chenlin/Mujoco_WASM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Align all arrays in mjData's buffer to 64 byte boundaries.

Browse Source 
The introduction of additional mjData arrays for the implicit integrator caused in a shift in data alignment in memory. In certain combinations of machines and models, this change resulted in ~5% performance regression that appears to be attributable to L2 cache misses.

This change aligns all arrays in mjData to the boundary of a typical L2 cache line size. Benchmarking shows that this resolves the performance regression on a Broadwell machine.

PiperOrigin-RevId: 451652783
Change-Id: I4ed2c2d03453a83a0b4a9b8e224e2fd7c0dadea9
This commit is contained in:
Saran Tunyasuvunakool
2022-05-28 13:02:49 -07:00
committed by Copybara-Service
parent 981fa22302
commit 5cc73c30bf
6 changed files with 28 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/engine/engine_forward.c
+3 -1
View File
@@ -626,7 +626,9 @@ void mj_RungeKutta(const mjModel* m, mjData* d, int N) {
// compute Xfinal
d->time = time + h;
mju_copy(d->qpos, X[0], nq+nv+na);
mju_copy(d->qpos, X[0], nq);
mju_copy(d->qvel, X[0]+nq, nv);
mju_copy(d->act, X[0]+nq+nv, na);
mj_integratePos(m, d->qpos, dX, h);
mju_addToScl(d->qvel, dX+nv, h, nv);
if (na) {