PNG  IHDRX cHRMz&u0`:pQ<bKGD pHYsodtIME MeqIDATxw]Wug^Qd˶ 6`!N:!@xI~)%7%@Bh&`lnjVF29gΨ4E$|>cɚ{gk= %,a KX%,a KX%,a KX%,a KX%,a KX%,a KX%, b` ǟzeאfp]<!SJmɤY޲ڿ,%c ~ع9VH.!Ͳz&QynֺTkRR.BLHi٪:l;@(!MԴ=žI,:o&N'Kù\vRmJ雵֫AWic H@" !: Cé||]k-Ha oݜ:y F())u]aG7*JV@J415p=sZH!=!DRʯvɱh~V\}v/GKY$n]"X"}t@ xS76^[bw4dsce)2dU0 CkMa-U5tvLƀ~mlMwfGE/-]7XAƟ`׮g ewxwC4\[~7@O-Q( a*XGƒ{ ՟}$_y3tĐƤatgvێi|K=uVyrŲlLӪuܿzwk$m87k( `múcE)"@rK( z4$D; 2kW=Xb$V[Ru819קR~qloѱDyįݎ*mxw]y5e4K@ЃI0A D@"BDk_)N\8͜9dz"fK0zɿvM /.:2O{ Nb=M=7>??Zuo32 DLD@D| &+֎C #B8ַ`bOb $D#ͮҪtx]%`ES`Ru[=¾!@Od37LJ0!OIR4m]GZRJu$‡c=%~s@6SKy?CeIh:[vR@Lh | (BhAMy=݃  G"'wzn޺~8ԽSh ~T*A:xR[ܹ?X[uKL_=fDȊ؂p0}7=D$Ekq!/t.*2ʼnDbŞ}DijYaȲ(""6HA;:LzxQ‘(SQQ}*PL*fc\s `/d'QXW, e`#kPGZuŞuO{{wm[&NBTiiI0bukcA9<4@SӊH*؎4U/'2U5.(9JuDfrޱtycU%j(:RUbArLֺN)udA':uGQN"-"Is.*+k@ `Ojs@yU/ H:l;@yyTn}_yw!VkRJ4P)~y#)r,D =ě"Q]ci'%HI4ZL0"MJy 8A{ aN<8D"1#IJi >XjX֔#@>-{vN!8tRݻ^)N_╗FJEk]CT՟ YP:_|H1@ CBk]yKYp|og?*dGvzنzӴzjֺNkC~AbZƷ`.H)=!QͷVTT(| u78y֮}|[8-Vjp%2JPk[}ԉaH8Wpqhwr:vWª<}l77_~{s۴V+RCģ%WRZ\AqHifɤL36: #F:p]Bq/z{0CU6ݳEv_^k7'>sq*+kH%a`0ԣisqにtү04gVgW΂iJiS'3w.w}l6MC2uԯ|>JF5`fV5m`Y**Db1FKNttu]4ccsQNnex/87+}xaUW9y>ͯ骵G{䩓Գ3+vU}~jJ.NFRD7<aJDB1#ҳgSb,+CS?/ VG J?|?,2#M9}B)MiE+G`-wo߫V`fio(}S^4e~V4bHOYb"b#E)dda:'?}׮4繏`{7Z"uny-?ǹ;0MKx{:_pÚmFמ:F " .LFQLG)Q8qN q¯¯3wOvxDb\. BKD9_NN &L:4D{mm o^tֽ:q!ƥ}K+<"m78N< ywsard5+вz~mnG)=}lYݧNj'QJS{S :UYS-952?&O-:W}(!6Mk4+>A>j+i|<<|;ر^߉=HE|V#F)Emm#}/"y GII웻Jі94+v뾧xu~5C95~ūH>c@덉pʃ1/4-A2G%7>m;–Y,cyyaln" ?ƻ!ʪ<{~h~i y.zZB̃/,雋SiC/JFMmBH&&FAbϓO^tubbb_hZ{_QZ-sύodFgO(6]TJA˯#`۶ɟ( %$&+V'~hiYy>922 Wp74Zkq+Ovn錄c>8~GqܲcWꂎz@"1A.}T)uiW4="jJ2W7mU/N0gcqܗOO}?9/wìXžΏ0 >֩(V^Rh32!Hj5`;O28؇2#ݕf3 ?sJd8NJ@7O0 b־?lldщ̡&|9C.8RTWwxWy46ah嘦mh٤&l zCy!PY?: CJyв]dm4ǜҐR޻RլhX{FƯanшQI@x' ao(kUUuxW_Ñ줮[w8 FRJ(8˼)_mQ _!RJhm=!cVmm ?sFOnll6Qk}alY}; "baӌ~M0w,Ggw2W:G/k2%R,_=u`WU R.9T"v,<\Ik޽/2110Ӿxc0gyC&Ny޽JҢrV6N ``یeA16"J³+Rj*;BϜkZPJaÍ<Jyw:NP8/D$ 011z֊Ⱳ3ι֘k1V_"h!JPIΣ'ɜ* aEAd:ݺ>y<}Lp&PlRfTb1]o .2EW\ͮ]38؋rTJsǏP@芎sF\> P^+dYJLbJ C-xϐn> ι$nj,;Ǖa FU *择|h ~izť3ᤓ`K'-f tL7JK+vf2)V'-sFuB4i+m+@My=O҈0"|Yxoj,3]:cо3 $#uŘ%Y"y죯LebqtҢVzq¼X)~>4L׶m~[1_k?kxֺQ`\ |ٛY4Ѯr!)N9{56(iNq}O()Em]=F&u?$HypWUeB\k]JɩSع9 Zqg4ZĊo oMcjZBU]B\TUd34ݝ~:7ڶSUsB0Z3srx 7`:5xcx !qZA!;%͚7&P H<WL!džOb5kF)xor^aujƍ7 Ǡ8/p^(L>ὴ-B,{ۇWzֺ^k]3\EE@7>lYBȝR.oHnXO/}sB|.i@ɥDB4tcm,@ӣgdtJ!lH$_vN166L__'Z)y&kH;:,Y7=J 9cG) V\hjiE;gya~%ks_nC~Er er)muuMg2;֫R)Md) ,¶ 2-wr#F7<-BBn~_(o=KO㭇[Xv eN_SMgSҐ BS헃D%g_N:/pe -wkG*9yYSZS.9cREL !k}<4_Xs#FmҶ:7R$i,fi!~' # !6/S6y@kZkZcX)%5V4P]VGYq%H1!;e1MV<!ϐHO021Dp= HMs~~a)ަu7G^];git!Frl]H/L$=AeUvZE4P\.,xi {-~p?2b#amXAHq)MWǾI_r`S Hz&|{ +ʖ_= (YS(_g0a03M`I&'9vl?MM+m~}*xT۲(fY*V4x@29s{DaY"toGNTO+xCAO~4Ϳ;p`Ѫ:>Ҵ7K 3}+0 387x\)a"/E>qpWB=1 ¨"MP(\xp߫́A3+J] n[ʼnӼaTbZUWb={~2ooKױӰp(CS\S筐R*JغV&&"FA}J>G֐p1ٸbk7 ŘH$JoN <8s^yk_[;gy-;߉DV{c B yce% aJhDȶ 2IdйIB/^n0tNtџdcKj4϶v~- CBcgqx9= PJ) dMsjpYB] GD4RDWX +h{y`,3ꊕ$`zj*N^TP4L:Iz9~6s) Ga:?y*J~?OrMwP\](21sZUD ?ܟQ5Q%ggW6QdO+\@ ̪X'GxN @'4=ˋ+*VwN ne_|(/BDfj5(Dq<*tNt1х!MV.C0 32b#?n0pzj#!38}޴o1KovCJ`8ŗ_"]] rDUy޲@ Ȗ-;xџ'^Y`zEd?0„ DAL18IS]VGq\4o !swV7ˣι%4FѮ~}6)OgS[~Q vcYbL!wG3 7띸*E Pql8=jT\꘿I(z<[6OrR8ºC~ډ]=rNl[g|v TMTղb-o}OrP^Q]<98S¤!k)G(Vkwyqyr޽Nv`N/e p/~NAOk \I:G6]4+K;j$R:Mi #*[AȚT,ʰ,;N{HZTGMoּy) ]%dHء9Պ䠬|<45,\=[bƟ8QXeB3- &dҩ^{>/86bXmZ]]yޚN[(WAHL$YAgDKp=5GHjU&99v簪C0vygln*P)9^͞}lMuiH!̍#DoRBn9l@ xA/_v=ȺT{7Yt2N"4!YN`ae >Q<XMydEB`VU}u]嫇.%e^ánE87Mu\t`cP=AD/G)sI"@MP;)]%fH9'FNsj1pVhY&9=0pfuJ&gޤx+k:!r˭wkl03׼Ku C &ѓYt{.O.zҏ z}/tf_wEp2gvX)GN#I ݭ߽v/ .& и(ZF{e"=V!{zW`, ]+LGz"(UJp|j( #V4, 8B 0 9OkRrlɱl94)'VH9=9W|>PS['G(*I1==C<5"Pg+x'K5EMd؞Af8lG ?D FtoB[je?{k3zQ vZ;%Ɠ,]E>KZ+T/ EJxOZ1i #T<@ I}q9/t'zi(EMqw`mYkU6;[t4DPeckeM;H}_g pMww}k6#H㶏+b8雡Sxp)&C $@'b,fPߑt$RbJ'vznuS ~8='72_`{q纶|Q)Xk}cPz9p7O:'|G~8wx(a 0QCko|0ASD>Ip=4Q, d|F8RcU"/KM opKle M3#i0c%<7׿p&pZq[TR"BpqauIp$ 8~Ĩ!8Սx\ւdT>>Z40ks7 z2IQ}ItԀ<-%S⍤};zIb$I 5K}Q͙D8UguWE$Jh )cu4N tZl+[]M4k8֦Zeq֮M7uIqG 1==tLtR,ƜSrHYt&QP윯Lg' I,3@P'}'R˪e/%-Auv·ñ\> vDJzlӾNv5:|K/Jb6KI9)Zh*ZAi`?S {aiVDԲuy5W7pWeQJk֤#5&V<̺@/GH?^τZL|IJNvI:'P=Ϛt"¨=cud S Q.Ki0 !cJy;LJR;G{BJy޺[^8fK6)=yʊ+(k|&xQ2`L?Ȓ2@Mf 0C`6-%pKpm')c$׻K5[J*U[/#hH!6acB JA _|uMvDyk y)6OPYjœ50VT K}cǻP[ $:]4MEA.y)|B)cf-A?(e|lɉ#P9V)[9t.EiQPDѠ3ϴ;E:+Օ t ȥ~|_N2,ZJLt4! %ա]u {+=p.GhNcŞQI?Nd'yeh n7zi1DB)1S | S#ًZs2|Ɛy$F SxeX{7Vl.Src3E℃Q>b6G ўYCmtկ~=K0f(=LrAS GN'ɹ9<\!a`)֕y[uՍ[09` 9 +57ts6}b4{oqd+J5fa/,97J#6yν99mRWxJyѡyu_TJc`~W>l^q#Ts#2"nD1%fS)FU w{ܯ R{ ˎ󅃏џDsZSQS;LV;7 Od1&1n$ N /.q3~eNɪ]E#oM~}v֯FڦwyZ=<<>Xo稯lfMFV6p02|*=tV!c~]fa5Y^Q_WN|Vs 0ҘދU97OI'N2'8N֭fgg-}V%y]U4 峧p*91#9U kCac_AFңĪy뚇Y_AiuYyTTYЗ-(!JFLt›17uTozc. S;7A&&<ԋ5y;Ro+:' *eYJkWR[@F %SHWP 72k4 qLd'J "zB6{AC0ƁA6U.'F3:Ȅ(9ΜL;D]m8ڥ9}dU "v!;*13Rg^fJyShyy5auA?ɩGHRjo^]׽S)Fm\toy 4WQS@mE#%5ʈfFYDX ~D5Ϡ9tE9So_aU4?Ѽm%&c{n>.KW1Tlb}:j uGi(JgcYj0qn+>) %\!4{LaJso d||u//P_y7iRJ߬nHOy) l+@$($VFIQ9%EeKʈU. ia&FY̒mZ=)+qqoQn >L!qCiDB;Y<%} OgBxB!ØuG)WG9y(Ą{_yesuZmZZey'Wg#C~1Cev@0D $a@˲(.._GimA:uyw֬%;@!JkQVM_Ow:P.s\)ot- ˹"`B,e CRtaEUP<0'}r3[>?G8xU~Nqu;Wm8\RIkբ^5@k+5(By'L&'gBJ3ݶ!/㮻w҅ yqPWUg<e"Qy*167΃sJ\oz]T*UQ<\FԎ`HaNmڜ6DysCask8wP8y9``GJ9lF\G g's Nn͵MLN֪u$| /|7=]O)6s !ĴAKh]q_ap $HH'\1jB^s\|- W1:=6lJBqjY^LsPk""`]w)󭃈,(HC ?䔨Y$Sʣ{4Z+0NvQkhol6C.婧/u]FwiVjZka&%6\F*Ny#8O,22+|Db~d ~Çwc N:FuuCe&oZ(l;@ee-+Wn`44AMK➝2BRՈt7g*1gph9N) *"TF*R(#'88pm=}X]u[i7bEc|\~EMn}P瘊J)K.0i1M6=7'_\kaZ(Th{K*GJyytw"IO-PWJk)..axӝ47"89Cc7ĐBiZx 7m!fy|ϿF9CbȩV 9V-՛^pV̌ɄS#Bv4-@]Vxt-Z, &ֺ*diؠ2^VXbs֔Ìl.jQ]Y[47gj=幽ex)A0ip׳ W2[ᎇhuE^~q흙L} #-b۸oFJ_QP3r6jr+"nfzRJTUqoaۍ /$d8Mx'ݓ= OՃ| )$2mcM*cЙj}f };n YG w0Ia!1Q.oYfr]DyISaP}"dIӗթO67jqR ҊƐƈaɤGG|h;t]䗖oSv|iZqX)oalv;۩meEJ\!8=$4QU4Xo&VEĊ YS^E#d,yX_> ۘ-e\ "Wa6uLĜZi`aD9.% w~mB(02G[6y.773a7 /=o7D)$Z 66 $bY^\CuP. (x'"J60׿Y:Oi;F{w佩b+\Yi`TDWa~|VH)8q/=9!g߆2Y)?ND)%?Ǐ`k/sn:;O299yB=a[Ng 3˲N}vLNy;*?x?~L&=xyӴ~}q{qE*IQ^^ͧvü{Huu=R|>JyUlZV, B~/YF!Y\u_ݼF{_C)LD]m {H 0ihhadd nUkf3oٺCvE\)QJi+֥@tDJkB$1!Đr0XQ|q?d2) Ӣ_}qv-< FŊ߫%roppVBwü~JidY4:}L6M7f٬F "?71<2#?Jyy4뷢<_a7_=Q E=S1И/9{+93֮E{ǂw{))?maÆm(uLE#lïZ  ~d];+]h j?!|$F}*"4(v'8s<ŏUkm7^7no1w2ؗ}TrͿEk>p'8OB7d7R(A 9.*Mi^ͳ; eeUwS+C)uO@ =Sy]` }l8^ZzRXj[^iUɺ$tj))<sbDJfg=Pk_{xaKo1:-uyG0M ԃ\0Lvuy'ȱc2Ji AdyVgVh!{]/&}}ċJ#%d !+87<;qN޼Nفl|1N:8ya  8}k¾+-$4FiZYÔXk*I&'@iI99)HSh4+2G:tGhS^繿 Kتm0 вDk}֚+QT4;sC}rՅE,8CX-e~>G&'9xpW,%Fh,Ry56Y–hW-(v_,? ; qrBk4-V7HQ;ˇ^Gv1JVV%,ik;D_W!))+BoS4QsTM;gt+ndS-~:11Sgv!0qRVh!"Ȋ(̦Yl.]PQWgٳE'`%W1{ndΗBk|Ž7ʒR~,lnoa&:ü$ 3<a[CBݮwt"o\ePJ=Hz"_c^Z.#ˆ*x z̝grY]tdkP*:97YľXyBkD4N.C_[;F9`8& !AMO c `@BA& Ost\-\NX+Xp < !bj3C&QL+*&kAQ=04}cC!9~820G'PC9xa!w&bo_1 Sw"ܱ V )Yl3+ס2KoXOx]"`^WOy :3GO0g;%Yv㐫(R/r (s } u B &FeYZh0y> =2<Ϟc/ -u= c&׭,.0"g"7 6T!vl#sc>{u/Oh Bᾈ)۴74]x7 gMӒ"d]U)}" v4co[ ɡs 5Gg=XR14?5A}D "b{0$L .\4y{_fe:kVS\\O]c^W52LSBDM! C3Dhr̦RtArx4&agaN3Cf<Ԉp4~ B'"1@.b_/xQ} _߃҉/gٓ2Qkqp0շpZ2fԫYz< 4L.Cyυι1t@鎫Fe sYfsF}^ V}N<_`p)alٶ "(XEAVZ<)2},:Ir*#m_YӼ R%a||EƼIJ,,+f"96r/}0jE/)s)cjW#w'Sʯ5<66lj$a~3Kʛy 2:cZ:Yh))+a߭K::N,Q F'qB]={.]h85C9cr=}*rk?vwV렵ٸW Rs%}rNAkDv|uFLBkWY YkX מ|)1!$#3%y?pF<@<Rr0}: }\J [5FRxY<9"SQdE(Q*Qʻ)q1E0B_O24[U'],lOb ]~WjHޏTQ5Syu wq)xnw8~)c 쫬gٲߠ H% k5dƝk> kEj,0% b"vi2Wس_CuK)K{n|>t{P1򨾜j>'kEkƗBg*H%'_aY6Bn!TL&ɌOb{c`'d^{t\i^[uɐ[}q0lM˕G:‚4kb祔c^:?bpg… +37stH:0}en6x˟%/<]BL&* 5&fK9Mq)/iyqtA%kUe[ڛKN]Ě^,"`/ s[EQQm?|XJ߅92m]G.E΃ח U*Cn.j_)Tѧj̿30ڇ!A0=͜ar I3$C^-9#|pk!)?7.x9 @OO;WƝZBFU keZ75F6Tc6"ZȚs2y/1 ʵ:u4xa`C>6Rb/Yм)^=+~uRd`/|_8xbB0?Ft||Z\##|K 0>>zxv8۴吅q 8ĥ)"6>~\8:qM}#͚'ĉ#p\׶ l#bA?)|g g9|8jP(cr,BwV (WliVxxᡁ@0Okn;ɥh$_ckCgriv}>=wGzβ KkBɛ[˪ !J)h&k2%07δt}!d<9;I&0wV/ v 0<H}L&8ob%Hi|޶o&h1L|u֦y~󛱢8fٲUsւ)0oiFx2}X[zVYr_;N(w]_4B@OanC?gĦx>мgx>ΛToZoOMp>40>V Oy V9iq!4 LN,ˢu{jsz]|"R޻&'ƚ{53ўFu(<٪9:΋]B;)B>1::8;~)Yt|0(pw2N%&X,URBK)3\zz&}ax4;ǟ(tLNg{N|Ǽ\G#C9g$^\}p?556]/RP.90 k,U8/u776s ʪ_01چ|\N 0VV*3H鴃J7iI!wG_^ypl}r*jɤSR 5QN@ iZ#1ٰy;_\3\BQQ x:WJv츟ٯ$"@6 S#qe딇(/P( Dy~TOϻ<4:-+F`0||;Xl-"uw$Цi󼕝mKʩorz"mϺ$F:~E'ҐvD\y?Rr8_He@ e~O,T.(ފR*cY^m|cVR[8 JҡSm!ΆԨb)RHG{?MpqrmN>߶Y)\p,d#xۆWY*,l6]v0h15M˙MS8+EdI='LBJIH7_9{Caз*Lq,dt >+~ّeʏ?xԕ4bBAŚjﵫ!'\Ը$WNvKO}ӽmSşذqsOy?\[,d@'73'j%kOe`1.g2"e =YIzS2|zŐƄa\U,dP;jhhhaxǶ?КZ՚.q SE+XrbOu%\GتX(H,N^~]JyEZQKceTQ]VGYqnah;y$cQahT&QPZ*iZ8UQQM.qo/T\7X"u?Mttl2Xq(IoW{R^ ux*SYJ! 4S.Jy~ BROS[V|žKNɛP(L6V^|cR7i7nZW1Fd@ Ara{詑|(T*dN]Ko?s=@ |_EvF]׍kR)eBJc" MUUbY6`~V޴dJKß&~'d3i WWWWWW
Current Directory: /usr/share/doc/openblas
Viewing File: /usr/share/doc/openblas/GotoBLAS_03FAQ.txt
GotoBLAS2 FAQ 1. General 1.1 Q Can I find useful paper about GotoBLAS2? A You may check following URL. http://www.cs.utexas.edu/users/flame/Publications/index.htm 11. Kazushige Goto and Robert A. van de Geijn, " Anatomy of High-Performance Matrix Multiplication," ACM Transactions on Mathematical Software, accepted. 15. Kazushige Goto and Robert van de Geijn, "High-Performance Implementation of the Level-3 BLAS." ACM Transactions on Mathematical Software, submitted. 1.2 Q Does GotoBLAS2 work with Hyperthread (SMT)? A Yes, it will work. GotoBLAS2 detects Hyperthread and avoid scheduling on the same core. 1.3 Q When I type "make", following error occured. What's wrong? $shell> make "./Makefile.rule", line 58: Missing dependency operator "./Makefile.rule", line 61: Need an operator ... A This error occurs because you didn't use GNU make. Some binary packages install GNU make as "gmake" and it's worth to try. 1.4 Q Function "xxx" is slow. Why? A Generally GotoBLAS2 has many well optimized functions, but it's far and far from perfect. Especially Level 1/2 function performance depends on how you call BLAS. You should understand what happends between your function and GotoBLAS2 by using profile enabled version or hardware performance counter. Again, please don't regard GotoBLAS2 as a black box. 1.5 Q I have a commercial C compiler and want to compile GotoBLAS2 with it. Is it possible? A All function that affects performance is written in assembler and C code is just used for wrapper of assembler functions or complicated functions. Also I use many inline assembler functions, unfortunately most of commercial compiler can't handle inline assembler. Therefore you should use gcc. 1.6 Q I use OpenMP compiler. How can I use GotoBLAS2 with it? A Please understand that OpenMP is a compromised method to use thread. If you want to use OpenMP based code with GotoBLAS2, you should enable "USE_OPENMP=1" in Makefile.rule. 1.7 Q Could you tell me how to use profiled library? A You need to build and link your application with -pg option. After executing your application, "gmon.out" is generated in your current directory. $shell> gprof <your application name> gmon.out Each sample counts as 0.01 seconds. % cumulative self self total time seconds seconds calls Ks/call Ks/call name 89.86 975.02 975.02 79317 0.00 0.00 .dgemm_kernel 4.19 1020.47 45.45 40 0.00 0.00 .dlaswp00N 2.28 1045.16 24.69 2539 0.00 0.00 .dtrsm_kernel_LT 1.19 1058.03 12.87 79317 0.00 0.00 .dgemm_otcopy 1.05 1069.40 11.37 4999 0.00 0.00 .dgemm_oncopy .... I think profiled BLAS library is really useful for your research. Please find bottleneck of your application and improve it. 1.8 Q Is number of thread limited? A Basically, there is no limitation about number of threads. You can specify number of threads as many as you want, but larger number of threads will consume extra resource. I recommend you to specify minimum number of threads. 1.9 Q I have segfaults when I compile with USE_OPENMP=1. What's wrong? A This may be related to a bug in the Linux kernel 2.6.32. Try applying the patch segaults.patch using patch < segfaults.patch and see if the crashes persist. Note that this patch will lead to many compiler warnings. 2. Architecture Specific issue or Implementation 2.1 Q GotoBLAS2 seems to support any combination with OS and architecture. Is it possible? A Combination is limited by current OS and architecture. For examble, the combination OSX with SPARC is impossible. But it will be possible with slight modification if these combination appears in front of us. 2.2 Q I have POWER architecture systems. Do I need extra work? A Although POWER architecture defined special instruction like CPUID to detect correct architecture, it's privileged and can't be accessed by user process. So you have to set the architecture that you have manually in getarch.c. 2.3 Q I can't create DLL on Cygwin (Error 53). What's wrong? A You have to make sure if lib.exe and mspdb80.dll are in Microsoft Studio PATH. The easiest way is to use 'which' command. $shell> which lib.exe /cygdrive/c/Program Files/Microsoft Visual Studio/VC98/bin/lib.exe