Codeforces Round #412 (rated, Div. 2, base on VK Cup 2017 Round 3) C

C. Success Rate

time limit per test

2 seconds

memory limit per test

256 megabytes

input

standard input

output

standard output

You are an experienced Codeforces user. Today you found out that during your activity on Codeforces you have made y submissions, out of which x have been successful. Thus, your current success rate on Codeforces is equal to x / y.

Your favorite rational number in the [0;1] range is p / q. Now you wonder: what is the smallest number of submissions you have to make if you want your success rate to be p / q?

Input

The first line contains a single integer t (1 ≤ t ≤ 1000) — the number of test cases.

Each of the next t lines contains four integers x, y, p and q (0 ≤ x ≤ y ≤ 109; 0 ≤ p ≤ q ≤ 109; y > 0; q > 0).

It is guaranteed that p / q is an irreducible fraction.

Hacks. For hacks, an additional constraint of t ≤ 5 must be met.

Output

For each test case, output a single integer equal to the smallest number of submissions you have to make if you want your success rate to be equal to your favorite rational number, or -1 if this is impossible to achieve.

Example

input

4
3 10 1 2
7 14 3 8
20 70 2 7
5 6 1 1

output

4
10
0
-1

Note

In the first example, you have to make 4 successful submissions. Your success rate will be equal to 7 / 14, or 1 / 2.

In the second example, you have to make 2 successful and 8 unsuccessful submissions. Your success rate will be equal to 9 / 24, or3 / 8.

In the third example, there is no need to make any new submissions. Your success rate is already equal to 20 / 70, or 2 / 7.

In the fourth example, the only unsuccessful submission breaks your hopes of having the success rate equal to 1.

想写一篇codeforces的题好久了。。可是无奈太菜。。div2场均不过三题

这题蒟蒻以为是二分。。然后写了很久没能写出来（可能我二分太菜。。？）

于是百度一下题解了解一下大神的做法

然后发现了某位大神的题解%%%：http://blog.csdn.net/dormousenone/article/details/71404361

于是发现这题数论可做

这里写一下我的理解：

一般来说，有不可行解的情况要先考虑

于是显然可以发现，

当p/q = 1且x/y != 1时（AC率一百%%%%），不可行

当p = 0且x != 0时，不可行

其他情况一概都是有解的

接下来考虑有解的情况

设a为后来A的题量，b为后来交的次数

显然有 

np = x + a

nq = y + b

其中n为自然数

移项

a = np - x

b = nq - y

由题意可知 0 <= a <= b

即0 <= np - x <= nq - y

于是我们可以得到三个不等式

根据不等式的传递性，去掉其中一个，得到：

n >= ceil(x / p)

n >= ceil((x - y) / (p - q))

解一下不等式就好

答案即为b

代码：

#include<cstdio>
#include<cmath>
#include<algorithm>
using namespace std;

typedef long long LL;

LL x,y,p,q,t,s;

inline LL getint()
{
	LL ret = 0;
	char c = getchar();
	while (c < '0' || c > '9') c = getchar();
	while (c >= '0' && c <= '9')
		ret = ret * 10 + c - '0' , c = getchar();
	return ret;
}

int main()
{
	t = getint();
	for (int i = 1; i <= t; i++)
	{
		x = getint(); y = getint(); p = getint(); q = getint();
		if (p == q) {printf("%d\n",x == y? 0 : -1); continue;}
		if (p == 0) {printf("%d\n",x == 0? 0 : -1); continue;}
		LL n = max(ceil((double)(x) / p),ceil((double)(x - y) / (p - q)));
		printf("%I64d\n",n * q - y);
	}
	return 0;
}