1 // Copyright (c) 2017-2024, Lawrence Livermore National Security, LLC and other CEED contributors. 2 // All Rights Reserved. See the top-level LICENSE and NOTICE files for details. 3 // 4 // SPDX-License-Identifier: BSD-2-Clause 5 // 6 // This file is part of CEED: http://github.com/ceed 7 8 #include <ceed/backend.h> 9 #include <ceed/ceed.h> 10 #include <ceed/jit-tools.h> 11 12 #include <map> 13 #include <string_view> 14 #include <sycl/sycl.hpp> 15 16 #include "../sycl/ceed-sycl-compile.hpp" 17 #include "ceed-sycl-shared.hpp" 18 19 //------------------------------------------------------------------------------ 20 // Compute the local range of for basis kernels 21 //------------------------------------------------------------------------------ 22 static int ComputeLocalRange(Ceed ceed, CeedInt dim, CeedInt thread_1d, CeedInt *local_range, CeedInt max_group_size = 256) { 23 local_range[0] = thread_1d; 24 local_range[1] = (dim > 1) ? thread_1d : 1; 25 const CeedInt min_group_size = local_range[0] * local_range[1]; 26 27 CeedCheck(min_group_size <= max_group_size, ceed, CEED_ERROR_BACKEND, "Requested group size is smaller than the required minimum."); 28 29 local_range[2] = max_group_size / min_group_size; // elements per group 30 return CEED_ERROR_SUCCESS; 31 } 32 33 //------------------------------------------------------------------------------ 34 // Apply basis 35 //------------------------------------------------------------------------------ 36 int CeedBasisApplyTensor_Sycl_shared(CeedBasis basis, const CeedInt num_elem, CeedTransposeMode t_mode, CeedEvalMode eval_mode, CeedVector u, 37 CeedVector v) { 38 Ceed ceed; 39 Ceed_Sycl *ceed_Sycl; 40 const CeedScalar *d_u; 41 CeedScalar *d_v; 42 CeedBasis_Sycl_shared *impl; 43 44 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 45 CeedCallBackend(CeedGetData(ceed, &ceed_Sycl)); 46 CeedCallBackend(CeedBasisGetData(basis, &impl)); 47 48 // Get read/write access to u, v 49 if (u != CEED_VECTOR_NONE) CeedCallBackend(CeedVectorGetArrayRead(u, CEED_MEM_DEVICE, &d_u)); 50 else CeedCheck(eval_mode == CEED_EVAL_WEIGHT, ceed, CEED_ERROR_BACKEND, "An input vector is required for this CeedEvalMode"); 51 CeedCallBackend(CeedVectorGetArrayWrite(v, CEED_MEM_DEVICE, &d_v)); 52 53 // Apply basis operation 54 switch (eval_mode) { 55 case CEED_EVAL_INTERP: { 56 CeedInt *lrange = impl->interp_local_range; 57 const CeedInt &elem_per_group = lrange[2]; 58 const CeedInt group_count = (num_elem / elem_per_group) + !!(num_elem % elem_per_group); 59 //----------- 60 sycl::range<3> local_range(lrange[2], lrange[1], lrange[0]); 61 sycl::range<3> global_range(group_count * lrange[2], lrange[1], lrange[0]); 62 sycl::nd_range<3> kernel_range(global_range, local_range); 63 //----------- 64 sycl::kernel *interp_kernel = (t_mode == CEED_TRANSPOSE) ? impl->interp_transpose_kernel : impl->interp_kernel; 65 66 std::vector<sycl::event> e; 67 68 if (!ceed_Sycl->sycl_queue.is_in_order()) e = {ceed_Sycl->sycl_queue.ext_oneapi_submit_barrier()}; 69 ceed_Sycl->sycl_queue.submit([&](sycl::handler &cgh) { 70 cgh.depends_on(e); 71 cgh.set_args(num_elem, impl->d_interp_1d, d_u, d_v); 72 cgh.parallel_for(kernel_range, *interp_kernel); 73 }); 74 75 } break; 76 case CEED_EVAL_GRAD: { 77 CeedInt *lrange = impl->grad_local_range; 78 const CeedInt &elem_per_group = lrange[2]; 79 const CeedInt group_count = (num_elem / elem_per_group) + !!(num_elem % elem_per_group); 80 //----------- 81 sycl::range<3> local_range(lrange[2], lrange[1], lrange[0]); 82 sycl::range<3> global_range(group_count * lrange[2], lrange[1], lrange[0]); 83 sycl::nd_range<3> kernel_range(global_range, local_range); 84 //----------- 85 sycl::kernel *grad_kernel = (t_mode == CEED_TRANSPOSE) ? impl->grad_transpose_kernel : impl->grad_kernel; 86 const CeedScalar *d_grad_1d = (impl->d_collo_grad_1d) ? impl->d_collo_grad_1d : impl->d_grad_1d; 87 88 std::vector<sycl::event> e; 89 90 if (!ceed_Sycl->sycl_queue.is_in_order()) e = {ceed_Sycl->sycl_queue.ext_oneapi_submit_barrier()}; 91 92 ceed_Sycl->sycl_queue.submit([&](sycl::handler &cgh) { 93 cgh.depends_on(e); 94 cgh.set_args(num_elem, impl->d_interp_1d, d_grad_1d, d_u, d_v); 95 cgh.parallel_for(kernel_range, *grad_kernel); 96 }); 97 } break; 98 case CEED_EVAL_WEIGHT: { 99 CeedInt *lrange = impl->weight_local_range; 100 const CeedInt &elem_per_group = lrange[2]; 101 const CeedInt group_count = (num_elem / elem_per_group) + !!(num_elem % elem_per_group); 102 //----------- 103 sycl::range<3> local_range(lrange[2], lrange[1], lrange[0]); 104 sycl::range<3> global_range(group_count * lrange[2], lrange[1], lrange[0]); 105 sycl::nd_range<3> kernel_range(global_range, local_range); 106 //----------- 107 std::vector<sycl::event> e; 108 109 if (!ceed_Sycl->sycl_queue.is_in_order()) e = {ceed_Sycl->sycl_queue.ext_oneapi_submit_barrier()}; 110 111 ceed_Sycl->sycl_queue.submit([&](sycl::handler &cgh) { 112 cgh.depends_on(e); 113 cgh.set_args(num_elem, impl->d_q_weight_1d, d_v); 114 cgh.parallel_for(kernel_range, *(impl->weight_kernel)); 115 }); 116 } break; 117 case CEED_EVAL_NONE: /* handled separately below */ 118 break; 119 // LCOV_EXCL_START 120 case CEED_EVAL_DIV: 121 case CEED_EVAL_CURL: 122 return CeedError(ceed, CEED_ERROR_BACKEND, "%s not supported", CeedEvalModes[eval_mode]); 123 // LCOV_EXCL_STOP 124 } 125 126 // Restore vectors, cover CEED_EVAL_NONE 127 CeedCallBackend(CeedVectorRestoreArray(v, &d_v)); 128 if (eval_mode == CEED_EVAL_NONE) CeedCallBackend(CeedVectorSetArray(v, CEED_MEM_DEVICE, CEED_COPY_VALUES, (CeedScalar *)d_u)); 129 if (eval_mode != CEED_EVAL_WEIGHT) CeedCallBackend(CeedVectorRestoreArrayRead(u, &d_u)); 130 131 return CEED_ERROR_SUCCESS; 132 } 133 134 //------------------------------------------------------------------------------ 135 // Destroy basis 136 //------------------------------------------------------------------------------ 137 static int CeedBasisDestroy_Sycl_shared(CeedBasis basis) { 138 Ceed ceed; 139 Ceed_Sycl *data; 140 CeedBasis_Sycl_shared *impl; 141 142 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 143 CeedCallBackend(CeedBasisGetData(basis, &impl)); 144 CeedCallBackend(CeedGetData(ceed, &data)); 145 CeedCallSycl(ceed, data->sycl_queue.wait_and_throw()); 146 CeedCallSycl(ceed, sycl::free(impl->d_q_weight_1d, data->sycl_context)); 147 CeedCallSycl(ceed, sycl::free(impl->d_interp_1d, data->sycl_context)); 148 CeedCallSycl(ceed, sycl::free(impl->d_grad_1d, data->sycl_context)); 149 CeedCallSycl(ceed, sycl::free(impl->d_collo_grad_1d, data->sycl_context)); 150 151 delete impl->interp_kernel; 152 delete impl->interp_transpose_kernel; 153 delete impl->grad_kernel; 154 delete impl->grad_transpose_kernel; 155 delete impl->weight_kernel; 156 delete impl->sycl_module; 157 158 CeedCallBackend(CeedFree(&impl)); 159 return CEED_ERROR_SUCCESS; 160 } 161 162 //------------------------------------------------------------------------------ 163 // Create tensor basis 164 // TODO: Refactor 165 //------------------------------------------------------------------------------ 166 int CeedBasisCreateTensorH1_Sycl_shared(CeedInt dim, CeedInt P_1d, CeedInt Q_1d, const CeedScalar *interp_1d, const CeedScalar *grad_1d, 167 const CeedScalar *q_ref_1d, const CeedScalar *q_weight_1d, CeedBasis basis) { 168 Ceed ceed; 169 Ceed_Sycl *data; 170 char *basis_kernel_source; 171 const char *basis_kernel_path; 172 CeedInt num_comp; 173 CeedBasis_Sycl_shared *impl; 174 175 CeedCallBackend(CeedBasisGetCeed(basis, &ceed)); 176 CeedCallBackend(CeedCalloc(1, &impl)); 177 CeedCallBackend(CeedGetData(ceed, &data)); 178 CeedCallBackend(CeedBasisGetNumComponents(basis, &num_comp)); 179 180 const CeedInt thread_1d = CeedIntMax(Q_1d, P_1d); 181 const CeedInt num_nodes = CeedIntPow(P_1d, dim); 182 const CeedInt num_qpts = CeedIntPow(Q_1d, dim); 183 184 CeedInt *interp_lrange = impl->interp_local_range; 185 186 CeedCallBackend(ComputeLocalRange(ceed, dim, thread_1d, interp_lrange)); 187 const CeedInt interp_group_size = interp_lrange[0] * interp_lrange[1] * interp_lrange[2]; 188 189 CeedInt *grad_lrange = impl->grad_local_range; 190 191 CeedCallBackend(ComputeLocalRange(ceed, dim, thread_1d, grad_lrange)); 192 const CeedInt grad_group_size = grad_lrange[0] * grad_lrange[1] * grad_lrange[2]; 193 194 CeedCallBackend(ComputeLocalRange(ceed, dim, Q_1d, impl->weight_local_range)); 195 196 std::vector<sycl::event> e; 197 198 if (!data->sycl_queue.is_in_order()) e = {data->sycl_queue.ext_oneapi_submit_barrier()}; 199 200 // Copy basis data to GPU 201 CeedCallSycl(ceed, impl->d_q_weight_1d = sycl::malloc_device<CeedScalar>(Q_1d, data->sycl_device, data->sycl_context)); 202 sycl::event copy_weight = data->sycl_queue.copy<CeedScalar>(q_weight_1d, impl->d_q_weight_1d, Q_1d, e); 203 204 const CeedInt interp_length = Q_1d * P_1d; 205 CeedCallSycl(ceed, impl->d_interp_1d = sycl::malloc_device<CeedScalar>(interp_length, data->sycl_device, data->sycl_context)); 206 sycl::event copy_interp = data->sycl_queue.copy<CeedScalar>(interp_1d, impl->d_interp_1d, interp_length, e); 207 208 CeedCallSycl(ceed, impl->d_grad_1d = sycl::malloc_device<CeedScalar>(interp_length, data->sycl_device, data->sycl_context)); 209 sycl::event copy_grad = data->sycl_queue.copy<CeedScalar>(grad_1d, impl->d_grad_1d, interp_length, e); 210 211 CeedCallSycl(ceed, sycl::event::wait_and_throw({copy_weight, copy_interp, copy_grad})); 212 213 // Compute collocated gradient and copy to GPU 214 impl->d_collo_grad_1d = NULL; 215 const bool has_collocated_grad = (dim == 3) && (Q_1d >= P_1d); 216 217 if (has_collocated_grad) { 218 CeedScalar *collo_grad_1d; 219 const CeedInt cgrad_length = Q_1d * Q_1d; 220 221 CeedCallBackend(CeedMalloc(Q_1d * Q_1d, &collo_grad_1d)); 222 CeedCallBackend(CeedBasisGetCollocatedGrad(basis, collo_grad_1d)); 223 CeedCallSycl(ceed, impl->d_collo_grad_1d = sycl::malloc_device<CeedScalar>(cgrad_length, data->sycl_device, data->sycl_context)); 224 CeedCallSycl(ceed, data->sycl_queue.copy<CeedScalar>(collo_grad_1d, impl->d_collo_grad_1d, cgrad_length, e).wait_and_throw()); 225 CeedCallBackend(CeedFree(&collo_grad_1d)); 226 } 227 228 // ---[Refactor into separate function]------> 229 // Define compile-time constants 230 std::map<std::string, CeedInt> jit_constants; 231 jit_constants["BASIS_DIM"] = dim; 232 jit_constants["BASIS_Q_1D"] = Q_1d; 233 jit_constants["BASIS_P_1D"] = P_1d; 234 jit_constants["T_1D"] = thread_1d; 235 jit_constants["BASIS_NUM_COMP"] = num_comp; 236 jit_constants["BASIS_NUM_NODES"] = num_nodes; 237 jit_constants["BASIS_NUM_QPTS"] = num_qpts; 238 jit_constants["BASIS_HAS_COLLOCATED_GRAD"] = has_collocated_grad; 239 jit_constants["BASIS_INTERP_SCRATCH_SIZE"] = interp_group_size; 240 jit_constants["BASIS_GRAD_SCRATCH_SIZE"] = grad_group_size; 241 242 // Load kernel source 243 CeedCallBackend(CeedGetJitAbsolutePath(ceed, "ceed/jit-source/sycl/sycl-shared-basis-tensor.h", &basis_kernel_path)); 244 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source -----\n"); 245 { 246 char *source; 247 248 CeedCallBackend(CeedLoadSourceToBuffer(ceed, basis_kernel_path, &source)); 249 basis_kernel_source = source; 250 } 251 CeedDebug256(ceed, CEED_DEBUG_COLOR_SUCCESS, "----- Loading Basis Kernel Source Complete -----\n"); 252 253 // Compile kernels into a kernel bundle 254 CeedCallBackend(CeedBuildModule_Sycl(ceed, basis_kernel_source, &impl->sycl_module, jit_constants)); 255 256 // Load kernel functions 257 CeedCallBackend(CeedGetKernel_Sycl(ceed, impl->sycl_module, "Interp", &impl->interp_kernel)); 258 CeedCallBackend(CeedGetKernel_Sycl(ceed, impl->sycl_module, "InterpTranspose", &impl->interp_transpose_kernel)); 259 CeedCallBackend(CeedGetKernel_Sycl(ceed, impl->sycl_module, "Grad", &impl->grad_kernel)); 260 CeedCallBackend(CeedGetKernel_Sycl(ceed, impl->sycl_module, "GradTranspose", &impl->grad_transpose_kernel)); 261 CeedCallBackend(CeedGetKernel_Sycl(ceed, impl->sycl_module, "Weight", &impl->weight_kernel)); 262 263 // Clean-up 264 CeedCallBackend(CeedFree(&basis_kernel_path)); 265 CeedCallBackend(CeedFree(&basis_kernel_source)); 266 // <---[Refactor into separate function]------ 267 268 CeedCallBackend(CeedBasisSetData(basis, impl)); 269 270 // Register backend functions 271 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Basis", basis, "Apply", CeedBasisApplyTensor_Sycl_shared)); 272 CeedCallBackend(CeedSetBackendFunctionCpp(ceed, "Basis", basis, "Destroy", CeedBasisDestroy_Sycl_shared)); 273 return CEED_ERROR_SUCCESS; 274 } 275 276 //------------------------------------------------------------------------------ 277