BoxF +1072 729 BoxF +1193 765 BoxF +1326 922 BoxF +1475 1020 BoxF +1640 1060 BoxF +1827 1127 BoxF +2038 1340 BoxF +2279 1474 BoxF +2558 1673 BoxF +2883 1795 BoxF +3267 1993 BoxF +3728 2225 BoxF +4291 2508 BoxF +4995 2924 BoxF +4779 3233 BoxF % End plot #2 1.000 UL LTb diff --git a/fig/lu_pre.pdf b/fig/lu_pre.pdf index 12823314faa98eb43ca2568ead3ec092663dd43d..68c80b2654833ebfc65e89ad590cbb6a50bd0697 100644 GIT binary patch delta 1484 zcmX?PzQufkTm9UrXZvou@f?3&)wOf0Wr>wc(Y|$@llCrQowS30L&2VI-8U-P`GF?> z@^v1Yw#cdkylHN84*WmggYWB$CU)^B_Ho~rKm7jv=-v3XC;wzGgnvISaOZ483t#ky zlXKa0j)ZUSbM|3P|9nL4=*@P$@NP3XNJN-L4sTS&AI_#(LPyK8NkZgVRgqhyB_b`!&_} z^=#*_x$$3ei+9(Z4gb6A68nK|D~@NJ6WJW?n6`eynRRb17#@6%nD8Qzx22Rd<(X(k zk@STpp$kra6?V9z7_Jb-oSG9}FJrjQ;F*!gXN`{B|150ZJ?Ckh>QZtO;)qpc{HFi?RYmpA9n(MAaMo*7*{iet?AtzP^3nH4^(Q~!H&#f_ zimXU@sx*1o&sp=-AD;gC;bY?eEgtP(o2TcmwtevTtU&T)V~K-O4J9Id-wG`qS8eI7 z@#oliZ0RYl)9nji?d9C;ESIq-EAm_S+vh!f2ae?)JQRK^WwQd?bcZnB6rE*C7X=!) zyKnY&)#`9{)EgbS6qC8jpXJxa%fXWamuEc)w3w^t@r_MTV34706^*0#HS-@C7F zUS6C2!or-(R_iX;+TD@6KC^A*0*UVaId*;`@d1;~%>LbNh_NX=ruO#p8P5qal|SG8zlzxS zrs*00fv#bynX#c+Qevv1x%uRIB1Ob0G!eCy2+1f_(DhB!tyBmxG__JNv@)_XGPbfZ znA|R^j9c=MXcKW(WQe(|5#wle{mF`A%0!vDSw=jFNf_jA-^5A*YssyNQLF;bcu|n^-qvS94<% zM@wfnHxn~6LjzN1BU3X|H**t912-oFS2sHwf+}LU?CiLTOA?DpDvDCmxXjEr{ zu{G}$=KAOo8H2P}4?7Px#47X0{y1@dr@S0{fc5o#-_*ZX9iMizK6U$UX&xq-DTWI6 zCcT!s?q>ctdaz@CSv`M#+2QI|&x7S6lH~%zB8Q)`9#XT<+v0EV?4)hn&goXWPw&`0 zd%-(1{dZ`9u`IAf87* z3X5fXrd^xvHx^~l344k!-eP?DZ0D7USIS_O0M`=sJ79 zjn|>sFJYqMjS6XIh3!+AxZ8G`J81Ek7;Z^ew4vZxMxk=%MYm}!%8YuEVm}=Y)ZfWm zw&3EH4VQg3Ui7(WCiSIMXUiF$vpQkSy-Qhzy;t|2yY%STrA56u6E9bm`n-%37Zq{& z^ee~r-_Hf#TWZeP@6kE(B)sr+a;atSguSL`?Dx;{-+sLPKi~X2mk+;Gm{+^|2dj44 zJ(bmWwm(1o*uMSz@BDY0o~zD(fA{X&e=kq}yj1UKekZEY?WK0%t}Xw5oc>$4XK&1% z*GDhcy}N6b|KM8FCEaCln=5v%e|~s*``^hue9w0m-g_CSmtTFLeKzC7_wpyM{c0+U z>Fqlm7HX3}BPPPJ>vW=?kzdv2-;3H^%qH(B{-mMmeltz@R6bKgxPMOS>K(gPiUds?UyPx}y){q%%)feSWe!ZfVq1$JztTI$0u5ay(-uT1zP2)O4Fv z=f?K!v#QL_3<=g5zg_#zx8Hue`OBd#?MGBktrf7e|CH`JVa41>yBmLGDkV*Q{K1FE z?u5>#v^_DBPJt8uTdZBP=9TBS+386=aYv_w9eGmf$x(RJICUv+yy=<_%N6M@rW+UR zbAEDv&zChFO3QX^6SB#AWOQg2SK$5q52B`pEQFlTeheLhZ*ECZ&8X)yjQdS?_T+T?DHMNx+YI#mSQrs+`N`~zkrHq znuW1>iiM$WYO<-ZuAyOSs;&VL=$cs?r5IW!nx-Thm`$D_Qbe3WHBoB`-~2oUUEf6A zB83nm6DtKnD^n|Db1N&u$(5qYxFxrUHW6nSQyV&*1B7-B}|nCgrT zFvN@u%*?@h`H|diVqtDNSz6jA*2vY>(bdV>#MRu<+}zB_!r8*q!r9Tu($&Ds&Be^c j(awgTidZf?JFeoA#G;alqSQ1lGc$8zE>%@me>W}w5GqVn diff --git a/paper.tex b/paper.tex index 508682e..2838f2d 100644 --- a/paper.tex +++ b/paper.tex @@ -28,7 +28,7 @@ \begin{document} \title{Dynamic Frequency Scaling for Energy Consumption - Reduction in Distributed MPI Programs} + Reduction in Synchronous Distributed Applications} \author{% \IEEEauthorblockN{% @@ -53,12 +53,10 @@ \begin{abstract} Dynamic Voltage Frequency Scaling (DVFS) can be applied to modern CPUs. This technique is usually used to reduce the energy consumed by a CPU while - computing. Indeed, power consumption by a processor at a given time is - exponentially related to its frequency. Thus, decreasing the frequency + computing. Thus, decreasing the frequency reduces the power consumed by the CPU. However, it can also significantly affect the performance of the executed program if it is compute bound and if a - low CPU frequency is selected. The performance degradation ratio can even be - higher than the saved energy ratio. Therefore, the chosen scaling factor must + low CPU frequency is selected. Therefore, the chosen scaling factor must give the best possible trade-off between energy reduction and performance. In this paper we present an algorithm that predicts the energy consumed with @@ -108,21 +106,18 @@ frequency scaling factor that simultaneously takes into consideration the energy consumption by the CPU and the performance of the application. The main objective of HPC systems is to execute as fast as possible the application. Therefore, our algorithm selects the scaling factor online with very small -footprint. The proposed algorithm takes into account the communication times of +overhead. The proposed algorithm takes into account the communication times of the MPI program to choose the scaling factor. This algorithm has the ability to predict both energy consumption and execution time over all available scaling factors. The prediction achieved depends on some computing time information, -gathered at the beginning of the runtime. We apply this algorithm to seven MPI -benchmarks. These MPI programs are the NAS parallel benchmarks (NPB v3.3) -developed by NASA~\cite{44}. Our experiments are executed using the simulator +gathered at the beginning of the runtime. We apply this algorithm to the NAS parallel benchmarks (NPB v3.3)~\cite{44}. Our experiments are executed using the simulator SimGrid/SMPI v3.10~\cite{Casanova:2008:SGF:1397760.1398183} over an homogeneous distributed memory architecture. Furthermore, we compare the proposed algorithm with Rauber and Rünger methods~\cite{3}. The comparison's results show that our algorithm gives better energy-time trade-off. This paper is organized as follows: Section~\ref{sec.relwork} presents some -related works from other authors. Section~\ref{sec.exe} explains the execution -of parallel tasks and the sources of slack times. It also presents an energy +related works from other authors. Section~\ref{sec.exe} presents an energy model for homogeneous platforms. Section~\ref{sec.mpip} describes how the performance of MPI programs can be predicted. Section~\ref{sec.compet} presents the energy-performance objective function that maximizes the reduction of energy @@ -162,11 +157,7 @@ part of it, is usually executed over all the available frequency gears and the execution time and the energy consumed with each frequency gear are measured. Then a heuristic or an exact method uses the retrieved information to compute the values of the scaling factor for the processors. -In~\cite{29}, Xie et al. use an exact exponential breadth-first search algorithm -to compute the scaling factor values that give the optimal energy reduction -while respecting a deadline for a sequential program. They also present a -linear heuristic that approximates the optimal solution. In~\cite{8} , Rountree -et al. use a linear programming algorithm, while in~\cite{38,34}, Cochran et +In~\cite{8} , Rountree et al. use a linear programming algorithm, while in~\cite{34}, Cochran et al. use a multi-logistic regression algorithm for the same goal. The main drawback of these methods is that they all require executing the whole program or, a part of it, on all frequency gears for each new instance of the same program. @@ -180,84 +171,40 @@ a lot of information is measured such as the execution time, the energy consumed using a multimeter, the slack times, \dots{} Then a method will exploit these measurements to compute the scaling factor values for each processor. This operation, measurements and computing new scaling factor, can be repeated as -much as needed if the iterations are not regular. Kimura, Peraza, Yu-Liang et -al.~\cite{11,2,31} used varied heuristics to select the appropriate scaling +much as needed if the iterations are not regular. Peraza, Yu-Liang et +al.~\cite{2,31} used varied heuristics to select the appropriate scaling factor values to eliminate the slack times during runtime. However, as seen -in~\cite{39,19}, machine learning methods can take a lot of time to converge +in~\cite{19}, machine learning method takes a lot of time to converge when the number of available gears is big. To reduce the impact of slack times, in~\cite{1}, Lim et al. developed an algorithm that detects the communication sections and changes the frequency during these sections only. This approach might change the frequency of each processor many times per iteration if an iteration contains more than one communication section. In~\cite{3}, Rauber and -Rünger used an analytical model that can predict the consumed energy and the -execution time for every frequency gear after measuring the consumed energy and -the execution time with the highest frequency gear. These predictions may be -used to choose the optimal gear for each processor executing the parallel -program to reduce energy consumption. To maintain the performance of the -parallel program , they set the processor with the biggest load to the highest -gear and then compute the scaling factor values for the rest of the processors. -Although this model was built for parallel architectures, it can be adapted to -distributed architectures by taking into account the communications. The -primary contribution of our paper is to present a new online scaling factor -selection method which has the following characteristics: -\begin{enumerate} -\item It is based on Rauber and Rünger analytical model to predict the energy - consumption of the application with different frequency gears. -\item It selects the frequency scaling factor for simultaneously optimizing - energy reduction and maintaining performance. -\item It is well adapted to distributed architectures because it takes into - account the communication time. -\item It is well adapted to distributed applications with imbalanced tasks. -\item It has a very small footprint when compared to other methods +Rünger used an analytical model that can predict the consumed energy for every frequency gear after measuring the consumed energy. They +maintain the performance as mush as possible by setting the highest frequency gear to the slowest task. + +The primary contribution of +our paper is to present a new online scaling factor selection method which has the + following characteristics:\\ +1) It is based on Rauber and Rünger analytical model to predict the energy + consumption of the application with different frequency gears. +2) It selects the frequency scaling factor for simultaneously optimizing + energy reduction and maintaining performance. +3) It is well adapted to distributed architectures because it takes into + account the communication time. +4) It is well adapted to distributed applications with imbalanced tasks. +5) It has a very small overhead when compared to other methods (e.g.,~\cite{19}) and does not require profiling or training as - in~\cite{38,34}. -\end{enumerate} + in~\cite{34}. -\section{Execution and energy of parallel tasks on homogeneous platform} -\label{sec.exe} - % \JC{The whole subsection ``Parallel Tasks Execution on Homogeneous Platform'', % can be deleted if we need space, we can just say we are interested in this % paper in homogeneous clusters} -\subsection{Parallel tasks execution on homogeneous platform} - -A homogeneous cluster consists in identical nodes in terms of hardware and -software. Each node has its own memory and at least one processor which can be -a multi-core. The nodes are connected via a high bandwidth network. Tasks -executed on this model can be either synchronous or asynchronous. In this paper -we consider execution of the synchronous tasks on distributed homogeneous -platform. These tasks can exchange the data via synchronous message passing. -\begin{figure*}[t] - \centering - \subfloat[Sync. imbalanced communications]{% - \includegraphics[scale=0.67]{fig/commtasks}\label{fig:h1}} - \subfloat[Sync. imbalanced computations]{% - \includegraphics[scale=0.67]{fig/compt}\label{fig:h2}} - \caption{Parallel tasks on homogeneous platform} - \label{fig:homo} -\end{figure*} -Therefore, the execution time of a task consists in the computation time and the -communication time. Moreover, the synchronous communications between tasks can -lead to slack times while tasks wait at the synchronization barrier for other -tasks to finish their tasks (see figure~(\ref{fig:h1})). The imbalanced -communications happen when nodes have to send/receive different amounts of data -or they communicate with different numbers of nodes. Other sources of slack -times are imbalanced computations. This happens when processing different -amounts of data on each processor (see figure~(\ref{fig:h2})). In this case the -fastest tasks have to wait at the synchronization barrier for the slowest ones -to begin the next task. In both cases the overall execution time of the program -is the execution time of the slowest task as in EQ~(\ref{eq:T1}). -\begin{equation} - \label{eq:T1} - \textit{Program Time} = \max_{i=1,2,\dots,N} T_i -\end{equation} -where $T_i$ is the execution time of task $i$ and all the tasks are executed -concurrently on different processors. - -\subsection{Energy model for homogeneous platform} +\section{Energy model for homogeneous platform} +\label{sec.exe} Many researchers~\cite{9,3,15,26} divide the power consumed by a processor into two power metrics: the static and the dynamic power. While the first one is consumed as long as the computing unit is on, the latter is only consumed during @@ -299,8 +246,7 @@ ratio between the maximum and the new frequency as in EQ~(\ref{eq:s}). \end{equation} The value of the scaling factor $S$ is greater than 1 when changing the frequency of the CPU to any new frequency value~(\emph{P-state}) in the -governor. The CPU governor is an interface driver supplied by the operating -system's kernel to lower a core's frequency. This factor reduces quadratically +governor. This factor reduces quadratically the dynamic power which may cause degradation in performance and thus, the increase of the static energy because the execution time is increased~\cite{36}. If the tasks are sorted according to their execution times before scaling in a @@ -315,7 +261,7 @@ function of the scaling factor $S$, as in EQ~(\ref{eq:energy}). P_\textit{static} \cdot T_1 \cdot S_1 \cdot N \hfill \end{equation} -where $N$ is the number of parallel nodes, $T_i$ and $S_i$ for $i=1,\dots,N$ are +where $N$ is the number of parallel nodes, $T_i$ for $i=1,\dots,N$ are the execution times and scaling factors of the sorted tasks. Therefore, $T1$ is the time of the slowest task, and $S_1$ its scaling factor which should be the highest because they are proportional to the time values $T_i$. The scaling @@ -341,7 +287,7 @@ EQ~(\ref{eq:sopt}). \label{sec.mpip} The performance (execution time) of parallel synchronous MPI applications depends -on the time of the slowest task as in figure~(\ref{fig:homo}). If there is no +on the time of the slowest task. If there is no communication and the application is not data bounded, the execution time of a parallel program is linearly proportional to the operational frequency and any DVFS operation for energy reduction increases the execution time of the parallel @@ -393,7 +339,7 @@ In the same way we can normalize the performance as follows: The second problem is that the optimization operation for both energy and performance is not in the same direction. In other words, the normalized energy and the performance curves are not at the same direction see -figure~(\ref{fig:r2}). While the main goal is to optimize the energy and +Figure~(\ref{fig:r2}). While the main goal is to optimize the energy and performance in the same time. According to the equations~(\ref{eq:enorm}) and~(\ref{eq:pnorm}), the scaling factor $S$ reduce both the energy and the performance simultaneously. But the main objective is to produce maximum energy @@ -411,21 +357,21 @@ performance as follows: T_\textit{Max Comm Old}}{T_\textit{Max Comp Old} \cdot S + T_\textit{Max Comm Old}} \end{equation} -\begin{figure*} +\begin{figure} \centering \subfloat[Converted relation.]{% - \includegraphics[width=.4\textwidth]{fig/file}\label{fig:r1}}% - \qquad% + \includegraphics[width=.24\textwidth]{fig/file}\label{fig:r1}}% +% \quad% \subfloat[Real relation.]{% - \includegraphics[width=.4\textwidth]{fig/file3}\label{fig:r2}} + \includegraphics[width=.24\textwidth]{fig/file3}\label{fig:r2}} \label{fig:rel} \caption{The energy and performance relation} -\end{figure*} -Then, we can modelize our objective function as finding the maximum distance +\end{figure} +Then, we can model our objective function as finding the maximum distance between the energy curve EQ~(\ref{eq:enorm}) and the inverse of performance curve EQ~(\ref{eq:pnorm_en}) over all available scaling factors. This represents the minimum energy consumption with minimum execution time (better -performance) at the same time, see figure~(\ref{fig:r1}). Then our objective +performance) at the same time, see Figure~(\ref{fig:r1}). Then our objective function has the following form: \begin{equation} \label{eq:max} @@ -475,35 +421,34 @@ objective function described above. \EndFor \State Return $S_{opt}$ \end{algorithmic} -\end{algorithm} +\end{algorithm} The proposed algorithm works online during the execution time of the MPI program. It selects the optimal scaling factor after gathering the computation and communication times from the program after one iteration. Then the program changes the new frequencies of the CPUs according to the computed scaling -factors. This algorithm has a small execution time: for a homogeneous cluster -composed of nodes having the characteristics presented in -table~\ref{table:platform}, it takes \np[ms]{0.00152} on average for 4 nodes and -\np[ms]{0.00665} on average for 32 nodes. The algorithm complexity is $O(F\cdot +factors. In our experiments over a homogeneous cluster described in section~\ref{sec.expe}, +this algorithm has a small execution time. It takes \np[$\mu$s]{1.52} on average for 4 nodes and +\np[$\mu$s]{6.65} on average for 32 nodes. The algorithm complexity is $O(F\cdot N)$, where $F$ is the number of available frequencies and $N$ is the number of computing nodes. The algorithm is called just once during the execution of the program. The DVFS algorithm~(\ref{dvfs}) shows where and when the algorithm is called in the MPI program. -\begin{table}[htb] - \caption{Platform file parameters} - % title of Table - \centering - \begin{tabular}{|*{7}{l|}} - \hline - Max & Min & Backbone & Backbone & Link & Link & Sharing \\ - Freq. & Freq. & Bandwidth & Latency & Bandwidth & Latency & Policy \\ - \hline - \np{2.5} & \np{800} & \np[GBps]{2.25} & \np[$\mu$s]{0.5} & \np[GBps]{1} & \np[$\mu$s]{50} & Full \\ - GHz & MHz & & & & & Duplex \\ - \hline - \end{tabular} - \label{table:platform} -\end{table} +%\begin{table}[htb] +% \caption{Platform file parameters} +% % title of Table +% \centering +% \begin{tabular}{|*{7}{l|}} +% \hline +% Max & Min & Backbone & Backbone & Link & Link & Sharing \\ +% Freq. & Freq. & Bandwidth & Latency & Bandwidth & Latency & Policy \\ +% \hline +% \np{2.5} & \np{800} & \np[GBps]{2.25} & \np[$\mu$s]{0.5} & \np[GBps]{1} & \np[$\mu$s]{50} & Full \\ +% GHz & MHz & & & & & Duplex \\ +% \hline +% \end{tabular} +% \label{table:platform} +%\end{table} \begin{algorithm}[tp] \caption{DVFS} @@ -517,7 +462,7 @@ called in the MPI program. communication from each node. \State Call algorithm~\ref{EPSA} with these times. \State Compute the new frequency from the\newline\hspace*{3em}% - returned optimal scaling factor. + returned optimal scaling factor. \State Set the new frequency to the CPU. \EndIf \EndFor @@ -539,12 +484,11 @@ CPU to the nodes' workloads to maintain the performance of the program. \section{Experimental results} \label{sec.expe} Our experiments are executed on the simulator SimGrid/SMPI v3.10. We configure -the simulator to use a homogeneous cluster with one core per node. The detailed -characteristics of our platform file are shown in -table~(\ref{table:platform}). Each node in the cluster has 18 frequency values +the simulator to use a homogeneous cluster with one core per node. +%The detailed characteristics of our platform file are shown in Table~(\ref{table:platform}). +Each node in the cluster has 18 frequency values from \np[GHz]{2.5} to \np[MHz]{800} with \np[MHz]{100} difference between each -two successive frequencies. The simulated network link is \np[GB]{1} Ethernet -(TCP/IP). The backbone of the cluster simulates a high performance switch. +two successive frequencies. The nodes are connected via an ethernet network with 1Gbit/s bandwidth. \subsection{Performance prediction verification} @@ -555,26 +499,25 @@ the predicted execution time. Each program runs offline with all available scaling factors on 8 or 9 nodes (depending on the benchmark) to produce real execution time values. These scaling factors are computed by dividing the maximum frequency by the new one see EQ~(\ref{eq:s}). -\begin{figure*}[t] +\begin{figure} \centering - \includegraphics[width=.328\textwidth]{fig/cg_per}\hfill% - \includegraphics[width=.328\textwidth]{fig/mg_pre}\hfill% + \includegraphics[width=.24\textwidth]{fig/cg_per}\hfill% + % \includegraphics[width=.328\textwidth]{fig/mg_pre}\hfill% % \includegraphics[width=.4\textwidth]{fig/bt_pre}\qquad% - \includegraphics[width=.328\textwidth]{fig/lu_pre}\hfill% - \caption{Comparing predicted to real execution time} + \includegraphics[width=.24\textwidth]{fig/lu_pre}\hfill% + \caption{Comparing predicted to real execution times} \label{fig:pred} -\end{figure*} +\end{figure} %see Figure~\ref{fig:pred} In our cluster there are 18 available frequency states for each processor. This leads to 18 run states for each program. We use seven MPI programs of the NAS parallel benchmarks: CG, MG, EP, FT, BT, LU and SP. Figure~(\ref{fig:pred}) -presents plots of the real execution times and the simulated ones. The maximum +presents plots of the real execution times and the simulated ones. The maximum normalized error between these two execution times varies between \np{0.0073} to \np{0.031} dependent on the executed benchmark. The smallest prediction error was for CG and the worst one was for LU. \subsection{The experimental results for the scaling algorithm } - The proposed algorithm was applied to seven MPI programs of the NAS benchmarks (EP, CG, MG, FT, BT, LU and SP) which were run with three classes (A, B and C). For each instance the benchmarks were executed on a number of processors @@ -595,57 +538,31 @@ performance curve is very close to the energy curve. Then the distance between the two curves is very small. This leads to small energy savings. The opposite happens when there are a lot of communication, the distance between the two curves is big. This leads to more energy savings (e.g. CG and FT), see -table~(\ref{table:factors results}). All discovered frequency scaling factors +Table~(\ref{table:compareC}). All discovered frequency scaling factors optimize both the energy and the performance simultaneously for all NAS -benchmarks. In table~(\ref{table:factors results}), we record all optimal +benchmarks. In Table~(\ref{table:compareC}), we record all optimal scaling factors results for each benchmark running class C. These scaling factors give the maximum energy saving percentage and the minimum performance degradation percentage at the same time from all available scaling factors. \begin{figure*}[t] \centering - \includegraphics[width=.328\textwidth]{fig/ep}\hfill% - \includegraphics[width=.328\textwidth]{fig/cg}\hfill% - \includegraphics[width=.328\textwidth]{fig/sp} - \includegraphics[width=.328\textwidth]{fig/lu}\hfill% - \includegraphics[width=.328\textwidth]{fig/bt}\hfill% - \includegraphics[width=.328\textwidth]{fig/ft} + \includegraphics[width=.33\textwidth]{fig/ep}\hfill% + \includegraphics[width=.33\textwidth]{fig/cg}\hfill% + % \includegraphics[width=.328\textwidth]{fig/sp} + % \includegraphics[width=.328\textwidth]{fig/lu}\hfill% + \includegraphics[width=.33\textwidth]{fig/bt}\hfill% + % \includegraphics[width=.328\textwidth]{fig/ft} \caption{Optimal scaling factors for the predicted energy and performance of NAS benchmarks} \label{fig:nas} \end{figure*} -\begin{table}[htb] - \caption{The scaling factors results} - % title of Table - \centering - \begin{tabular}{|l|*{4}{r|}} - \hline - Program & Optimal & Energy & Performance & Energy-Perf. \\ - Name & Scaling Factor & Saving \% & Degradation \% & Distance \\ - \hline - CG & 1.56 & 39.23 & 14.88 & 24.35 \\ - \hline - MG & 1.47 & 34.97 & 21.70 & 13.27 \\ - \hline - EP & 1.04 & 22.14 & 20.73 & 1.41 \\ - \hline - LU & 1.38 & 35.83 & 22.49 & 13.34 \\ - \hline - BT & 1.31 & 29.60 & 21.28 & 8.32 \\ - \hline - SP & 1.38 & 33.48 & 21.36 & 12.12 \\ - \hline - FT & 1.47 & 34.72 & 19.00 & 15.72 \\ - \hline - \end{tabular} - \label{table:factors results} - % is used to refer this table in the text -\end{table} -As shown in table~(\ref{table:factors results}), when the optimal scaling +As shown in Table~(\ref{table:compareC}), when the optimal scaling factor has a big value we can gain more energy savings as in CG and -FT benchmarks. The opposite happens when the optimal scaling factor has a small value as in BT and EP benchmarks. Our algorithm selects a big scaling factor value when the -communication and the other slacks times are big and smaller ones in opposite -cases. In EP there are no communication inside the iterations. This leads our -algorithm to select smaller scaling factor values (inducing smaller energy +FT benchmarks. The opposite happens when the optimal scaling factor has a +small value as in BT and EP benchmarks. Our algorithm selects a big scaling +factor value when the communication and the other slacks times are big and smaller +ones in opposite cases. In EP there are no communication inside the iterations. +This leads our algorithm to select smaller scaling factor values (inducing smaller energy savings). \subsection{Results comparison} @@ -658,101 +575,17 @@ is similar to the first except setting the slower task to the maximum frequency (when the scale $S=1$) to keep the performance from degradation as mush as possible. We refer to this scenario as $R_{E-P}$. While we refer to our algorithm as EPSA (Energy to Performance Scaling Algorithm). The comparison is -made in tables \ref{table:compareA}, \ref{table:compareB}, -and~\ref{table:compareC}. These tables show the results of our method and -Rauber and Rünger scenarios for all the NAS benchmarks programs for classes A, B -and C. -\begin{table}[p] - \caption{Comparing results for the NAS class A} - % title of Table - \centering - \begin{tabular}{|l|l|*{4}{r|}} - \hline - Method & Program & Factor & Energy & Performance & Energy-Perf. \\ - Name & Name & Value & Saving \% & Degradation \% & Distance \\ - \hline - % \rowcolor[gray]{0.85} - $EPSA$ & CG & 1.56 & 37.02 & 13.88 & 23.14 \\ \hline - $R_{E-P}$ & CG & 2.14 & 42.77 & 25.27 & 17.50 \\ \hline - $R_{E}$ & CG & 2.14 & 42.77 & 26.46 & 16.31 \\ \hline - - $EPSA$ & MG & 1.47 & 27.66 & 16.82 & 10.84 \\ \hline - $R_{E-P}$ & MG & 2.14 & 34.45 & 31.84 & 2.61 \\ \hline - $R_{E}$ & MG & 2.14 & 34.48 & 33.65 & 0.80 \\ \hline +made in Table ~\ref{table:compareC}. This table shows the results of our method and +Rauber and Rünger scenarios for all the NAS benchmarks programs for class C. - $EPSA$ & EP & 1.19 & 25.32 & 20.79 & 4.53 \\ \hline - $R_{E-P}$ & EP & 2.05 & 41.45 & 55.67 & -14.22 \\ \hline - $R_{E}$ & EP & 2.05 & 42.09 & 57.59 & -15.50 \\ \hline - - $EPSA$ & LU & 1.56 & 39.55 & 19.38 & 20.17 \\ \hline - $R_{E-P}$ & LU & 2.14 & 45.62 & 27.00 & 18.62 \\ \hline - $R_{E}$ & LU & 2.14 & 45.66 & 33.01 & 12.65 \\ \hline - - $EPSA$ & BT & 1.31 & 29.60 & 20.53 & 9.07 \\ \hline - $R_{E-P}$ & BT & 2.10 & 45.53 & 49.63 & -4.10 \\ \hline - $R_{E}$ & BT & 2.10 & 43.93 & 52.86 & -8.93 \\ \hline - - $EPSA$ & SP & 1.38 & 33.51 & 15.65 & 17.86 \\ \hline - $R_{E-P}$ & SP & 2.11 & 45.62 & 42.52 & 3.10 \\ \hline - $R_{E}$ & SP & 2.11 & 45.78 & 43.09 & 2.69 \\ \hline - - $EPSA$ & FT & 1.25 & 25.00 & 10.80 & 14.20 \\ \hline - $R_{E-P}$ & FT & 2.10 & 39.29 & 34.30 & 4.99 \\ \hline - $R_{E}$ & FT & 2.10 & 37.56 & 38.21 & -0.65 \\ \hline - \end{tabular} - \label{table:compareA} - % is used to refer this table in the text -\end{table} -\begin{table}[p] - \caption{Comparing results for the NAS class B} - % title of Table - \centering - \begin{tabular}{|l|l|*{4}{r|}} - \hline - Method & Program & Factor & Energy & Performance & Energy-Perf. \\ - Name & Name & Value & Saving \% & Degradation \% & Distance \\ - \hline - % \rowcolor[gray]{0.85} - $EPSA$ & CG & 1.66 & 39.23 & 16.63 & 22.60 \\ \hline - $R_{E-P}$ & CG & 2.15 & 45.34 & 27.60 & 17.74 \\ \hline - $R_{E}$ & CG & 2.15 & 45.34 & 28.88 & 16.46 \\ \hline - - $EPSA$ & MG & 1.47 & 34.98 & 18.35 & 16.63 \\ \hline - $R_{E-P}$ & MG & 2.14 & 43.55 & 36.42 & 7.13 \\ \hline - $R_{E}$ & MG & 2.14 & 43.56 & 37.07 & 6.49 \\ \hline - - $EPSA$ & EP & 1.08 & 20.29 & 17.15 & 3.14 \\ \hline - $R_{E-P}$ & EP & 2.00 & 42.38 & 56.88 & -14.50 \\ \hline - $R_{E}$ & EP & 2.00 & 39.73 & 59.94 & -20.21 \\ \hline - - $EPSA$ & LU & 1.47 & 38.57 & 21.34 & 17.23 \\ \hline - $R_{E-P}$ & LU & 2.10 & 43.62 & 36.51 & 7.11 \\ \hline - $R_{E}$ & LU & 2.10 & 43.61 & 38.54 & 5.07 \\ \hline - - $EPSA$ & BT & 1.31 & 29.59 & 20.88 & 8.71 \\ \hline - $R_{E-P}$ & BT & 2.10 & 44.53 & 53.05 & -8.52 \\ \hline - $R_{E}$ & BT & 2.10 & 42.93 & 52.80 & -9.87 \\ \hline - - $EPSA$ & SP & 1.38 & 33.44 & 19.24 & 14.20 \\ \hline - $R_{E-P}$ & SP & 2.15 & 45.69 & 43.20 & 2.49 \\ \hline - $R_{E}$ & SP & 2.15 & 45.41 & 44.47 & 0.94 \\ \hline - - $EPSA$ & FT & 1.38 & 34.40 & 14.57 & 19.83 \\ \hline - $R_{E-P}$ & FT & 2.13 & 42.98 & 37.35 & 5.63 \\ \hline - $R_{E}$ & FT & 2.13 & 43.04 & 37.90 & 5.14 \\ \hline - \end{tabular} - \label{table:compareB} - % is used to refer this table in the text -\end{table} - -\begin{table}[p] +\begin{table} \caption{Comparing results for the NAS class C} % title of Table \centering \begin{tabular}{|l|l|*{4}{r|}} \hline Method & Program & Factor & Energy & Performance & Energy-Perf. \\ - Name & Name & Value & Saving \% & Degradation \% & Distance \\ + Name & Name & Value & Saving \% & Degradation \% & Distance \\ \hline % \rowcolor[gray]{0.85} $EPSA$ & CG & 1.56 & 39.23 & 14.88 & 24.35 \\ \hline @@ -786,28 +619,28 @@ and C. \label{table:compareC} % is used to refer this table in the text \end{table} -As shown in tables~\ref{table:compareA},~\ref{table:compareB} -and~\ref{table:compareC}, the ($R_{E-P}$) method outperforms the ($R_{E}$) +As shown in Table~\ref{table:compareC}, the ($R_{E-P}$) method outperforms the ($R_{E}$) method in terms of performance and energy reduction. The ($R_{E-P}$) method also gives better energy savings than our method. However, although our scaling -factor is not optimal for energy reduction, the results in these tables prove +factor is not optimal for energy reduction, the results in this table prove that our algorithm returns the best scaling factor that satisfy our objective method: the largest distance between energy reduction and performance -degradation. Figure~\ref{fig:compare} illustrates even better the distance between the energy reduction and performance degradation. The negative values mean that one of +degradation. Figure~\ref{fig:compare} illustrates even better the distance between +the energy reduction and performance degradation. The negative values mean that one of the two objectives (energy or performance) have been degraded more than the other. The positive trade-offs with the highest values lead to maximum energy savings while keeping the performance degradation as low as possible. Our algorithm always gives the highest positive energy to performance trade-offs while Rauber and Rünger's method, ($R_{E-P}$), gives sometimes negative trade-offs such as in BT and EP. -\begin{figure*}[t] +\begin{figure}[t] \centering - \includegraphics[width=.328\textwidth]{fig/compare_class_A} - \includegraphics[width=.328\textwidth]{fig/compare_class_B} - \includegraphics[width=.328\textwidth]{fig/compare_class_C} +% \includegraphics[width=.328\textwidth]{fig/compare_class_A} +% \includegraphics[width=.328\textwidth]{fig/compare_class_B} + \includegraphics[width=.49\textwidth]{fig/compare_class_C} \caption{Comparing our method to Rauber and Rünger's methods} \label{fig:compare} -\end{figure*} +\end{figure} \section{Conclusion} \label{sec.concl} @@ -823,7 +656,7 @@ the energy and the inverted performance curves. To evaluate this method, we have applied it to the NAS benchmarks and it was compared to Rauber and Rünger methods while being executed on the simulator SimGrid. The results showed that our method, outperforms Rauber and Rünger's methods in terms of energy-performance -ratio. +ratio. In the near future, we would like to adapt this scaling factor selection method to heterogeneous platforms where each node has different characteristics. In -- 2.39.5