wu-arch/kernels
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

simulation framework refactoring

Browse Source
This commit is contained in:
Blaise Tine
2021-10-09 10:20:42 -04:00
parent 51673665b5
commit 54bddeee9c
89 changed files with 1217 additions and 1471 deletions
Download Patch File Download Diff File Expand all files Collapse all files

442
sim/vlsim/opae_sim.cpp Normal file
View File

@@ -0,0 +1,442 @@
#include "opae_sim.h"
#include <verilated.h>
#include "Vvortex_afu_shim.h"
#include "Vvortex_afu_shim__Syms.h"
#ifdef VCD_OUTPUT
#include <verilated_vcd_c.h>
#endif
#include <iostream>
#include <fstream>
#include <iomanip>
#include <mem.h>
#define CCI_LATENCY 8
#define CCI_RAND_MOD 8
#define CCI_RQ_SIZE 16
#define CCI_WQ_SIZE 16
#define ENABLE_MEM_STALLS
#ifndef TRACE_START_TIME
#define TRACE_START_TIME 0ull
#endif
#ifndef TRACE_STOP_TIME
#define TRACE_STOP_TIME -1ull
#endif
#ifndef MEM_LATENCY
#define MEM_LATENCY 24
#endif
#ifndef MEM_RQ_SIZE
#define MEM_RQ_SIZE 16
#endif
#ifndef MEM_STALLS_MODULO
#define MEM_STALLS_MODULO 16
#endif
#ifndef VERILATOR_RESET_VALUE
#define VERILATOR_RESET_VALUE 2
#endif
using namespace vortex;
static uint64_t timestamp = 0;
double sc_time_stamp() {
return timestamp;
}
static void *__aligned_malloc(size_t alignment, size_t size) {
// reserve margin for alignment and storing of unaligned address
size_t margin = (alignment-1) + sizeof(void*);
void *unaligned_addr = malloc(size + margin);
void **aligned_addr = (void**)((uintptr_t)(((uint8_t*)unaligned_addr) + margin) & ~(alignment-1));
aligned_addr[-1] = unaligned_addr;
return aligned_addr;
}
static void __aligned_free(void *ptr) {
// retreive the stored unaligned address and use it to free the allocation
void* unaligned_addr = ((void**)ptr)[-1];
free(unaligned_addr);
}
///////////////////////////////////////////////////////////////////////////////
static bool trace_enabled = false;
static uint64_t trace_start_time = TRACE_START_TIME;
static uint64_t trace_stop_time = TRACE_STOP_TIME;
bool sim_trace_enabled() {
if (timestamp >= trace_start_time
&& timestamp < trace_stop_time)
return true;
return trace_enabled;
}
void sim_trace_enable(bool enable) {
trace_enabled = enable;
}
///////////////////////////////////////////////////////////////////////////////
namespace vortex {
class VL_OBJ {
public:
#ifdef AXI_BUS
VVortex_axi *device;
#else
Vvortex_afu_shim *device;
#endif
#ifdef VCD_OUTPUT
VerilatedVcdC *trace;
#endif
VL_OBJ() {
// force random values for unitialized signals
Verilated::randReset(VERILATOR_RESET_VALUE);
Verilated::randSeed(50);
// Turn off assertion before reset
Verilated::assertOn(false);
#ifdef AXI_BUS
this->device = new Vvortex_afu_shim();
#else
this->device = new Vvortex_afu_shim();
#endif
#ifdef VCD_OUTPUT
Verilated::traceEverOn(true);
this->trace = new VerilatedVcdC();
this->device->trace(this->trace, 99);
this->trace->open("trace.vcd");
#endif
}
~VL_OBJ() {
#ifdef VCD_OUTPUT
this->trace->close();
delete this->trace;
#endif
delete this->device;
}
};
}
///////////////////////////////////////////////////////////////////////////////
opae_sim::opae_sim()
: stop_(false)
, host_buffer_ids_(0) {
vl_obj_ = new VL_OBJ();
ram_ = new RAM((1<<12), (1<<20));
// reset the device
this->reset();
// launch execution thread
future_ = std::async(std::launch::async, [&]{
while (!stop_) {
std::lock_guard<std::mutex> guard(mutex_);
this->step();
}
});
}
opae_sim::~opae_sim() {
stop_ = true;
if (future_.valid()) {
future_.wait();
}
for (auto& buffer : host_buffers_) {
__aligned_free(buffer.second.data);
}
delete vl_obj_;
delete ram_;
}
int opae_sim::prepare_buffer(uint64_t len, void **buf_addr, uint64_t *wsid, int flags) {
auto alloc = __aligned_malloc(CACHE_BLOCK_SIZE, len);
if (alloc == NULL)
return -1;
host_buffer_t buffer;
buffer.data = (uint64_t*)alloc;
buffer.size = len;
buffer.ioaddr = uintptr_t(alloc);
auto buffer_id = host_buffer_ids_++;
host_buffers_.emplace(buffer_id, buffer);
*buf_addr = alloc;
*wsid = buffer_id;
return 0;
}
void opae_sim::release_buffer(uint64_t wsid) {
auto it = host_buffers_.find(wsid);
if (it != host_buffers_.end()) {
__aligned_free(it->second.data);
host_buffers_.erase(it);
}
}
void opae_sim::get_io_address(uint64_t wsid, uint64_t *ioaddr) {
*ioaddr = host_buffers_[wsid].ioaddr;
}
void opae_sim::read_mmio64(uint32_t mmio_num, uint64_t offset, uint64_t *value) {
std::lock_guard<std::mutex> guard(mutex_);
vl_obj_->device->vcp2af_sRxPort_c0_mmioRdValid = 1;
vl_obj_->device->vcp2af_sRxPort_c0_ReqMmioHdr_address = offset / 4;
vl_obj_->device->vcp2af_sRxPort_c0_ReqMmioHdr_length = 1;
vl_obj_->device->vcp2af_sRxPort_c0_ReqMmioHdr_tid = 0;
this->step();
vl_obj_->device->vcp2af_sRxPort_c0_mmioRdValid = 0;
assert(vl_obj_->device->af2cp_sTxPort_c2_mmioRdValid);
*value = vl_obj_->device->af2cp_sTxPort_c2_data;
}
void opae_sim::write_mmio64(uint32_t mmio_num, uint64_t offset, uint64_t value) {
std::lock_guard<std::mutex> guard(mutex_);
vl_obj_->device->vcp2af_sRxPort_c0_mmioWrValid = 1;
vl_obj_->device->vcp2af_sRxPort_c0_ReqMmioHdr_address = offset / 4;
vl_obj_->device->vcp2af_sRxPort_c0_ReqMmioHdr_length = 1;
vl_obj_->device->vcp2af_sRxPort_c0_ReqMmioHdr_tid = 0;
memcpy(vl_obj_->device->vcp2af_sRxPort_c0_data, &value, 8);
this->step();
vl_obj_->device->vcp2af_sRxPort_c0_mmioWrValid = 0;
}
///////////////////////////////////////////////////////////////////////////////
void opae_sim::reset() {
cci_reads_.clear();
cci_writes_.clear();
vl_obj_->device->vcp2af_sRxPort_c0_mmioRdValid = 0;
vl_obj_->device->vcp2af_sRxPort_c0_mmioWrValid = 0;
vl_obj_->device->vcp2af_sRxPort_c0_rspValid = 0;
vl_obj_->device->vcp2af_sRxPort_c1_rspValid = 0;
vl_obj_->device->vcp2af_sRxPort_c0_TxAlmFull = 0;
vl_obj_->device->vcp2af_sRxPort_c1_TxAlmFull = 0;
for (int b = 0; b < MEMORY_BANKS; ++b) {
mem_reads_[b].clear();
vl_obj_->device->avs_readdatavalid[b] = 0;
vl_obj_->device->avs_waitrequest[b] = 0;
}
vl_obj_->device->reset = 1;
for (int i = 0; i < RESET_DELAY; ++i) {
vl_obj_->device->clk = 0;
this->eval();
vl_obj_->device->clk = 1;
this->eval();
}
vl_obj_->device->reset = 0;
// Turn on assertion after reset
Verilated::assertOn(true);
}
void opae_sim::step() {
this->sRxPort_bus();
this->sTxPort_bus();
this->avs_bus();
vl_obj_->device->clk = 0;
this->eval();
vl_obj_->device->clk = 1;
this->eval();
#ifndef NDEBUG
fflush(stdout);
#endif
}
void opae_sim::eval() {
vl_obj_->device->eval();
#ifdef VCD_OUTPUT
if (sim_trace_enabled()) {
vl_obj_->trace->dump(timestamp);
}
#endif
++timestamp;
}
void opae_sim::sRxPort_bus() {
// check mmio request
bool mmio_req_enabled = vl_obj_->device->vcp2af_sRxPort_c0_mmioRdValid
|| vl_obj_->device->vcp2af_sRxPort_c0_mmioWrValid;
// schedule CCI read responses
std::list<cci_rd_req_t>::iterator cci_rd_it(cci_reads_.end());
for (auto it = cci_reads_.begin(), ie = cci_reads_.end(); it != ie; ++it) {
if (it->cycles_left > 0)
it->cycles_left -= 1;
if ((cci_rd_it == ie) && (it->cycles_left == 0)) {
cci_rd_it = it;
}
}
// schedule CCI write responses
std::list<cci_wr_req_t>::iterator cci_wr_it(cci_writes_.end());
for (auto it = cci_writes_.begin(), ie = cci_writes_.end(); it != ie; ++it) {
if (it->cycles_left > 0)
it->cycles_left -= 1;
if ((cci_wr_it == ie) && (it->cycles_left == 0)) {
cci_wr_it = it;
}
}
// send CCI write response
vl_obj_->device->vcp2af_sRxPort_c1_rspValid = 0;
if (cci_wr_it != cci_writes_.end()) {
vl_obj_->device->vcp2af_sRxPort_c1_rspValid = 1;
vl_obj_->device->vcp2af_sRxPort_c1_hdr_resp_type = 0;
vl_obj_->device->vcp2af_sRxPort_c1_hdr_mdata = cci_wr_it->mdata;
cci_writes_.erase(cci_wr_it);
}
// send CCI read response (ensure mmio disabled)
vl_obj_->device->vcp2af_sRxPort_c0_rspValid = 0;
if (!mmio_req_enabled
&& (cci_rd_it != cci_reads_.end())) {
vl_obj_->device->vcp2af_sRxPort_c0_rspValid = 1;
vl_obj_->device->vcp2af_sRxPort_c0_hdr_resp_type = 0;
memcpy(vl_obj_->device->vcp2af_sRxPort_c0_data, cci_rd_it->data.data(), CACHE_BLOCK_SIZE);
vl_obj_->device->vcp2af_sRxPort_c0_hdr_mdata = cci_rd_it->mdata;
/*printf("%0ld: [sim] CCI Rd Rsp: addr=%ld, mdata=%d, data=", timestamp, cci_rd_it->addr, cci_rd_it->mdata);
for (int i = 0; i < CACHE_BLOCK_SIZE; ++i)
printf("%02x", cci_rd_it->data[CACHE_BLOCK_SIZE-1-i]);
printf("\n");*/
cci_reads_.erase(cci_rd_it);
}
}
void opae_sim::sTxPort_bus() {
// process read requests
if (vl_obj_->device->af2cp_sTxPort_c0_valid) {
assert(!vl_obj_->device->vcp2af_sRxPort_c0_TxAlmFull);
cci_rd_req_t cci_req;
cci_req.cycles_left = CCI_LATENCY + (timestamp % CCI_RAND_MOD);
cci_req.addr = vl_obj_->device->af2cp_sTxPort_c0_hdr_address;
cci_req.mdata = vl_obj_->device->af2cp_sTxPort_c0_hdr_mdata;
auto host_ptr = (uint64_t*)(vl_obj_->device->af2cp_sTxPort_c0_hdr_address * CACHE_BLOCK_SIZE);
memcpy(cci_req.data.data(), host_ptr, CACHE_BLOCK_SIZE);
//printf("%0ld: [sim] CCI Rd Req: addr=%ld, mdata=%d\n", timestamp, vl_obj_->device->af2cp_sTxPort_c0_hdr_address, cci_req.mdata);
cci_reads_.emplace_back(cci_req);
}
// process write requests
if (vl_obj_->device->af2cp_sTxPort_c1_valid) {
assert(!vl_obj_->device->vcp2af_sRxPort_c1_TxAlmFull);
cci_wr_req_t cci_req;
cci_req.cycles_left = CCI_LATENCY + (timestamp % CCI_RAND_MOD);
cci_req.mdata = vl_obj_->device->af2cp_sTxPort_c1_hdr_mdata;
auto host_ptr = (uint64_t*)(vl_obj_->device->af2cp_sTxPort_c1_hdr_address * CACHE_BLOCK_SIZE);
memcpy(host_ptr, vl_obj_->device->af2cp_sTxPort_c1_data, CACHE_BLOCK_SIZE);
cci_writes_.emplace_back(cci_req);
}
// check queues overflow
vl_obj_->device->vcp2af_sRxPort_c0_TxAlmFull = (cci_reads_.size() >= (CCI_RQ_SIZE-1));
vl_obj_->device->vcp2af_sRxPort_c1_TxAlmFull = (cci_writes_.size() >= (CCI_WQ_SIZE-1));
}
void opae_sim::avs_bus() {
for (int b = 0; b < MEMORY_BANKS; ++b) {
// update memory responses schedule
for (auto& rsp : mem_reads_[b]) {
if (rsp.cycles_left > 0)
rsp.cycles_left -= 1;
}
// schedule memory responses in FIFO order
std::list<mem_rd_req_t>::iterator mem_rd_it(mem_reads_[b].end());
if (!mem_reads_[b].empty()
&& (0 == mem_reads_[b].begin()->cycles_left)) {
mem_rd_it = mem_reads_[b].begin();
}
// send memory response
vl_obj_->device->avs_readdatavalid[b] = 0;
if (mem_rd_it != mem_reads_[b].end()) {
vl_obj_->device->avs_readdatavalid[b] = 1;
memcpy(vl_obj_->device->avs_readdata[b], mem_rd_it->data.data(), MEM_BLOCK_SIZE);
uint32_t addr = mem_rd_it->addr;
mem_reads_[b].erase(mem_rd_it);
/*printf("%0ld: [sim] MEM Rd Rsp: bank=%d, addr=%x, pending={", timestamp, b, addr * MEM_BLOCK_SIZE);
for (auto& req : mem_reads_[b]) {
if (req.cycles_left != 0)
printf(" !%0x", req.addr * MEM_BLOCK_SIZE);
else
printf(" %0x", req.addr * MEM_BLOCK_SIZE);
}
printf("}\n");*/
}
// handle memory stalls
bool mem_stalled = false;
#ifdef ENABLE_MEM_STALLS
if (0 == ((timestamp/2) % MEM_STALLS_MODULO)) {
mem_stalled = true;
} else
if (mem_reads_[b].size() >= MEM_RQ_SIZE) {
mem_stalled = true;
}
#endif
// process memory requests
if (!mem_stalled) {
assert(!vl_obj_->device->avs_read[b] || !vl_obj_->device->avs_write[b]);
if (vl_obj_->device->avs_write[b]) {
uint64_t byteen = vl_obj_->device->avs_byteenable[b];
unsigned base_addr = vl_obj_->device->avs_address[b] * MEM_BLOCK_SIZE;
uint8_t* data = (uint8_t*)(vl_obj_->device->avs_writedata[b]);
for (int i = 0; i < MEM_BLOCK_SIZE; i++) {
if ((byteen >> i) & 0x1) {
(*ram_)[base_addr + i] = data[i];
}
}
/*printf("%0ld: [sim] MEM Wr Req: bank=%d, addr=%x, data=", timestamp, b, base_addr);
for (int i = 0; i < MEM_BLOCK_SIZE; i++) {
printf("%02x", data[(MEM_BLOCK_SIZE-1)-i]);
}
printf("\n");*/
}
if (vl_obj_->device->avs_read[b]) {
mem_rd_req_t mem_req;
mem_req.addr = vl_obj_->device->avs_address[b];
ram_->read(mem_req.data.data(), vl_obj_->device->avs_address[b] * MEM_BLOCK_SIZE, MEM_BLOCK_SIZE);
mem_req.cycles_left = MEM_LATENCY;
for (auto& rsp : mem_reads_[b]) {
if (mem_req.addr == rsp.addr) {
// duplicate requests receive the same cycle delay
mem_req.cycles_left = rsp.cycles_left;
break;
}
}
mem_reads_[b].emplace_back(mem_req);
/*printf("%0ld: [sim] MEM Rd Req: bank=%d, addr=%x, pending={", timestamp, b, mem_req.addr * MEM_BLOCK_SIZE);
for (auto& req : mem_reads_[b]) {
if (req.cycles_left != 0)
printf(" !%0x", req.addr * MEM_BLOCK_SIZE);
else
printf(" %0x", req.addr * MEM_BLOCK_SIZE);
}
printf("}\n");*/
}
}
vl_obj_->device->avs_waitrequest[b] = mem_stalled;
}
}