diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/cuda-sim/instructions.cc | 16 | ||||
| -rw-r--r-- | src/cuda-sim/ptx_ir.h | 2 | ||||
| -rw-r--r-- | src/gpgpu-sim/gpu-sim.h | 6 | ||||
| -rw-r--r-- | src/gpgpu-sim/shader.h | 4 |
4 files changed, 14 insertions, 14 deletions
diff --git a/src/cuda-sim/instructions.cc b/src/cuda-sim/instructions.cc index 2314bef..108de97 100644 --- a/src/cuda-sim/instructions.cc +++ b/src/cuda-sim/instructions.cc @@ -5442,33 +5442,33 @@ void shfl_impl(const ptx_instruction *pI, core_t *core, warp_inst_t inst) { } void shf_impl(const ptx_instruction *pI, ptx_thread_info *thread) { - ptx_reg_t a,b,c,d; + ptx_reg_t a, b, c, d; const operand_info &dst = pI->dst(); const operand_info &src1 = pI->src1(); const operand_info &src2 = pI->src2(); const operand_info &src3 = pI->src3(); - // Only b32 is allowed + // Only b32 is allowed unsigned i_type = pI->get_type(); a = thread->get_operand_value(src1, dst, i_type, thread, 1); b = thread->get_operand_value(src2, dst, i_type, thread, 1); c = thread->get_operand_value(src3, dst, i_type, thread, 1); - if(i_type != B32_TYPE) + if (i_type != B32_TYPE) printf("Only the b32 data_type is allowed per the ISA\n"); unsigned clamp_mode = pI->clamp_mode(); unsigned n = c.u32 & 0x1f; - if(clamp_mode) { - if(c.u32 < 32) + if (clamp_mode) { + if (c.u32 < 32) n = c; else n = 32; } - if(pI->left_mode()) - d.u32 = (b.u32 << n) | (a.u32 >> (32-n)); + if (pI->left_mode()) + d.u32 = (b.u32 << n) | (a.u32 >> (32 - n)); else - d.u32 = (b.u32 << (32-n)) | (a.u32 >> n); + d.u32 = (b.u32 << (32 - n)) | (a.u32 >> n); thread->set_operand_value(dst, d, i_type, thread, pI); } diff --git a/src/cuda-sim/ptx_ir.h b/src/cuda-sim/ptx_ir.h index 46f183b..d253866 100644 --- a/src/cuda-sim/ptx_ir.h +++ b/src/cuda-sim/ptx_ir.h @@ -1085,7 +1085,7 @@ class ptx_instruction : public warp_inst_t { unsigned cache_option() const { return m_cache_option; } unsigned rounding_mode() const { return m_rounding_mode; } unsigned saturation_mode() const { return m_saturation_mode; } - unsigned clamp_mode() const {return m_clamp_mode;} + unsigned clamp_mode() const { return m_clamp_mode; } unsigned left_mode() const { return m_left_mode; } unsigned dimension() const { return m_geom_spec; } unsigned barrier_op() const { return m_barrier_op; } diff --git a/src/gpgpu-sim/gpu-sim.h b/src/gpgpu-sim/gpu-sim.h index a24ffd3..d43b399 100644 --- a/src/gpgpu-sim/gpu-sim.h +++ b/src/gpgpu-sim/gpu-sim.h @@ -132,9 +132,9 @@ struct power_config { // NOTE: After changing the nonlinear model to only scaling idle core, // NOTE: The min_inc_per_active_sm is not used any more - if (g_use_nonlinear_model) - sscanf(gpu_nonlinear_model_config, "%lf:%lf", &gpu_idle_core_power, - &gpu_min_inc_per_active_sm); + // if (g_use_nonlinear_model) + // sscanf(gpu_nonlinear_model_config, "%lf:%lf", &gpu_idle_core_power, + // &gpu_min_inc_per_active_sm); } void reg_options(class OptionParser *opp); diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h index b1f904f..92691d3 100644 --- a/src/gpgpu-sim/shader.h +++ b/src/gpgpu-sim/shader.h @@ -134,7 +134,7 @@ class shd_warp_t { m_waiting_ldgsts = false; // Ni: Clear m_ldgdepbar_buf - for (int i = 0; i < m_ldgdepbar_buf.size(); i++) { + for (unsigned i = 0; i < m_ldgdepbar_buf.size(); i++) { m_ldgdepbar_buf[i].clear(); } m_ldgdepbar_buf.clear(); @@ -165,7 +165,7 @@ class shd_warp_t { m_waiting_ldgsts = false; // Ni: Clear m_ldgdepbar_buf - for (int i = 0; i < m_ldgdepbar_buf.size(); i++) { + for (unsigned i = 0; i < m_ldgdepbar_buf.size(); i++) { m_ldgdepbar_buf[i].clear(); } m_ldgdepbar_buf.clear(); |