Architecture Hierarchy

The GPU is modeled as follows. The class trace_simt_core_cluster models the whole GPU. Its member, trace_gpgpu_sim->m_cluster, is a vector of trace_simt_core_cluster objects. Each object models an SM cluster. In each cluster, there is memeber, trace_simt_core_cluster->m_core, that models the SM cores in this cluster. Each SM core is modeled with an object of trace_shader_core_ctx. Within the SM core, there is a set of hardware warps, trace_shader_core_ctx->m_warp, modeled by class trace_shd_warp_t. The number of hardware warps in each SM equals to the maximum number of threads (2048 for V100) divided by the warp size 32.