chenlin/Mujoco_WASM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
2,536 Commits 2 Branches 1 Tag
376dc59ea07e005448f7ec159fb0589eaf940683
Commit Graph

13 Commits

Author SHA1 Message Date
Taylor Howell a36f2cccb6 Add nnz argument for size of allocated memory to mju_dense2sparse and check if this number is too small for number of non-zeros. 
PiperOrigin-RevId: 690545825
Change-Id: I0e31cb907a1151cb2e4766d5f0cdfecb5629a2a6
 2024-10-28 03:46:50 -07:00
Taylor Howell 94e9912d28 Add mju_mulMatTVecSparse to multiply transposed sparse matrix and dense vector. 
PiperOrigin-RevId: 688535240
Change-Id: I0bacf1e911cde99014b2dcb0a1434e61a4f41d51
 2024-10-22 07:06:29 -07:00
Yuval Tassa 0ffaeb4d56 Improve comments in mju_cholFactorNNZ, clarify that this function reads from the upper triangle of the source matrix. 
PiperOrigin-RevId: 686113654
Change-Id: I029f9279d177f4cb9ff60396a26d64821aee3a6c
 2024-10-15 08:26:07 -07:00
Yuval Tassa 7ec94f46d4 Refactor Newton solver: move Hessian from arena back to the stack. 
The changes in 9a0dc20821, moving the Hessian from the stack to the arena, should be rolled back: they prevent future threading over islands. Unlike the stack, arena allocations are not thread-friendly. The reason for the original move was to save memory, but the savings are small: `O(ctx->nH)` and only linear in `nv`. The significant reduction of the Cholesky factor, from quadratic in `nv` to quadratic in the largest dof island — the reduction afforded by 2dd518734f — remains in place.

Also refactor and improve readability.

PiperOrigin-RevId: 685717356
Change-Id: Ia9ec3e44a62d459a3b9cffb578cf6479d5aa1d7f
 2024-10-14 08:34:16 -07:00
Yuval Tassa 3f3fe92860 Remove unused argument in mju_sqrMatTDSparseInit 
PiperOrigin-RevId: 682375566
Change-Id: If05f5d15a2f8925eee9dd0ebc5d485164d90d787
 2024-10-04 10:45:33 -07:00
Taylor Howell 6a6e10d779 Add mju_cholFactorNNZ to compute the number of non-zeros per row for sparse Cholesky factorization. 
PiperOrigin-RevId: 679553197
Change-Id: I6f92deec347dae97d2e219cfb5376f92e777da1a
 2024-09-27 05:39:08 -07:00
Yuval Tassa 29aa5e4a41 Allow dot products with sparse vectors to specify that one vector uses uncompressed memory. 
PiperOrigin-RevId: 561951337
Change-Id: I1310d7c09d9a85ad9b43877155844a9d0ce6edab
 2023-09-01 07:40:16 -07:00
Kyle Bayes ea5e00cad8 Move mju_combineSparseCount into engine_util_sparse. 
PiperOrigin-RevId: 561025741
Change-Id: I0bdc48554928fc4545a38a8cea3bdeb593bd7ecd
 2023-08-29 07:40:53 -07:00
Kyle Bayes 8ea690ed70 Remove unused parameters from mju_sqrMatTDSparseInit. 
PiperOrigin-RevId: 557114227
Change-Id: Ia7d3fedfba612a68ff01d68f1b9dadb45bd76e85
 2023-08-15 06:30:19 -07:00
Kyle Bayes 9924cce4b7 Add nnz per row precount method for mju_sqrMatTDSparse. 
PiperOrigin-RevId: 521744893
Change-Id: Ie0fbfab552a680127f4acf9041b07916f8a2a490
 2023-04-04 06:21:22 -07:00
Kyle Bayes 056e849273 Provide improved mju_sqrMatTDSparse implementation that doesn't require dense memory allocation for sparse matrices. 
PiperOrigin-RevId: 516812783
Change-Id: Ieb43337831d8d3b2c7f18a7facd8e0a0f2b0eff5
 2023-03-15 06:56:16 -07:00
Kyle Bayes c741dfce7d Implement a more performant mju_transposeSparse that doesn't require dense memory allocation. 
PiperOrigin-RevId: 505044194
Change-Id: Ibe0e39e3ad711b2b5bdad45180ca4f83b5c8fc82
 2023-01-27 00:20:01 -08:00
Alessio Quaglino 7b0fbc63f8 Speed improvement with no AVX. 40% on Linux Intel Xeon and 60% on ARM Mac. 
PiperOrigin-RevId: 488360660
Change-Id: I423269cd362fdcc2434393b691638ac6ed778ef2
 2022-11-14 07:31:32 -08:00