Last active
October 26, 2021 15:35
-
-
Save aprell/9e704b06b1d82de656c05b28799dd9fd to your computer and use it in GitHub Desktop.
Different loop unrolling strategies
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| // CFLAGS="-Wall -Wextra -Wno-implicit-fallthrough" | |
| #include <assert.h> | |
| #include <stdio.h> | |
| // 1) (n - (n % 4)) / 4 iterations | |
| // 2) n % 4 iterations | |
| void loop1(int n) | |
| { | |
| int i, j; | |
| for (i = 0; i < n - 3; i += 4) { | |
| printf("loop1-main: %d\n", i + 0); | |
| printf("loop1-main: %d\n", i + 1); | |
| printf("loop1-main: %d\n", i + 2); | |
| printf("loop1-main: %d\n", i + 3); | |
| } | |
| assert(n - 3 <= i && i <= n); | |
| // Post loop iterates between 0 and n - (n - 3) == 3 times | |
| assert(n - 3 <= n - (n % 4)); | |
| assert(i % 4 == 0); | |
| assert(i == n - (n % 4)); | |
| // n - i == n - (n - (n % 4)) == n % 4 | |
| assert(n - i == n % 4); | |
| for (j = i; j < n; j++) { | |
| printf("loop1-post: %d\n", j); | |
| } | |
| assert(j == n); | |
| } | |
| // 1) (n - (n % 4)) / 4 iterations | |
| // 2) n % 4 iterations | |
| void loop2(int n) | |
| { | |
| int i, j; | |
| for (i = 0; i < n - (n % 4); i += 4) { | |
| printf("loop2-main: %d\n", i + 0); | |
| printf("loop2-main: %d\n", i + 1); | |
| printf("loop2-main: %d\n", i + 2); | |
| printf("loop2-main: %d\n", i + 3); | |
| } | |
| assert(i % 4 == 0); | |
| assert(i == n - (n % 4)); | |
| // n - i == n - (n - (n % 4)) == n % 4 | |
| assert(n - i == n % 4); | |
| for (j = i; j < n; j++) { | |
| printf("loop2-post: %d\n", j); | |
| } | |
| assert(j == n); | |
| } | |
| // 1) n % 4 iterations | |
| // 2) (n - (n % 4)) / 4 iterations | |
| void loop3(int n) | |
| { | |
| int i, j; | |
| for (i = 0; i < n % 4; i++) { | |
| printf("loop3-pre: %d\n", i); | |
| } | |
| assert(i == n % 4); | |
| // n - i == n - (n % 4) | |
| assert(n - i == n - (n % 4)); | |
| for (j = i; j < n; j += 4) { | |
| printf("loop3-main: %d\n", j + 0); | |
| printf("loop3-main: %d\n", j + 1); | |
| printf("loop3-main: %d\n", j + 2); | |
| printf("loop3-main: %d\n", j + 3); | |
| } | |
| assert(j == n); | |
| } | |
| // Duff's device | |
| void loop4(int n) | |
| { | |
| int i = 0; | |
| if (n > 0) { | |
| switch (n % 4) { | |
| case 0: do { printf("loop4-main: %d\n", i++); | |
| case 3: printf("loop4-main: %d\n", i++); | |
| case 2: printf("loop4-main: %d\n", i++); | |
| case 1: printf("loop4-main: %d\n", i++); | |
| } while (i < n); | |
| } | |
| } | |
| assert(i == n); | |
| } | |
| void test(int n) | |
| { | |
| printf(" n | n - 3 | n - (n %% 4)\n"); | |
| printf("---+-------+-------------\n"); | |
| int i; | |
| for (i = 0; i <= n; i++) { | |
| printf("%2d | %5d | %10d\n", i, i - 3, i - (i % 4)); | |
| } | |
| } | |
| int main(void) | |
| { | |
| int i; | |
| for (i = 0; i <= 10; i++) { | |
| loop1(i); | |
| printf("\n"); | |
| } | |
| for (i = 0; i <= 10; i++) { | |
| loop2(i); | |
| printf("\n"); | |
| } | |
| for (i = 0; i <= 10; i++) { | |
| loop3(i); | |
| printf("\n"); | |
| } | |
| for (i = 0; i <= 10; i++) { | |
| loop4(i); | |
| printf("\n"); | |
| } | |
| //test(10); | |
| return 0; | |
| } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment