diff options
| -rw-r--r-- | configs/tested-cfgs/SM75_RTX2060/gpgpusim.config | 6 | ||||
| -rw-r--r-- | configs/tested-cfgs/SM7_QV100_SASS/gpgpusim.config | 4 | ||||
| -rw-r--r-- | src/gpgpu-sim/gpu-sim.cc | 3 | ||||
| -rw-r--r-- | src/gpgpu-sim/shader.cc | 2 |
4 files changed, 8 insertions, 7 deletions
diff --git a/configs/tested-cfgs/SM75_RTX2060/gpgpusim.config b/configs/tested-cfgs/SM75_RTX2060/gpgpusim.config index e8329dd..b89971e 100644 --- a/configs/tested-cfgs/SM75_RTX2060/gpgpusim.config +++ b/configs/tested-cfgs/SM75_RTX2060/gpgpusim.config @@ -33,9 +33,9 @@ # volta clock domains #-gpgpu_clock_domains <Core Clock>:<Interconnect Clock>:<L2 Clock>:<DRAM Clock> --gpgpu_clock_domains 1365.0:1365.0:1365.0:7000.0 +-gpgpu_clock_domains 1365.0:1365.0:1365.0:3500.0 # boost mode -# -gpgpu_clock_domains 1680.0:1680.0:1680.0:7000.0 +# -gpgpu_clock_domains 1680.0:1680.0:1680.0:3500.0 # shader core pipeline config -gpgpu_shader_registers 65536 @@ -156,7 +156,7 @@ -gpgpu_n_mem_per_ctrlr 1 -gpgpu_dram_buswidth 2 -gpgpu_dram_burst_length 16 --dram_data_command_freq_ratio 2 # GDDR6 is configured as DDR in Turing +-dram_data_command_freq_ratio 4 -gpgpu_mem_address_mask 1 -gpgpu_mem_addr_mapping dramid@8;00000000.00000000.00000000.00000000.0000RRRR.RRRRRRRR.RBBBCCCC.BCCSSSSS diff --git a/configs/tested-cfgs/SM7_QV100_SASS/gpgpusim.config b/configs/tested-cfgs/SM7_QV100_SASS/gpgpusim.config index 0255f76..0df3eec 100644 --- a/configs/tested-cfgs/SM7_QV100_SASS/gpgpusim.config +++ b/configs/tested-cfgs/SM7_QV100_SASS/gpgpusim.config @@ -74,8 +74,8 @@ -ptx_opcode_initiation_dp 4,4,4,4,130 -ptx_opcode_latency_sfu 100 -ptx_opcode_initiation_sfu 8 --ptx_opcode_latency_tesnor 6 --ptx_opcode_initiation_tensor 2 +-ptx_opcode_latency_tesnor 8 +-ptx_opcode_initiation_tensor 4 # Volta has sub core model, in which each scheduler has its own register file and EUs # i.e. schedulers are isolated diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc index e5b9c9d..cd5fa56 100644 --- a/src/gpgpu-sim/gpu-sim.cc +++ b/src/gpgpu-sim/gpu-sim.cc @@ -1828,7 +1828,8 @@ void shader_core_ctx::dump_warp_state( FILE *fout ) const void gpgpu_sim::perf_memcpy_to_gpu( size_t dst_start_addr, size_t count ) { if (m_memory_config->m_perf_sim_memcpy) { - assert (dst_start_addr % 32 == 0); + //if(!m_config.trace_driven_mode) //in trace-driven mode, CUDA runtime can start nre data structure at any position + // assert (dst_start_addr % 32 == 0); for ( unsigned counter = 0; counter < count; counter += 32 ) { const unsigned wr_addr = dst_start_addr + counter; diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc index ed0c25e..65ec113 100644 --- a/src/gpgpu-sim/shader.cc +++ b/src/gpgpu-sim/shader.cc @@ -3091,7 +3091,7 @@ unsigned int shader_core_config::max_cta( const kernel_info_t &k ) const case VOLTA: { //For Volta, we assign the remaining shared memory to L1 cache //For more info about adaptive cache, see https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#shared-memory-7-x - assert(gpgpu_shmem_size == 98304); //Volta has 96 KB shared + //assert(gpgpu_shmem_size == 98304); //Volta has 96 KB shared //To Do: make it flexible and not tuned to 9KB share memory unsigned max_assoc = m_L1D_config.get_max_assoc(); |