22 [](
const sham::DeviceScheduler_ptr &dev_sched,
auto &self) {
23 if (dev_sched->ctx->device->prop.type == sham::DeviceType::GPU) {
Name-keyed registry of the implementation selectors (IImplVariant) of every algorithm.
#define SHAMALGS_REGISTER_IMPL(name, impl)
Register the implementation selector impl under name at static initialization.
namespace to control implementation behavior
shamalgs::ImplVariantGlobal< Reference, NaiveGpu, GpuTeamFetching, GpuOversubscribe > ComputeHistogramImpl
Implementation selector type for compute_histogram.
constexpr std::string_view compute_histogram_impl_name
Registry name, shared by its registration and the dispatch site(s).
ComputeHistogramImpl compute_histogram_impl
Implementation selector for compute_histogram (defined in compute_histogram.cpp).
GPU kernel oversubscribing a work-group per bin, reducing locally.
Portable GPU kernel, one work-item per bin.