Submit | All submissions | Best solutions | Back to list |
MATRMUL - Matrix Multiplication 4K |
Consider the following C++ code, it constructs n×n matrices A,B and vector V from matrix C=A×B which you need to calculate.
uint32_t n, i, j, d1, p1, r1, m1, d2, p2, m2, A[n][n], B[n][n]; uint64_t C[n][n], V[n]; //here you need to read n, p1, d1, r1, m1, p2, d2, r2, m2 from input. for (i=0; i<n; ++i) for (j=0; j<n; ++j) { d1 = d1 * p1 + r1; d2 = d2 * p2 + r2; A[i][j] = d1 >> (32 - m1); B[i][j] = d2 >> (32 - m2); } //here you need to calculate C=A*B for (i=0; i<n; ++i) { V[i] = 0; for (j=0; j<n; ++j) V[i] ^= C[i][j]; } //here you need to output V[0], V[1], ..., V[n-1], separated by spaces.
Input
You are given integers n, p1, d1, r1, m1, p2, d2, r2, m2, separated by spaces.
1≤n≤4096, 1≤m1,m2≤26, 0≤p1,d1,r1,p2,d2,r2<232.
Output
You need to output V[0], ..., V[n-1], separated by spaces.
Example
Input: 4 1664525 0 1013904223 26 1664525 1 1013904223 26 Output: 3929555766216722 770418013451752 7738542081270672 4286515685761206
P.S.
It is recommended to solve MATRMUL0 before this task, to get AC here your time there should be less than 14s for cubic algorithm and 16s for Strassen's.
There are 4 tests with n≈4096. Score is average time in seconds, lower is better. My best score — ~5.6s which is 5 times faster than time limit.
Added by: | Michael Kharitonov |
Date: | 2017-04-08 |
Time limit: | 28s |
Source limit: | 50000B |
Memory limit: | 1536MB |
Cluster: | Cube (Intel G860) |
Languages: | All |
Resource: | fastcomputing.org |
hide comments
2017-12-31 12:17:52 Michael Kharitonov
tip: use Function Attributes or Function Specific Option Pragmas: https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html https://gcc.gnu.org/onlinedocs/gcc/Function-Specific-Option-Pragmas.html Last edit: 2018-07-25 22:34:23 |