[英]Compute Levenshtein distance with aparapi
I'm looking at the possibility of implementing a Levenshtein distance algorithm using APARAPI, but I'm running into some problems with the limitations posed - specifically that I need to create an array in the kernel which is prohibited. 我期待在实现使用APARAPI一个Levenshtein距离算法的可能性,但我遇到的一些问题与局限构成-特别是,我需要创造一种被禁止内核的数组。
Is there a way around this, or better has anyone got a method for Levenshtein distance that works with APARAPI? 有没有办法绕过这个,或者更好的是有没有人有一个与APARAPI一起使用的Levenshtein距离的方法?
The attached code is just in place to try to sort the APARAPI stuff out, I know that I'm not doing anything with the result and I'm just executing once at the moment. 附加的代码就是为了尝试对APARAPI的东西进行排序,我知道我没有对结果做任何事情,而我现在只执行一次。
Kernel kernel = new Kernel() {
@Override
public void run() {
ld("hello", "heya");
}
public int ld(String s, String t) {
int d[]; // matrix
int n; // length of s
int m; // length of t
int i; // iterates through s
int j; // iterates through t
int s_i; // ith character of s
int t_j; // jth character of t
int cost; // cost
// Step 1
n = s.length();
m = t.length();
if (n == 0) {
return m;
}
if (m == 0) {
return n;
}
int firstSize = n+1;
d = new int[firstSize*(m + 1)]; //THIS ISN'T ALLOWED
// Step 2
for (i = 0; i <= n; i++) {
d[firstSize*i+0] = i;
}
for (j = 0; j <= m; j++) {
d[firstSize*0+j] = j;
}
// Step 3
for (i = 1; i <= n; i++) {
s_i = s.charAt(i - 1);
// Step 4
for (j = 1; j <= m; j++) {
t_j = t.charAt(j - 1);
// Step 5
if (s_i == t_j) {
cost = 0;
} else {
cost = 1;
}
// Step 6
int a = d[firstSize*(i - 1)+j] + 1;
int b = d[firstSize*i+(j - 1)] + 1;
int c = d[firstSize*(i - 1)+(j - 1)] + cost;
int mi;
mi = a;
if (b < mi) {
mi = b;
}
if (c < mi) {
mi = c;
}
d[firstSize*i+j] = mi;
}
}
return d[firstSize*n+m];
}
};
kernel.execute(1);
As you indicate Aparapi does not allow any form of new in the Kernel body, however you could pre-allocate a buffer of ints which the Kernel code can access. 正如您所指出的,Aparapi不允许在Kernel主体中使用任何形式的new,但是您可以预先分配内核代码可以访问的int的缓冲区。
Furthermore because you can determine the group size at runtime your buffer does not have to be huge, but some reasonable proportion/ratio of Kernel.getGroupSize(). 此外,因为您可以在运行时确定组大小,所以缓冲区不必非常大,而是Kernel.getGroupSize()的一些合理比例/比例。
Of course you will need to convert the arguments from String to char[] to satisfy Aparapi's Object restriction (Strings are not allowed), but I think from a similar thread on aparapi discussion lists you had already found a workaround for that. 当然,您需要将参数从String转换为char []以满足Aparapi的Object限制(不允许使用字符串),但我想从aparapi讨论列表上的类似线程中您已经找到了解决方法。
If you are prepared to experiment with some 'experimental code' I have a branch in SVN tagged SupportLocalMemory which will allow you to place your temporary int[] buffer in local memory which should also be more performant. 如果您准备尝试一些“实验代码”,我在SVN标记的SupportLocalMemory中有一个分支,它允许您将临时int []缓冲区放在本地内存中,这也应该更高效。
Gary 加里
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.