chenlin/Mujoco_WASM
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Switch mjData.{qH,qLD} from reduced ("C") to full ("M") inertia matrix structure. No performance impact of extra zeros because of existing "simple dof" skipping mechanism.

Browse Source 
PiperOrigin-RevId: 733523931
Change-Id: Ic8d8a152dda5532331c239cb6b4ce7d8d09b7fff
This commit is contained in:
Yuval Tassa
2025-03-04 17:33:48 -08:00
committed by Copybara-Service
parent 51f6aa8b43
commit 0f563ecf31
20 changed files with 95 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/engine/engine_support.c
+3 -3
View File
@@ -1097,14 +1097,14 @@ void mj_mulM2(const mjModel* m, const mjData* d, mjtNum* res, const mjtNum* vec)
// non-simple: add off-diagonals
if (!m->dof_simplenum[i]) {
int adr = d->C_rowadr[i];
res[i] += mju_dotSparse(qLD+adr, vec, d->C_rownnz[i] - 1, d->C_colind+adr, /*flg_unc1=*/0);
int adr = d->M_rowadr[i];
res[i] += mju_dotSparse(qLD+adr, vec, d->M_rownnz[i] - 1, d->M_colind+adr, /*flg_unc1=*/0);
}
}
// res *= sqrt(D)
for (int i=0; i < nv; i++) {
int diag = d->C_rowadr[i] + d->C_rownnz[i] - 1;
int diag = d->M_rowadr[i] + d->M_rownnz[i] - 1;
res[i] *= mju_sqrt(qLD[diag]);
}
}