removed creation of class every CPU loop. It should get better performance

This commit is contained in:
Màrius Montón 2020-05-28 17:18:50 +02:00
parent 4838205aba
commit 95b9685ad9
7 changed files with 86 additions and 38 deletions

View File

@ -57,6 +57,14 @@ public:
*/
A_Instruction(sc_uint<32> instr);
/**
* @brief Sets instruction
* @param p_instr instruction to decode
*/
void setInstr(uint32_t p_instr) {
a_instr = sc_uint<32> (p_instr);
}
/**
* @brief Access to opcode field
* @return return opcode field

View File

@ -13,6 +13,7 @@
#include "tlm.h"
#include "tlm_utils/simple_initiator_socket.h"
#include "tlm_utils/tlm_quantumkeeper.h"
#include "memory.h"
#include "Execute.h"
@ -66,11 +67,19 @@ private:
Registers *register_bank;
Performance *perf;
Log *log;
Instruction * inst;
C_Instruction *c_inst;
M_Instruction *m_inst;
A_Instruction *a_inst;
tlm_utils::tlm_quantumkeeper *m_qk;
bool interrupt;
uint32_t int_cause;
bool irq_already_down;
sc_time default_time;
bool dmi_ptr_valid;
/**

View File

@ -105,6 +105,13 @@ public:
*/
C_Instruction(sc_uint<32> instr);
/**
* @brief Sets instruction
* @param p_instr instruction to decode
*/
void setInstr(uint32_t p_instr) {
m_instr = sc_uint<32> (p_instr);
}
/**
* @brief Access to opcode field
* @return return opcode field

View File

@ -443,6 +443,10 @@ public:
*/
extension_t check_extension();
void setInstr(uint32_t p_instr) {
m_instr = p_instr;
}
/**
* @brief return instruction
* @return all instruction bits (31:0)

View File

@ -51,6 +51,14 @@ public:
*/
M_Instruction(sc_uint<32> instr);
/**
* @brief Sets instruction
* @param p_instr instruction to decode
*/
void setInstr(uint32_t p_instr) {
m_instr = sc_uint<32> (p_instr);
}
/**
* @brief Access to opcode field
* @return return opcode field

View File

@ -25,7 +25,7 @@ using namespace sc_dt;
using namespace std;
/**
* @brief Basic TLm-2 memory
* @brief Basic TLM-2 memory
*/
class Memory: sc_module {
public:

View File

@ -2,7 +2,7 @@
SC_HAS_PROCESS(CPU);
CPU::CPU(sc_module_name name, uint32_t PC) :
sc_module(name), instr_bus("instr_bus") {
sc_module(name), instr_bus("instr_bus"), default_time(10, SC_NS) {
register_bank = new Registers();
exec = new Execute("Execute", register_bank);
perf = Performance::getInstance();
@ -20,7 +20,15 @@ CPU::CPU(sc_module_name name, uint32_t PC) :
irq_already_down = false;
dmi_ptr_valid = false;
instr_bus.register_invalidate_direct_mem_ptr( this, &CPU::invalidate_direct_mem_ptr);
instr_bus.register_invalidate_direct_mem_ptr(this,
&CPU::invalidate_direct_mem_ptr);
inst = new Instruction(0);
c_inst = new C_Instruction(0);
m_inst = new M_Instruction(0);
a_inst = new A_Instruction(0);
m_qk = new tlm_utils::tlm_quantumkeeper();
SC_THREAD(CPU_thread);
}
@ -40,7 +48,7 @@ bool CPU::cpu_process_IRQ() {
if (interrupt == true) {
csr_temp = register_bank->getCSR(CSR_MSTATUS);
if ( (csr_temp & MSTATUS_MIE) == 0) {
if ((csr_temp & MSTATUS_MIE) == 0) {
log->SC_log(Log::DEBUG) << "interrupt delayed" << endl;
return ret_value;
}
@ -87,9 +95,9 @@ bool CPU::cpu_process_IRQ() {
bool CPU::process_c_instruction(Instruction &inst) {
bool PC_not_affected = true;
C_Instruction c_inst(inst.getInstr());
c_inst->setInstr(inst.getInstr());
switch (c_inst.decode()) {
switch (c_inst->decode()) {
case OP_C_ADDI4SPN:
PC_not_affected = exec->C_ADDI4SPN(inst);
break;
@ -183,9 +191,9 @@ bool CPU::process_c_instruction(Instruction &inst) {
bool CPU::process_m_instruction(Instruction &inst) {
bool PC_not_affected = true;
M_Instruction m_inst(inst.getInstr());
m_inst->setInstr(inst.getInstr());
switch (m_inst.decode()) {
switch (m_inst->decode()) {
case OP_M_MUL:
exec->M_MUL(inst);
break;
@ -223,9 +231,9 @@ bool CPU::process_m_instruction(Instruction &inst) {
bool CPU::process_a_instruction(Instruction inst) {
bool PC_not_affected = true;
A_Instruction a_inst(inst.getInstr());
a_inst->setInstr(inst.getInstr());
switch (a_inst.decode()) {
switch (a_inst->decode()) {
case OP_A_LR:
exec->A_LR(inst);
break;
@ -451,13 +459,13 @@ bool CPU::process_base_instruction(Instruction &inst) {
*/
void CPU::CPU_thread(void) {
tlm::tlm_generic_payload* trans = new tlm::tlm_generic_payload;
tlm::tlm_generic_payload *trans = new tlm::tlm_generic_payload;
uint32_t INSTR;
sc_time delay = SC_ZERO_TIME;
bool PC_not_affected = false;
bool incPCby2 = false;
tlm::tlm_dmi dmi_data;
unsigned char* dmi_ptr = NULL;
unsigned char *dmi_ptr = NULL;
trans->set_command(tlm::TLM_READ_COMMAND);
trans->set_data_ptr(reinterpret_cast<unsigned char*>(&INSTR));
@ -467,17 +475,11 @@ void CPU::CPU_thread(void) {
trans->set_dmi_allowed(false); // Mandatory initial value
trans->set_response_status(tlm::TLM_INCOMPLETE_RESPONSE);
//if (trans->is_dmi_allowed() ) {
// if (true) {
// dmi_ptr_valid = instr_bus->get_direct_mem_ptr(*trans, dmi_data);
// dmi_ptr = dmi_data.get_dmi_ptr();
// }
//register_bank->dump();
//Instruction inst(0);
m_qk->reset();
while (1) {
/* Get new PC value */
//cout << "CPU: PC 0x" << hex << (uint32_t) register_bank->getPC() << endl;
//dmi_ptr_valid = true;
if (dmi_ptr_valid == true) {
/* if memory_offset at Memory module is set, this won't work */
memcpy(&INSTR, dmi_ptr + register_bank->getPC(), 4);
@ -490,12 +492,12 @@ void CPU::CPU_thread(void) {
}
if (trans->is_dmi_allowed()) {
dmi_ptr_valid = instr_bus->get_direct_mem_ptr(*trans, dmi_data);
if (dmi_ptr_valid) {
std::cout << "Get DMI_PTR " << std::endl;
dmi_ptr = dmi_data.get_dmi_ptr();
}
}
dmi_ptr_valid = instr_bus->get_direct_mem_ptr(*trans, dmi_data);
if (dmi_ptr_valid) {
std::cout << "Get DMI_PTR " << std::endl;
dmi_ptr = dmi_data.get_dmi_ptr();
}
}
}
perf->codeMemoryRead();
@ -503,30 +505,30 @@ void CPU::CPU_thread(void) {
log->SC_log(Log::INFO) << "PC: 0x" << hex << register_bank->getPC()
<< ". ";
Instruction inst(INSTR);
inst->setInstr(INSTR);
/* check what type of instruction is and execute it */
switch (inst.check_extension()) {
switch (inst->check_extension()) {
case BASE_EXTENSION:
PC_not_affected = process_base_instruction(inst);
PC_not_affected = process_base_instruction(*inst);
incPCby2 = false;
break;
case C_EXTENSION:
PC_not_affected = process_c_instruction(inst);
PC_not_affected = process_c_instruction(*inst);
incPCby2 = true;
break;
case M_EXTENSION:
PC_not_affected = process_m_instruction(inst);
PC_not_affected = process_m_instruction(*inst);
incPCby2 = false;
break;
case A_EXTENSION:
PC_not_affected = process_a_instruction(inst);
PC_not_affected = process_a_instruction(*inst);
incPCby2 = false;
break;
default:
std::cout << "Extension not implemented yet" << std::endl;
inst.dump();
exec->NOP(inst);
inst->dump();
exec->NOP(*inst);
} // switch (inst.check_extension())
perf->instructionsInc();
@ -539,7 +541,18 @@ void CPU::CPU_thread(void) {
cpu_process_IRQ();
/* Fixed instruction time to 10 ns (i.e. 100 MHz)*/
sc_core::wait(10, SC_NS);
//#define USE_QK
#ifdef USE_QK
// Model time used for additional processing
m_qk->inc(default_time);
if (m_qk->need_sync()) {
m_qk->sync();
}
#else
sc_core::wait(10, SC_NS);
#endif
} // while(1)
} // CPU_thread
@ -549,7 +562,6 @@ void CPU::call_interrupt(tlm::tlm_generic_payload &trans, sc_time &delay) {
memcpy(&int_cause, trans.get_data_ptr(), sizeof(uint32_t));
}
void CPU::invalidate_direct_mem_ptr(sc_dt::uint64 start, sc_dt::uint64 end)
{
void CPU::invalidate_direct_mem_ptr(sc_dt::uint64 start, sc_dt::uint64 end) {
dmi_ptr_valid = false;
}