simX within 30% cycles of the RTL for matAdd

This commit is contained in:
felsabbagh3 2019-11-11 00:37:34 -05:00
parent 92e88a7bb2
commit f717ff56da
9 changed files with 605 additions and 73 deletions

View file

@ -13,8 +13,9 @@ LIB=
CF=-CFLAGS '-std=c++11 -fPIC -O3'
LIGHTW=-Wno-UNOPTFLAT -Wno-BLKLOOPINIT
DEB=--trace --prof-cfuncs -DVL_DEBUG=1
EXE=--exe $(LIB_OBJS)
all: simX
# simX:
@ -22,7 +23,7 @@ all: simX
simX:
verilator $(COMP) -cc $(FILE) $(INCLUDE) $(EXE) $(LIB) $(CF) $(LIGHTW)
verilator $(COMP) -cc $(FILE) $(INCLUDE) $(EXE) $(LIB) $(CF) $(LIGHTW) $(DEB)
(cd obj_dir && make -j -f Vcache_simX.mk OPT='-DVL_DEBUG' VL_DEBUG=1 DVL_DEBUG=1)
clean:

View file

@ -93,25 +93,26 @@ module cache_simX (
always @(posedge clk, posedge reset) begin
if (reset)
begin
icache_i_m_ready <= 0;
dcache_i_m_ready <= 0;
icache_i_m_ready = 0;
dcache_i_m_ready = 0;
end else begin
if (VX_dram_req_rsp_icache.o_m_valid) begin
icache_i_m_ready <= 1;
icache_i_m_ready = 1;
// $display("cache_simX.v: setting icache_i_m_ready = %d", icache_i_m_ready);
end else if (icache_i_m_ready) begin
icache_i_m_ready <= 0;
icache_i_m_ready = 0;
end else begin
icache_i_m_ready <= 0;
icache_i_m_ready = 0;
end
if (VX_dram_req_rsp.o_m_valid) begin
dcache_i_m_ready <= 1;
dcache_i_m_ready = 1;
end else if (dcache_i_m_ready) begin
dcache_i_m_ready <= 0;
dcache_i_m_ready = 0;
end else begin
dcache_i_m_ready <= 0;
dcache_i_m_ready = 0;
end
end

View file

@ -20,9 +20,74 @@
#include "include/qsim-harp.h"
#endif
#define NO_MEM_READ 7
#define LB_MEM_READ 0
#define LH_MEM_READ 1
#define LW_MEM_READ 2
#define LBU_MEM_READ 4
#define LHU_MEM_READ 5
#define NO_MEM_WRITE 7
#define SB_MEM_WRITE 0
#define SH_MEM_WRITE 1
#define SW_MEM_WRITE 2
#define INIT_TRACE(trace_inst) \
trace_inst.valid_inst = false; \
trace_inst.pc = 0; \
trace_inst.wid = schedule_w; \
trace_inst.rs1 = -1; \
trace_inst.rs2 = -1; \
trace_inst.rd = -1; \
trace_inst.is_lw = false; \
trace_inst.is_sw = false; \
trace_inst.mem_addresses = new unsigned[a.getNThds()]; \
for (int tid = 0; tid < a.getNThds(); tid++) trace_inst.mem_addresses[tid] = 0xdeadbeef; \
trace_inst.mem_stall_cycles = 0; \
trace_inst.fetch_stall_cycles = 0; \
trace_inst.stall_warp = false; \
trace_inst.wspawn = false; \
trace_inst.stalled = false;
#define CPY_TRACE(drain, source) \
drain.valid_inst = source.valid_inst; \
drain.pc = source.pc; \
drain.wid = source.wid; \
drain.rs1 = source.rs1; \
drain.rs2 = source.rs2; \
drain.rd = source.rd; \
drain.is_lw = source.is_lw; \
drain.is_sw = source.is_sw; \
for (int tid = 0; tid < a.getNThds(); tid++) drain.mem_addresses[tid] = source.mem_addresses[tid]; \
drain.mem_stall_cycles = source.mem_stall_cycles; \
drain.fetch_stall_cycles = source.fetch_stall_cycles; \
drain.stall_warp = source.stall_warp; \
drain.wspawn = source.wspawn; \
drain.stalled = false;
using namespace Harp;
using namespace std;
void printTrace(trace_inst_t * trace, const char * stage_name)
{
cout << "********************************** " << stage_name << " *********************************\n";
cout << "valid: " << trace->valid_inst << '\n';
cout << "PC: " << hex << trace->pc << dec << '\n';
cout << "wid: " << trace->wid << '\n';
cout << "rd: " << trace->rd << "\trs1: " << trace->rs1 << "\trs2: " << trace->rs2 << '\n';
cout << "is_lw: " << trace->is_lw << '\n';
cout << "is_sw: " << trace->is_sw << '\n';
cout << "fetch_stall_cycles: " << trace->fetch_stall_cycles << '\n';
cout << "mem_stall_cycles: " << trace->mem_stall_cycles << '\n';
cout << "stall_warp: " << trace->stall_warp << '\n';
cout << "wspawn: " << trace->wspawn << '\n';
cout << "stalled: " << trace->stalled << '\n';
}
#ifdef EMU_INSTRUMENTATION
void Harp::reg_doRead(Word cpuId, Word regNum) {
Harp::OSDomain::osDomain->do_reg(cpuId, regNum, 8, true);
@ -34,10 +99,45 @@ void Harp::reg_doWrite(Word cpuId, Word regNum) {
#endif
Core::Core(const ArchDef &a, Decoder &d, MemoryUnit &mem, Word id):
a(a), iDec(d), mem(mem), steps(0)
a(a), iDec(d), mem(mem), steps(4)
{
release_warp = false;
foundSchedule = true;
schedule_w = 0;
INIT_TRACE(inst_in_fetch);
INIT_TRACE(inst_in_decode);
INIT_TRACE(inst_in_scheduler);
INIT_TRACE(inst_in_exe);
INIT_TRACE(inst_in_lsu);
INIT_TRACE(inst_in_wb);
for (int i = 0; i < 32; i++)
{
stallWarp[i] = false;
for (int j = 0; j < 32; j++)
{
renameTable[i][j] = true;
}
}
cache_simulator = new Vcache_simX;
m_trace = new VerilatedVcdC;
cache_simulator->trace(m_trace, 99);
m_trace->open("simXtrace.vcd");
cache_simulator->reset = 1;
cache_simulator->clk = 0;
cache_simulator->eval();
// m_trace->dump(10);
cache_simulator->reset = 1;
cache_simulator->clk = 1;
cache_simulator->eval();
// m_trace->dump(11);
cache_simulator->reset = 0;
cache_simulator->clk = 0;
for (unsigned i = 0; i < a.getNWarps(); ++i)
w.push_back(Warp(this, i));
@ -51,54 +151,270 @@ bool Core::interrupt(Word r0) {
void Core::step()
{
cout << "\n\n\n------------------------------------------------------\n";
steps++;
cout << "CYCLE: " << steps << '\n';
cout << "Stalled Warps:\n";
for (int widd = 0; widd < a.getNWarps(); widd++)
{
cout << stallWarp[widd] << " ";
}
cout << '\n';
// cout << "Rename table\n";
// for (int regii = 0; regii < 32; regii++)
// {
// cout << regii << ": " << renameTable[0][regii] << '\n';
// }
cout << '\n';
this->writeback();
this->load_store();
this->execute_unit();
this->scheduler();
this->decode();
this->fetch();
if (release_warp)
{
release_warp = false;
stallWarp[release_warp_num] = false;
}
}
void Core::getCacheDelays(trace_inst_t * trace_inst)
{
static int curr_cycle = 0;
if (trace_inst->valid_inst)
{
bool in_dcache_in_valid[a.getNThds()];
unsigned in_dcache_in_address[a.getNThds()];
unsigned in_dcache_mem_read;
unsigned in_dcache_mem_write;
if (trace_inst->is_lw)
{
in_dcache_mem_read = LW_MEM_READ;
in_dcache_mem_write = NO_MEM_WRITE;
}
else if (trace_inst->is_sw)
{
in_dcache_mem_read = NO_MEM_READ;
in_dcache_mem_write = SW_MEM_WRITE;
}
else
{
in_dcache_mem_read = NO_MEM_READ;
in_dcache_mem_write = NO_MEM_WRITE;
}
for (int j = 0; j < a.getNThds(); j++)
{
if ((w[trace_inst->wid].tmask[j]) && (trace_inst->is_sw || trace_inst->is_lw))
{
in_dcache_in_valid[j] = true;
in_dcache_in_address[j] = trace_inst->mem_addresses[j];
}
else
{
in_dcache_in_valid[j] = false;
in_dcache_in_address[j] = 0xdeadbeef;
}
}
cache_simulator->clk = 1;
cache_simulator->eval();
m_trace->dump(2*curr_cycle);
cache_simulator->in_icache_pc_addr = trace_inst->pc;
cache_simulator->in_icache_valid_pc_addr = 1;
// DCache start
cache_simulator->in_dcache_mem_read = in_dcache_mem_read;
cache_simulator->in_dcache_mem_write = in_dcache_mem_write;
for (int cur_t = 0; cur_t < a.getNThds(); cur_t++)
{
cache_simulator->in_dcache_in_valid[cur_t] = in_dcache_in_valid[cur_t];
cache_simulator->in_dcache_in_address[cur_t] = in_dcache_in_address[cur_t];
}
// DCache end
cache_simulator->clk = 0;
cache_simulator->eval();
m_trace->dump(2*curr_cycle+1);
curr_cycle++;
while((cache_simulator->out_icache_stall || cache_simulator->out_dcache_stall))
{
////////// Feed input
if (cache_simulator->out_icache_stall)
{
cache_simulator->in_icache_pc_addr = trace_inst->pc;
cache_simulator->in_icache_valid_pc_addr = 1;
trace_inst->fetch_stall_cycles++;
}
else
{
cache_simulator->in_icache_valid_pc_addr = 0;
}
if (cache_simulator->out_dcache_stall)
{
cache_simulator->in_dcache_mem_read = in_dcache_mem_read;
cache_simulator->in_dcache_mem_write = in_dcache_mem_write;
for (int cur_t = 0; cur_t < a.getNThds(); cur_t++)
{
cache_simulator->in_dcache_in_valid[cur_t] = in_dcache_in_valid[cur_t];
cache_simulator->in_dcache_in_address[cur_t] = in_dcache_in_address[cur_t];
}
trace_inst->mem_stall_cycles++;
}
else
{
cache_simulator->in_dcache_mem_read = NO_MEM_READ;
cache_simulator->in_dcache_mem_write = NO_MEM_WRITE;
for (int cur_t = 0; cur_t < a.getNThds(); cur_t++)
{
cache_simulator->in_dcache_in_valid[cur_t] = 0;
}
}
cache_simulator->clk = 1;
cache_simulator->eval();
m_trace->dump(2*curr_cycle);
//////// Feed input
if (cache_simulator->out_icache_stall)
{
cache_simulator->in_icache_pc_addr = trace_inst->pc;
cache_simulator->in_icache_valid_pc_addr = 1;
}
else
{
cache_simulator->in_icache_valid_pc_addr = 0;
}
if (cache_simulator->out_dcache_stall)
{
cache_simulator->in_dcache_mem_read = in_dcache_mem_read;
cache_simulator->in_dcache_mem_write = in_dcache_mem_write;
for (int cur_t = 0; cur_t < a.getNThds(); cur_t++)
{
cache_simulator->in_dcache_in_valid[cur_t] = in_dcache_in_valid[cur_t];
cache_simulator->in_dcache_in_address[cur_t] = in_dcache_in_address[cur_t];
}
}
else
{
cache_simulator->in_dcache_mem_read = NO_MEM_READ;
cache_simulator->in_dcache_mem_write = NO_MEM_WRITE;
for (int cur_t = 0; cur_t < a.getNThds(); cur_t++)
{
cache_simulator->in_dcache_in_valid[cur_t] = 0;
}
}
cache_simulator->clk = 0;
cache_simulator->eval();
m_trace->dump(2*curr_cycle+1);
curr_cycle++;
}
}
}
void Core::warpScheduler()
{
int numSteps = 0;
bool cont;
do
{
numSteps++;
schedule_w = (schedule_w+1) % w.size();
bool has_active_threads = (w[schedule_w].activeThreads > 0);
bool stalled = stallWarp[schedule_w];
cont = ((!has_active_threads) || (stalled)) && (numSteps <= w.size());
// cout << "&&&&&&&WID: " << schedule_w << '\n';
// cout << "activeThreads: " << w[schedule_w].activeThreads << "\t!has_active_threads: " << (!has_active_threads) << '\n';
// cout << "stalled: " << stalled << '\n';
// cout << "numSteps: " << numSteps << " CONT: " << cont << '\n';
} while (cont);
if (numSteps > w.size())
{
this->foundSchedule = false;
}
else
{
this->foundSchedule = true;
}
}
void Core::fetch()
{
++steps;
#ifdef PRINT_ACTIVE_THREADS
cout << endl << "Threads:";
#endif
// D(-1, "Found schedule: " << foundSchedule);
for (unsigned i = 0; i < w.size(); ++i) {
if (w[i].activeThreads) {
if ((!inst_in_scheduler.stalled) && (inst_in_fetch.fetch_stall_cycles == 0))
{
// CPY_TRACE(inst_in_decode, inst_in_fetch);
// if (w[schedule_w].activeThreads)
{
trace_inst_t trace_inst;
trace_inst.valid_inst = false;
trace_inst.pc = 0;
trace_inst.wid = i;
trace_inst.rs1 = -1;
trace_inst.rs2 = -1;
trace_inst.rd = -1;
trace_inst.is_lw = false;
trace_inst.is_sw = false;
trace_inst.mem_addresses = new unsigned[a.getNThds()];
trace_inst.mem_stall_cycles = 0;
trace_inst.fetch_stall_cycles = 0;
trace_inst.stall_warp = false;
D(3, "Core step stepping warp " << i << '[' << w[i].activeThreads << ']');
w[i].step(&trace_inst);
D(3, "Now " << w[i].activeThreads << " active threads in " << i);
D(-1, "********************************");
D(-1, "*** valid: " << trace_inst.valid_inst << " pc: " << hex << trace_inst.pc << dec << " rs1..rs2..rd " << trace_inst.rs1 << ".." << trace_inst.rs2 << ".." << trace_inst.rd << "\n");
D(-1, "********************************");
INIT_TRACE(inst_in_fetch);
if (foundSchedule)
{
D(3, "Core step stepping warp " << schedule_w << '[' << w[schedule_w].activeThreads << ']');
w[schedule_w].step(&inst_in_fetch);
D(3, "Now " << w[schedule_w].activeThreads << " active threads in " << schedule_w);
this->getCacheDelays(&inst_in_fetch);
if (inst_in_fetch.stall_warp)
{
stallWarp[inst_in_fetch.wid] = true;
}
}
warpScheduler();
}
}
else
{
inst_in_fetch.stalled = false;
if (inst_in_fetch.fetch_stall_cycles > 0) inst_in_fetch.fetch_stall_cycles--;
}
printTrace(&inst_in_fetch, "Fetch");
#ifdef PRINT_ACTIVE_THREADS
for (unsigned j = 0; j < w[i].tmask.size(); ++j) {
if (w[i].activeThreads > j && w[i].tmask[j]) cout << " 1";
else cout << " 0";
if (j != w[i].tmask.size()-1 || i != w.size()-1) cout << ',';
}
#endif
}
// #ifdef PRINT_ACTIVE_THREADS
// for (unsigned j = 0; j < w[schedule_w].tmask.size(); ++j) {
// if (w[schedule_w].activeThreads > j && w[schedule_w].tmask[j]) cout << " 1";
// else cout << " 0";
// if (j != w[schedule_w].tmask.size()-1 || schedule_w != w.size()-1) cout << ',';
// }
// #endif
#ifdef PRINT_ACTIVE_THREADS
cout << endl;
#endif
@ -107,31 +423,198 @@ void Core::fetch()
void Core::decode()
{
if ((inst_in_fetch.fetch_stall_cycles == 0) && !inst_in_scheduler.stalled)
{
CPY_TRACE(inst_in_decode, inst_in_fetch);
INIT_TRACE(inst_in_fetch);
}
printTrace(&inst_in_decode, "Decode");
}
void Core::scheduler()
{
}
void Core::gpr_read()
{
}
void Core::execute_unit()
{
if (!inst_in_scheduler.stalled)
{
CPY_TRACE(inst_in_scheduler, inst_in_decode);
INIT_TRACE(inst_in_decode);
}
printTrace(&inst_in_scheduler, "scheduler");
}
void Core::load_store()
{
bool do_nothing = false;
if ((inst_in_lsu.mem_stall_cycles > 0) || (inst_in_lsu.stalled))
{
// LSU currently busy
if ((inst_in_scheduler.is_lw || inst_in_scheduler.is_sw))
{
inst_in_scheduler.stalled = true;
}
do_nothing = true;
}
else
{
// LSU not busy
if (inst_in_scheduler.is_lw || inst_in_scheduler.is_sw)
{
// Scheduler has LSU inst
bool scheduler_srcs_ready = true;
if (inst_in_scheduler.rs1 > 0)
{
scheduler_srcs_ready = scheduler_srcs_ready && renameTable[inst_in_scheduler.wid][inst_in_scheduler.rs1];
}
if (inst_in_scheduler.rs2 > 0)
{
scheduler_srcs_ready = scheduler_srcs_ready && renameTable[inst_in_scheduler.wid][inst_in_scheduler.rs2];
}
if (scheduler_srcs_ready)
{
if (inst_in_scheduler.rd != -1) renameTable[inst_in_scheduler.wid][inst_in_scheduler.rd] = false;
CPY_TRACE(inst_in_lsu, inst_in_scheduler);
INIT_TRACE(inst_in_scheduler);
}
else
{
inst_in_scheduler.stalled = true;
// INIT_TRACE(inst_in_lsu);
do_nothing = true;
}
}
else
{
// INIT_TRACE(inst_in_lsu);
do_nothing = true;
}
}
if (inst_in_lsu.mem_stall_cycles > 0) inst_in_lsu.mem_stall_cycles--;
printTrace(&inst_in_lsu, "LSU");
}
void Core::execute_unit()
{
// cout << "$$$$$$$$$$$$$$$$$$$ EXE START\n";
bool do_nothing = false;
// EXEC is always not busy
if (inst_in_scheduler.is_lw || inst_in_scheduler.is_sw)
{
// Not an execute instruction
// INIT_TRACE(inst_in_exe);
do_nothing = true;
}
else
{
bool scheduler_srcs_ready = true;
if (inst_in_scheduler.rs1 > 0)
{
scheduler_srcs_ready = scheduler_srcs_ready && renameTable[inst_in_scheduler.wid][inst_in_scheduler.rs1];
// cout << "Rename RS1: " << inst_in_scheduler.rs1 << " is " << renameTable[inst_in_scheduler.wid][inst_in_scheduler.rs1] << " wid: " << inst_in_scheduler.wid << '\n';
}
if (inst_in_scheduler.rs2 > 0)
{
scheduler_srcs_ready = scheduler_srcs_ready && renameTable[inst_in_scheduler.wid][inst_in_scheduler.rs2];
// cout << "Rename RS2: " << inst_in_scheduler.rs1 << " is " << renameTable[inst_in_scheduler.wid][inst_in_scheduler.rs2] << " wid: " << inst_in_scheduler.wid << '\n';
}
if (scheduler_srcs_ready)
{
if (inst_in_scheduler.rd != -1) {
// cout << "rename setting rd: " << inst_in_scheduler.rd << " to not useabel wid: " << inst_in_scheduler.wid << '\n';
renameTable[inst_in_scheduler.wid][inst_in_scheduler.rd] = false;
}
CPY_TRACE(inst_in_exe, inst_in_scheduler);
INIT_TRACE(inst_in_scheduler);
}
else
{
cout << "&&&&&&&&&&&&&&&&&&&&&&&& EXECUTE SRCS NOT READY\n";
inst_in_scheduler.stalled = true;
// INIT_TRACE(inst_in_exe);
do_nothing = true;
}
}
// if (!do_nothing)
// {
// }
printTrace(&inst_in_exe, "execute_unit");
// INIT_TRACE(inst_in_exe);
}
void Core::writeback()
{
if (inst_in_wb.rd > 0) renameTable[inst_in_wb.wid][inst_in_wb.rd] = true;
if (inst_in_wb.stall_warp)
{
// stallWarp[inst_in_wb.wid] = false;
release_warp = true;
release_warp_num = inst_in_wb.wid;
}
INIT_TRACE(inst_in_wb);
bool serviced_exe = false;
bool serviced_mem = false;
if ((inst_in_exe.rd > 0) || (inst_in_exe.stall_warp))
{
CPY_TRACE(inst_in_wb, inst_in_exe);
INIT_TRACE(inst_in_exe);
serviced_exe = true;
// cout << "WRITEBACK SERVICED EXE\n";
}
if ((inst_in_lsu.rd > 0) && (inst_in_lsu.mem_stall_cycles == 0))
{
if (serviced_exe)
{
cout << "$$$$$$$$$$$$$$$$$$$$ Stalling LSU because EXE is being used\n";
inst_in_lsu.stalled = true;
}
else
{
serviced_mem = true;
CPY_TRACE(inst_in_wb, inst_in_lsu);
INIT_TRACE(inst_in_lsu);
}
}
// if (!serviced_exe && !serviced_mem) INIT_TRACE(inst_in_wb);
printTrace(&inst_in_wb, "Writeback");
}
bool Core::running() const {
bool stages_have_valid = inst_in_fetch.valid_inst || inst_in_decode.valid_inst || inst_in_scheduler.valid_inst ||
inst_in_lsu.valid_inst || inst_in_exe.valid_inst || inst_in_wb.valid_inst;
if (stages_have_valid) return true;
for (unsigned i = 0; i < w.size(); ++i)
if (w[i].running()) return true;
if (w[i].running())
{
cout << "Warp ID " << i << " is running\n";
return true;
}
return false;
}
@ -140,8 +623,7 @@ void Core::printStats() const {
for (unsigned i = 0; i < w.size(); ++i)
insts += w[i].insts;
cout << "Total steps: " << steps << endl;
cout << "Total insts: " << insts << endl;
cerr << "Total steps: " << steps << endl;
for (unsigned i = 0; i < w.size(); ++i) {
cout << "=== Warp " << i << " ===" << endl;
w[i].printStats();
@ -190,7 +672,7 @@ void Warp::step(trace_inst_t * trace_inst) {
if (activeThreads == 0) return;
++steps;
// ++steps;
D(3, "in step pc=0x" << hex << pc);
@ -219,7 +701,7 @@ void Warp::step(trace_inst_t * trace_inst) {
// Execute
inst->executeOn(*this);
inst->executeOn(*this, trace_inst);
// At Debug Level 3, print debug info after each instruction.
@ -274,13 +756,13 @@ bool Warp::interrupt(Word r0) {
}
void Warp::printStats() const {
cerr << "Steps : " << steps << endl
cout << "Steps : " << steps << endl
<< "Insts : " << insts << endl
<< "Loads : " << loads << endl
<< "Stores: " << stores << endl;
unsigned const grade = reg[0][28];
if (grade == 1) cerr << "GRADE: PASSED\n";
else cerr << "GRADE: FAILED " << (grade >> 1) << "\n";
if (grade == 1) cout << "GRADE: PASSED\n";
else cout << "GRADE: FAILED " << (grade >> 1) << "\n";
}

View file

@ -19,9 +19,9 @@
#include "Vcache_simX.h"
#include "verilated.h"
#ifdef VCD_OUTPUT
// #ifdef VCD_OUTPUT
#include <verilated_vcd_c.h>
#endif
// #endif
#include "trace.h"
@ -92,17 +92,34 @@ namespace Harp {
public:
Core(const ArchDef &a, Decoder &d, MemoryUnit &mem, Word id=0);
Vcache_simX * cache_simulator;
Vcache_simX * cache_simulator;
VerilatedVcdC * m_trace;
bool renameTable[32][32];
bool stallWarp[32];
bool foundSchedule;
trace_inst_t inst_in_fetch;
trace_inst_t inst_in_decode;
trace_inst_t inst_in_scheduler;
trace_inst_t inst_in_exe;
trace_inst_t inst_in_lsu;
trace_inst_t inst_in_wb;
bool release_warp;
int release_warp_num;
bool interrupt(Word r0);
bool running() const;
void getCacheDelays(trace_inst_t *);
void warpScheduler();
void fetch();
void decode();
void scheduler();
void gpr_read();
void execute_unit();
void load_store();
void writeback();
void step();
@ -117,6 +134,7 @@ namespace Harp {
unsigned long steps;
std::vector<Warp> w;
std::map<Word, std::set<Warp *> > b; // Barriers
int schedule_w;
};
class Warp {

View file

@ -8,7 +8,7 @@
#include <iostream>
#include "types.h"
#include "trace.h"
namespace Harp {
class Warp;
class Ref;
@ -79,7 +79,7 @@ namespace Harp {
{
}
void executeOn(Warp &warp);
void executeOn(Warp &warp, trace_inst_t *);
friend std::ostream &operator<<(std::ostream &, Instruction &);
/* Setters used to "craft" the instruction. */

View file

@ -28,6 +28,9 @@ namespace Harp {
// Instruction execute
bool stall_warp;
bool wspawn;
bool stalled;
} trace_inst_t;
}

View file

@ -78,7 +78,7 @@ Word signExt(Word w, Size bit, Word mask) {
return w;
}
void Instruction::executeOn(Warp &c) {
void Instruction::executeOn(Warp &c, trace_inst_t * trace_inst) {
D(3, "Begin instruction execute.");
/* If I try to execute a privileged instruction in user mode, throw an
@ -357,14 +357,14 @@ void Instruction::executeOn(Warp &c) {
break;
case L_INST:
//std::cout << "L_INST\n";
memAddr = ((reg[rsrc[0]] + immsrc) & 0xFFFFFFFC);
shift_by = ((reg[rsrc[0]] + immsrc) & 0x00000003) * 8;
data_read = c.core->mem.read(memAddr, c.supervisorMode);
trace_inst->is_lw = true;
trace_inst->mem_addresses[t] = memAddr;
// //std::cout <<std::hex<< "EXECUTE: " << reg[rsrc[0]] << " + " << immsrc << " = " << memAddr << " -> data_read: " << data_read << "\n";
#ifdef EMU_INSTRUMENTATION
Harp::OSDomain::osDomain->
do_mem(0, memAddr, c.core->mem.virtToPhys(memAddr), 8, true);
#endif
switch (func3)
{
@ -484,6 +484,8 @@ void Instruction::executeOn(Warp &c) {
++c.stores;
memAddr = reg[rsrc[0]] + immsrc;
std::cout << "STORE MEM ADDRESS: " << std::hex << reg[rsrc[0]] << " + " << immsrc << "\n";
trace_inst->is_sw = true;
trace_inst->mem_addresses[t] = memAddr;
// //std::cout << "FUNC3: " << func3 << "\n";
if ((memAddr == 0x00010000) && (t == 0))
{
@ -517,6 +519,7 @@ void Instruction::executeOn(Warp &c) {
break;
case B_INST:
//std::cout << "B_INST\n";
trace_inst->stall_warp = true;
switch (func3)
{
case 0:
@ -579,6 +582,7 @@ void Instruction::executeOn(Warp &c) {
break;
case JAL_INST:
//std::cout << "JAL_INST\n";
trace_inst->stall_warp = true;
if (!pcSet) nextPc = (c.pc - 4) + immsrc;
if (!pcSet) {/*std::cout << "JAL... SETTING PC: " << nextPc << "\n"; */}
if (rdest != 0)
@ -589,6 +593,7 @@ void Instruction::executeOn(Warp &c) {
break;
case JALR_INST:
std::cout << "JALR_INST\n";
trace_inst->stall_warp = true;
if (!pcSet) nextPc = reg[rsrc[0]] + immsrc;
if (!pcSet) {/*std::cout << "JALR... SETTING PC: " << nextPc << "\n";*/ }
if (rdest != 0)
@ -704,6 +709,7 @@ void Instruction::executeOn(Warp &c) {
case 1:
// WSPAWN
std::cout << "WSPAWN\n";
trace_inst->wspawn = true;
if (sjOnce)
{
sjOnce = false;
@ -745,6 +751,7 @@ void Instruction::executeOn(Warp &c) {
{
// SPLIT
//std::cout << "SPLIT\n";
trace_inst->stall_warp = true;
if (sjOnce)
{
sjOnce = false;
@ -812,12 +819,13 @@ void Instruction::executeOn(Warp &c) {
}
break;
case 4:
trace_inst->stall_warp = true;
// is_barrier
break;
case 0:
// TMC
//std::cout << "JALRS\n";
trace_inst->stall_warp = true;
nextActiveThreads = reg[rsrc[0]];
{
for (int ff = 0; ff < c.tmask.size(); ff++)

View file

@ -104,6 +104,24 @@ int emu_main(int argc, char **argv) {
// core.w[0].pc = 0x8000007c; // If I want to start at a specific location
std::cout << "ABOUT TO START\n";
// bool count_down = false;
// int cycles_left;
// while (!count_down || (count_down && (cycles_left == 0)))
// {
// if (count_down)
// {
// cycles_left--;
// }
// console.poll();
// core.step();
// bool run = core.running();
// if (!run)
// {
// count_down = true;
// }
// }
while (core.running()) { console.poll(); core.step(); }
if (showStats) core.printStats();

View file

@ -3,4 +3,5 @@ echo start > results.txt
# echo ../kernel/vortex_test.hex
make clean
make
echo "Fasten your seatbelts ladies and gentelmen!!"
cd obj_dir && ./Vcache_simX -E -a rv32i --core ../../runtime/mains/dev/vx_dev_main.hex -s -b 1> emulator.debug