Vortex 2.0 changes:

+ Microarchitecture optimizations
+ 64-bit support
+ Xilinx FPGA support
+ LLVM-16 support
+ Refactoring and quality control fixes
This commit is contained in:
Blaise Tine
2023-10-19 20:51:22 -07:00
parent d69a64c32c
commit d47cccc157
1300 changed files with 247321 additions and 311189 deletions

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
`timescale 1ns/1ps
module VX_tb_divide();
@@ -17,7 +30,7 @@ module VX_tb_divide();
wire [31:0] o_div[0:7], o_rem[0:7];
for (genvar i = 0; i < 8; i++) begin
for (genvar i = 0; i < 8; ++i) begin
VX_divide#(
.WIDTHN(32),
.WIDTHD(32),

View File

@@ -1,4 +1,6 @@
PARAMS += -DCACHE_SIZE=4096 -DCACHE_WORD_SIZE=4 -DCACHE_LINE_SIZE=16 -DCACHE_NUM_BANKS=4 -DCACHE_CREQ_SIZE=4 -DMRVQ_SIZE=16 -DDFPQ_SIZE=16 -DSNRQ_SIZE=16 -DCWBQ_SIZE=4 -DDWBQ_SIZE=4 -DFQQ_SIZE=4
PARAMS += -DCACHE_SIZE=4096 -DCACHE_WORD_SIZE=4 -DCACHE_LINE_SIZE=16 -DCACHE_NUM_BANKS=4 -DMRVQ_SIZE=16 -DDFPQ_SIZE=16 -DSNRQ_SIZE=16 -DCWBQ_SIZE=4 -DDWBQ_SIZE=4 -DFQQ_SIZE=4
TOP = VX_cache
# control RTL debug tracing states
DBG_TRACE_FLAGS = -DDBG_TRACE_CORE_PIPELINE \
@@ -31,7 +33,7 @@ VF += -exe $(SRCS) $(INCLUDE)
VF += $(PARAMS)
gen:
verilator $(VF) -cc $(TOP).v -CFLAGS '$(CF)' --exe $(SRCS)
verilator $(VF) -cc $(TOP).sv -CFLAGS '$(CF)' --exe $(SRCS)
build: gen
(cd obj_dir && make -j -f V$(TOP).mk)

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "cachesim.h"
#include <fstream>
#include <iomanip>
@@ -235,7 +248,7 @@ void CacheSim::eval_mem_bus() {
if (cache_->mem_req_valid) {
if (cache_->mem_req_rw) { //write = 1
uint64_t byteen = cache_->mem_req_byteen;
unsigned base_addr = (cache_->mem_req_addr * MEM_BLOCK_SIZE);
uint64_t base_addr = (cache_->mem_req_addr * MEM_BLOCK_SIZE);
uint8_t* data = (uint8_t*)(cache_->mem_req_data);
for (int i = 0; i < MEM_BLOCK_SIZE; i++) {
if ((byteen >> i) & 0x1) {

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include "VVX_cache.h"

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <stdio.h>

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "cachesim.h"
#include <iostream>
#include <fstream>

View File

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <array>

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "vl_simulator.h"
#include "VVX_fifo_queue.h"
#include <iostream>

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
`timescale 1ns/1ns
`include "VX_fifo_queue.v"
@@ -15,8 +28,8 @@ module testbench();
wire empty;
VX_fifo_queue #(
.DATAW(4),
.SIZE(4)
.DATAW (4),
.DEPTH (4)
) dut (
.clk(clk),
.reset(reset),

View File

@@ -1,3 +1,16 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <array>

View File

@@ -0,0 +1,65 @@
#---------------------------------------------------------
# Makefile to compile and test the memory stream unit
#---------------------------------------------------------
TOP = VX_mem_scheduler
PARAMS += -GNUM_REQS=4 -GADDRW=8 -GDATAW=8 -GTAGW=8 -GWORD_SIZE=1 -GQUEUE_SIZE=4
ifdef RSP_PARTIAL
PARAMS += -GRSP_PARTIAL=$(RSP_PARTIAL)
endif
ifdef DUPLICATE_ADDR
PARAMS += -GDUPLICATE_ADDR=$(DUPLICATE_ADDR)
endif
RTL_DIR = ../../../rtl
DPI_DIR = ../../../dpi
RTL_INCLUDE = -I$(RTL_DIR) -I$(RTL_DIR)/libs -I$(DPI_DIR)
VERILATOR_ROOT = /opt/verilator
VERILATOR ?= $(VERILATOR_ROOT)/bin/verilator
VL_FLAGS += --exe --cc $(TOP).sv --top-module $(TOP)
VL_FLAGS += --language 1800-2009 --assert -Wall
VL_FLAGS += -Wno-DECLFILENAME -Wno-REDEFMACRO
VL_FLAGS += --x-initial unique --x-assign unique
VL_FLAGS += --trace
VL_FLAGS += $(RTL_INCLUDE)
VL_FLAGS += $(PARAMS)
SRCS += memsim.cpp ram.cpp $(DPI_DIR)/util_dpi.cpp
CXXFLAGS += -std=c++11 -Wall -Wextra -Wfatal-errors -Wno-array-bounds -Wno-maybe-uninitialized
CXXFLAGS += -I../../../../dpi -I../../../../
default: run
gen: $(SRCS)
@echo
@echo "### VERILATE ###"
$(VERILATOR) $(VL_FLAGS) $^ -CFLAGS '$(CXXFLAGS)'
build: gen
@echo
@echo "### BUILD ###"
$(MAKE) -C obj_dir -j 4 -f V$(TOP).mk
run: build
@echo
@echo "### RUN ###"
obj_dir/V$(TOP)
waves: trace.vcd
@echo
@echo "### TRACE ###"
gtkwave -o trace.vcd
clean:
@echo
@echo "### CLEAN ###"
-rm -rf obj_dir *.vcd *.log
#---------------------------------------------------------

View File

@@ -0,0 +1,176 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include <random>
#include "memsim.h"
#include "ram.h"
static bool trace_enabled = false;
static uint64_t trace_start_time = 0;
static uint64_t trace_stop_time = -1ull;
static uint64_t timestamp = 0;
double sc_time_stamp() {
return timestamp;
}
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;
}
//////////////////////////////////////////////////////
int generate_rand (int min, int max) {
int range = max - min + 1;
return rand() % range + min;
}
//////////////////////////////////////////////////////
int generate_rand_mask (int mask) {
int result = 0;
int m = mask;
for (int i = 0; i < 4; i++) {
int bit = m & 0b1;
int rand_bit = generate_rand (0, bit);
result |= (rand_bit << i);
m = m >> 1;
}
return result;
}
//////////////////////////////////////////////////////
MemSim::MemSim() {
msu_ = new VVX_mem_streamer();
// Enable tracing
Verilated::traceEverOn(true);
trace_ = new VerilatedVcdC;
msu_->trace(trace_, 99);
trace_->open("trace.vcd");
}
//////////////////////////////////////////////////////
MemSim::~MemSim() {
trace_->close();
delete msu_;
}
//////////////////////////////////////////////////////
void MemSim::eval() {
msu_->eval();
trace_->dump(timestamp++);
}
//////////////////////////////////////////////////////
void MemSim::step() {
msu_->clk = 0;
this->eval();
msu_->clk = 1;
this->eval();
}
//////////////////////////////////////////////////////
void MemSim::reset() {
msu_->reset = 1;
this->step();
msu_->reset = 0;
this->step();
}
//////////////////////////////////////////////////////
void MemSim::attach_core() {
if (msu_->req_ready) {
msu_->req_valid = generate_rand(0, 1);
msu_->req_rw = generate_rand(0, 1);
msu_->req_mask = generate_rand(0b0001, 0b1111);
msu_->req_byteen = 0b1;
msu_->req_addr = generate_rand(0, 0x10000000);
msu_->req_data = generate_rand(0x60000000, 0x80000000);
msu_->req_tag = generate_rand(0x00, 0xFF);
}
msu_->rsp_ready = true;
}
//////////////////////////////////////////////////////
void MemSim::attach_ram (RAM *ram) {
req_t req;
req.valid = msu_->mem_req_valid;
req.rw = msu_->mem_req_rw;
req.byteen = msu_->mem_req_byteen;
req.addr = msu_->mem_req_addr;
req.data = msu_->mem_req_data;
req.tag = msu_->mem_req_tag;
msu_->mem_req_ready = ram->is_ready();
ram->insert_req(req);
rsp_t rsp;
rsp = ram->schedule_rsp();
msu_->mem_rsp_valid = rsp.valid;
msu_->mem_rsp_mask = rsp.mask;
msu_->mem_rsp_data = rsp.data;
msu_->mem_rsp_tag = rsp.tag;
rsp.ready = msu_->mem_rsp_ready;
std::cout<<"MEMSIM: mem_rsp_ready: "<<rsp.ready<<"\n";
ram->halt_rsp(rsp);
}
//////////////////////////////////////////////////////
void MemSim::run(RAM *ram) {
this->reset();
while (sc_time_stamp() < SIM_TIME) {
this->step();
std::cout<<"========================="<<"\n";
std::cout<<"Cycle: "<<sc_time_stamp()<<"\n";
this->attach_core();
this->attach_ram(ram);
}
}
//////////////////////////////////////////////////////
int main (int argc, char** argv, char** env) {
Verilated::commandArgs(argc, argv);
MemSim memsim;
RAM ram;
memsim.run(&ram);
return 0;
}
//////////////////////////////////////////////////////

View File

@@ -0,0 +1,47 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <iostream>
#include <unordered_map>
#include <vector>
#include <verilated.h>
#include <verilated_vcd_c.h>
#include "VVX_mem_streamer.h"
#include "VVX_mem_streamer__Syms.h"
#include "ram.h"
#define SIM_TIME 5000
int generate_rand (int min, int max);
int generate_rand_mask (int mask);
class MemSim {
private:
VVX_mem_streamer *msu_;
VerilatedVcdC *trace_;
void eval();
void step();
void reset();
void attach_core();
void attach_ram(RAM *ram);
public:
MemSim();
virtual ~MemSim();
void run(RAM *ram);
};

View File

@@ -0,0 +1,137 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "ram.h"
#include "memsim.h"
RAM::RAM() {
ram_.clear();
is_rsp_active_ = false;
is_rsp_stall_ = false;
}
//////////////////////////////////////////////////////
bool RAM::check_duplicate_req(req_t req) {
for(int i = 0; i < ram_.size(); i++) {
if (ram_[i].addr == req.addr) {
std::cout<<"RAM: Duplicate entry. Do not insert..."<<std::endl;
return true;
}
}
return false;
}
//////////////////////////////////////////////////////
int RAM::simulate_cycle_delay() {
std::cout<<"RAM: # entries: "<<ram_.size()<<std::endl;
int dequeue_index = -1;
for (int i = 0; i < ram_.size(); i++) {
if (!is_rsp_stall_) {
if (ram_[i].cycles_left > 0) {
ram_[i].cycles_left -= 1;
}
}
std::cout<<"RAM: # cycles left: "<<ram_[i].cycles_left<<std::endl;
if (ram_[i].cycles_left == 0) {
dequeue_index = i;
}
}
return dequeue_index;
}
//////////////////////////////////////////////////////
void RAM::insert_req(req_t req) {
if ( !(this->check_duplicate_req(req)) && req.valid && !req.rw) {
req_t r;
r.valid = req.valid;
r.rw = req.rw;
r.byteen = req.byteen;
r.addr = req.addr;
r.data = req.data;
r.tag = req.tag & 0b11;
// Store metadata
r.cycles_left = MEM_LATENCY;
std::cout<<"RAM: Insert entry... "<<std::endl;
std::cout<<"Write? : "<<req.rw<<std::endl;
ram_.push_back(r);
}
}
//////////////////////////////////////////////////////
uint8_t RAM::is_ready() {
// return generate_rand(0b1000, 0b1111);
return 0b1111;
}
//////////////////////////////////////////////////////
rsp_t RAM::schedule_rsp() {
rsp_t rsp;
int dequeue_index = this->simulate_cycle_delay();
if (!is_rsp_active_) {
if (dequeue_index != -1) {
std::cout<<"RAM: Scheduling response... "<<std::endl;
is_rsp_active_ = true;
rsp.valid = 1;
rsp.mask = generate_rand_mask(ram_[dequeue_index].valid);
rsp.data = generate_rand(0x20000000, 0x30000000);
rsp.tag = ram_[dequeue_index].tag;
std::cout<<std::hex;
std::cout<<"RAM: Response mask: "<<+rsp.mask<<" | Required mask: "<<+ram_[dequeue_index].valid<<std::endl;
ram_[dequeue_index].rsp_sent_mask = rsp.mask;
ram_[dequeue_index].valid = ram_[dequeue_index].valid & ~ram_[dequeue_index].rsp_sent_mask;
if (0 == ram_[dequeue_index].valid) {
ram_.erase(ram_.begin() + dequeue_index);
is_rsp_stall_ = false;
std::cout<<"RAM: Clear entry... "<<std::endl;
} else {
is_rsp_stall_ = true;
std::cout<<"RAM: Stall... "<<std::endl;
}
} else {
rsp.valid = false;
}
}
return rsp;
}
//////////////////////////////////////////////////////
// Schedule response for only one cycle
void RAM::halt_rsp(rsp_t rsp) {
if (is_rsp_active_ && rsp.valid && rsp.ready) {
std::cout<<"RAM: Halt response..."<<std::endl;
is_rsp_active_ = false;
}
}
//////////////////////////////////////////////////////

View File

@@ -0,0 +1,64 @@
// Copyright © 2019-2023
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <iostream>
#include <vector>
#define MEM_LATENCY 4
typedef struct {
uint8_t valid;
bool rw;
uint8_t byteen;
uint32_t addr;
uint32_t data;
uint8_t tag;
uint8_t ready;
// Metadata
uint8_t rsp_sent_mask;
double cycles_left;
} req_t;
typedef struct {
bool valid;
uint8_t mask;
uint32_t data;
uint8_t tag;
bool ready;
} rsp_t;
class RAM {
private:
std::vector<req_t> ram_;
bool is_rsp_active_;
bool is_rsp_stall_;
bool check_duplicate_req(req_t req);
int simulate_cycle_delay();
public:
RAM();
uint8_t is_ready();
void insert_req(req_t req);
rsp_t schedule_rsp();
void halt_rsp(rsp_t rsp);
};
//////////////////////////////////////////////////////

View File

@@ -1,30 +0,0 @@
TOP = VX_tex_sampler
PARAMS ?=
INCLUDE = -I../../../rtl/ -I../../../rtl/libs -I../../../rtl/tex_unit
SRCS = main.cpp
all: build
CF += -std=c++11 -fms-extensions -I../..
VF += $(PARAMS)
VF += --language 1800-2009 --assert -Wall --trace
VF += -Wno-DECLFILENAME
VF += --x-initial unique
VF += -exe $(SRCS) $(INCLUDE)
VF += $(PARAMS)
gen:
verilator $(VF) -cc $(TOP).v -CFLAGS '$(CF)' --exe $(SRCS)
build: gen
(cd obj_dir && make -j -f V$(TOP).mk)
run: build
(cd obj_dir && ./V$(TOP))
clean:
rm -rf obj_dir

View File

@@ -1,215 +0,0 @@
#include "vl_simulator.h"
#include "VVX_tex_sampler.h"
#include <iostream>
#include <map>
#define MAX_TICKS 20
#define MAX_UNIT_CYCLES 5
#define NUM_THREADS
#define CHECK(x) \
do { \
if (x) \
break; \
std::cout << "FAILED: " << #x << std::endl; \
std::abort(); \
} while (false)
uint64_t ticks = 0;
// using Device = VVX_tex_sampler;
template <typename T>
class testbench
{
private:
vl_simulator<T> sim;
std::map<int, struct Input> input_map;
std::map<int, struct Output> output_map;
public:
struct UnitTest {
bool use_reset;
unsigned int num_cycles;
bool use_cmodel;
struct Output outputs[MAX_UNIT_CYCLES];
struct Input inputs[MAX_UNIT_CYCLES];
unsigned int num_output_check;
unsigned int check_output_cycle[MAX_UNIT_CYCLES];
}
struct Input {
bool req_valid;
unsigned int req_wid;
unsigned int req_tmask;
unsigned int req_PC;
unsigned int req_rd;
unsigned int req_wb;
unsigned int req_filter;
unsigned int req_format;
unsigned int req_u[NUM_THREADS];
unsigned int req_v[NUM_THREADS];
unsigned int req_texels[NUM_THREADS][4];
bool rsp_ready;
}
struct Output {
int output_cycle;
// outputs
bool req_ready;
bool rsp_valid;
unsigned int rsp_wid;
unsigned int rsp_tmask;
unsigned int rsp_PC;
unsigned int rsp_rd;
bool rsp_wb;
unsigned int rsp_data[NUM_THREADS];
}
testbench(/* args */){
}
~testbench(){
}
void unittest_Cmodel(struct UnitTest * test){
int cycles = test->num_cycles;
int num_outputs = test->num_output_check;
// struct Input* inputs = new (struct Input)[cycles];
struct Output* outputs = new (struct Output)[num_outputs];
// implement c model and assign outputs to struct
if (test->inputs[0]->req_filter == 0){
for (int i = 0; i < NUM_THREADS; i++)
outputs[0]->rsp_data[0] = test->inputs->req_texels[i][0];
} else {
// for (int i = 0; i < NUM_THREADS; i++){
// uint32_t low[4], high[4];
// for (int j = 0; j < 4; j++){
// low[j] = test->inputs->req_texels[i][j] & 0x00ff00ff;
// high[j] = (test->inputs->req_texels[i][j] >> 8) & 0x00ff00ff;
// }
// }
}
outputs[0]->output_cycle = 1;
test->num_cycles = 1;
test->outputs = &outputs;
}
void generate_test_vectors(struct UnitTest * tests, int num_tests, bool is_pipe){
// for all unit tests create output test vectors (w w/o c-model)
int prev_test_cycle = 0;
for (int i = 0; i < num_tests; i++)
{
int op_counter = 0;
int ip_counter = 0;
int test_cycle = 0;
int last_ip_cycle = 0;
struct UnitTest curr_test = tests[i];
if (curr_test->use_cmodel){
unittest_Cmodel(&curr_test);
}
for (int j = 0; j < curr_test->num_cycles; j++)
{
if (curr_test->inputs[ip_counter]->input_cycle == test_cycle){
input_map.insert(std::make_pair(prev_test_cycle + test_cycle, curr_test->inputs[j]));
last_ip_cycle = prev_test_cycle + test_cycle;
ip_counter++;
}
if (curr_test->outputs[op_counter]->output_cycle == test_cycle){
output_map.insert(std::make_pair(prev_test_cycle + test_cycle, curr_test->outputs[op_counter]));
op_counter++;
}
test_cycle++;
}
if(!is_pipe){
prev_test_cycle += (test_cycle - 1);
}
else{
prev_test_cycle = last_ip_cycle + 1;
}
}
}
void run(){
ticks = sim.reset(0);
int cycle = 0;
while (ticks < MAX_TICKS) {
auto input = input_map.find(cycle);
auto output = output_map.find(cycle);
if (input != input_map.end()){
sim->req_valid = input->req_valid;
sim->req_wid = input->req_wid;
sim->req_tmask = input->req_tmask;
sim->req_PC = input->req_PC;
sim->req_rd = input->req_rd;
sim->req_wb = input->req_wb;
sim->req_filter = input->req_filter;
sim->req_format = input->req_format;
// sim->req_u = input->req_u[NUM_THREADS];
// sim->req_v = input->req_v[NUM_THREADS];
vl_setw(sim->req_texels, input->req_texels)
// sim->req_texels = input->req_texels[NUM_THREADS][4];
sim->rsp_ready = input->rsp_ready;
} else{
std::cout << "Warning! No Input on Cycle " << cycle << std::endl;
}
if(output != output_map.end()){
CHECK(sim->req_ready == output->req_ready);
CHECK(sim->rsp_valid == output->rsp_valid);
CHECK(sim->rsp_wid == output->rsp_wid);
CHECK(sim->rsp_tmask == output->rsp_tmask);
CHECK(sim->rsp_PC == output->rsp_PC);
CHECK(sim->rsp_rd == output->rsp_rd);
CHECK(sim->rsp_wb == output->rsp_wb);
CHECK(vl_cmpw(sim->rsp_data, output->rsp_data));
}
cycle++;
ticks = sim.step(ticks,2);
}
}
std::cout << "PASSED!" << std::endl;
std::cout << "Simulation time: " << std::dec << ticks/2 << " cycles" << std::endl;
};
double sc_time_stamp() {
return ticks;
}
int main(int argc, char **argv) {
// Initialize Verilators variables
Verilated::commandArgs(argc, argv);
testbench<VVX_tex_sampler> sampler_testbench;
sampler_testbench.generate_test_vectors(tests, 1, 0);
sampler_test_bench.run();
return 0;
}