Continuous explosion, quite enjoyable.
6th: Rank # 4
7th: Rank # 9
For the sixth test, personally satisfied, because the only difference between a T1 card often.
Because just two days before the exam told matrix, mind is full matrix, also thought of cycling matrix optimization.
The whole idea is very clear there is no problem, but the complexity of the theory of ultra-micro data points in addition to too small a range of 20% given the title did not get.
Finally int128, unified modulo after the matrix multiplication, constant reduced hundreds of times? ! pass.
Correct answer is a combination of several, if not long ago wanted to learn matrix words may come out, but the idea is confined.
T2 data points rub too much water has passed. n times Dijk no quality.
Correct answer is two branch bit pairs each time divided into two groups, at least once to ensure that all the different groups are assigned to all points between pairs.
Needs a total log times grouping, ran point on the shortest of the two groups connected to the node No. 1, update the answer.
Well T3 good question. . However, in the examination room I vaguely aware that this is a variant of graph theory
As the number of points, as the card side, I also thought
I feel that they can not think, and finally just stop at this point, to stop violence
That examination did not feel anything. I have no idea. In addition to T1 bit helpless?
Divergent thinking! Do not want to learn anything recently. Try a variety of methods, especially in the complexity of the overrun of the time.
Daily cards often! Matrix multiplication can be stored in a high-level data type, only after completion of all of the multiplication modulo once.
Then talk about the powerful explosion 7th examination of it.
In fact, I should be 80 points Rank # 20, a long story.
I think the title T3 face ambiguous, ask the coach
The results point he was wrong, I do not care to see my score of T1: 30? !
I do not know if I did not see, I will not go and change T1
T1 special sentenced good question. You never know what you overlooked something.
At first glance, ex_gcd template ah!
How to fight a template? how to prove? No solution under what circumstances? Infinite Solutions under what circumstances?
Usually you really learn to understand it?
When school backplane daughter, do question when the sticky board. All exposed.
Know, know why!
I spent almost an hour on the test to prove the correctness of ex_gcd, and push its template.
Although lead to waste a lot of time during the exam, but there will be no second time.
I really got it! What push again to remember!
And I did not lose on the test Naju special sentence, but the wrong one.
如果a=0，b×c>0，b|c，那么有1组解。
傻子吧！b有1组解，a有无穷组啊！
在考场上一定要保证思路清晰，码出的每一句话都要过脑子。
然而T2什么的刚开始出了一道原题，心态很平和没有着急去码。
（因为我感觉就算是原题我也码不对）
结果换题了，考试时间延长10分钟。
新题是一个明显的组合数学，我的思路虽说清奇但是是对的。见题解。
然而忘了用Lucas了，用朴素Lucas就能减小很大的运算量。
曾经以为用不用Lucas只是运行时间的问题，结果竟然是WA的问题。
对于小质数，阶乘需要特殊处理。否则输出全是0。
而且，更滑稽的是，考场上一看到数据范围，明显CRT啊。
CRT怎么打？公式是啥？会证明吗？
1小时现场发明CRT。
最后因为没用Lucas爆成40分，套上ex或者动用Lucas直接A。
然而自己测试时怕大质数的分解打错，手膜的测试点都用1e9+7做的模数，没有暴露问题。
手模测试点在精不再多。要覆盖所有的数据类型。
T3心态比较平稳，我以为前两题挺稳的。
一看，纯粹模拟有60分，便着急地码了一个模拟。因为剩下的时间已经不多了。
结果东南西北没分清，还不知道怎么就过样例了。10分。。。
考后把两个const数组的两个数字交换了一下，60分。
然而考场上我还有最后30分钟，似乎想到了正解。
花了10分钟证明了结论1，也证明了我的做法的正确性。还剩20分钟。
在想要不要去拿20%的k=0的分值。放弃了，估计匆匆忙忙写不完，也调不出来。
决策是对的，4k是闹着玩的吗？
死而无憾。苟在Rank#9。
上去分享爆零经验还把T3正解讲了。我，，，，有什么用？
暑假起步非常差劲，下次，还能对得起自己吗？