Added support for -gpgpu_registers_per_block config

3 files changed, 10 insertions, 0 deletions

diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 5d63ab7..b283964 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -247,6 +247,9 @@ void shader_core_config::reg_options(class OptionParser * opp)
     option_parser_register(opp, "-gpgpu_shader_registers", OPT_UINT32, &gpgpu_shader_registers, 
                  "Number of registers per shader core. Limits number of concurrent CTAs. (default 8192)",
                  "8192");
+    option_parser_register(opp, "-gpgpu_registers_per_block", OPT_UINT32, &gpgpu_registers_per_block,
+                 "Maximum number of registers per thread block. (default 8192)",
+                 "8192");
     option_parser_register(opp, "-gpgpu_shader_cta", OPT_UINT32, &max_cta_per_core, 
                  "Maximum number of concurrent CTAs in shader (default 8)",
                  "8");
@@ -674,6 +677,11 @@ int gpgpu_sim::num_registers_per_core() const
    return m_shader_config->gpgpu_shader_registers;
 }
 
+int gpgpu_sim::num_registers_per_block() const
+{
+   return m_shader_config->gpgpu_registers_per_block;
+}
+
 int gpgpu_sim::wrp_size() const
 {
    return m_shader_config->warp_size;
diff --git a/src/gpgpu-sim/gpu-sim.h b/src/gpgpu-sim/gpu-sim.h
index 8d1c4fc..5908fd5 100644
--- a/src/gpgpu-sim/gpu-sim.h
+++ b/src/gpgpu-sim/gpu-sim.h
@@ -391,6 +391,7 @@ public:
    int shared_mem_size() const;
    int shared_mem_per_block() const;
    int num_registers_per_core() const;
+   int num_registers_per_block() const;
    int wrp_size() const;
    int shader_clock() const;
    const struct cudaDeviceProp *get_prop() const;
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index fbddd18..6201840 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -1286,6 +1286,7 @@ struct shader_core_config : public core_config
     unsigned max_barriers_per_cta;
     char * gpgpu_scheduler_string;
     unsigned gpgpu_shmem_per_block;
+    unsigned gpgpu_registers_per_block;
     char* pipeline_widths_string;
     int pipe_widths[N_PIPELINE_STAGES];