diff options
| author | Tim Rogers <[email protected]> | 2019-05-13 19:42:43 -0400 |
|---|---|---|
| committer | GitHub <[email protected]> | 2019-05-13 19:42:43 -0400 |
| commit | cbb929b6f15f63f0b104e2acac7a17f02c0208fe (patch) | |
| tree | ed2bf825d27848c818a9293075cf9b9d73a26864 /src/gpgpu-sim/gpu-sim.cc | |
| parent | f507979bcf0f14d1e5843c9b08613d6b0a4bb7a2 (diff) | |
| parent | 059dabc1af44b8eb60f0cacc8d5c2d06f1e85a00 (diff) | |
Merge pull request #118 from mkhairy/dev
Integrating Mahmoud's changes that improve simulation speed by 2-3x.
Also includes a change that mitigates memory channel conflicts by using a random hashing function.
Diffstat (limited to 'src/gpgpu-sim/gpu-sim.cc')
| -rw-r--r-- | src/gpgpu-sim/gpu-sim.cc | 13 |
1 files changed, 8 insertions, 5 deletions
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc index 4b68e2e..c1ba934 100644 --- a/src/gpgpu-sim/gpu-sim.cc +++ b/src/gpgpu-sim/gpu-sim.cc @@ -99,6 +99,8 @@ unsigned long long gpu_tot_sim_cycle_parition_util = 0; unsigned long long partiton_replys_in_parallel = 0; unsigned long long partiton_replys_in_parallel_total = 0; +tr1_hash_map<new_addr_type,unsigned> address_random_interleaving; + /* Clock Domains */ #define CORE 0x01 @@ -476,6 +478,7 @@ void shader_core_config::reg_options(class OptionParser * opp) option_parser_register(opp, "-gpgpu_concurrent_kernel_sm", OPT_BOOL, &gpgpu_concurrent_kernel_sm, "Support concurrent kernels on a SM (default = disabled)", "0"); + } void gpgpu_sim_config::reg_options(option_parser_t opp) @@ -1181,19 +1184,19 @@ void gpgpu_sim::gpu_print_stat() m_memory_sub_partition[i]->accumulate_L2cache_stats(l2_stats); m_memory_sub_partition[i]->get_L2cache_sub_stats(l2_css); - fprintf( stdout, "L2_cache_bank[%d]: Access = %u, Miss = %u, Miss_rate = %.3lf, Pending_hits = %u, Reservation_fails = %u\n", + fprintf( stdout, "L2_cache_bank[%d]: Access = %llu, Miss = %llu, Miss_rate = %.3lf, Pending_hits = %llu, Reservation_fails = %llu\n", i, l2_css.accesses, l2_css.misses, (double)l2_css.misses / (double)l2_css.accesses, l2_css.pending_hits, l2_css.res_fails); total_l2_css += l2_css; } if (!m_memory_config->m_L2_config.disabled() && m_memory_config->m_L2_config.get_num_lines()) { //L2c_print_cache_stat(); - printf("L2_total_cache_accesses = %u\n", total_l2_css.accesses); - printf("L2_total_cache_misses = %u\n", total_l2_css.misses); + printf("L2_total_cache_accesses = %llu\n", total_l2_css.accesses); + printf("L2_total_cache_misses = %llu\n", total_l2_css.misses); if(total_l2_css.accesses > 0) printf("L2_total_cache_miss_rate = %.4lf\n", (double)total_l2_css.misses/(double)total_l2_css.accesses); - printf("L2_total_cache_pending_hits = %u\n", total_l2_css.pending_hits); - printf("L2_total_cache_reservation_fails = %u\n", total_l2_css.res_fails); + printf("L2_total_cache_pending_hits = %llu\n", total_l2_css.pending_hits); + printf("L2_total_cache_reservation_fails = %llu\n", total_l2_css.res_fails); printf("L2_total_cache_breakdown:\n"); l2_stats.print_stats(stdout, "L2_cache_stats_breakdown"); printf("L2_total_cache_reservation_fail_breakdown:\n"); |