Measuring Gauss-Seidel loop-carried dependency and fixing it via loop unrolling

Automated excerpt

Last time, we explored the Gauss-Seidel method. Gauss-Seidel: hard dependency, compiler’s hands are tied. Same workflow as before: compile with -O3 -march=native -S, mark the inner loop, and let OSACA loose on it. 74 | | | | | | | | || | | . L5: 75 | | | 0. 50 0. 50 | 0. 50 0. 50 | | | | || | | vmovsd 16(%rax), %xmm1 76 | | | 0. 50 0. 50 | 0. 50 0. 50 | | | | || | | vmovsd 8(%rax), %xmm0 77 | 0. 00 | 0.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.