diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/gpgpu-sim/shader.cc | 12 |
1 files changed, 10 insertions, 2 deletions
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc index 814311d..6f68283 100644 --- a/src/gpgpu-sim/shader.cc +++ b/src/gpgpu-sim/shader.cc @@ -2092,10 +2092,11 @@ bool ldst_unit::constant_cycle(warp_inst_t &inst, mem_stage_stall_type &rc_fail, mem_stage_stall_type fail; if (m_config->perfect_inst_const_cache) { fail = NO_RC_FAIL; + unsigned access_count = inst.accessq_count(); while (inst.accessq_count() > 0) inst.accessq_pop_back(); if (inst.is_load()) { for (unsigned r = 0; r < MAX_OUTPUT_VALUES; r++) - if (inst.out[r] > 0) m_pending_writes[inst.warp_id()][inst.out[r]]--; + if (inst.out[r] > 0) m_pending_writes[inst.warp_id()][inst.out[r]] -= access_count; } } else { fail = process_memory_access_queue(m_L1C, inst); @@ -4241,12 +4242,19 @@ bool opndcoll_rfu_t::collector_unit_t::allocate(register_set *pipeline_reg_set, warp_inst_t **pipeline_reg = pipeline_reg_set->get_ready(); if ((pipeline_reg) and !((*pipeline_reg)->empty())) { m_warp_id = (*pipeline_reg)->warp_id(); + std::vector<int> prev_regs; // remove duplicate regs within same instr for (unsigned op = 0; op < MAX_REG_OPERANDS; op++) { int reg_num = (*pipeline_reg) ->arch_reg.src[op]; // this math needs to match that used in // function_info::ptx_decode_inst - if (reg_num >= 0) { // valid register + bool new_reg = true; + for (auto r : prev_regs) { + if (r == reg_num) + new_reg = false; + } + if (reg_num >= 0 && new_reg) { // valid register + prev_regs.push_back(reg_num); m_src_op[op] = op_t(this, op, reg_num, m_num_banks, m_bank_warp_shift, m_sub_core_model, m_num_banks_per_sched, (*pipeline_reg)->get_schd_id()); |