Yu-Hsiang M. Tsai
Yu-Hsiang M. Tsai
This PR ports the Jacobi to dpcpp including scaler and adaptive-block. Currently, dpcpp can not use different subgroup size inside the kernel, so we do not have the concept, which...
~~IR use the mutable workspace to reuse the memory when the input #rows of rhs does not change.~~ Also, splitting the crierion in IR to eliminate one reduction computation in...
This PR adds transposition LinOp such that build the transposition in lazy way or now. When the behavior is lazy, building transposition will happen in first apply, get_transposition or prepare_transposition....
This PR makes the preconditioner IC wrapper without SolverType. It moves the solver type into the factory operator. When we provide the solver, the SolverType is useless but we still...
This PR adds the extension resource manager. The purpose is to add a way to search/build/insert data from json (possible yaml or command-line support in future). Use json as the...
This pr is mainly for the status and the collection of recent multigrid improvement or fix. the updates are splitted into #981 #980. Some changes related multigrid itself will be...
Track the possible interface breaking we find now. 1. csr strategy (#320 ) 2. csr clac_size -> calc_size 3. Rename / Rebrand / Add executor information (e.g. change `warp_size` so...
Due to the limitation of block/grid size, the max of gridDim.x is `2^31 - 1` and max of blockDim.x is `1024`. (in cuda) We usually use this two parameter to...
When we add more function into dpcpp module, the debug build will give some compilation issue. It will throw `relocation truncated to fit: R_X86_64_GOTPCREL....` and `PC-relative offset overflows in PLT...
# Description From https://github.com/intel/llvm/pull/6407, it moves almost all headers from CL/sycl to sycl I followed https://github.com/oneapi-src/oneMKL/pull/199 way make the header can use sycl/* if they exist and allow the old...