29 auto edges = get_edges();
31 edges.spans_x.check_sizes(edges.sizes.indexes);
32 edges.spans_y.check_sizes(edges.sizes.indexes);
33 edges.spans_z.ensure_sizes(edges.sizes.indexes);
35 Tscal alpha_0 = edges.alpha_0.data;
36 Tscal alpha_1 = edges.alpha_1.data;
37 Tscal alpha_2 = edges.alpha_2.data;
40 bool read_z = alpha_0 != 0;
43 = edges.sizes.indexes.template map<u32>([&](
u64 id,
u32 obj_cnt) {
44 return obj_cnt * nvar;
48 shamsys::instance::get_compute_scheduler_ptr(),
52 [alpha_0, alpha_1, alpha_2, read_z](
53 u32 i,
const T *__restrict x,
const T *__restrict y, T *__restrict z) {
56 z[i] = read_z ? alpha_0 * z[i] + alpha_1 * x[i] + alpha_2 * y[i]
57 : alpha_1 * x[i] + alpha_2 * y[i];
void distributed_data_kernel_call(sham::DeviceScheduler_ptr dev_sched, RefIn in, RefOut in_out, const shambase::DistributedData< index_t > &thread_counts, Functor &&func)
A variant of sham::kernel_call for distributed data.