From 9ff4c950269bad47dde6971485511977380cda6e Mon Sep 17 00:00:00 2001
From: Caballo <1076318970@qq.com>
Date: Wed, 20 Sep 2023 11:54:18 +0800
Subject: [PATCH] Finished
---
README.md | 249 ++++++++++++++++++++++++++-
result/blocksize.png | Bin 0 -> 50758 bytes
result/compactNPot.png | Bin 0 -> 50040 bytes
result/compactPot.png | Bin 0 -> 54846 bytes
result/scanNPot.png | Bin 0 -> 53998 bytes
result/scanPot.png | Bin 0 -> 57018 bytes
result/sort.png | Bin 0 -> 54371 bytes
src/main.cpp | 66 ++++++-
stream_compaction/common.cu | 9 +
stream_compaction/cpu.cu | 46 ++++-
stream_compaction/cpu.h | 2 +
stream_compaction/efficient.cu | 306 ++++++++++++++++++++++++++++++++-
stream_compaction/efficient.h | 3 +
stream_compaction/naive.cu | 50 +++++-
stream_compaction/thrust.cu | 10 +-
15 files changed, 723 insertions(+), 18 deletions(-)
create mode 100644 result/blocksize.png
create mode 100644 result/compactNPot.png
create mode 100644 result/compactPot.png
create mode 100644 result/scanNPot.png
create mode 100644 result/scanPot.png
create mode 100644 result/sort.png
diff --git a/README.md b/README.md
index 0e38ddb..f97b136 100644
--- a/README.md
+++ b/README.md
@@ -3,12 +3,249 @@ CUDA Stream Compaction
**University of Pennsylvania, CIS 565: GPU Programming and Architecture, Project 2**
-* (TODO) YOUR NAME HERE
- * (TODO) [LinkedIn](), [personal website](), [twitter](), etc.
-* Tested on: (TODO) Windows 22, i7-2222 @ 2.22GHz 22GB, GTX 222 222MB (Moore 2222 Lab)
+* Xiaoyue Ma
+ * [LinkedIn](https://www.linkedin.com/in/xiaoyue-ma-6b268b193/)
+* Tested on: Windows 10, i7-12700H @ 2.30 GHz 16GB, GTX3060 8GB
-### (TODO: Your README)
+# Table of Contents
+[Overview](#overview)
-Include analysis, etc. (Remember, this is public, so don't put
-anything here that you don't want to share with the world.)
+[Features](#features)
+[Implementation](#implementation)
+
+[Result](#result)
+
+[Performance Analysis](#performanceAnalysis)
+
+[Extra Credit](#extraCredit)
+
+
+
+# Overview
+
+This project is mainly about Parallel Scan(prefix sum) and Stream Compaction algorithms implemented with CUDA. Scan is about computing the prefix sum of an array, and stream compaction is about deleting all elements in an array that meet certain condition. These algorithms seem to be inherently sequential at the first glance, but with GPU we can convert these algorithms into very efficient parallel algorithms.
+
+One application of Parallel Scan is Summed Area Table, which is a very important algorithm real-time rendering, especially for pre-computation. Another one is Radix Sort, which is a sorting algorithm that can run in parallel. Stream Compaction is very important in ray tracing, which can help delete unnecessory rays.
+
+The data is computed in the Release Mode.
+
+
+# Features
+* CPU Version Scan
+* CPU Version Compaction with/without Scan
+* GPU Naive Version Scan
+* GPU Work-efficient Version Scan
+* GPU Work-efficient Version Compaction with Scan
+* Thrust Version Scan/Compaction
+* GPU Version Scan Optimization (Extra Credit)
+* Radix Sort (Extra Credit)
+
+
+
+# Implementation
+**TBD**
+
+
+
+# Result
+
+* Input array Size 2^22
+* For Scan and compaction, max element value is 50
+* For Sort, max element value is 7
+```
+
+****************
+** SCAN TESTS **
+****************
+ [ 7 5 8 15 24 5 45 33 9 19 33 37 40 ... 18 0 ]
+==== cpu scan, power-of-two ====
+ elapsed time: 7.7637ms (std::chrono Measured)
+ [ 0 7 12 20 35 59 64 109 142 151 170 203 240 ... 102732720 102732738 ]
+==== cpu scan, non-power-of-two ====
+ elapsed time: 7.9926ms (std::chrono Measured)
+ [ 0 7 12 20 35 59 64 109 142 151 170 203 240 ... 102732598 102732638 ]
+ passed
+==== naive scan, power-of-two ====
+ elapsed time: 2.83472ms (CUDA Measured)
+ passed
+==== naive scan, non-power-of-two ====
+ elapsed time: 2.75386ms (CUDA Measured)
+ passed
+==== work-efficient scan, power-of-two ====
+ elapsed time: 2.58877ms (CUDA Measured)
+ passed
+==== work-efficient scan, non-power-of-two ====
+ elapsed time: 2.49437ms (CUDA Measured)
+ passed
+==== work-efficient scanOptimized, power-of-two ====
+ elapsed time: 1.47747ms (CUDA Measured)
+ passed
+==== work-efficient scanOptimized, non-power-of-two ====
+ elapsed time: 1.34262ms (CUDA Measured)
+ passed
+==== thrust scan, power-of-two ====
+ elapsed time: 0.615424ms (CUDA Measured)
+ passed
+==== thrust scan, non-power-of-two ====
+ elapsed time: 0.500736ms (CUDA Measured)
+ passed
+
+*****************************
+** STREAM COMPACTION TESTS **
+*****************************
+ [ 1 3 0 1 2 3 1 1 1 3 1 3 0 ... 2 0 ]
+==== cpu compact without scan, power-of-two ====
+ elapsed time: 10.3321ms (std::chrono Measured)
+ [ 1 3 1 2 3 1 1 1 3 1 3 2 2 ... 1 2 ]
+ passed
+==== cpu compact without scan, non-power-of-two ====
+ elapsed time: 10.7476ms (std::chrono Measured)
+ [ 1 3 1 2 3 1 1 1 3 1 3 2 2 ... 2 1 ]
+ passed
+==== cpu compact with scan ====
+ elapsed time: 22.0692ms (std::chrono Measured)
+ [ 1 3 1 2 3 1 1 1 3 1 3 2 2 ... 1 2 ]
+ passed
+==== work-efficient compact, power-of-two ====
+ elapsed time: 3.81203ms (CUDA Measured)
+ passed
+==== work-efficient compact, non-power-of-two ====
+ elapsed time: 4.11203ms (CUDA Measured)
+ passed
+==== work-efficient compactOptimized, power-of-two ====
+ elapsed time: 2.52003ms (CUDA Measured)
+ passed
+==== work-efficient compactOptimized, non-power-of-two ====
+ elapsed time: 2.6936ms (CUDA Measured)
+ passed
+
+*****************************
+** RADIX SORT TESTS **
+*****************************
+ [ 4 3 2 6 4 3 1 5 2 5 6 6 2 ... 6 0 ]
+==== cpu std::sort, power-of-two ====
+ elapsed time: 46.5172ms (std::chrono Measured)
+ [ 0 0 0 0 0 0 0 0 0 0 0 0 0 ... 6 6 ]
+==== cpu std::sort, non-power-of-two ====
+ elapsed time: 45.6927ms (std::chrono Measured)
+ [ 0 0 0 0 0 0 0 0 0 0 0 0 0 ... 6 6 ]
+==== radix sort, power-of-two ====
+ elapsed time: 15.2273ms (CUDA Measured)
+ passed
+==== radix sort, non-power-of-two ====
+ elapsed time: 14.2143ms (CUDA Measured)
+ passed
+```
+
+
+
+# Performance Analysis
+
+As seen in this graph, **blockSize = 128** has the general best performance:
+
+
+
+
+
+
+Thrust's implementation consistently delivers the top performance. Its sort function is not just a simple quicksort or mergesort. Behind the scenes, Thrust might choose the best sorting strategy based on the input data and the specific GPU hardware. Like remove function in Thrust container, it re-partition the element order instead of deleting the element directly. So I guess Thrust's implementation has great Optimized Algorithms, Memory Management and Parallel Patterns.
+
+
+
+In the Scan operation, when handling smaller arrays (below 2^12), the CPU consistently delivers better performance than the GPU, irrespective of the GPU's scanning method, be it naive or work-efficient. Yet, as array sizes surpass 2^16, the GPU, with its naive scan, takes the lead, capitalizing on the strengths of parallel computation for handling large datasets. In such cases, following the GPU's naive scan in performance ranking is its work-efficient scan, with the CPU trailing noticeably with extremely large datasets. Interestingly, performance doesn't noticeably differ whether the array sizes are power-of-two or not. This uniformity is attributed to the GPU's approach of padding non-power-of-two arrays to their closest power-of-two counterpart, and the CPU maintaining its efficiency even with slight variations in array sizes.
+
+
+
+
+
+
+With small datasets, the CPU is faster than the GPU. But as data size increases, the GPU's efficiency, especially using the Work Efficient Scan, surpasses the CPU. This is accentuated in tasks like compaction where parallelism shines. Conversely, the CPU's Compact With Scan lags because it tackles more tasks without benefiting from parallel processing.
+
+
+
+The CPU employs std::sort with an O(N log(N)) complexity. Yet, for inputs beyond 2^17, the GPU's Radix sort outperforms it(Still slower than thrust).
+
+
+
+# Extra Credit
+### Why is My GPU Approach So Slow?
+
+* Global Memory Access: A primary factor causing our GPU to lag behind the CPU in speed is its reliance on global memory for computations. Accessing data from global memory can significantly hamper performance, even if the GPU's parallel processing capability offers advantages with large arrays. To enhance the performance, we should transition to using shared memory rather than global memory and employ an algorithm designed to minimize bank conflicts.
+
+* More Threads Than Needed: The original launches more threads than necessary, especially in the deeper levels of the sweep. Many of these threads do not perform any useful work but still contribute to launch overhead.
+
+* Overhead of Modulo Operation: The use of the modulo operation (index % (2 * path) == 0) in the original methods is computationally more expensive than a simple boundary check.
+
+
+```
+__global__ void kernUpSweepOptimized(int n, int d, int* odata) {
+ int path = 1 << d;
+ int index = (threadIdx.x + blockIdx.x * blockDim.x) * 2 * path;
+
+ if (index + 2 * path - 1 < n) {
+ odata[index + 2 * path - 1] += odata[index + path - 1];
+ }
+ }
+
+__global__ void kernDownSweepOptimized(int n, int d, int* odata) {
+ //same optimization as above
+}
+
+void scanUpDownSweepOptimized(int n, int* odata) {
+ for (int d = 0; d <= ilog2ceil(n) - 1; ++d) {
+ int threadsNeeded = n / (2 << d);
+ dim3 BlockDim = (threadsNeeded + BlockSize - 1) / BlockSize;
+
+ kernUpSweepOptimized << < BlockDim, BlockSize >> > (n, d, odata);
+ }
+ ...
+
+ //Same code
+
+ ...
+}
+ ```
+
+**For my optimization:**
+
+* Thread Index Calculation:
+
+ * Original (kernUpSweep & kernDownSweep): The thread index is calculated as a combination of the block's index and the thread's index within that block: int index = threadIdx.x + (blockIdx.x * blockDim.x);
+ * Optimized (kernUpSweepOptimized & kernDownSweepOptimized): The thread index is calculated differently: int index = (threadIdx.x + blockIdx.x * blockDim.x) * 2 * path;. Each thread in the optimized version is responsible for twice as many elements as in the original, which means fewer threads are needed to handle the same amount of data.
+
+* Index Conditions for Operations:
+
+ * Original: Operations are conditioned on index % (2 * path) == 0. This is a more general case that applies to each thread.
+ * Optimized: Operations are conditioned on boundary checks, i.e., index + 2 * path - 1 < n. Given the adjusted index calculation, the optimized version has a more straightforward condition that checks if operations will stay within the array bounds.
+
+* Thread and Block Launch Calculations:
+
+ * Original (scanUpDownSweep): The block dimensions are calculated based on the BlockSize, without considering the depth d of the sweep.
+ * Optimized (scanUpDownSweepOptimized): The block dimensions are calculated based on both the BlockSize and the depth d. Specifically, int threadsNeeded = n / (2 << d); determines the actual number of threads required for each depth, potentially reducing the number of threads launched, especially in the deeper levels of the sweep where fewer updates are needed.
+
+In general, Even though I didn't use share memory, I still slowed it down to a much lower speed than the GPU and the original version.
+
+### Radix Sort
+I've crafted a radix sort algorithm in the efficient.cu file and benchmarked it against the std::sort function on the CPU. The results show that my implementation begins to outpace the CPU-based sort when the input exceeds 2^16 elements.
+
+This is the result for ArraySize = 2^22:
+```
+*****************************
+** RADIX SORT TESTS **
+*****************************
+ [ 4 3 2 6 4 3 1 5 2 5 6 6 2 ... 6 0 ]
+==== cpu std::sort, power-of-two ====
+ elapsed time: 46.5172ms (std::chrono Measured)
+ [ 0 0 0 0 0 0 0 0 0 0 0 0 0 ... 6 6 ]
+==== cpu std::sort, non-power-of-two ====
+ elapsed time: 45.6927ms (std::chrono Measured)
+ [ 0 0 0 0 0 0 0 0 0 0 0 0 0 ... 6 6 ]
+==== radix sort, power-of-two ====
+ elapsed time: 15.2273ms (CUDA Measured)
+ passed
+==== radix sort, non-power-of-two ====
+ elapsed time: 14.2143ms (CUDA Measured)
+ passed
+```
+The max element value is 7, cuz Radix Sort is only suitable for small numbers sort. If I set the max element value to be 50, Radix Sort will be slow.
diff --git a/result/blocksize.png b/result/blocksize.png
new file mode 100644
index 0000000000000000000000000000000000000000..905a8e68ee21173b06552d502ee312240155d677
GIT binary patch
literal 50758
zcmeFZbyQqmvnWUcfe;`#ArK^JaCdhLkPzGhA-F@(MnWLCO9*ZO0*$*g65QPycWWB$
z1{#?r-|xF`=FWO|-kLRk+*vbcb+04as%qD+s$Es*guPQ!z#=$@j2)73ArWtA>I!O4-P(J>(6VrIe}^3QA=R*3HK!$a_pDMLky(6cYM}
z|05+0x?>cSZm^Q9l%}Wg;Vq7@=2pi2J;MLyF6Pd=nh#RP4mtX=SHHx+w6t(ocbMW=
zYf=rd`Je;Vv?n(xDX!%O&Q1I#?naQKK^5ibze=J7I?=i?q7yy+N{UCf7#g-1jF4==
zTw_?1zr7jR+59<_Dl|Kko;pOAuz7K~arTD6l7VL@UC6Zz-Zmth)#`lW%9_{a)m5k}Xp8Af=H5Nn>>q9`Si{@
zSW}?pXYah`0QJF!xmH}_blP{a3k$jAtz71|ci}1_N_tvRyxl%RN^O)tsgXPaEBn~}
z(~laOv5bv?kvP{8uB6FXfIWRdbw2;m!81(Q^|SDyrgHV;{G_167!Dv87M(Vq;L7Q%
zI|z}Ay~l~>~gV+{pjQ+V_}
zwcMWp!b7NhuU_H=9(qcu2}Z
zcdANza~ni5w;UNO9l&Uww@>6WG`JO&2TSr9?j98ytzZ25`pPGdWxx8zA*hh}Li#a<
zYwMN|?Veb)#r6e}irh_5=sg#Ok@~W-SsV7X$0=4s!KZh{nwkQ#zMpC5VU;K=&Rlz3
z)rKL&#EQlaS;rnLLP8TdqlC)J7TOgY5+ySsHpOq);yQhy9=25d?#3Qcd-Vy^7u&{B+b~SG(v>i4WH`8S4j8
z0tt5BSJo)2dR1=v$b8jJn$$VP&`}QZ$~USiSDytGYj%5U6gpc23|9geLsf}263x9@
zEJWs~`zr40db4N&07h;u-4&T=qHYmzy=ad$yk}a|h*{2SnQ&Qmx^ARogOA?~2-i(A
z+4tRl{F@``X17-aJo&Rt6DJ-tJ}5lx5i15v)mk^Ht6+86sRB^a&e}6)RXWLurdWt=
z_b#2V`irExA1JT!?D8@&6Yz!;lnW#sR_gqZIf)nFzFt=JjxkU5788;?`KmBzF5Hya
zlGNpwIUAuuZ?ieebLY*ziF!J
zXK>i>V9niUESzu@PN@yJq=y7U$i?Sp^N0^V{gxki=E^qMTb_QWW-hkfsLEVET5Pz(
zw~)!Rtt7A3iC*)YgQY@Ln`Ko=c9?U|5zz8MN+v=Vn&Vh`E5vr4)@N?-ir
z7KyT6X*5}X<|6%c5iq&KYkYLg4p(J^dfT$_o^#N0n&b!#Au3-6CW3$bfIn)X7c>8Y
z^c%CzvkVFvZ9BZ8_puX|0ygCI%}{H(*4X@*Mvy@K@`moVvAPLUGyv~J;#b~5zroAr
z*o#$VqvvSoOcXq>*Mr$}CVn6JEHr-DdigHeAa3TLh_VZ==zut9yq0yvs(z-A>T4WO
zhuwnPunLbGXZDXRc~+6Eh=tS-_!?N)HP^Uh{G<3252R5axE|zOxss_WR~Y=lU9d!w
zfQuz03$B)iK4S_uW_PTR_0J{DTKomV
zwf)OZpB+uJiCK!U=iVsp&4Gg+qZZor7DR$Fe{|$+{awZTk?rlpcOZ_AXbf};;msx76c-gal^8->Bgcou>hmWi=2UO4ZAw@Q
zhU$*&quLu7Up_g3sz#1yv;x1T%XSIpJkwbp>%@nPbu~y^c|}57l%`XZa%x42zwmG*
zf3#?#7%Xsm7q=CSUKc?uf3SGUGz@kxgyoP_mG-K7H8?k~Lx+454c
z=H&%6u*`h8SYd9@M@8SEDCQNx=MPeI^TKn0QdFOku(d;<&A+)t@RUn!e%AloHj-@&
zQ4WmPFo+SbypItb^5EnwVoUM`JIgsXvalzej#=inXeyfl0l_m_cJ;0I2IC;ORi)hpm^~!KezARgKsXz>#&IRkQE1S_a-2YmSh*
z1eEHWv2%m4cFtm7v_=-5=6a(}8;D!hQBns=#Ncb#<UJdpgguMYLSr3|jy&1>$g`u|HUvo2s>W{Q7i&VwT*QfYk
z*ZE&5CAhw-dV)OzrCjj{(T?5}>195^pVynMJ3%d7v!3E!zA&Sxr`CpY<>i!ex_AM1
zqjFnG9%IwYo!o*!MmJ*OT4SRO0agz(aFvm$PHKXtZCChii|A&Mw{K&|9+0^GNcb+vSw(ePs`RsLfGkxgQToy_gbmDQWw4itr&&8
zrD1~g)uKK14y8~o*zIq7yDXf+FGG@fuj4IV1r-J2+9MQXg%C;bmg#%Fx^rFU{4XwG
z9#qs6cZvOj0}DL6){DA!>baExBpKLEP{&VyQ2XS@L<1p_N|mv=8HzJ~&`H)@(P(r`vwzrRcHhBAWoDniJn=mQx
zdoY{nKy$r|{5FoW)8@Jl+ti}ympWXuFcoW*PuwhH5&^F@R4~O8M
zprW>369Cr9o-z1};SyvJEvU6)q+F|?x4^p}wu40w`_M%Y>&rx
z&&4bcr5+TXyjs1D8BNF~r4j|(M(NGCCm@yNXX8;*rVD_jC-kaLl;I@W!RS8qDfacI
zy#1)*y{=GWutoA1$3=TP|D26;EQYXMj=n6E+x}Ai*0|onJGw8~e!fjt`P5#F%+4HZ
zzKhbagKeiJ*MH5FsU)1Q+f~=5S5s@lZg6)N&oNuxLhpq$^V|W7H^yUB>aUP*G`d&i
z{BB6LI_cLa1DF=PD&#s60V}RwxUI_0jm*%}Hm%uv-R(-!_N@BUji*2u%Prmq;ma=t
zQ30r|rBV~Gno4@HuvuF&X<4#it=LF?-c@?%a|OK$Y29msuNpFrX|X5@+{>M#Q+
z1CXp1Yi4RnL07b4qNbF^+DKet@R6XLF_lpL3@rPZSV4?m(G)M+EOWO@&?^eGumK6f
z7sJaoexkfDJMMBIxB2#Q@;srLo4-f5=%ad7d+Rd12Oe@P@Qx=@B#5d@L`4M)qHnuo
zIikq8QdL8lv*y|F9^taJYDG@X7~tY!KVE!Jj1(`c?bMRp!Vl`~$CepJOmZ)pkfiR6
ze<23C+QoRsg^k^8T{8ox6X+(FG7S(YCCgEpt#4b9OJ{O@?LBbG!suTy*E6ZQ_d}m&
zs0~!gBP*+NRGnIgTzN(d@SG-ulE(|ZhF$r1ct=+XR==Ob0l*gyn!=^
zaSXOIc1n@^GA+mBmAjngn;w_WE%BRYv&8_ZI=b4#18*aNmbf5@Bh{es_I_BObb7+QLig!valP|u`XQ|L++)BtLsZfT0wB9%J9h6#E^{=s5C+3PCa
zF*RIhoH~N5mS53bu9)aW*Y1fGbFY{BIKuL@cgkjvx0dS{DXg+If^M`wU>Uik4~@Od
zuZ84`A$O+VmAu?ryu3Fuik_GwxFIIGH1xrDgt4yA1_^!VJV{Y-2Z~B4C
zCZ>4)Vqb1N69!!8&@rQyvx0#2Q{=9Y|AUe~ahPP-amBxFN;eO6jLiI(mTHH#fegb5
zV9Ux#sjS*c2Gzy?On%Pk&453MjT>0@1LNb$9UouvBUHlZM|c4t7+XOF?cU8_{mGU;
z!J)sU?x6LnQVPP+agtD3QxsXnKT5XE(f&f+ciiq|^2zwo=CvYVx0{Q(vP3j%s|C;|
zYSfjfahCl)9e6tYsk53ffY$g$qN0-QA1+aZ%xZi4mjXVkU~eH=70<^GSykLkzKo`I
z4x1y0aDnu%cffG7d#xXLL3v-A{t^T}vZL6~3BG*t`py~qnTkr@jLdnMZ~hdSqnHq9
zg%RW|<`Mqi<&%xtTlp50iFYvNKTf4#eb(P;)tVUifuN_QfJ+HjXCCCA-S?@k+?`DE0Fqsqzd?$5
zuJ*8ujtEG1`=!sVZq)V+`|R856#t|Y01E?*lbqD-%r0t*Jd6vWDtGC3$t<=f##yZ&
z&C*CJjam|P2B*H`Z|&XU@p4uKGj*`^Ncu=2SwQD$-oX8$_g9<7$N7vVT-O~lf{ZD5
z5Yp$`3l`bqraMrIv3f^MPXgl4S7byZr*Sw1mpB0ri1Tt<0h3N?{PC=?+K>w+lZrGS
zsgMs3Ffs8&(XaAeULG=@-Q7+CP%xFby`l|Vey~_#*V?H^P(-|ht6ZlVzjShZF!1R8!AQ4vT~#yepDx>dDE(vi
zVt&W3&IssCFYHzO)_G1l4z7p(44ZW{X|hdGSy|aEam(*B_zRZrW6JwyA4^>U;3>_(
zLE)kwe@=gHJfE?oC}N}DaqRZH^|~`=oiX6*Oe%6t(BA;2VW2B%mimMtB?3}vnmy+(
zN}3`zN?LSgJu7EqbvXXvsh~c19YeBv+xz3!g#X)Gh&h067Gaoq3RG739-SK2gfQqO
zeyU&DL-n-c*8FB^&Kz~lA}9TygAL?oue)F
z2^FbYSF^6DVTO0F?|LtrZ5iAjQ||Cb%#|1Pdr{mbNIV+pj{vOOdJQ|XOatCs^>7_)
zb|ED}-Ka;=dW=c-=58zbCDF%UHan2
z^MB$cd(j8|h>O1ZlB_3ht)q;oJecN92BxbOs{M)Qq>qY=JV3o?8)|t(>^t9n6-^1_
zy6VQx78()W;y^T;&%iwtZ$gD_)nAbcmWy1YjV|a$nH{X*Ezyhj6FMOP6=7txKn;WU
z1bGpSZ7yab&JJ`wF0(lRPBqkaZ*WB2bS{QG53y<5lWWoI^W*kxbx@+_R8J}!!>;#%
zT%Z&|Z{>)W7-i+x-YyYu{Tct7{FOEXp$Y$Y1zTrX(e2r)t!KswLZ6c#yi#NO2CwO_
z0Hvc_CF!C}%n@QC?~Wl@`JqQgPXH8<46Q7iNwkN|3L-+IMUGN7Da;cqg?RS>gDLeyV@
z_EFysIPuPiJqYWJZ3%I~qf6agP)&EDW6KSe=z)Zse5SD~iaYIa2Hi&t=Kp>Ti@;s>K-$uSIJ3Qz3ny(E`xY(fxLt%X65zW2vKO
z7|s^;d?g+Hqts4ranIlU-w0IvcktQ&4QTxTaSW9I2!H**CAKrRF|F+NK6CTDKcz!U
zb=%UYyx68<6W$-!lIyZV{Mp!6CQ4IFQ&CQ?Ky#K~vE<6@R}KHPvU|Z@>$BqD*0H|=
z{41(WEN4o}zs!D=vjDnXFB=%J%XtY+Ah#Ny#uW-(g%HjK%@W!wtylOc*>Gx3l(8sY
zC7^YHmgdFjhuInZVrGscDSCR1^%+j4lHaug?6xZO_vx;iuYmD-+uw~GwZptrcQ#e5
z(FucQ7M?NnL(T7!3$Hg-`b()1?>*Z}o>zIwBqiEN65hg*`;np=;m>*X8|6Ggs
zSq0(04;`M4Fk2hoUKL~zuzF1=Z`V71XT9EaD}!w)gngQ1wz4lu9;o`beYrE0tx;UW
zV2T>xhV;UX0eBOYa)i7j7z-7dz4wUQepb=+4{{y6ZgRmYs1f$=o>JEPXwtuzz!=0_
zZCSvl^*p{sTxbHRV_b1M$O)}i24XeZyarD7>xKhXg=d2VyeS8R`R%1E
zpdjX9Tyv|R^z@DPgWE`lD@`H(hq4Z}_n>pWa$1jfNqTd%6y-cjI3wcYF!*_6MAg-w
zso=oCDdFC9kkAPKOFF&t@jXMm7o@yOTdTfjjmJ7_{yXlU-5;Z)EMWyc3UJ5+mA1Q@
z{n(01$F_DMQ*S&%j3#~U*m6j@2no?IE+gIklJR9$Rqb>6oCg&$;x<;-wwXM1YBcZo
zQ;QX_bV1zqT2ri{ZGOjXUOH^nq5RdBzP@Jj0=h
zJtN^_sO-Bo+_7tkvT`adL~UsK%`~@QxqTLM*tBzYv2^$%%l&P}Z5bQ(uA>3GOqq#m
zaN&h@e0XjhB*A1~iDKeJIDzQdic9*ahJiWf%u4hhyZdihj1e;ox7bg74bf49wCgM9
zPs9X`MiWRVme|K*RA@MzW@VoF&nE3Q9JoX~A6b@`6OC3KR%B)h2V3B5f3M8kXu6d!jyQZk6EWPk
zhTd>jRNSN)e!6h<+JkCiTAzhD2#cP>?aAPSohdly=I@@la;tZ_?>x-uYSDDhc|ZL*
z29`s%)1RK)=K
zA@2!98X?MCaW5W0wly?%8e2%6KNqH7}7S&Lz(B63h;8
zDag*3wz)vfpysqHVoq59y4YucHsoDYb)=#6$|LW|Q;y$*LqE{B>hc^)rnIq>JB!R!
zzzDXKpMoolM~KL3KX7kz=t8XZygbTSFfugnhLwYKEcL0X+9o)?&8wp!eMfVLd;LKvQvfdK4w#IQ2bb+582
zA98|^(D`m+8NcwSz77I@r@brg(bCR*uD5SuC$7?T3qYtG+D~eV&7HlRg4j@|*`j+5
zu_adNrzPeegY_-W+(Ufi6RJNH;FF#u9P@rXMB%HHUqEjU`x>
zSk9hN|2^ZFaSkCL8^zO-mL`51o95%N5}!_O!MC5MWj!j(<34^)O>Z${^5AF;uI1P4
zU-jL|FLzw2>rv{OzIQ_9**5ip=m0U4)|a=}e3o2KF$(XQf3e*7(PHel)&=@#H9^$pm&KVt_k#t)b>f@b7nClv*Bf+JXHIN*&@h{VUZg}FC?BY%~6be~k=CcU?wKhAuB5C)N
zzWa!OTGILl^mp}UU9ZOy7s76cppPL5n8KX5eoO0(H6&!UP6HXF!Qie^JWE8#yL
zrKqYFNtM42g`#NFS_Ct^?W5Ry72LVdizieVLxf!tzPLP6QpU#NGV5$&dGAg|MayaB
zGpyu5i_-TPX+l<+Bw6Qskxys(44P-8)Mpi4uthu9j=HY@baX`FPkiw+Vck=wrZdXg
zB#&jPJ(A1RsOW^G)OqcAF|S1D_i2PXbLBM}Be(-K=vc{pQ^^(Ky-U^H3}|pBF2R4%
zJ@Ygiyd~H}io*1)bv9iZoOGcpCegBVl_1jS=;$d61k0|sjeH75O?3((b};N+!2Od|
zpxdFHP#xL0PgR?sg~N|i>4{K7?C7ybnyzAwmg*G9yjJYm;lJByCzq}(w>-G}(RtQI
zbX=YKhI98EqCp?Dv@CEUjhp-xR-`t!`t6B1d&XW@7RNn5PeY;
zGkc_*N0xsOnbOe7E4!|}nQ6*QD>>j!v>lQ%=ns@JE3kh*4gkxnjERbif
z8Bh`M?fzkm9M_p#=i`*o8GOuNfdd9q*l#c;D(m6sDUc12ar>=Ag
z5^7wR5a|hf3D0ijm1HAh*h&t3y5QIxrtd$#
zfWC3QNJ-?K<&0`hEIcAYOv|b0NB&6i3)1?E-AWHh2Ad9?3H(S;u&VFCGmmW?3)4SD
zm=R`_`jqwbh_QTmuzWxyD)mROj*48eA35mi10NP@XH<|L(nKQp=PldI+@>Vfg!%6Y
z4z$sTjdA7vFz4see8J(@0iX^~;N{_K9`M4XQHvUbn3A=WRD~6?;k?Udw*sspogWhc
zE0|Bjq{vVBMy?%wK>0@>)Gmb+1O|EC&TgoF&1LZ@tdGXiPBp+BMq#6J0bnNSsOo?~
zsY^Z|AC;na8^aTbqPn-wq~WvmCK=I_6yn-3DW9?!e<70wGFpu~agGU?>R(|jl$18G
zafV)sJX&!@0jdgTtaCDjle%#!T#JZ&gS*c7VJXFK#48{WZqdbIIv>C*p~R$gtx+h{
zD#+Y{KAV$r-yYH@OS_(?tGtTmj=hcbBCI1-kRU2=AA)aW1+dvnQHtAU+3U1b>ZZ(exUA_3PdAB7-vOro6BeZS=O^jvLG74rI{aEm*?YSiettNjN0P
zi2>VZKLfHjb0JcQ49Y+INrL8;(XZ@8uHL{H;}0%a+zp#vtqZHQOTX>9{wcFT-nIW|
zeZ(Qw^5t}!CUYJY9`xctgSx*r9qWKhP<};c-)WcwsZG2fl||-0{hk=}L=n>1-K!(E
z64UGYdDYI)$BM}{Ux$zK2@{#grpM}AQ13X-c|cq$1<|SbV7*Bv?2O|9R-LrYBKBa8b|4i{RpWyedjWmso`!pX
zIP@9rM8)&df1^&2sb?PRFHxmk%h=NY&eZf-nA5TSt$9Pb)Dl
zp2d8Oow|jGp*_W>Y2eYQX^oaNr}LZ^M8v;r8S)FxZsG_?z
zSl7Y{B+DUTEWPzv`OCY-{5&%G{@6kMYaO^?7^Z8Qbf?cq9-rd0scY~Y_*N7dGR}-tEmS7;+
zolhE14qvytzr2Yh!k+7jyJ!(6WK&YLURP2*Ua#+#KR&s<`u2bo^CJZDYm&5->%I>D
zyf{(n=muWb2&1TTYKK*8W->;@Sk}47tJg*y&xLoJAOzx>#Jz{Y8!vAq%>4HyP4R5n
zdxd&xsAyaKs{!{h0!N|MMyfGY$0vNR(A?aT(f*AxD>2ArH$!2G!EHUwmzJ?F%=Yq`
zubOdXq5^K%@;mxJ&#o0cpI_+)N(rL$_Yb?LsQOj>ImWycYLwH!Z`CLJoGNviyW$!N
zpZ8;tj_p8n9&|{gG~~VZm1}i$ZV$Y*!|8zSjk+a&D-xI^^TX-zCPikbSzEDRiZa1~
z73a22%O5fk*X|>*{Q$5)?d;3_#}uC=Y;acy3_nf-XqqD0_TS|(w)@_9dyRl5k+uOl
z=Tm`m7y8+ydviVp^8A*u!@LpkSB;eE_IS@m31FOxYhLS^K-a>YpHN>s+Iov1&FA}6
z4w3S}pt41F$mF-}&B4`Oj05jCSW%3zRJCHlc%+_`X$ca#zw^G_Xp8+y-B3`{jD647
z_m+0=l?b5sWJaG!z&uJeW@F+`d}-88Yl`1KY2nli&a*PBr+&17iH&m3`Je_rAi}%@
z-nW3&8xrg{(L|kr<*_uAc(V!DAg?DHni@zW@AD^z!Z;o7644WsFnpZ@jaP^76h>wf
zJmEsuiVl7`L_Q`FAMOy#5GYTLpIbIR{5PmPoW@7eb2jMg$6au@hK2?Kmvke9PW%@s
zlBhk^9{?`yMk$I)l{C3=4r
z=u7%aW4SUhs*XTOYt&d@YAg*5q@CYU%KpXM2B7NOOWP4QasCL}Z1Vr(RTh8TdcN#F
zRsfd!neXj%>Fy&@a*GIwe&0CaN7TwEOh78wa$p=|^h9pt`lRLa4r#`YXr({odoPjm
zN(n-6=YoXz$HieH$i!rh=t|t(XXl`D)XOl_E6aN*gZNoE(g^++r8P>8QNO^JpIT#K
zYkr7IOxaX;YqA@HaYe7OW(*AHQTeK{3xYq#`7)<}ke*KBr8WQ_$dJosy?*vX8IQD!
zbcv?Ufg1|bS$
z7OgNMepIcxgf&%3w)5yQ>W*9*`%W0NzkbAF
zRWqxFhAC>f(X5prY3Ph@ihoc^%3)I+DxOl7PI^Gi2JXHVe{A>S&uu9Tmv2U9ON?Ls
zz|=IG?A-HE;(6m_h4z%NEe}cIiHSDr)%)I&x_Pz%-&WR^eU|T6F-n;ml+|=7yTwSN
zP%{DTgUp18$zHHDtk&zH&5-da>J(ZhF;r8Mk#S$v%RonqNBq9ev+{--EWq+crWO+K
z?Dt=4@_P1W99uS;H#;?D_CAH2WwieBsZ&w@PB>~=%Y&-`K_zCm5&%=pjwMd{k@^{0
zI#Q$Tk|8;LOfSA=0&6aHUSuu?d$6Ve1%;Xo2}X9pH;3v_kv56gci^eOLj`$)4<8?+
z1Z*M=0Lj9qQJa)b15(t)o${l{ha4p{b7FvLHD66qfLGX`!9FRLPq4biHMt-wy4bRDh{uC
z*dI4alZob)T$botI~kNEOr#Drc3dhd*=phN`yhVWU1G7ReJ0~Zj)V1mpnK;4HrZsf
z{UfVcZMuK=fIWeup$+lNP}HUwvGFUoesFdv%ZLW`1*NNha+B~&FKh5wmdBpwPTLr5
z_gsp!8RYw(C@9~fO-|<(q8dP6mgW|N-O^pWrjp-rgDnY_H$}-#0WUAlO*Yv~NS(R5
zVHxxec7Fz219Dd0>6!Y`yS&|AbbF70BEgE-{OxQX)gmMiOm%kgx@0MLaEaOe;PXA;
zV{1{k7GhM~G4hNVI?2$aNduDm6$o8{^eHd+L`nK^w!3avP)THs*}%NU?DxOVn#+wD
zX$4ZV3WluTF1w`q6jZISspR+H)Tnrpm5uAPkK;ST=oMOSIf1bI3Qv58BaK~ZThq|dp^PVOxas?Dn6QEd
z#7A7>H=I_6)ToHyoN3yn%<>qew@20f?FSqzF<#Rom(W;qD9I_S#jq&QQ@j5>#L|;#g;Sv0T@Sfbm6qSc3>0-}Q6buZ55XCBxdKD1mZCKmZQ#ZP~
zo5qfH57@btEpOHZGmLxd?YAWFw!J*_Vf<-N?hR5gOM;8Au%>Q=_ZFL1jri6#AwqYW
zkwJP%UL=D!L&TgX$ewkTq#Ar^scJli!4i5DcGO_$vLOd8SQrDlx{Q;~O(0maLsco5
z2&63Nl;agi>r2e+TVU70mV-#b7^op-++K6&o9(xP_NTvmHN5ZyUZOMD-dIVNG@rLM
z_0skYUb%zkr{7*~Hu?gp$+qk@r96c-gQu^c!qH8ZqK5W3(Z|!HMNQU9;0-av&-;N5
z<5btdX-E)QZpOl8xM=$Q>-Zhx-LW&DhM+a;7oS#K_xyy6wXb32^xR`Nck62-N`P^C
zP;XRWcdl7rIsU87}
zk8PXv7z7b}9=`uhK@zIcLv=#j=9myK6{n
zY{OQ246a)_s}>LfG`)_FOxlix8@HPd@=e+tef-3!s{)Un#b~Z*
zzi14yF*@5A+&tLc+->no>C+j45=t
znwPb~175c)5p2FvI>
zqLAOQ`gRNX)dOFQ!oR}#ykl@jEHz{dpwKpg1&2p(e*
z4NdHvp_HPBiKXO}KOYYYU19OvL?*$%m%&6Btx!M9ewi1qHDbUHZBb6+!_v0J!QFpy`H(b9F{mt0HX-Eak?xTrZloq;^t;ZgG^=)P1T!>VrG&
z!E%<57rBqrV`ScwPsWwk6lRwB)@_cDd8-bJn9iTvvdvnsm>#>ppsaDh2zv$@sUWx&
zSm_#9ZFFdq@&`%9cj_Ldt4nQW*v9qlWs9Xc4jJBVA9ufjDM{&UN{f!Qv=Dwo2l-+*
zizC;^(KTwOc;78Wy6v!o%Oe$drzR;#YoKS9=$oPpDhjm_OH{{HeDPoO3yulCCd(m}
z5P;la1S(DMn=o=x&*&}KPOVk?Yn{~IZ*U{zyD_h;uAGT*672ms#x2L1>|ZGXQ*`Tr
zY^}Y^D#J}Ud61-Wxttx3WXAG2F`{z
zr~8?`6R&d!UyovD?hI(QG}c@&8-mTvGc@HhF1^X%DKTd-5+oN50eVduyIr63`UKle
z7+YJ3S`|;Pu|(8mV(I1bpawZza=&9ZtON2C{-&~pj;}5CSmR)e>a1QbWxf4jj)fiF
zUvn@t3WEP&t!-rlDhZ^`t}X3|b$rlN|M?};;4{rKd7zYCCIpdz(3z|M{(%s^hK+FC
zhjutH&?Bp!pWj@mVub+#>MXYph>UIsr*z?5bc*tJw8q7z6Ll}(<7^lasgoj7pIf~>
zWUw*p6^40yYW+69978K^dqV0EazHO48IaKYZZ7LG$(q%pz9Tlae_3pssg1Jd%lQb!oa4)cWIR|fRB~S
z28(dh)>i63)irQSC&J;f8c(9@3=K=^{50(1PNd4X)66%T5HRXhSGXLGceRKF`Rlp#
zd#SH@|4f;;rYW9zqbK8cRUQX^9q-Qa;?@+IE_?jvqi6NFktF5lLj~UhX%~-=pJ$RZ
zJN$^tp+wKznj|OvhTa*WgIw9u+c{|Hm?e2gvP$;-D9~{lmf1K`iS|U~AkB;VN(%L_
zxBT<)>9`Q*Mfhs#=<;Mh45KeNuC;#HtG>%8Gn0VS6JxmcR(yIxd90m!*9FC!@Imvt
zN3^%QcD{l#YkHD(tp*0dmsjc%K)d&_j_63BdW`HbfImu2lT8xJAxzdMd2N2h=bhmNWgm1
zzH`B?^GS5|av(m^twV~`n(O-{fq|}7^Wu4d;xqi5$#B`e;?+Q@PcdJUAgFp*7}AlR
z6kO;gcKX80*q^?{M3%v#VHrlk?qf2ek%Ypu0V`a3W8x@_pY-IOwl_kVl64ZzV`^}0nv4`Z%vZ?6W7w#^G{rQ3e+d^h;oRqTM{
zFQ?x}*Tz9#srcEI5s380!$Lmh-tzbVosNb_WE+*4?w`qrC0Jc%0{bea2}!%IUImvo
zDfoHNJ^O6ddLLbyQbC7`3bo^O?y*(8Xdv)g6aY-xhIm?%BZVLyavM%vc7#tV$S8va
z#Z(7QI|hh0RgAoq^HAQ;kiXW-X)I8sCnWu*epH?q!NIpCg~SkgaOpknsn;Yj)g(rb
zqarq`EjMJF2@vMv9FwHxO}To(5mzD)j17TJ*XblLQ-StLgl3(A5**&SId=`f&w+_!fHZ{7n#N4wB-!Q=*qDuOw`^g1YK%y+3q(XvRyqsc_qNT
zY2c-cD#Dp&|FX=4-MUHh1Awi*H$!JS&qc2n%e1I6)l6N!CMezy+d=0&>#JzG4x
z!xqy1^QQvV?N~X@sFL=eGM%k{*IHxgAp|P1BOSTA?&N5C3!gJ_s;A+XX(EatVa-$cXs}pB`JMd0#PYKb?!&u
z*Q~r{$#JtrBv*GJ?^Hv+Pg0&FX>ab_Cy=p+q;uzPD`U1i$EGGgPj+3CB0I<;a7rlc
z^$Ftf7#L}s?8enJ+iJUYi(42gXzCbs&Du0Q7vS|cK^Ju&vTJh^*q+LW|22l#-@hfW
zpJ)PzFaQz0B&{4-W42i5;^w?f;1#Og>8Ti&!Xe?IP5_#u_Fo5@B+0WeFgi-?`@dea
zK569XZ>j+bO6)_*M|R2I??zB#Ay1Mz$bB{niX@5tYXK9_)N|MUpUw}}iIMdlj{g>J
zs0rXdo)^(t(@Cx~l|NOHwZGv%G5$h0*7y{A!v4Us@iJm1F~)O${XKY?180zXrpY?Y
z>u(K|jEHvXrFd*?&;*ow&*NhxSyiX#-;#h4In!7!l~1Of?W;B~8>esnA|1dgjeLrR
zZnPGJE1nqgr}xurAVa;rdKO>Xqlad6LG9G$@i^FON*^|)tOYIZ2$fj;x7eV?@-CW%$8J_HHz5ji_
z*WVAJP~$%iio_%#_1(EcreskLkC=}-B5rji1Z&stdSt|^w%H6Uah~QLzV-~y
zee?I{?FYDA`~J(~TlND=NVB#*{gHjlt>QSgHK)Wu5jurtCC1E+0P>Xt{kq-{L~`#R
ze$!`I{j{8YKc1tb3#7KMPMmZqqMz8$yN;1PpHs$A_zMo5Jd)s*s6oF}Sw1Z+K2ZZE
zd!JP+_I_G$M;fXQWxT0bokxcVs=h*YaxPfy3z2J9ylO=e26eMJar<-YCga}OVehbAb84+t!?VS+6-3+&qv
zeJ?l?Y(qHW+_ynL-LVt~v~Gu>BW*_8ylLO5bLIyH23Cld&5;8qok_@iS~FNY{>uma
z$Vo~t#SU6>-TL{C(-g?EwOx?BjG<^X-7HdI10m7Vuhx1lkmYs{Tnk|19rOM)V0NbX
zFuLT^K&c)y((~h~FR48HUm9ChJ1
za)_(W-yt2PO#_UrJAMma$#y_y#w2ZTSW1TCABx6A@R2?K275_f-yKTXnr8D=xc$=m
zOImN_XK2oCRC^6Jef-HRkX@eMU)H=5_U>fGd1%>+5%_4e;c8E4rd{#?OTtT%-@%bn
z=~D;%157;TIPDX+K;Mck^8MFY26obydPvH!qHiF^w)
z`hB&|!*H$~tuiESt@Esy53R(I37)Z%7r!V@Q>m6MaBylUQ>*6wDMEse*M8K)a8=ef5P7EX;@h3wzjBM*=FYQjzKuHGRjTQJG}mpUGDrYLTbXJOh`
z%USYzNSGb(@06_VgdP6QaKKSeJBnfucwqaXXw<~HycAqD`!AT6P?6{`@RY<+n=
zl)LF|+0gdQ-BX2aqCNcZti
z>wose3DUjZXE(sNvl_mU2bvPZZ7jCC1`AlZ{zOM7%fjr4JVd^wOKw<-7coB!|0pe=
zLu!DH^^uj}GeVHs?up!#2LT2K#o)fcUwBQta1&-d1K;+20~;-l3!nkJc$HRw@hGw&
zF2Y|2I#NojcdFkE7T%gOs<<&{Qg);Kq@6^~)bP@Z|37kGTA?WCO&++etmE^2jU%#>
zift#ZNPX}&DZ3Y(AZHgjJ)B>aGX!6pGJP4OR^PgN^})*#56$UU3?4u7gtXZxnmRtC
zUxFe_*mj)(bzLUw4)fSZ&I|8ubsE+(8-LlkkL~caeiIvs_sS?A2A$r<T@S;h@n-f?TuA2g}Q-rDyjXM{mMwU7AImcTG>I3=J*HgKuRP
zL${}V`c?iZ+(+uIrHtD
z@38UoyYD)9z>gY>GpY|UqD@;A2lc{GL~Dct;|jS%(LG<%AHRldRM)LK~R6~
zEPN5Ox%ngSIL1~|m3*&~NvyHoXE_tLg56nrzI#|N!CJ|FseBPTCTf;|q1&50E$Nc3
zkA*=2XajHb{2pXGYqMtK*P5+L#E=vF4$H0o?piv4oBd(d{-z96FGaH2CUS0#GiEjB
zKbEW0@$2sHyx{=C`c7WHP@`A){Wz;98dY)9)=FG;*>f>xB%)>wvDH723-OSI`Dg8*
zSLobyt&DN4_Q;K^{_Mx#IxI1jNi#Rd?>s%VS0s^BH`Fvd3J!fr)8S*}6iC
z{s*^5a-A2CkKD8PU$+H38SxLQu$K7g2Pi>{gGZEMLbMrj6kv#Fq<#Zo
zEEE*#Ksym`ltBOdRIm6#r>cEP0Nvyk(OzE*gQsjPNs?Pgq2PCYzwRoHzeZ$GR6saS
z>tS5;V5`U2d_h$^8`~qbdmjgPpT(SkXB}sjPOQzm-&fu7Bj$M^JlpQaT|X}GV~AXA
zy2!TO`$h}`CA7Jnjh}(BH99jSLZTn(1;y;|cywzy5sCXKSJrnJI38Tn-fqhc8vmhJ
zTCENq1w1B{U2p&4Ava6o*>|-D%>bHk(M37N6K?fyadC@R!!}l}I%`W0-3*lcLyVka
zTRgpR6eFwPpa3p(vRqPUoyI>zR#eB8L*71;h?{0-YVObHum|5V$0|CqyQK_d``u0}
zxt{hj*GwKcj&DAPSI<6a=H)#p(gxS?=#Z2_j7Dip8scx3wBZ0EEbIw$u&cf$hB2Lj#2G=DX@aKyZY|!}?30
zZE3XJUmO5pAa2#t`$-hPVNL7q9;6}QIH_Az)p4Aa*b^wVmwR(Kn{duY_elt-E_yvp
za`#}cd^EJ5ib$)E*S?FIc_n}Tag?w1@XK$p+
zuG_PX1v4vIFP(gM$G3>{TA5FLSC>AzYx$8p^p7y<$Jv5HA5GV77Y)BItoB?)Ox+sa
zT!f3^N({w5ppNgM^R0jFwc!QJPsO%c?EAVfI87cKQk^dhpfdnmv+K7|^x#O|negj@
zl;xx?oS*tSYgY;}~x*K+vfBxuAM
zz?>4W6C)el7p$a)A5cUdng^Q-I-}
z5)MpiB>ZGJu&x$wwBUNw>B~
z19(8+bUM9QcHJA61^`}%n(D~!y>yBE?1J=$2Cav0>X4>N7P>un5FG;ld1-@5JeG0`
zQRiWWNoE=C4VLwngYcOhKcX0--k%H2e!(MvKnTRZWeu0s?&@&~?EBV}Wcw13ra@6i
z``^l~1>U-CxFsg!f&q5WuXGO?lJ@p9Y}c)R$cfxvE*99_StB8@azUYn_{+VU$2f`}
zwkxhpnj8Q>F*&v~KrRt9+6v|O%xRNhgMXkP1?b{dwshAR$c)p=BVM0|qto-l-AmOK
zEytO@=41uhw4@*gBo#rqPiv}*psv+(Cb|gGGkA76s<^_=@6+||guR{16?}8zAvExC
z=p|5@{RV9b5$!?oL;y=vE
z#wB~CR2(!4pVOAJ343L^AAx6lQ2@fEepHH1uF?1l%F2I(O}SK!zY{-M*gL9~(~<(u
zP^AKD?V!d7OKi~4=56Ta39BZh$oYK3_4#sPjZHAEGRU$hU*kZvH}yhzQo9E=L>Rr8
z2U=z2um^d^l#(e`>gVHi8Vh^r(3Lb&bLJ$gHHdu*1NIDKy`Ma%1h$mbbp43ns6^~}
zuHCZ5KGZWN@6}Q1dP9F2hgRe2&fRY*gJ7)|Mx&0e)6mYkCus?tk#&`nxYau-VDB}$
zUJv#av05|oo-LG6Cy5(QPHezs^CKc%a8WBqb6O7H%ovQPhY#s|&P!>f1WZy-0lBB#
zx3Nb4ln{2lWKY-BD=XE!BF*}IS20qmbT0M`2#!+}fh2TFw`2A3Ph%;i#9l`eRK6L_~
z3fL^O@CT(H1V;~Qff~~bj3cL1!U7R57m0i{^Z?ux|F%S#+T*NCu~OZzhjP=q@$iq<
z7rPHgAqzW;a4s<<^Up(jMOw8OBsfHC-azF--=#35a)Z{P6Mg%J;1CyMMoU5GY@S>P
z5msAg2i}W|;7$H|z~+%=S)?KlEZx`qAZseV?~l`;<2Age{`7oHDUUTlixq^Pj`Eg%
zP_eQ;-78{ll^aS;Xe~G{unadHU4Ab{Jsu2ZJa9Gn$9SDOUhWR7UbRQ@qS93qCF
zOLUG?*NTzRLU7~}MT`fTfd}}?bYNnrXJs+u-eql_m_&DhkK-T8M|}RcMw@vD%xTa|
zgOx*RUNdi#%=c-5+r3^aomMpUad!#)2v5X^$&t6@0qOlKN|Z0VNdk@KGRfuUUz10zx)^N@tDM9WmdiA0Favb
zz!SGPTJySa!Apzahf*w@n-O5SyYRmb&^)JpTqm!9Nf+hzbZyxKIA`|DIg5SfUwjzR
zbs(#D|NqfvsOM>5rJ?<5-Y)c;Lw=?MBA
zby{4rliz9kEwxy3A@)X-+fcpnSxG^0d8H8N<^^jL)89pwbt<)u;THQ+@nL2H3jiQM
z`H(;#&_^FX5|Akz0(vg@(~tMJEAwY3PgPKuEVTG-wvLqzL8fM^Lc;!vio|_KUR&JsU*komPZ^f4fmFa&
z#GDJ=C%NU44oFG}x%CR{K=Zv4b-Z7)DD7MCU;kMD8N~XVe?jPhl&GNz2F7ih4pEBH
zOmc&?SaCzhga818!9|4((`yrNARY(342r90p_sxYIWUFCI)xcN08Eh9$rVxM(kfL}
z1cAshTe|`Hf_+ip5V!j+i-O*4}s6ABcyms;={(LIy|Y?Zn}0-fJD^HjfZlQtmnt@=U(`
zceN`Tzz9H4am9804am9!-~i%Eky|2v>SA8T2~R}i2sIa)#-^U|4kDGl<7}H&1F;|t
zm49>#7-})f;$#ux=-Z7RTad&*FjH%LbIBGy3Nn39cDLJ1vJBQvY?@6+16?A9Y6_{a{;r
z9xk`I-J%ZiMN)#&!mFovh#2-t2LH2182iu>6esVFShS(_Iaf@_>4{$Tyi
z?&1(cM`)O)4MJ|;#6t(TO}F-9q)7O~Y&~-_rj)ygW&(;UcFn8T2_ysz2fPQd#)wr0
z2Z@;IU16LAVMunPrb{Vyk#g=J{;rvEb|j?WFq?>_oO=Qc0U`z@_>g5`C3m;&LE_O%
zIropkMfg^S-d)xfBC{Tsf=hoW%Id+zRia)R7wNM;MGFh{Nw1-W^G4Gk>yrB3b#d3f
zdt=O_;a#b?eTe>2hp0(Xj+?-ah$3jPun}~9%65N4Rn5EBvH&^;oyRr^1BL}|Dk)al
zXW;;>M_y@ib+D`!_T?^*H4Hv!BfR?#7I;UimBUjah$)(J9uV>^5(Y7T2G5o1DCos^
z4u~`&HEIZXre3n+F;l2}$ZdUFM#j3D;wFlOQ1T!l<|I*W;(>jg$$o9AzHN{H3xnb}
zLf%sZGzUn8Yj3`)t|WBerYcnn#T-3
zJ;=fONY+461MyY6AG+g~ifmc7+7$<4(Q(WvX%k5HV+uw#jp9Mx;k;NZ30tjCIKjur
zqk)_~&}0Owqbans@!p5vF`_!C8|TFLDbah1qTJ>a8@}e2Q8&
z0H>qCEC%`rGAuKfB>a+YcHoS+U}S_~X7?Bdf?!w4)zZ!hppjjRS}hJ?R*>Ckmfu>U
zMg-qIf*3s{Vk9P)bJ(rG-kEdJc=%ZN@SaXkpn`zAyYiU)iL-Y-7^!K&9l{M;J~#xDn3i2mq~cS8)4KAzRH
zpaiu22f(rY_8fK;r7uP<5W_vG$)m71S!VE9V4u2g+%~a>18vEXps2+Fa#DMN6G1o(
zdy}Fje!n)y|DNc7fPn8z*b203VXt1ywoc9J-Bd70%SuO{7&w>
z&ud;oJHeUASe_g!pWEYwG>y%#4AMhNZd;XowQnj_na#b!%i9uFRaO0VR3=@+K4EP9
zxXLYLcKX39RGt5})WicU$MX^KBUXx{x8^Uz=;_Mt#5k~RK0kNoZO+adcZE84^~Y4b
z{fo|N8Xgxy`OZc3-qAU@|47{ULL8k>zNBeG*k+ez;g7q5N3T~TjN3-=_*aOtzAS`*
z3gjtXVIQbZ6Q2w=5=@FW34$y*`;KK1(5dlbx7_WMA
zje(6yTA{v#4*1kSl;}9IDrFs4hHS?;G
z5yf!;O8dugxc+}|J=Pc|I-KNvZ9i2B<9RQXwp@9urF|ban*{SQ_`7xK?na+tD|y{I
zY8F6~*VV+a&-X3>poPZbKX4`?1qDK?Nuo#&1QcLE+ljzQhy?&(>h>T
zvN?{0Tia2aIZ~$N_T#TC&+Bhp&M0hI6QpXqm^XmvoWw~f+eih=RRW7>Zn9#pf&C%4
z2@3tq_e?`grZ=OE?5wKy9ay+tf1@-TMh~;Fjz!&A;m&01>6ekyeSPucAF7+z=e}lD
ze@YtZiJrg^59DREH%0l#^?=WWbW_^e9zrT2-7=8wHCKk02bm5$AYc7WM|ZB1LA+Ij
z_M-LpO=B46{*sN4TEP>DY4CCeW|H4j{)FzcnRZ3Mr-_`{z^(-gYS~Lyt>dkCS?Izo
z0U^!hlGL--O?7f>-|p!2a$TcW5UMdQMLrCE%u32-5V;TJk1GGVogHgf2h$(c5rg`*
zsTTxAE;k79hqK+Get
z)m*ncam6_rTs)h2EUzE}y4KKo$UxikTUzcXI|8RruMq>);4-8%GVmY3R6ciQ7s^U0
zw=*2brdLha_;`{xOH7_NwBu8@Bt+Xh&2_32FpZ+)DER?A+wbReBWyxnb|oGBW-p!l
z<2ibwrfHJ@=@n$;_WrE=kdp(2Wz{Nlld+EvIK58&(6~vXa24X;Fo=VH0Ii|_@jJ*|
zXpD5H`@lcEAGG
z+QqpSlyF4`^!(GV2AzbmR|S;HH^3tq#yz~OofW8-ry3@%8Nj8~Er;x)mSCjkqyZ^w
zg3QhGb2`kW0tql_2x|m)haj8Zwfi#2H2HoCD98dpW&o*!mczu`HJ*=(#l`fH^atX)
zQ^9pB+;QSj2UK+YcA}kkU|Y}s1?m-d5kt?R1d`(Y$I`IaNFtKT?z2Z;UE}c}fPyG)
zB1OBYMkS6#U<4pV(gnB@4tA*j<+S_HQrh
zqjxWfs3w)fKe{5Yv($~yiQBh*ptK21KdWfUo;heZNb*gf`wb{9R41HNu27-;HF^RjQqeEd~8Ug^ox)N!wvR_Qwd%oHQpI5-;34G_f+#dWS3}UvFUH7
z%^8lw>*hLncy5D(SqR)4I9@L(_)gUV$G;{z*;JzzV=t>5nExksy2XSZq-p*MNu*VK
zBFJdb5cT|jVum|ZRE0KjUIy7KGcGnG`|JhZq?42#Du>Xe102m(4{NF2jBqnz6B*J$
zYYT3qF$fkK%p3~BfdLNt%%{_IDzf@C_(<*i{Cj8RV=xMrT-g>!qJ@in`s{OGQ;x8pclWGAg)d
z!(Ac$NXo@vCE+r!VO-tjy8nx`L8}%U7*c_EYv^e;B;*h^3Vf6~^-I)|#QOF^oyrMd
zz3%M{-XZHWHBag^k&R<`AbqV`E$kk!SjWpQ;E+KsNSqf7wHx_qfZpp_Uli;WMH>0;wfU^7AEj;0#@q-g
zrtvZ(5A)#e_lXO`$4*pc>A-J;T&V5u`Ixdcf~_(@^7pSz^4=y^FT~uN5Syrhm|=6?
zO0cn?N6pPccv``GQcik!Jtk{RrC*X#>C26ncfmJ_V;hIDz?1-dc&^MBsi)-srvOL{
z(5sTS=bAl72%Tv`i!XEwA}VuqE4%+#9bX|C8~rkl?weoRiMgV??L4hIC%d@QY}K+u
zaArA^{yU|(@|ft{5g9}uT7X&a6Ff7^^fKT1zYzbQpbghKMV1c@krVA(C#XFv7t`!C
z9SKCJMya&0D1Bn;_k`(A9@_
zQ@Iu|@n^Qq|BgrMEmuzIOOU)F1YSNZ?>>H)G{W3VwrrPFrLjX+^Un3en{AP<1o~k?
zLDVHiI*R!ayGbRY$bNdcwSGuyP;JhK4$X~I<64)FWfMSBL0$>5#NqWMi~=$5AliFT
zUfs-?{RcYd*|5lCVD~V>^0sP0G8OhI9yZU19|(IG;`j&4l!w3f{}_InjCnxB$_=4a
z3`TrLa<|%2sGrSx9cxJt;5H9d%8^%}%NjLn|B8`MSj-5{Zh0C)qP@a9dV^Qm^@Af;?7VeNoTwN7Y
z%=hrAo;MH~jtA2xPnr(}`063Tifd!A-FVwCL~WEoi8zw{xoJ4kxU+mp9;7dq6r|~r
zmcE`;aD89yp5wVvv9GV)COfT!pV&epWt61OzUq_S{&ojGAs?qZrOubx%`jZL#8}a4
z)j}%u52?nr#Rq6+aQF0!I&7%BSQ9?BqE-7|(1p
z*-&)M&4A*PP4-AxcaZoz6W55pSgBArS^)v4Z_Ny~iLE?|L91~A)1xVMi3#xW5W_%mb0MtBj~`f-mL$e0$1S4#qQ_FtUVdH^ehR63k*Bm$;(x
zwTz5I9WHea+2I-?moVNEqY>(Ueq`6MjtY{Ie{-w5DaN~nm8AYD`plsSx>MNY+_UUQ-1
zSp+i16TR1c0WBv^&HI|mZm&pmOm4y090Odg&=KaNL6)zWJuwEHPlEBi22M@32
z4^=t^8A_G=4m(jLxj{`(M;2#bXK2UgHytjek_Z903|RmS)5&ZbW
z;tSe_lN{#SlPef6I;Aum6ACL&8Liqeh2W6*RN}1zScG|kQURbJfp?VV5L25x9kO~)
zernC_Qx`rGOr=%AQ3WjVz@GBxVjLhQqdla@lMy`#o68RI73bR7*nUTASigd#Jlfm+Tw2;B0UvUnNVTBB1Bev
zDbcf4E#sqQHDB+)xrRX^%32D8UYh)Fh|c-!qwdv6cJ><56_5?8T6PFeZ(hJLP83kz
zywN;R;sn=}WXY!u7x{vq{eV3uHI*yn{S^EYy_0R;`1T(cfF
z-=xQcE#{qfIS#i6HT^YP)`Nha{C#LqD=eeG#!Qy!iK>Ik<(@%nPHL{OhaT*y4r}J5
zNnY*s3Mq7qFQ>u-}zz%#33pQ4!e|w+h@xoF?jcc5$XAJ8w5#r95)tKP0HxD`o3V9U1Cmz}@b5cJCK#Z0L)e#DRguoQc%UrjR$(01c$p{0=SM5h|V
z;ug23DP(cdtAkjbp9eF3TI^%&b4XLQ(MUInbueB@pLafCv_2p^gl>|#=mk8{C2;wl
z)+Lpe-P+YcA=+x?_8p#l4%R32ZUKXd8GO;-FpzJpbI4Qbp+r7zymlS37c`>CXk0ZyHn&thNtE2biV$09$joY
zX}<;~EPee4vl-IcBeC3S6NePqGiXD+eO5e4lc%-OHX{K_FQB%X%LS^x@BgQ!{A_mY
zBmQLa1s_Zsx~t^DsrtW=TuvDc7eYD
zll!6oiS(xrA~If{b|)VRv;+tL<<6EfO1cH&soLuJrXE
zyEq$^L@P!k8>FtbK08&CyWy3&M3$QSiz*e`s&pSQAAkl0!`61Hj|3h&9>oy!gtK!0
zBPfwUqD>GPgr0n@Jghb!D%rr}A#??4@{+a`@xFxk)${#!VeLUdKF!xY&#c`U@obTy
zAr1%RWrc;*MYkA3x=
ztEO8UWa#!x6nF3{@+__@EgP0OQwjX9kKTH~vdaqQff*+?N}jD}06pwZdaEy3
zX~ssSkbV#(Wn=2762r!wQ^4&FxzGULknu<|dayHlD8MLtwq-G3dxK9v;8
zG#214sJvbMW82Lz-fg>A#Bt6wIb0B(V-*Y4|DlnNU5Q7lA8(m~$hxBlyhz>1xzHdK
zS1D4JxMll%`gPQC*XEoEsQ2UZk(SzCazen$fXCAWl?Hq9pcg|Ym~1y3GcUy&^3od^
zw1AqGARo5us9ls%R74rcydGo4^}AWNEykbqnp)!pYCr`k>r)o2XoU<;f0896Le4Li
z&%YDY$o%b-h;
z{8)Rkr^ZBn2uQQB8ixp8L0$c*45^G``FiK%6A$^Ui|5U2#GEdP(lty>;pqt~nQX34
z@NQclR=O+X9?&xTndI9vhaL~QyHLBw%yCnB?FcVmOr(F|Z2_~~>Zj5-2aotEM#hMK
zT2i38lAG4!>o
zWNzDaq4WnFkAL-5gDZPM}u+
z$xLKc3ce+&8`GlUH+;cCWwY~xQrYPD$BTcV%Kyb-DWnr+TP1Y_Fy7~@mOU&_%=sLv
z02YR{{wFfO-Jzq%~m-(bQ;
zwb43AXB8jyKkXWilwJRybog_WhwEtNWA;-9gCqryQ(U>#YwtY;KbwXQI>u^lB)ISC
z5SF?-1)v@?XOp}>{^i_|F+CR_7|2(%A`5rA{eJ{THHz0pHDK%fyMFXwpL36Adr!Dc
zj$Cqn-**!C-gvwuUhiAB&q?`PbNR+<4*r)k%UlDIMC4R~DAar%=kVeByPX{;xeg9y
z;!t69(L=*8e>hQZ-eI|y7L}$G27U)iv%j+2>FR3F4z`47`t_%zoS6i^{%23AZH#y&
z1p5k_!n&J~H5mD+>3)DvXcAH^Lp
z%E0m;eCKUkBP+Z8@yt2Z$i|{>OXTiD*99s~3;(b!6FD8Xx^GGyGVMa)6wK!K7`|OuT8Fj104j4~-qUn$Y
zKBB=I?VBOdhGb#zHf&a@qAmDLyO&3yYwFB{u!{Zgnyy!P~V_$t&>GUetNG%f1X-i|F;F!9*Z>o>u-l79O3K_PlWTjpm*6MUI6LiwVLkL7IlFKUf_
z#tcWw)@!`CRFp!hOCb3U*Seuv>CG5h`_DeDxg#&jy?PCG+f0bTRlzO#Wjnr}+@E_N
z%n}rTf3LhNc`{V+$_RJ)7Pqz(s~{Ch7jPf?B<|{ECqmsb?ndQ)W{Y+=Wp(9xJA(7g
ze{<;yr5SxS+>>(b9%3$e2#b;Cu@bGEBuXqg+v!?iy`57`{(5WiuO&Juz55>==SMQC
zF3fXUL^4t#xmjP%SqOROl3qX&(-8WqB_Hm^T6^;_ivFt7^o>?ysCOt$@O%uJU6qVm
zo0wbOc^eNhx92MMiP&q!yYG=QMA9*fdjoUv=e4C$0*B8$PteFv?3p@fx8WB#T)fc0
z28{x;P>knmYz7zH?UM+8pK)&2MuW10XH-_f9c{>qa;6{^ahgynr2m0D=yv8Yx3&$0
zX5V*;zmhyw8l!l$^)OnQ(BCMgUlih;lR#|9(%R?X(Vca~Mddmbkr%?CB`K-JFGjr%
zjq<^X=CVIadh3;g!hh-Jyx-4{Rze+I*kVGUtbf3YkUM%y8j5;}SC9`{M@YLi
zBKcS}4Qo!d)Z^>}Zfu^2lP%#&_;;-ZIwXqZ)+EpK@oL`UmWCyNP%Z{ajCtS7j$j#;
z(lA$%%(x5C>Jq+2O6GRz@RFQ{kYjnrcsyZrsxT2US_6A4uU3
zN5>@tjiaXNEcX1hw%A<)GJ1MfHp6fpsL78dHs4oXX--jDNior(&oc(D%Sd)&XO&za
zi+`ICa~*Pk{5x{v*Gc93dj*eQdTgmeg?^eQZaBVxuUFYVl^8icvXI+i^f@WM5TYgD
zv#t1QE$_{RtE7DobHTJE=v5jsiCAMjX9s@*RD<>&0p(zq`8k{vQ{8+de~G=1!`6e(
z58SxKmu%6b&yau0-SEZO8-d$My(BT#ia~A<+Y!N`0qHtJ#rA&3)%TmfgNv%?@8<<>
z|1O<*#MN*O`3?7*VG`S?JUrq(XP;a3g}3Or^ls{oL+w#Z$&wt@LzE?C-jiFs0tMY?
zltbKVG{+J4b0pY%7>iS)1hXpW*d-|vGp8%PZ0RgZau=_$$xtyt1`;+NK3?ZO=bmGu
z;Y5fH8sc{Kc8+r1$misBs1%0YCxrFlV$_+z3nCfDS!;XgsnTBgHy)okj@C#rxsrdU
zUjNDRF7URyx?VN9RVdnID_>vA16Ne=nUTg8FR9IT_ED1Wa2^!SxM_RgduJB0;1dVJc)gRX{$9#s*pE!@hp8r)Tp)-YQfYwE7s?;!70ZvR-8%}<}
z^)Oowiqio%Um>{$HL2A}Tv%`UU)M`euAPSay{k>bTp*G+ZxkWx>{Z~fwu=BOkh`Ih
ziTZ@|fNH%E1*`zXy|bP74av#ahwf$iY{V18_}&X<6L*_yMt*->u9U#qbBOoL90roc
z_f6-a?wy6Yrf2?%?j-1bX%4#E{IJz8LEdr$KR+9`lQ@3u9M;${-iW(Nz6zb^`9OO!
z;j9$#C)WQ(3%BphL2h2C5Fu5wWD7=WM1E5rJvNo>f7H0r@s*3wi)j<%_vgifJncKT
z?7o+4fU!NTJ0kzL+(t@e-bmh#`-sE%anXWs(u5ITO`KPrr{yZu+K$PlCD{$>&U8Ny
z+!K}5H{eB`)`nC_x2$uUVnvh?E>|rBJyA8QQL_{tPYEK;Nug!?Dqyz0p(h`UAhuXfLvnL<%VG(}4V7YK@
zJ@8b*1KH?T`8Bvr^$NF58u$A@*{tL`*J#g7zl+(~>XIS3{)u!o>+?wQGk?SiO)IM<
z`HwP_e$u>6!7v#`a%J{r9T|}w0{Q~Q>`;R-+?E=3saVK&TI+eJlsgh%q@qM^sCO{i
zui5)&05e5ZP+|S#8RvP+?I%z9-V9G0%IVJ;mfdjWnBS&7N5^>J0|^>S&uLVDyuBjE
zDc=6>#Ng_wst0K<(ogt&ChiRky&LF%n&9xcjWB2T7yLfBHk#dkxXt%#UDaa#m<+o|
z=q(PuUt!|UgovDdJ+qNO^)w-^2}nulI)OE{dOK=FU+F2)plIG@F$$)5QW2f%+v6T<
z*w@HMsuo2$AgJ_z}kh7U%F>c
z@_N6C`~{!5OxkkpLRWq#FA@@*4_>QAfk$JN8
z3@Ovi<&n(lfzIR6wrnA{>(g4S)y}IgrA?FVWFb^`M%Or#Hs)iMsi)Gk_<(+PAnaUC
zqfHH_c(Ca|Wu$Sl$6r9tgDhWbnWr6PClRIs1aWq6JU{$!CAr-bi6E2eA9j5Gd!Yxf
z(bSEnGlV&3cZTA?~#vT=|G}3!-FGc4UeBIPgJ{j7I*NQh*_d$qc
zBak<6kt|=MxxLZ;aN*B9bGdiOIjx;F{GGLChS7&DgPkI-^;>-hxuRn#cdjg|AZONq
z&SHAETB2+0JRjKRdNi=lfZh)4wbVQ}-)}$64KVU<^;#^CXY(Eb*=Y2mUpB?|fj%)<
zAoE67x(f4ZHK(v9J+*Z4*T{|wwuMDvD5<`QZdvCW5l@h|D6HJoW#F5dzYaogXwqU?
zoD?^-KWNN(&K+H=k(7P76|7QTJu1^4AD-T@NIKt?Y828`YnYFb*wo5ntk-$3%M55#
z0aI-2;J-OtmixRwrx3xzq6v73HGt&
z;h9{J#9(ep3M(=H`qX+2L31R2Z6B>V;!*<>y0Yu~zPZ+Z!)UbX=o3h>t^okn?^ntc
z6O(PwV{RO=`F)MxiZ+YuY(;L
z$}rq3JZ|gv8+S=6FyAOuXohv3?>R54`3u+vXZj`5KK1g1Na&LfHDSyNZ9o6sP0GS+
zUgyf>adOiYjNHXZ8RKI``GBtVm`Y3E*U*{MZ*6+p&wPF9C5}aU7#La6A0kUE#{sIA
zR?ll%IW9PPq3N&geSlPlZe8l)T9>!KiQ~DK>9lS6(nFQN-F>WK#?2&bjlOAoYTYm|
zXd_A7V$!IX_(MC(Vc@KC+e}o~tWV*k48^a>q8lyJv`EraK{BdI;guBOJ2>%h%!*`;
z<<$1@_d+9kzE~cl+iLkNzoyZZYjSS*aFMN9gT-8hl}{JjV5;G-{=XqkFQXi*Bg<5a%B{Px~Lsk`hybpPfm1gLg^X*a>7-3y)-Gouq{-
zK&lS^mchtPg;&s?bvjL)eaLSI{d#tTsTmq$#y-f)*vTp~+%dsfydPRihgIL^6E(^8LwOiYzX_zs}vJr2@x2FGz9RRH^u3|5g!bFsk?&mbDOJB4fK%eKVcX
z)+osNL@^_)b&E}-f(Svw2BDeccf%i&Fc+z&3f^=Fas|f|3s(tf$;l0Wx;aMKa4TS=ddh9!j*54ydq}%^It8!VoYkyjy
z4ss6)%*eU#2+i$1u`pp+Jo~7`J#uawC~YSaa48
zMtLRSLq9vPYt&Rrhm-vd?`XWGJvZGj!(u%`lF2h`LWfN+aoA5II6{rsG}E
zWH5<@$pW9
zsBskE99!bLZ$(1ad4RRiZxLl1uw~MBorUs`%U7D9_kN$)ur)P3E>&
zQ{ckfee?>~Wk%;yd3wLvWRgS=eh(%z_CO`}JG^oYe?hKiKkmAnvT7c`oS0Dp-W^l+
z^Hk}Lz7hUK<~LD;qd5e{eSWrxR^sBoHgBGwSJOzXe?rpqC!Gyuw;GT7&Mhm}-$+QD_(IbQ
zIxnS(R=OS*wbFAoMYtPzAt7u$8kg%&LaK*^X;fk{30Jju*V&$+#~6mRE0WCG9&c6x
zKp_X4a6yYAKU5Q{)MjxM|7wtNzuzw8?zrgESLYLo_QcPRrv+G=8!eKYHodM2nG1s7
z@+B##1|HD3pY-)qO_k<^-F%GSVPyb>NNB%~y}JBuAvgqn7dr_P97Af0GnLLda0iNK
zt|D|Gp5qwtcES7rPsKjDem0uv*#q%v^|#hspW;)u=L4P1>x)Jqb@5)}N7-qg_KzxS
z1&Zw-74v3S*-Mdjh>E3hVG-!?)h{ZV^%x}%_!mzQ&qPQGg0P)MH4q)JKK?BOcwwX)a&wT=nLCwKf1b*PXgXa!6yH
z8J~IEcBjaL4K+l{U4Hu$!OZeY%NVYd+PfsOQ~G|Xgr^iuCn}s!G}7hOtza0=N74>OMwEw=!VWZ
zvs~nG)RZSj7eitH75UGf-bA(Lz4PK(m05CC4#70VNe+j_CMIxic}ssmA5cr*-fr3$
zo8I8Tj`k8~{+e!ikzv00icP=c^4|#qk7LXJoGMKqxWA}!b$o3rUD4@fpTq;Gk800h
zwhKk=&VMYqe)o;M3ko+8G`PsO;<8ye;dj@8r%Gx)V&BnXltzE+tvr9zt+`hx7MOJj
zma_VcchJ4It7G4avl8Yl&D);FR}%0pwuvd4TSz%bdaZH`cb8IzP1#nS=9EINmZYN2>^EkOu*ZwKjghkSttN*C2JZ0OePR5rsB*4M
zV$;7b9KUl`@DwZ0Jg3RIyi*4qli>&Ap`%YA!td}h=d^NPu5-Q0&4BH^o#o9VBmO|3{wzwD)F
zBPg`5Y*M&hr?DPhgb`zW5z;u;KO8wCCt+0WI_pS!EEL@}dBNQ?U?XmQ==6rh>^sSj
zmZ)1E0nJLWfq94~fA1SJH-D&=CYE6F
z$pHCHY%1T<4!^?%K6X}jAUe32L>oKc9hC{r;f8^|3(;2UQU6l5%R=*u^UWj1-Wkhh
zFaM2T6#~l~g5JXOGFnAxIYhmhCDvX%7Sxk>NbWLKqz&MLk}*;Mb@}AJUI>{9n$)sd
zvu(5@=(*U1Kzid)pUuXtzEORV-W@j|7ys_4k)gA#JYW2s%RL?M4PP-9a3+U@)!6T(
zAS&JDMi4AG<$cQ)(^bNtNm;Lob<$U!#IhY3buSoSf%H2r_G4>+upv<6w%3j;<-u={
zsAL?fwEgWYVqsh+CLz3hQtAf%MH~eD%O#U}mDwXsQT^>3;9i<=eEkg_y*3Zq|FAeM
zEm^rm9>tGdO~ah9Y5<=t8#BjI^C{0t)w)H)%n(Kh$%s2w+;br>+80_$Tb$E6zV*|J
z+?TJbTf-xe@dfiOZtZue?Uf4SV-o6bi#afc?AwZ}rwZCjdfnx`jH$v9lXpT3%8QWF
zuPgIV>~ce?cscfZ0h#C#H$o^H;rLz1-Ti!P{_UL;>tAe&JKs<-rCz0%u7X?DEXS#Ribya4q>$sE&K$xh%BSzmn%fz^24ppgY{*xmU#R7UjHPY$zjnBnn9An~@>
z%v<~rzIv)`<9
zwC%WB^l8XFwuo{YPNDpz2syLw4`0MBBu;A=Crd#s?ii42SU-O{=1!)#;d8xe1r1}e
zAs@~5eP`QIobJr4i8!%wqiJyBqMI!(RUA$_88)`g;o~MHF+U4Y4u1pvpK=8Y1m)H(nuzX1#tHX)P`XQ$|^4hnF
z_6U7Ij@*lLIUUm`l@zpooGd#owxEi*sk9QV+k=ZQwNJT!W8|PhK@}-glB!JMw4F`u
z87xwuHgV`LB=disUvGApCyuCMn!CO&(B9N7h&#Q(w3q!8lb-XBB+ncuY;tDHxfCaC
zIBX+joj`4rRsFM2+hOo0Q{;L}uH@L<cF}3c5Bqs
z_j3*WjLc=i`4|p0v`Y}4VXE(MXM+B|>Wz%Q7XuQB>XOV+^?+(O7_+4PfbP
z^L}!pJYk5dfHRK2eePrJ;BRJw6xX5i3~VN$tvuFaEiabd(}+MXmh6`)W`@qq7{8M#j34gd(5AL)4sE*a({dM0d?l}Y1EQd*x;#eps5)V-jLE5VYJeaVenMfK3cXj8}Gu?5ZkR9vAI}DRUbJ+^ittPQZ@uabYSI;hj*EnC<>(ThH?}e>>C(^7__isT
z`f}a}tC(vdl#G7TyzDfy#SJ|Fz?hrr4EY8*Q9_jo<20Vhz2r(ei@pa53v~gW%~#Xm
z@*kCufjs1<@y^koSK2!(Wy$J!%cAT?H3bItoPME3^fBgX6#@zl`>@a>#q0g#wY2o7
z8e{U!jAgo)hR2`pZ0&6SXzwCpBrBFBpZlMQ?Do@75l{ruPltX85k&3T5VK|P0A3~R=APXcoXC}n{~60APO3~AUq
zahEg(*@AIV`6|9&U*#WdK|((jwN5Hd@uMo2ly+aSO^Rk&Do+`Hu=CSwPC!<$rsCF(
zb}y|tK9dVjlQyTWwmL5V<`lWN$)4yw6
zmG*c!Os{UJ73aH0OKbk*HuPaHvK3kWF<~?QT(;KyTUb=u&mJan1+_UnGJgqO={MWV
zgXP4pNF%YV#7V=`YL>ThsF&;u3h?JKGNQ`W3xO}Z7+-KLFPq3-RH|!nD~S9+b}zYK
zVff*(-DL|kY_Uho39T<$oiGg67wuNEdVEGx}8hZUl>iEaQ^);T7NVw24BOL61F9CCqbo4CAx5y>
z@`9-)Bu;XbEd&s2%~ZvQVBnK1$D#g}>+dpAO$iW1ezcNEnWiDOLW8d3e*csNiMB)kGJ@RG9=0L(n?s!`nU(-w~Im16NnO+wJQObOw!=zt*HR`$bNGZc)$N;8c%E@0^B_?ICiBeBS`nsq@9?}W6WZC>ZX
ziG?l%vp7yV7nFvu{C4=JFQTpx*VP_RcMNfX73