riscv
/
OpenFPGA
mirror of https://github.com/lnis-uofu/OpenFPGA.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Merge branch 'master' of https://github.com/lnis-uofu/OpenFPGA into vtr_upgrade

This commit is contained in:
tangxifan 2022-09-16 16:47:21 -07:00
parent e98d022d3a 30988d7072
commit 373566416c
30 changed files with 1780 additions and 305 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
VERSION.md
View File

@ -1 +1 @@
1.1.489
1.1.525

8
docs/source/manual/openfpga_shell/openfpga_commands/fpga_bitstream_commands.rst
View File

@ -25,6 +25,14 @@ Repack's functionality are in the following aspects:
.. warning:: Design constraints are designed to help repacker to identify which clock net to be mapped to which pin, so that multi-clock benchmarks can be correctly implemented, in the case that VPR may not have sufficient vision on clock net mapping. **Try not to use design constraints to remap any other types of nets!!!**
.. option:: --ignore_global_nets_on_pins
Specify the mapping results of global nets should be ignored on which pins of a ``pb_type``. For example, ``--ignore_global_nets_on_pins clb.I[0:11]``. Once specified, the mapping results on the pins for all the global nets, such as clock, reset *etc.*, are ignored. Routing traces will be appeneded to other pins where the same global nets are mapped to.
.. note:: This option is designed for global nets which are applied to both data path and global networks. For example, a reset signal is mapped to both a LUT input and the reset pin of a FF. Suggest not to use the option in other purposes!
.. warning:: Users must specify the size/width of the pin. Currently, OpenFPGA cannot infer the pin size from the architecture!!!
.. option:: --verbose
Show verbose log

6
openfpga/src/base/openfpga_bitstream_command.cpp
View File

@ -21,9 +21,15 @@ ShellCommandId add_openfpga_repack_command(openfpga::Shell<OpenfpgaContext>& she
const ShellCommandClassId& cmd_class_id,
const std::vector<ShellCommandId>& dependent_cmds) {
Command shell_cmd("repack");
/* Add an option '--design_constraints' */
CommandOptionId opt_design_constraints = shell_cmd.add_option("design_constraints", false, "file path to the design constraints");
shell_cmd.set_option_require_value(opt_design_constraints, openfpga::OPT_STRING);
/* Add an option '--ignore_global_nets_on_pins' */
CommandOptionId opt_ignore_global_nets = shell_cmd.add_option("ignore_global_nets_on_pins", false, "Specify the pins where global nets will be ignored. Routing traces are merged to other pins");
shell_cmd.set_option_require_value(opt_ignore_global_nets, openfpga::OPT_STRING);
/* Add an option '--verbose' */
shell_cmd.add_option("verbose", false, "Enable verbose output");

17
openfpga/src/base/openfpga_repack.cpp
View File

@ -30,6 +30,7 @@ int repack(OpenfpgaContext& openfpga_ctx,
const Command& cmd, const CommandContext& cmd_context) {
CommandOptionId opt_design_constraints = cmd.option("design_constraints");
CommandOptionId opt_ignore_global_nets = cmd.option("ignore_global_nets_on_pins");
CommandOptionId opt_verbose = cmd.option("verbose");
/* Load design constraints from file */
@ -40,22 +41,32 @@ int repack(OpenfpgaContext& openfpga_ctx,
repack_design_constraints = read_xml_repack_design_constraints(dc_fname.c_str());
}
/* Setup repacker options */
RepackOption options;
options.set_design_constraints(repack_design_constraints);
options.set_ignore_global_nets_on_pins(cmd_context.option_value(cmd, opt_ignore_global_nets));
options.set_verbose_output(cmd_context.option_enable(cmd, opt_verbose));
if (!options.valid()) {
VTR_LOG("Detected errors when parsing options!\n");
return CMD_EXEC_FATAL_ERROR;
}
pack_physical_pbs(g_vpr_ctx.device(),
g_vpr_ctx.atom(),
g_vpr_ctx.clustering(),
openfpga_ctx.mutable_vpr_device_annotation(),
openfpga_ctx.mutable_vpr_clustering_annotation(),
openfpga_ctx.vpr_bitstream_annotation(),
repack_design_constraints,
openfpga_ctx.arch().circuit_lib,
cmd_context.option_enable(cmd, opt_verbose));
options);
build_physical_lut_truth_tables(openfpga_ctx.mutable_vpr_clustering_annotation(),
g_vpr_ctx.atom(),
g_vpr_ctx.clustering(),
openfpga_ctx.vpr_device_annotation(),
openfpga_ctx.arch().circuit_lib,
cmd_context.option_enable(cmd, opt_verbose));
options.verbose_output());
/* TODO: should identify the error code from internal function execution */
return CMD_EXEC_SUCCESS;

118
openfpga/src/fabric/build_fabric_io_location_map.cpp
View File

@ -28,6 +28,8 @@ namespace openfpga {
/********************************************************************
* Find all the GPIO ports in the grid module
* and cache their port/pin index in the top-level module
*
* .. note:: The I/O sequence(indexing) is already determined in the io_children() list of top-level module. Here we just build a fast lookup from (x, y, z) coordinate to the actual indices
*******************************************************************/
IoLocationMap build_fabric_io_location_map(const ModuleManager& module_manager,
const DeviceGrid& grids) {
@ -37,59 +39,54 @@ IoLocationMap build_fabric_io_location_map(const ModuleManager& module_manager,
std::map<std::string, size_t> io_counter;
/* Create the coordinate range for each side of FPGA fabric */
std::map<e_side, std::vector<vtr::Point<size_t>>> io_coordinates = generate_perimeter_grid_coordinates( grids);
std::string top_module_name = generate_fpga_top_module_name();
ModuleId top_module = module_manager.find_module(top_module_name);
VTR_ASSERT(true == module_manager.valid_module_id(top_module));
/* Walk through the I/O child list */
for (size_t ichild = 0; ichild < module_manager.io_children(top_module).size(); ++ichild) {
ModuleId child = module_manager.io_children(top_module)[ichild];
vtr::Point<int> coord = module_manager.io_child_coordinates(top_module)[ichild];
/* Walk through all the grids on the perimeter, which are I/O grids */
for (const e_side& io_side : FPGA_SIDES_CLOCKWISE) {
for (const vtr::Point<size_t>& io_coordinate : io_coordinates[io_side]) {
/* Bypass EMPTY grid */
if (true == is_empty_type(grids[io_coordinate.x()][io_coordinate.y()].type)) {
if (true == is_empty_type(grids[coord.x()][coord.y()].type)) {
continue;
}
/* Skip width or height > 1 tiles (mostly heterogeneous blocks) */
if ( (0 < grids[io_coordinate.x()][io_coordinate.y()].width_offset)
|| (0 < grids[io_coordinate.x()][io_coordinate.y()].height_offset)) {
if ( (0 < grids[coord.x()][coord.y()].width_offset)
|| (0 < grids[coord.x()][coord.y()].height_offset)) {
continue;
}
t_physical_tile_type_ptr grid_type = grids[io_coordinate.x()][io_coordinate.y()].type;
/* Find the module name for this type of grid */
std::string grid_module_name_prefix(GRID_MODULE_NAME_PREFIX);
std::string grid_module_name = generate_grid_block_module_name(grid_module_name_prefix, std::string(grid_type->name), is_io_type(grid_type), io_side);
ModuleId grid_module = module_manager.find_module(grid_module_name);
VTR_ASSERT(true == module_manager.valid_module_id(grid_module));
VTR_ASSERT_SAFE(true == module_manager.valid_module_id(child));
/* Find all the GPIO ports in the grid module */
/* MUST DO: register in io location mapping!
* I/O location mapping is a critical look-up for testbench generators
* As we add the I/O grid instances to top module by following order:
* TOP -> RIGHT -> BOTTOM -> LEFT
* The I/O index will increase in this way as well.
* This organization I/O indices is also consistent to the way
* that GPIOs are wired in function connect_gpio_module()
*
* Note: if you change the GPIO function, you should update here as well!
*/
for (int z = 0; z < grids[io_coordinate.x()][io_coordinate.y()].type->capacity; ++z) {
if (size_t(grids[coord.x()][coord.y()].type->capacity) != module_manager.io_children(child).size()) {
VTR_LOG("%s[%ld][%ld] capacity: %d while io_child number is %d", grids[coord.x()][coord.y()].type->name, coord.x(), coord.y(), grids[coord.x()][coord.y()].type->capacity, module_manager.io_children(child).size());
}
VTR_ASSERT(size_t(grids[coord.x()][coord.y()].type->capacity) == module_manager.io_children(child).size());
for (size_t isubchild = 0; isubchild < module_manager.io_children(child).size(); ++isubchild) {
vtr::Point<int> subchild_coord = module_manager.io_child_coordinates(child)[isubchild];
for (const ModuleManager::e_module_port_type& module_io_port_type : MODULE_IO_PORT_TYPES) {
for (const ModulePortId& gpio_port_id : module_manager.module_port_ids_by_type(grid_module, module_io_port_type)) {
for (const ModulePortId& gpio_port_id : module_manager.module_port_ids_by_type(child, module_io_port_type)) {
/* Only care mappable I/O */
if (false == module_manager.port_is_mappable_io(grid_module, gpio_port_id)) {
if (false == module_manager.port_is_mappable_io(child, gpio_port_id)) {
continue;
}
const BasicPort& gpio_port = module_manager.module_port(grid_module, gpio_port_id);
const BasicPort& gpio_port = module_manager.module_port(child, gpio_port_id);
auto curr_io_index = io_counter.find(gpio_port.get_name());
/* Index always start from zero */
if (curr_io_index == io_counter.end()) {
io_counter[gpio_port.get_name()] = 0;
}
io_location_map.set_io_index(io_coordinate.x(), io_coordinate.y(), z,
io_location_map.set_io_index(coord.x(), coord.y(), subchild_coord.x(),
gpio_port.get_name(),
io_counter[gpio_port.get_name()]);
io_counter[gpio_port.get_name()]++;
@ -97,73 +94,8 @@ IoLocationMap build_fabric_io_location_map(const ModuleManager& module_manager,
}
}
}
}
/* Walk through all the center grids, which may include I/O grids */
for (size_t ix = 1; ix < grids.width() - 1; ++ix) {
for (size_t iy = 1; iy < grids.height() - 1; ++iy) {
/* Bypass EMPTY grid */
if (true == is_empty_type(grids[ix][iy].type)) {
continue;
}
/* Skip width or height > 1 tiles (mostly heterogeneous blocks) */
if ( (0 < grids[ix][iy].width_offset)
|| (0 < grids[ix][iy].height_offset)) {
continue;
}
t_physical_tile_type_ptr grid_type = grids[ix][iy].type;
/* Find the module name for this type of grid */
std::string grid_module_name_prefix(GRID_MODULE_NAME_PREFIX);
std::string grid_module_name = generate_grid_block_module_name(grid_module_name_prefix, std::string(grid_type->name), is_io_type(grid_type), NUM_SIDES);
ModuleId grid_module = module_manager.find_module(grid_module_name);
VTR_ASSERT(true == module_manager.valid_module_id(grid_module));
/* Find all the GPIO ports in the grid module */
/* MUST DO: register in io location mapping!
* I/O location mapping is a critical look-up for testbench generators
* As we add the I/O grid instances to top module by following order:
* TOP -> RIGHT -> BOTTOM -> LEFT
* The I/O index will increase in this way as well.
* This organization I/O indices is also consistent to the way
* that GPIOs are wired in function connect_gpio_module()
*
* Note: if you change the GPIO function, you should update here as well!
* FIXME: The codes should be adapt to support sub tiles!!!
*/
for (int z = 0; z < grids[ix][iy].type->capacity; ++z) {
for (const ModuleManager::e_module_port_type& module_io_port_type : MODULE_IO_PORT_TYPES) {
for (const ModulePortId& gpio_port_id : module_manager.module_port_ids_by_type(grid_module, module_io_port_type)) {
/* Only care mappable I/O */
if (false == module_manager.port_is_mappable_io(grid_module, gpio_port_id)) {
continue;
}
const BasicPort& gpio_port = module_manager.module_port(grid_module, gpio_port_id);
auto curr_io_index = io_counter.find(gpio_port.get_name());
/* Index always start from zero */
if (curr_io_index == io_counter.end()) {
io_counter[gpio_port.get_name()] = 0;
}
io_location_map.set_io_index(ix, iy, z,
gpio_port.get_name(),
io_counter[gpio_port.get_name()]);
io_counter[gpio_port.get_name()]++;
}
}
}
}
}
/* Check all the GPIO ports in the top-level module has been mapped */
std::string top_module_name = generate_fpga_top_module_name();
ModuleId top_module = module_manager.find_module(top_module_name);
VTR_ASSERT(true == module_manager.valid_module_id(top_module));
for (const ModuleManager::e_module_port_type& module_io_port_type : MODULE_IO_PORT_TYPES) {
for (const ModulePortId& gpio_port_id : module_manager.module_port_ids_by_type(top_module, module_io_port_type)) {
/* Only care mappable I/O */

4
openfpga/src/fabric/build_grid_modules.cpp
View File

@ -1026,7 +1026,9 @@ void build_physical_tile_module(ModuleManager& module_manager,
/* Add all the sub modules */
size_t pb_instance_id = module_manager.num_instance(grid_module, pb_module);
module_manager.add_child_module(grid_module, pb_module);
module_manager.add_child_module(grid_module, pb_module, false);
/* Add a custom I/O child with coordinate 'z' */
module_manager.add_io_child(grid_module, pb_module, pb_instance_id, vtr::Point<int>(iz, 0));
/* Give the child module with a unique instance name */
std::string instance_name = generate_physical_block_instance_name(lb_type->pb_graph_head->pb_type, iz);

182
openfpga/src/fabric/build_top_module.cpp
View File

@ -50,7 +50,7 @@ size_t add_top_module_grid_instance(ModuleManager& module_manager,
/* Record the instance id */
size_t grid_instance = module_manager.num_instance(top_module, grid_module);
/* Add the module to top_module */
module_manager.add_child_module(top_module, grid_module);
module_manager.add_child_module(top_module, grid_module, false);
/* Set an unique name to the instance
* Note: it is your risk to gurantee the name is unique!
*/
@ -102,33 +102,6 @@ vtr::Matrix<size_t> add_top_module_grid_instances(ModuleManager& module_manager,
vtr::Matrix<size_t> grid_instance_ids({grids.width(), grids.height()});
grid_instance_ids.fill(size_t(-1));
/* Instanciate core grids */
for (size_t ix = 1; ix < grids.width() - 1; ++ix) {
for (size_t iy = 1; iy < grids.height() - 1; ++iy) {
/* Bypass EMPTY grid */
if (true == is_empty_type(grids[ix][iy].type)) {
continue;
}
/* Skip width or height > 1 tiles (mostly heterogeneous blocks) */
if ( (0 < grids[ix][iy].width_offset)
|| (0 < grids[ix][iy].height_offset)) {
/* Find the root of this grid, the instance id should be valid.
* We just copy it here
*/
vtr::Point<size_t> root_grid_coord(ix - grids[ix][iy].width_offset,
iy - grids[ix][iy].height_offset);
VTR_ASSERT(size_t(-1) != grid_instance_ids[root_grid_coord.x()][root_grid_coord.y()]);
grid_instance_ids[ix][iy] = grid_instance_ids[root_grid_coord.x()][root_grid_coord.y()];
continue;
}
/* Add a grid module to top_module*/
vtr::Point<size_t> grid_coord(ix, iy);
grid_instance_ids[ix][iy] = add_top_module_grid_instance(module_manager, top_module,
grids[ix][iy].type,
NUM_SIDES, grid_coord);
}
}
/* Instanciate I/O grids */
/* Create the coordinate range for each side of FPGA fabric */
std::map<e_side, std::vector<vtr::Point<size_t>>> io_coordinates = generate_perimeter_grid_coordinates( grids);
@ -157,6 +130,37 @@ vtr::Matrix<size_t> add_top_module_grid_instances(ModuleManager& module_manager,
}
}
/* Instanciate core grids
* IMPORTANT: sequence matters here, it impacts the I/O indexing.
* We should follow the same sequence as the build_io_location_map()!
* If you change the sequence of walking through grids here, you should change it in the build_io_location map()!
*/
for (size_t ix = 1; ix < grids.width() - 1; ++ix) {
for (size_t iy = 1; iy < grids.height() - 1; ++iy) {
/* Bypass EMPTY grid */
if (true == is_empty_type(grids[ix][iy].type)) {
continue;
}
/* Skip width or height > 1 tiles (mostly heterogeneous blocks) */
if ( (0 < grids[ix][iy].width_offset)
|| (0 < grids[ix][iy].height_offset)) {
/* Find the root of this grid, the instance id should be valid.
* We just copy it here
*/
vtr::Point<size_t> root_grid_coord(ix - grids[ix][iy].width_offset,
iy - grids[ix][iy].height_offset);
VTR_ASSERT(size_t(-1) != grid_instance_ids[root_grid_coord.x()][root_grid_coord.y()]);
grid_instance_ids[ix][iy] = grid_instance_ids[root_grid_coord.x()][root_grid_coord.y()];
continue;
}
/* Add a grid module to top_module*/
vtr::Point<size_t> grid_coord(ix, iy);
grid_instance_ids[ix][iy] = add_top_module_grid_instance(module_manager, top_module,
grids[ix][iy].type,
NUM_SIDES, grid_coord);
}
}
return grid_instance_ids;
}
@ -201,7 +205,7 @@ vtr::Matrix<size_t> add_top_module_switch_block_instances(ModuleManager& module_
/* Record the instance id */
sb_instance_ids[rr_gsb.get_sb_x()][rr_gsb.get_sb_y()] = module_manager.num_instance(top_module, sb_module);
/* Add the module to top_module */
module_manager.add_child_module(top_module, sb_module);
module_manager.add_child_module(top_module, sb_module, false);
/* Set an unique name to the instance
* Note: it is your risk to gurantee the name is unique!
*/
@ -257,7 +261,7 @@ vtr::Matrix<size_t> add_top_module_connection_block_instances(ModuleManager& mod
/* Record the instance id */
cb_instance_ids[rr_gsb.get_cb_x(cb_type)][rr_gsb.get_cb_y(cb_type)] = module_manager.num_instance(top_module, cb_module);
/* Add the module to top_module */
module_manager.add_child_module(top_module, cb_module);
module_manager.add_child_module(top_module, cb_module, false);
/* Set an unique name to the instance
* Note: it is your risk to gurantee the name is unique!
*/
@ -271,6 +275,118 @@ vtr::Matrix<size_t> add_top_module_connection_block_instances(ModuleManager& mod
return cb_instance_ids;
}
/********************************************************************
* Add the I/O children to the top-level module, which impacts the I/O indexing
* This is the default function to build the I/O sequence/indexing
* The I/O children is added in a maze shape
* The function supports I/Os in the center of grids, starting from the bottom-left corner and ending at the center
*
* +----------------------+
* |+--------------------+|
* ||+------------------+||
* |||+----------------+|||
* ||||+-------------->||||
* ||||+---------------+|||
* |||+-----------------+||
* ||+-------------------+|
* |+---------------------+
* ^
* io[0]
*******************************************************************/
static
void add_top_module_io_children(ModuleManager& module_manager,
const ModuleId& top_module,
const DeviceGrid& grids,
const vtr::Matrix<size_t>& grid_instance_ids) {
/* Create the coordinate range for the perimeter I/Os of FPGA fabric */
std::map<e_side, std::vector<vtr::Point<size_t>>> io_coordinates = generate_perimeter_grid_coordinates( grids);
for (const e_side& io_side : FPGA_SIDES_CLOCKWISE) {
for (const vtr::Point<size_t>& io_coord : io_coordinates[io_side]) {
/* Bypass EMPTY grid */
if (true == is_empty_type(grids[io_coord.x()][io_coord.y()].type)) {
continue;
}
/* Skip width, height > 1 tiles (mostly heterogeneous blocks) */
if ( (0 < grids[io_coord.x()][io_coord.y()].width_offset)
|| (0 < grids[io_coord.x()][io_coord.y()].height_offset)) {
continue;
}
/* Find the module name for this type of grid */
t_physical_tile_type_ptr grid_type = grids[io_coord.x()][io_coord.y()].type;
std::string grid_module_name_prefix(GRID_MODULE_NAME_PREFIX);
std::string grid_module_name = generate_grid_block_module_name(grid_module_name_prefix, std::string(grid_type->name), is_io_type(grid_type), io_side);
ModuleId grid_module = module_manager.find_module(grid_module_name);
VTR_ASSERT(true == module_manager.valid_module_id(grid_module));
/* Add a I/O children to top_module*/
module_manager.add_io_child(top_module, grid_module, grid_instance_ids[io_coord.x()][io_coord.y()], vtr::Point<int>(io_coord.x(), io_coord.y()));
}
}
/* Walk through the center grids */
size_t xmin = 1;
size_t xmax = grids.width() - 2;
size_t ymin = 1;
size_t ymax = grids.height() - 2;
std::vector<vtr::Point<size_t>> coords;
while (xmin < xmax && ymin < ymax) {
for (size_t iy = ymin; iy < ymax + 1; iy++) {
coords.push_back(vtr::Point<size_t>(xmin, iy));
}
for (size_t ix = xmin + 1; ix < xmax + 1; ix++) {
coords.push_back(vtr::Point<size_t>(ix, ymax));
}
for (size_t iy = ymax - 1; iy > ymin; iy--) {
coords.push_back(vtr::Point<size_t>(xmax, iy));
}
for (size_t ix = xmax; ix > xmin; ix--) {
coords.push_back(vtr::Point<size_t>(ix, ymin));
}
xmin++;
ymin++;
xmax--;
ymax--;
}
/* If height is odd, add the missing horizental line */
if ((grids.height() - 2) % 2 == 1) {
if (ymin == ymax) {
for (size_t ix = xmin; ix < xmax + 1; ix++) {
coords.push_back(vtr::Point<size_t>(ix, ymin));
}
}
}
/* If width is odd, add the missing vertical line */
if ((grids.width() - 2) % 2 == 1) {
if (xmin == xmax) {
for (size_t iy = ymin; iy < ymax + 1; iy++) {
coords.push_back(vtr::Point<size_t>(xmin, iy));
}
}
}
/* Now walk through the coordinates */
for (vtr::Point<size_t> coord : coords) {
/* Bypass EMPTY grid */
if (true == is_empty_type(grids[coord.x()][coord.y()].type)) {
continue;
}
/* Skip width or height > 1 tiles (mostly heterogeneous blocks) */
if ( (0 < grids[coord.x()][coord.y()].width_offset)
|| (0 < grids[coord.x()][coord.y()].height_offset)) {
continue;
}
/* Find the module name for this type of grid */
t_physical_tile_type_ptr grid_type = grids[coord.x()][coord.y()].type;
std::string grid_module_name_prefix(GRID_MODULE_NAME_PREFIX);
std::string grid_module_name = generate_grid_block_module_name(grid_module_name_prefix, std::string(grid_type->name), is_io_type(grid_type), NUM_SIDES);
ModuleId grid_module = module_manager.find_module(grid_module_name);
VTR_ASSERT(true == module_manager.valid_module_id(grid_module));
/* Add a I/O children to top_module*/
module_manager.add_io_child(top_module, grid_module, grid_instance_ids[coord.x()][coord.y()], vtr::Point<int>(coord.x(), coord.y()));
}
}
/********************************************************************
* Print the top-level module for the FPGA fabric in Verilog format
* This function will
@ -323,6 +439,9 @@ int build_top_module(ModuleManager& module_manager,
cb_instance_ids[CHANX] = add_top_module_connection_block_instances(module_manager, top_module, device_rr_gsb, CHANX, compact_routing_hierarchy);
cb_instance_ids[CHANY] = add_top_module_connection_block_instances(module_manager, top_module, device_rr_gsb, CHANY, compact_routing_hierarchy);
/* Update I/O children list */
add_top_module_io_children(module_manager, top_module, grids, grid_instance_ids);
/* Add nets when we need a complete fabric modeling,
* which is required by downstream functions
*/
@ -350,8 +469,7 @@ int build_top_module(ModuleManager& module_manager,
}
/* Add GPIO ports from the sub-modules under this Verilog module
* This is a much easier job after adding sub modules (instances),
* we just need to find all the I/O ports from the child modules and build a list of it
* For top-level module, we follow a special sequencing for I/O modules. So we rebuild the I/O children list here
*/
add_module_gpio_ports_from_child_modules(module_manager, top_module);

2
openfpga/src/fabric/build_top_module_directs.cpp
View File

@ -131,7 +131,7 @@ void add_module_nets_tile_direct_connection(ModuleManager& module_manager,
/* Add a submodule of direct connection module to the top-level module */
size_t direct_instance_id = module_manager.num_instance(top_module, direct_module);
module_manager.add_child_module(top_module, direct_module);
module_manager.add_child_module(top_module, direct_module, false);
/* Create the 1st module net */
ModuleNetId net_direct_src = module_manager.create_module_net(top_module);

6
openfpga/src/fabric/build_top_module_memory.cpp
View File

@ -1055,7 +1055,7 @@ void add_top_module_nets_cmos_memory_bank_config_bus(ModuleManager& module_manag
}
VTR_ASSERT(ModuleId::INVALID() != bl_decoder_module);
size_t curr_bl_decoder_instance_id = module_manager.num_instance(top_module, bl_decoder_module);
module_manager.add_child_module(top_module, bl_decoder_module);
module_manager.add_child_module(top_module, bl_decoder_module, false);
/**************************************************************
* Add the WL decoder module
@ -1083,7 +1083,7 @@ void add_top_module_nets_cmos_memory_bank_config_bus(ModuleManager& module_manag
}
VTR_ASSERT(ModuleId::INVALID() != wl_decoder_module);
size_t curr_wl_decoder_instance_id = module_manager.num_instance(top_module, wl_decoder_module);
module_manager.add_child_module(top_module, wl_decoder_module);
module_manager.add_child_module(top_module, wl_decoder_module, false);
/**************************************************************
* Add module nets from the top module to BL decoder's inputs
@ -1531,7 +1531,7 @@ void add_top_module_nets_cmos_memory_frame_decoder_config_bus(ModuleManager& mod
/* Instanciate the decoder module here */
size_t decoder_instance = module_manager.num_instance(parent_module, decoder_module);
module_manager.add_child_module(parent_module, decoder_module);
module_manager.add_child_module(parent_module, decoder_module, false);
/* Connect the enable (EN) port of memory modules under the parent module
* to the frame decoder inputs

8
openfpga/src/fabric/build_top_module_memory_bank.cpp
View File

@ -507,7 +507,7 @@ void add_top_module_nets_cmos_ql_memory_bank_bl_decoder_config_bus(ModuleManager
}
VTR_ASSERT(ModuleId::INVALID() != bl_decoder_module);
size_t curr_bl_decoder_instance_id = module_manager.num_instance(top_module, bl_decoder_module);
module_manager.add_child_module(top_module, bl_decoder_module);
module_manager.add_child_module(top_module, bl_decoder_module, false);
/**************************************************************
* Add module nets from the top module to BL decoder's inputs
@ -705,7 +705,7 @@ void add_top_module_nets_cmos_ql_memory_bank_wl_decoder_config_bus(ModuleManager
}
VTR_ASSERT(ModuleId::INVALID() != wl_decoder_module);
size_t curr_wl_decoder_instance_id = module_manager.num_instance(top_module, wl_decoder_module);
module_manager.add_child_module(top_module, wl_decoder_module);
module_manager.add_child_module(top_module, wl_decoder_module, false);
/**************************************************************
* Add module nets from the top module to WL decoder's inputs
@ -1471,7 +1471,7 @@ void add_top_module_nets_cmos_ql_memory_bank_bl_shift_register_config_bus(Module
VTR_ASSERT(sr_bank_module);
size_t cur_inst = module_manager.num_instance(top_module, sr_bank_module);
module_manager.add_child_module(top_module, sr_bank_module);
module_manager.add_child_module(top_module, sr_bank_module, false);
sr_banks.link_bl_shift_register_bank_to_module(config_region, sr_bank, sr_bank_module);
sr_banks.link_bl_shift_register_bank_to_instance(config_region, sr_bank, cur_inst);
@ -1565,7 +1565,7 @@ void add_top_module_nets_cmos_ql_memory_bank_wl_shift_register_config_bus(Module
VTR_ASSERT(sr_bank_module);
size_t cur_inst = module_manager.num_instance(top_module, sr_bank_module);
module_manager.add_child_module(top_module, sr_bank_module);
module_manager.add_child_module(top_module, sr_bank_module, false);
sr_banks.link_wl_shift_register_bank_to_module(config_region, sr_bank, sr_bank_module);
sr_banks.link_wl_shift_register_bank_to_instance(config_region, sr_bank, cur_inst);

75
openfpga/src/fabric/module_manager.cpp
View File

@ -91,6 +91,29 @@ std::vector<vtr::Point<int>> ModuleManager::configurable_child_coordinates(const
return configurable_child_coordinates_[parent_module];
}
/* Find all the configurable child modules under a parent module */
std::vector<ModuleId> ModuleManager::io_children(const ModuleId& parent_module) const {
/* Validate the module_id */
VTR_ASSERT(valid_module_id(parent_module));
return io_children_[parent_module];
}
/* Find all the instances of configurable child modules under a parent module */
std::vector<size_t> ModuleManager::io_child_instances(const ModuleId& parent_module) const {
/* Validate the module_id */
VTR_ASSERT(valid_module_id(parent_module));
return io_child_instances_[parent_module];
}
std::vector<vtr::Point<int>> ModuleManager::io_child_coordinates(const ModuleId& parent_module) const {
/* Validate the module_id */
VTR_ASSERT(valid_module_id(parent_module));
return io_child_coordinates_[parent_module];
}
/* Find the source ids of modules */
ModuleManager::module_net_src_range ModuleManager::module_net_sources(const ModuleId& module, const ModuleNetId& net) const {
/* Validate the module_id */
@ -562,6 +585,10 @@ ModuleId ModuleManager::add_module(const std::string& name) {
config_region_ids_.emplace_back();
config_region_children_.emplace_back();
io_children_.emplace_back();
io_child_instances_.emplace_back();
io_child_coordinates_.emplace_back();
port_ids_.emplace_back();
ports_.emplace_back();
port_types_.emplace_back();
@ -680,7 +707,7 @@ void ModuleManager::set_port_preproc_flag(const ModuleId& module, const ModulePo
}
/* Add a child module to a parent module */
void ModuleManager::add_child_module(const ModuleId& parent_module, const ModuleId& child_module) {
void ModuleManager::add_child_module(const ModuleId& parent_module, const ModuleId& child_module, const bool& is_io_child) {
/* Validate the id of both parent and child modules */
VTR_ASSERT ( valid_module_id(parent_module) );
VTR_ASSERT ( valid_module_id(child_module) );
@ -693,19 +720,27 @@ void ModuleManager::add_child_module(const ModuleId& parent_module, const Module
}
std::vector<ModuleId>::iterator child_it = std::find(children_[parent_module].begin(), children_[parent_module].end(), child_module);
int child_instance_id = -1;
if (child_it == children_[parent_module].end()) {
/* Update the child module of parent module */
children_[parent_module].push_back(child_module);
num_child_instances_[parent_module].push_back(1); /* By default give one */
child_instance_id = 0;
/* Update the instance name list */
child_instance_names_[parent_module].emplace_back();
child_instance_names_[parent_module].back().emplace_back();
} else {
/* Increase the counter of instances */
child_instance_id = num_child_instances_[parent_module][child_it - children_[parent_module].begin()];
num_child_instances_[parent_module][child_it - children_[parent_module].begin()]++;
child_instance_names_[parent_module][child_it - children_[parent_module].begin()].emplace_back();
}
/* Add to I/O child if needed */
if (is_io_child) {
add_io_child(parent_module, child_module, child_instance_id);
}
/* Update fast look-up for nets */
size_t instance_id = net_lookup_[parent_module][child_module].size();
net_lookup_[parent_module][child_module].emplace_back();
@ -815,6 +850,36 @@ void ModuleManager::add_configurable_child_to_region(const ModuleId& parent_modu
config_region_children_[parent_module][config_region].push_back(config_child_id);
}
void ModuleManager::add_io_child(const ModuleId& parent_module,
const ModuleId& child_module,
const size_t& child_instance,
const vtr::Point<int> coord) {
/* Validate the id of both parent and child modules */
VTR_ASSERT ( valid_module_id(parent_module) );
VTR_ASSERT ( valid_module_id(child_module) );
/* Ensure that the instance id is in range */
VTR_ASSERT ( child_instance < num_instance(parent_module, child_module));
io_children_[parent_module].push_back(child_module);
io_child_instances_[parent_module].push_back(child_instance);
io_child_coordinates_[parent_module].push_back(coord);
}
void ModuleManager::reserve_io_child(const ModuleId& parent_module,
const size_t& num_children) {
VTR_ASSERT ( valid_module_id(parent_module) );
/* Do reserve when the number of children is larger than current size of lists */
if (num_children > io_children_[parent_module].size()) {
io_children_[parent_module].reserve(num_children);
}
if (num_children > io_child_instances_[parent_module].size()) {
io_child_instances_[parent_module].reserve(num_children);
}
if (num_children > io_child_coordinates_[parent_module].size()) {
io_child_coordinates_[parent_module].reserve(num_children);
}
}
void ModuleManager::reserve_module_nets(const ModuleId& module,
const size_t& num_nets) {
/* Validate the module id */
@ -1020,6 +1085,14 @@ void ModuleManager::clear_config_region(const ModuleId& parent_module) {
config_region_children_[parent_module].clear();
}
void ModuleManager::clear_io_children(const ModuleId& parent_module) {
VTR_ASSERT(valid_module_id(parent_module));
io_children_[parent_module].clear();
io_child_instances_[parent_module].clear();
io_child_coordinates_[parent_module].clear();
}
/******************************************************************************
* Private validators/invalidators
******************************************************************************/

60
openfpga/src/fabric/module_manager.h
View File

@ -151,6 +151,14 @@ class ModuleManager {
std::vector<size_t> configurable_child_instances(const ModuleId& parent_module) const;
/* Find the coordindate of a configurable child module under a parent module */
std::vector<vtr::Point<int>> configurable_child_coordinates(const ModuleId& parent_module) const;
/* Find all the I/O child modules under a parent module */
std::vector<ModuleId> io_children(const ModuleId& parent_module) const;
/* Find all the instances of I/O child modules under a parent module */
std::vector<size_t> io_child_instances(const ModuleId& parent_module) const;
/* Find the coordindate of an I/O child module under a parent module */
std::vector<vtr::Point<int>> io_child_coordinates(const ModuleId& parent_module) const;
/* Find the source ids of modules */
module_net_src_range module_net_sources(const ModuleId& module, const ModuleNetId& net) const;
/* Find the sink ids of modules */
@ -255,8 +263,13 @@ class ModuleManager {
void set_port_is_register(const ModuleId& module, const std::string& port_name, const bool& is_register);
/* Set the preprocessing flag for a port */
void set_port_preproc_flag(const ModuleId& module, const ModulePortId& port, const std::string& preproc_flag);
/* Add a child module to a parent module */
void add_child_module(const ModuleId& parent_module, const ModuleId& child_module);
/** @brief Add a child module to a parent module.
* By default, it considers the child module as an I/O child, and update the children list of I/O modules inside
* It not needed, just turn it off. Then you need to call add_io_child() API to update child list
*
* .. note:: By default, we assume the I/O indexing to the same as sequence when adding child modules to a parent. However, it may not be true all the time, especially for the top-level module, where customization is needed.
*/
void add_child_module(const ModuleId& parent_module, const ModuleId& child_module, const bool& is_io_child = true);
/* Set the instance name of a child module */
void set_child_instance_name(const ModuleId& parent_module, const ModuleId& child_module, const size_t& instance_id, const std::string& instance_name);
/* Add a configurable child module to module
@ -266,9 +279,7 @@ class ModuleManager {
* By default, it is an invalid coordinate
*/
void add_configurable_child(const ModuleId& module, const ModuleId& child_module, const size_t& child_instance, const vtr::Point<int> coord = vtr::Point<int>(-1, -1));
/* Reserved a number of configurable children
* for memory efficiency
*/
/* Reserved a number of configurable children for memory efficiency */
void reserve_configurable_child(const ModuleId& module, const size_t& num_children);
/* Create a new configurable region under a module */
@ -283,10 +294,18 @@ class ModuleManager {
const ModuleId& child_module,
const size_t& child_instance,
const size_t& config_child_id);
/* Reserved a number of module nets for a given module
* for memory efficiency
/** @brief Add a I/O child to module
* This function also set the coordinate of a configurable child
* The coordinate is used for build I/O location map. So it is consistent with the VPR coordinate system
* By default, it is an invalid coordinate
*
* .. note:: I/O child does not necessary have to be a I/O block. It just provide a sequence for other functions, e.g., connect_gpio_module() to index the I/Os from each child module/instance.
*/
void add_io_child(const ModuleId& module, const ModuleId& child_module, const size_t& child_instance, const vtr::Point<int> coord = vtr::Point<int>(-1, -1));
/** @brief Reserved a number of I/O children for memory efficiency */
void reserve_io_child(const ModuleId& module, const size_t& num_children);
/* Reserved a number of module nets for a given module for memory efficiency */
void reserve_module_nets(const ModuleId& module, const size_t& num_nets);
/* Add a net to the connection graph of the module */
@ -295,9 +314,7 @@ class ModuleManager {
void set_net_name(const ModuleId& module, const ModuleNetId& net,
const std::string& name);
/* Reserved a number of sources for a module net for a given module
* for memory efficiency
*/
/* Reserved a number of sources for a module net for a given module for memory efficiency */
void reserve_module_net_sources(const ModuleId& module, const ModuleNetId& net,
const size_t& num_sources);
@ -306,9 +323,7 @@ class ModuleManager {
const ModuleId& src_module, const size_t& instance_id,
const ModulePortId& src_port, const size_t& src_pin);
/* Reserved a number of sinks for a module net for a given module
* for memory efficiency
*/
/* Reserved a number of sinks for a module net for a given module for memory efficiency */
void reserve_module_net_sinks(const ModuleId& module, const ModuleNetId& net,
const size_t& num_sinks);
@ -330,6 +345,14 @@ class ModuleManager {
* Do NOT use unless you know what you are doing!!!
*/
void clear_config_region(const ModuleId& parent_module);
/* This is a strong function which will remove all the io children
* under a given parent module
* It is mainly used by other functions which want to force an I/O sequence
* Do NOT use unless you know what you are doing!!!
*/
void clear_io_children(const ModuleId& parent_module);
public: /* Public validators/invalidators */
bool valid_module_id(const ModuleId& module) const;
bool valid_module_port_id(const ModuleId& module, const ModulePortId& port) const;
@ -371,6 +394,15 @@ class ModuleManager {
vtr::vector<ModuleId, vtr::vector<ConfigRegionId, ConfigRegionId>> config_region_ids_;
vtr::vector<ModuleId, vtr::vector<ConfigRegionId, std::vector<size_t>>> config_region_children_;
/* I/O child modules are used to record the position of I/O modules in GPIO indexing
* The sequence of children in the list denotes which one is indexed in the GPIO first, etc.
* Note that the sequence can be totally different from the children_ list
* This is really dependent how the I/O indexing is organized which should be made by users/designers
*/
vtr::vector<ModuleId, std::vector<ModuleId>> io_children_;
vtr::vector<ModuleId, std::vector<size_t>> io_child_instances_;
vtr::vector<ModuleId, std::vector<vtr::Point<int>>> io_child_coordinates_;
/* Port-level data */
vtr::vector<ModuleId, vtr::vector<ModulePortId, ModulePortId>> port_ids_; /* List of ports for each Module */
vtr::vector<ModuleId, vtr::vector<ModulePortId, BasicPort>> ports_; /* List of ports for each Module */

87
openfpga/src/repack/repack.cpp
View File

@ -388,10 +388,11 @@ void add_lb_router_nets(LbRouter& lb_router,
const VprDeviceAnnotation& device_annotation,
const ClusteringContext& clustering_ctx,
const VprClusteringAnnotation& clustering_annotation,
const RepackDesignConstraints& design_constraints,
const ClusterBlockId& block_id,
const bool& verbose) {
const RepackOption& options) {
size_t net_counter = 0;
bool verbose = options.verbose_output();
RepackDesignConstraints design_constraints = options.design_constraints();
/* Two spots to find source nodes for each nets
* - nets that appear in the inputs of a clustered block
@ -437,6 +438,54 @@ void add_lb_router_nets(LbRouter& lb_router,
pb_pin_mapped_nets[pb_pin] = atom_net_id;
}
/* Cache the sink nodes/routing traces for the global nets which is specifed to be ignored on given pins */
std::map<AtomNetId, std::vector<LbRRNodeId>> ignored_global_net_sinks;
std::map<AtomNetId, bool> ignored_atom_nets;
for (int j = 0; j < lb_type->pb_type->num_pins; j++) {
/* Get the source pb_graph pin and find the rr_node in logical block routing resource graph */
const t_pb_graph_pin* source_pb_pin = get_pb_graph_node_pin_from_block_pin(block_id, j);
VTR_ASSERT(source_pb_pin->parent_node == pb->pb_graph_node);
/* Bypass output pins */
if (OUT_PORT == source_pb_pin->port->type) {
continue;
}
/* Find the net mapped to this pin in clustering results*/
ClusterNetId cluster_net_id = clustering_ctx.clb_nlist.block_net(block_id, j);
/* Get the actual net id because it may be renamed during routing */
if (true == clustering_annotation.is_net_renamed(block_id, j)) {
cluster_net_id = clustering_annotation.net(block_id, j);
}
/* Bypass unmapped pins */
if (ClusterNetId::INVALID() == cluster_net_id) {
continue;
}
/* Only for global net which should be ignored, cache the sink nodes */
BasicPort curr_pin(std::string(source_pb_pin->port->name), source_pb_pin->pin_number, source_pb_pin->pin_number);
if ( (clustering_ctx.clb_nlist.net_is_ignored(cluster_net_id))
&& (clustering_ctx.clb_nlist.net_is_global(cluster_net_id))
&& (options.is_pin_ignore_global_nets(std::string(lb_type->pb_type->name), curr_pin))) {
/* Find the net mapped to this pin in clustering results*/
AtomNetId atom_net_id = pb_pin_mapped_nets[source_pb_pin];
std::vector<int> pb_route_indices = find_pb_route_by_atom_net(pb, source_pb_pin, atom_net_id);
VTR_ASSERT(1 == pb_route_indices.size());
int pb_route_index = pb_route_indices[0];
t_pb_graph_pin* packing_source_pb_pin = get_pb_graph_node_pin_from_block_pin(block_id, pb_route_index);
VTR_ASSERT(nullptr != packing_source_pb_pin);
/* Find all the sink pins in the pb_route, we walk through the input pins and find the pin */
std::vector<t_pb_graph_pin*> sink_pb_graph_pins = find_routed_pb_graph_pins_atom_net(pb, source_pb_pin, packing_source_pb_pin, atom_net_id, device_annotation, pb_pin_mapped_nets, pb_graph_pin_lookup_from_index);
std::vector<LbRRNodeId> sink_lb_rr_nodes = find_lb_net_physical_sink_lb_rr_nodes(lb_rr_graph, sink_pb_graph_pins, device_annotation);
VTR_ASSERT(sink_lb_rr_nodes.size() == sink_pb_graph_pins.size());
ignored_global_net_sinks[atom_net_id].insert(ignored_global_net_sinks[atom_net_id].end(), sink_lb_rr_nodes.begin(), sink_lb_rr_nodes.end());
ignored_atom_nets[atom_net_id] = true;
}
}
/* Cache all the source nodes and sinks node for each net
* net_terminal[net][0] is the list of source nodes
* net_terminal[net][1] is the list of sink nodes
@ -460,6 +509,12 @@ void add_lb_router_nets(LbRouter& lb_router,
/* Find the net mapped to this pin in clustering results*/
AtomNetId atom_net_id = pb_pin_mapped_nets[source_pb_pin];
BasicPort curr_pin(std::string(source_pb_pin->port->name), source_pb_pin->pin_number, source_pb_pin->pin_number);
if ( (ignored_atom_nets[atom_net_id])
&& (options.is_pin_ignore_global_nets(std::string(lb_type->pb_type->name), curr_pin))) {
continue;
}
/* Check if the net information is constrained or not */
std::string constrained_net_name = design_constraints.find_constrained_pin_net(std::string(lb_type->pb_type->name), BasicPort(std::string(source_pb_pin->port->name), source_pb_pin->pin_number, source_pb_pin->pin_number));
@ -573,6 +628,10 @@ void add_lb_router_nets(LbRouter& lb_router,
sink_pb_pin->to_string().c_str());
}
/* Append sink nodes from ignored global net cache */
sink_lb_rr_nodes.insert(sink_lb_rr_nodes.end(), ignored_global_net_sinks[atom_net_id_to_route].begin(), ignored_global_net_sinks[atom_net_id_to_route].end());
VTR_LOGV(verbose, "Append %ld sinks from the routing traces of ignored global nets\n", ignored_global_net_sinks[atom_net_id_to_route].size());
/* Add the net */
add_lb_router_net_to_route(lb_router, lb_rr_graph,
std::vector<LbRRNodeId>(1, source_lb_rr_node),
@ -671,13 +730,13 @@ void repack_cluster(const AtomContext& atom_ctx,
const VprDeviceAnnotation& device_annotation,
VprClusteringAnnotation& clustering_annotation,
const VprBitstreamAnnotation& bitstream_annotation,
const RepackDesignConstraints& design_constraints,
const ClusterBlockId& block_id,
const bool& verbose) {
const RepackOption& options) {
/* Get the pb graph that current clustered block is mapped to */
t_logical_block_type_ptr lb_type = clustering_ctx.clb_nlist.block_type(block_id);
t_pb_graph_node* pb_graph_head = lb_type->pb_graph_head;
VTR_ASSERT(nullptr != pb_graph_head);
bool verbose = options.verbose_output();
/* We should get a non-empty graph */
const LbRRGraph& lb_rr_graph = device_annotation.physical_lb_rr_graph(pb_graph_head);
@ -693,8 +752,7 @@ void repack_cluster(const AtomContext& atom_ctx,
/* Add nets to be routed with source and terminals */
add_lb_router_nets(lb_router, lb_type, lb_rr_graph, atom_ctx, device_annotation,
clustering_ctx, const_cast<const VprClusteringAnnotation&>(clustering_annotation),
design_constraints,
block_id, verbose);
block_id, options);
/* Initialize the modes to expand routing trees with the physical modes in device annotation
* This is a must-do before running the routeri in the purpose of repacking!!!
@ -740,8 +798,7 @@ void repack_clusters(const AtomContext& atom_ctx,
const VprDeviceAnnotation& device_annotation,
VprClusteringAnnotation& clustering_annotation,
const VprBitstreamAnnotation& bitstream_annotation,
const RepackDesignConstraints& design_constraints,
const bool& verbose) {
const RepackOption& options) {
vtr::ScopedStartFinishTimer timer("Repack clustered blocks to physical implementation of logical tile");
for (auto blk_id : clustering_ctx.clb_nlist.blocks()) {
@ -749,8 +806,8 @@ void repack_clusters(const AtomContext& atom_ctx,
device_annotation,
clustering_annotation,
bitstream_annotation,
design_constraints,
blk_id, verbose);
blk_id,
options);
}
}
@ -808,22 +865,20 @@ void pack_physical_pbs(const DeviceContext& device_ctx,
VprDeviceAnnotation& device_annotation,
VprClusteringAnnotation& clustering_annotation,
const VprBitstreamAnnotation& bitstream_annotation,
const RepackDesignConstraints& design_constraints,
const CircuitLibrary& circuit_lib,
const bool& verbose) {
const RepackOption& options) {
/* build the routing resource graph for each logical tile */
build_physical_lb_rr_graphs(device_ctx,
device_annotation,
verbose);
options.verbose_output());
/* Call the LbRouter to re-pack each clustered block to physical implementation */
repack_clusters(atom_ctx, clustering_ctx,
const_cast<const VprDeviceAnnotation&>(device_annotation),
clustering_annotation,
bitstream_annotation,
design_constraints,
verbose);
options);
/* Annnotate wire LUTs that are ONLY created by repacker!!!
* This is a MUST RUN!
@ -833,7 +888,7 @@ void pack_physical_pbs(const DeviceContext& device_ctx,
clustering_ctx,
device_annotation,
circuit_lib,
verbose);
options.verbose_output());
}
} /* end namespace openfpga */

5
openfpga/src/repack/repack.h
View File

@ -9,8 +9,8 @@
#include "vpr_clustering_annotation.h"
#include "vpr_routing_annotation.h"
#include "vpr_bitstream_annotation.h"
#include "repack_design_constraints.h"
#include "circuit_library.h"
#include "repack_option.h"
/********************************************************************
* Function declaration
@ -25,9 +25,8 @@ void pack_physical_pbs(const DeviceContext& device_ctx,
VprDeviceAnnotation& device_annotation,
VprClusteringAnnotation& clustering_annotation,
const VprBitstreamAnnotation& bitstream_annotation,
const RepackDesignConstraints& design_constraints,
const CircuitLibrary& circuit_lib,
const bool& verbose);
const RepackOption& options);
} /* end namespace openfpga */

127
openfpga/src/repack/repack_option.cpp Normal file
View File

@ -0,0 +1,127 @@
/******************************************************************************
* Memember functions for data structure RepackOption
******************************************************************************/
#include <map>
#include <array>
#include "vtr_assert.h"
#include "vtr_log.h"
#include "repack_option.h"
#include "openfpga_tokenizer.h"
#include "openfpga_port_parser.h"
/* begin namespace openfpga */
namespace openfpga {
/**************************************************
* Public Constructors
*************************************************/
RepackOption::RepackOption() {
verbose_output_ = false;
num_parse_errors_ = 0;
}
/**************************************************
* Public Accessors
*************************************************/
RepackDesignConstraints RepackOption::design_constraints() const {
return design_constraints_;
}
bool RepackOption::is_pin_ignore_global_nets(const std::string& pb_type_name, const BasicPort& pin) const {
auto result = ignore_global_nets_on_pins_.find(pb_type_name);
if (result == ignore_global_nets_on_pins_.end()) {
/* Not found, return false */
return false;
} else {
/* If the pin is contained by the ignore list, return true */
for (BasicPort existing_port : result->second) {
if (existing_port.mergeable(pin) && existing_port.contained(pin)) {
return true;
}
}
}
return false;
}
bool RepackOption::verbose_output() const {
return verbose_output_;
}
/******************************************************************************
* Private Mutators
******************************************************************************/
void RepackOption::set_design_constraints(const RepackDesignConstraints& design_constraints) {
design_constraints_ = design_constraints;
}
void RepackOption::set_ignore_global_nets_on_pins(const std::string& content) {
num_parse_errors_ = 0;
/* Split the content using a tokenizer */
StringToken tokenizer(content);
std::vector<std::string> tokens = tokenizer.split(',');
/* Parse each token */
for (std::string token : tokens) {
/* Extract the pb_type name and port name */
StringToken pin_tokenizer(token);
std::vector<std::string> pin_info = pin_tokenizer.split('.');
/* Expect two contents, otherwise error out */
if (pin_info.size() != 2) {
std::string err_msg = std::string("Invalid content '") + token + std::string("' to skip, expect <pb_type_name>.<pin>\n");
VTR_LOG_ERROR(err_msg.c_str());
num_parse_errors_++;
continue;
}
std::string pb_type_name = pin_info[0];
PortParser port_parser(pin_info[1]);
BasicPort curr_port = port_parser.port();
if (!curr_port.is_valid()) {
std::string err_msg = std::string("Invalid pin definition '") + token + std::string("', expect <pb_type_name>.<pin_name>[int:int]\n");
VTR_LOG_ERROR(err_msg.c_str());
num_parse_errors_++;
continue;
}
/* Check if the existing port already in the ignore list or not */
auto result = ignore_global_nets_on_pins_.find(pb_type_name);
if (result == ignore_global_nets_on_pins_.end()) {
/* Not found, push the port */
ignore_global_nets_on_pins_[pb_type_name].push_back(curr_port);
} else {
/* Already a list of ports. Check one by one.
* - It already contained, do nothing but throw a warning.
* - If we can merge, merge it.
* - Otherwise, create it */
bool included_by_existing_port = false;
for (BasicPort existing_port : result->second) {
if (existing_port.mergeable(curr_port)) {
if (!existing_port.contained(curr_port)) {
result->second.push_back(curr_port);
included_by_existing_port = true;
break;
} else {
std::string warn_msg = std::string("Pin definition '") + token + std::string("' is already included by other pin\n");
VTR_LOG_WARN(warn_msg.c_str());
}
}
}
if (!included_by_existing_port) {
result->second.push_back(curr_port);
}
}
}
}
void RepackOption::set_verbose_output(const bool& enabled) {
verbose_output_ = enabled;
}
bool RepackOption::valid() const {
if (num_parse_errors_) {
return false;
}
return true;
}
} /* end namespace openfpga */

52
openfpga/src/repack/repack_option.h Normal file
View File

@ -0,0 +1,52 @@
#ifndef REPACK_OPTION_H
#define REPACK_OPTION_H
/********************************************************************
* Include header files required by the data structure definition
*******************************************************************/
#include <string>
#include <vector>
#include "repack_design_constraints.h"
/* Begin namespace openfpga */
namespace openfpga {
/********************************************************************
* Options for RRGSB writer
*******************************************************************/
class RepackOption {
public: /* Public constructor */
/* Set default options */
RepackOption();
public: /* Public accessors */
RepackDesignConstraints design_constraints() const;
/* Identify if a pin should ignore all the global nets */
bool is_pin_ignore_global_nets(const std::string& pb_type_name, const BasicPort& pin) const;
bool verbose_output() const;
public: /* Public mutators */
void set_design_constraints(const RepackDesignConstraints& design_constraints);
void set_ignore_global_nets_on_pins(const std::string& content);
void set_verbose_output(const bool& enabled);
public: /* Public validators */
/* Check if the following internal data is valid or not:
* - no parsing errors
*/
bool valid() const;
private: /* Internal Data */
RepackDesignConstraints design_constraints_;
/* The pin information on which global nets should be mapped to: [pb_type_name][0..num_ports]
* For example:
* - clb.I[0:1], clb.I[5:6] -> ["clb"][BasicPort(I, 0, 1), BasicPort(I, 5, 6)]
* - clb.I[0:1], clb.I[2:6] -> ["clb"][BasicPort(I, 0, 6)]
*/
std::map<std::string, std::vector<BasicPort>> ignore_global_nets_on_pins_;
bool verbose_output_;
/* A flag to indicate if the data parse is invalid or not */
int num_parse_errors_;
};
} /* End namespace openfpga*/
#endif

18
openfpga/src/utils/module_manager_utils.cpp
View File

@ -1326,7 +1326,7 @@ void add_module_nets_cmos_memory_frame_decoder_config_bus(ModuleManager& module_
/* Instanciate the decoder module here */
VTR_ASSERT(0 == module_manager.num_instance(parent_module, decoder_module));
module_manager.add_child_module(parent_module, decoder_module);
module_manager.add_child_module(parent_module, decoder_module, false);
/* Connect the enable (EN) port of memory modules under the parent module
* to the frame decoder inputs
@ -1802,10 +1802,9 @@ void add_module_io_ports_from_child_modules(ModuleManager& module_manager,
std::vector<BasicPort> gpio_ports_to_add;
std::vector<bool> mappable_gpio_ports;
/* Iterate over the child modules */
for (const ModuleId& child : module_manager.child_modules(module_id)) {
/* Iterate over the child instances */
for (size_t i = 0; i < module_manager.num_instance(module_id, child); ++i) {
/* Iterate over the child modules and instances */
for (size_t i = 0; i < module_manager.io_children(module_id).size(); ++i) {
ModuleId child = module_manager.io_children(module_id)[i];
/* Find all the global ports, whose port type is special */
for (const ModulePortId& gpio_port_id : module_manager.module_port_ids_by_type(child, module_port_type)) {
const BasicPort& gpio_port = module_manager.module_port(child, gpio_port_id);
@ -1837,7 +1836,6 @@ void add_module_io_ports_from_child_modules(ModuleManager& module_manager,
}
}
}
}
/* Record the port id for each type of GPIO port */
std::vector<ModulePortId> gpio_port_ids;
@ -1854,9 +1852,9 @@ void add_module_io_ports_from_child_modules(ModuleManager& module_manager,
/* Set up a counter for each type of GPIO port */
std::vector<size_t> gpio_port_lsb(gpio_ports_to_add.size(), 0);
/* Add module nets to connect the GPIOs of the module to the GPIOs of the sub module */
for (const ModuleId& child : module_manager.child_modules(module_id)) {
/* Iterate over the child instances */
for (const size_t& child_instance : module_manager.child_module_instances(module_id, child)) {
for (size_t i = 0; i < module_manager.io_children(module_id).size(); ++i) {
ModuleId child = module_manager.io_children(module_id)[i];
size_t child_instance = module_manager.io_child_instances(module_id)[i];
/* Find all the global ports, whose port type is special */
for (ModulePortId child_gpio_port_id : module_manager.module_port_ids_by_type(child, module_port_type)) {
BasicPort child_gpio_port = module_manager.module_port(child, child_gpio_port_id);
@ -1892,11 +1890,9 @@ void add_module_io_ports_from_child_modules(ModuleManager& module_manager,
}
}
}
}
/* Check: all the lsb should now match the size of each GPIO port */
for (size_t iport = 0; iport < gpio_ports_to_add.size(); ++iport) {
if (gpio_ports_to_add[iport].get_width() != gpio_port_lsb[iport])
VTR_ASSERT(gpio_ports_to_add[iport].get_width() == gpio_port_lsb[iport]);
}
}

26
openfpga_flow/benchmarks/micro_benchmark/rst_on_lut/rst_on_lut.v Normal file
View File

@ -0,0 +1,26 @@
/////////////////////////////////////////
// Functionality: A register driven by a combinational logic with reset signal
// Author: Xifan Tang
////////////////////////////////////////
`timescale 1ns / 1ps
module rst_on_lut(a, b, q, out, clk, rst);
input wire rst;
input wire clk;
input wire a;
input wire b;
output reg q;
output wire out;
always @(posedge rst or posedge clk) begin
if (rst) begin
q <= 0;
end else begin
q <= a;
end
end
assign out = b & ~rst;
endmodule

76
openfpga_flow/openfpga_shell_scripts/ignore_global_nets_on_pins_example_script.openfpga Normal file
View File

@ -0,0 +1,76 @@
# Run VPR for the 'and' design
#--write_rr_graph example_rr_graph.xml
vpr ${VPR_ARCH_FILE} ${VPR_TESTBENCH_BLIF} --clock_modeling ideal
# Read OpenFPGA architecture definition
read_openfpga_arch -f ${OPENFPGA_ARCH_FILE}
# Read OpenFPGA simulation settings
read_openfpga_simulation_setting -f ${OPENFPGA_SIM_SETTING_FILE}
# Annotate the OpenFPGA architecture to VPR data base
# to debug use --verbose options
link_openfpga_arch --sort_gsb_chan_node_in_edges
# Check and correct any naming conflicts in the BLIF netlist
check_netlist_naming_conflict --fix --report ./netlist_renaming.xml
# Apply fix-up to clustering nets based on routing results
pb_pin_fixup --verbose
# Apply fix-up to Look-Up Table truth tables based on packing results
lut_truth_table_fixup
# Build the module graph
# - Enabled compression on routing architecture modules
# - Enable pin duplication on grid modules
build_fabric --compress_routing #--verbose
# Write the fabric hierarchy of module graph to a file
# This is used by hierarchical PnR flows
write_fabric_hierarchy --file ./fabric_hierarchy.txt
# Repack the netlist to physical pbs
# This must be done before bitstream generator and testbench generation
# Strongly recommend it is done after all the fix-up have been applied
repack --ignore_global_nets_on_pins clb.I[0:11] #--verbose
# Build the bitstream
# - Output the fabric-independent bitstream to a file
build_architecture_bitstream --verbose --write_file fabric_independent_bitstream.xml
# Build fabric-dependent bitstream
build_fabric_bitstream --verbose
# Write fabric-dependent bitstream
write_fabric_bitstream --file fabric_bitstream.bit --format plain_text
# Write the Verilog netlist for FPGA fabric
# - Enable the use of explicit port mapping in Verilog netlist
write_fabric_verilog --file ./SRC --explicit_port_mapping --include_timing --print_user_defined_template --verbose
# Write the Verilog testbench for FPGA fabric
# - We suggest the use of same output directory as fabric Verilog netlists
# - Must specify the reference benchmark file if you want to output any testbenches
# - Enable top-level testbench which is a full verification including programming circuit and core logic of FPGA
# - Enable pre-configured top-level testbench which is a fast verification skipping programming phase
# - Simulation ini file is optional and is needed only when you need to interface different HDL simulators using openfpga flow-run scripts
write_full_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --include_signal_init --pin_constraints_file ${OPENFPGA_PIN_CONSTRAINTS_FILE} --bitstream fabric_bitstream.bit
write_preconfigured_fabric_wrapper --embed_bitstream iverilog --file ./SRC --pin_constraints_file ${OPENFPGA_PIN_CONSTRAINTS_FILE}
write_preconfigured_testbench --file ./SRC --reference_benchmark_file_path ${REFERENCE_VERILOG_TESTBENCH} --pin_constraints_file ${OPENFPGA_PIN_CONSTRAINTS_FILE}
# Write the SDC files for PnR backend
# - Turn on every options here
write_pnr_sdc --file ./SDC
# Write SDC to disable timing for configure ports
write_sdc_disable_timing_configure_ports --file ./SDC/disable_configure_ports.sdc
# Write the SDC to run timing analysis for a mapped FPGA fabric
write_analysis_sdc --file ./SDC_analysis
# Finish and exit OpenFPGA
exit
# Note :
# To run verification at the end of the flow maintain source in ./SRC directory

5
openfpga_flow/regression_test_scripts/basic_reg_test.sh
View File

@ -129,6 +129,11 @@ echo -e "Testing K4N5 with pattern based local routing";
run-task basic_tests/k4_series/k4n5_pattern_local_routing $@
echo -e "Testing K4N4 with custom I/O location syntax";
run-task basic_tests/k4_series/k4n4_custom_io_loc $@
run-task basic_tests/k4_series/k4n4_custom_io_loc_center $@
run-task basic_tests/k4_series/k4n4_custom_io_loc_center_height_odd $@
run-task basic_tests/k4_series/k4n4_custom_io_loc_center_width_odd $@
echo -e "Testing K4N4 with a local routing where reset can driven LUT inputs";
run-task basic_tests/k4_series/k4n4_rstOnLut $@
echo -e "Testing different tile organizations";
echo -e "Testing tiles with pins only on top and left sides";

109
openfpga_flow/scripts/io_sequence_visualizer.py Normal file
View File

@ -0,0 +1,109 @@
"""
=========================================
Represetes IO Sequence in OpenFPGA Engine
=========================================
This example demonstrates the ``OpenFPGA_Arch`` class which parses the
`VPR` and `OpenFPGA` Architecture file and provides logical information.
.. image:: ../../../examples/OpenFPGA_basic/_sample_io_sequence.svg
:width: 60%
:align: center
Author: Ganesh Gore
"""
import math
import svgwrite
from svgwrite.container import Group
def draw_connections(width, height, connections):
"""
Draw connection sequence
"""
dwg = svgwrite.Drawing()
DRAW_WIDTH = (width + 2) * SCALE
DRAW_HEIGHT = (height + 2) * SCALE
# set user coordinate space
dwg.viewbox(width=DRAW_WIDTH, height=DRAW_HEIGHT, miny=-1 * DRAW_HEIGHT)
dwg_main = Group(id="Main", transform="scale(1,-1)")
dwg.add(dwg_main)
for w in range(1, width + 2):
dwg_main.add(
dwg.line(
(w * SCALE, SCALE), (w * SCALE, (height + 1) * SCALE), stroke="red"
)
)
for h in range(1, height + 2):
dwg_main.add(
dwg.line((SCALE, h * SCALE), ((width + 1) * SCALE, h * SCALE), stroke="red")
)
path = "M "
for point in connections:
path += " %d %d " % ((point[0] + 0.5) * SCALE, (point[1] + 0.5) * SCALE)
dwg_main.add(dwg.path(path, stroke="blue", fill="none", stroke_width="2px"))
dwg.saveas("_sample_io_sequence.svg", pretty=True)
SCALE = 20
FPGA_WIDTH = 40
FPGA_HEIGHT = 15
W = max(FPGA_WIDTH, FPGA_HEIGHT)
W2 = math.floor(W / 2) + 1
connections = []
xmin, xmax = 1, FPGA_WIDTH
ymin, ymax = 1, FPGA_HEIGHT
while (xmin < xmax) and (ymin < ymax):
print(xmin, ymin, end=" -> ")
print(xmax, ymax)
x = xmin
for y in range(ymin, ymax + 1):
connections.append((x, y))
y = ymax
for x in range(xmin, xmax + 1):
connections.append((x, y))
x = xmax
for y in range(ymin, ymax + 1)[::-1]:
connections.append((x, y))
y = ymin
for x in range(xmin, xmax + 1)[::-1][:-1]:
connections.append((x, y))
xmin += 1
ymin += 1
xmax -= 1
ymax -= 1
if FPGA_HEIGHT % 2 == 1: # if height is odd
if ymin == ymax: # if touching vertically
y = ymin
for x in range(xmin, xmax + 1):
connections.append((x, y))
if FPGA_WIDTH % 2 == 1: # if width is odd
if xmin == xmax: # if touching horizontally
x = xmin
for y in range(ymin, ymax + 1):
connections.append((x, y))
# print(connections)
if connections:
draw_connections(FPGA_WIDTH, FPGA_HEIGHT, connections)
else:
# Dummy draw
draw_connections(FPGA_WIDTH, FPGA_HEIGHT, [(1, 1)])

2
openfpga_flow/tasks/basic_tests/k4_series/k4n4_custom_io_loc/config/task.conf
View File

@ -19,7 +19,7 @@ fpga_flow=vpr_blif
openfpga_shell_template=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_shell_scripts/write_full_testbench_example_script.openfpga
openfpga_arch_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_arch/k4_N4_40nm_cc_openfpga.xml
openfpga_sim_setting_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_simulation_settings/auto_sim_openfpga.xml
openfpga_vpr_device_layout=--device 2x2
openfpga_vpr_device_layout=--device 4x4
openfpga_fast_configuration=
[ARCHITECTURES]

37
openfpga_flow/tasks/basic_tests/k4_series/k4n4_custom_io_loc_center/config/task.conf Normal file
View File

@ -0,0 +1,37 @@
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# Configuration file for running experiments
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# timeout_each_job : FPGA Task script splits fpga flow into multiple jobs
# Each job execute fpga_flow script on combination of architecture & benchmark
# timeout_each_job is timeout for each job
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
[GENERAL]
run_engine=openfpga_shell
power_tech_file = ${PATH:OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.xml
power_analysis = true
spice_output=false
verilog_output=true
timeout_each_job = 20*60
fpga_flow=vpr_blif
[OpenFPGA_SHELL]
openfpga_shell_template=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_shell_scripts/write_full_testbench_example_script.openfpga
openfpga_arch_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_arch/k4_N4_40nm_cc_openfpga.xml
openfpga_sim_setting_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_simulation_settings/auto_sim_openfpga.xml
openfpga_vpr_device_layout=--device 4x4_io_center
openfpga_fast_configuration=
[ARCHITECTURES]
arch0=${PATH:OPENFPGA_PATH}/openfpga_flow/vpr_arch/k4_N4_tileable_customIoLoc_40nm.xml
[BENCHMARKS]
bench0=${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.blif
[SYNTHESIS_PARAM]
bench0_top = and2
bench0_act = ${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.act
bench0_verilog = ${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.v
[SCRIPT_PARAM_MIN_ROUTE_CHAN_WIDTH]
end_flow_with_test=

37
openfpga_flow/tasks/basic_tests/k4_series/k4n4_custom_io_loc_center_height_odd/config/task.conf Normal file
View File

@ -0,0 +1,37 @@
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# Configuration file for running experiments
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# timeout_each_job : FPGA Task script splits fpga flow into multiple jobs
# Each job execute fpga_flow script on combination of architecture & benchmark
# timeout_each_job is timeout for each job
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
[GENERAL]
run_engine=openfpga_shell
power_tech_file = ${PATH:OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.xml
power_analysis = true
spice_output=false
verilog_output=true
timeout_each_job = 20*60
fpga_flow=vpr_blif
[OpenFPGA_SHELL]
openfpga_shell_template=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_shell_scripts/write_full_testbench_example_script.openfpga
openfpga_arch_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_arch/k4_N4_40nm_cc_openfpga.xml
openfpga_sim_setting_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_simulation_settings/auto_sim_openfpga.xml
openfpga_vpr_device_layout=--device 4x3_io_center
openfpga_fast_configuration=
[ARCHITECTURES]
arch0=${PATH:OPENFPGA_PATH}/openfpga_flow/vpr_arch/k4_N4_tileable_customIoLoc_40nm.xml
[BENCHMARKS]
bench0=${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.blif
[SYNTHESIS_PARAM]
bench0_top = and2
bench0_act = ${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.act
bench0_verilog = ${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.v
[SCRIPT_PARAM_MIN_ROUTE_CHAN_WIDTH]
end_flow_with_test=

37
openfpga_flow/tasks/basic_tests/k4_series/k4n4_custom_io_loc_center_width_odd/config/task.conf Normal file
View File

@ -0,0 +1,37 @@
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# Configuration file for running experiments
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# timeout_each_job : FPGA Task script splits fpga flow into multiple jobs
# Each job execute fpga_flow script on combination of architecture & benchmark
# timeout_each_job is timeout for each job
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
[GENERAL]
run_engine=openfpga_shell
power_tech_file = ${PATH:OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.xml
power_analysis = true
spice_output=false
verilog_output=true
timeout_each_job = 20*60
fpga_flow=vpr_blif
[OpenFPGA_SHELL]
openfpga_shell_template=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_shell_scripts/write_full_testbench_example_script.openfpga
openfpga_arch_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_arch/k4_N4_40nm_cc_openfpga.xml
openfpga_sim_setting_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_simulation_settings/auto_sim_openfpga.xml
openfpga_vpr_device_layout=--device 3x4_io_center
openfpga_fast_configuration=
[ARCHITECTURES]
arch0=${PATH:OPENFPGA_PATH}/openfpga_flow/vpr_arch/k4_N4_tileable_customIoLoc_40nm.xml
[BENCHMARKS]
bench0=${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.blif
[SYNTHESIS_PARAM]
bench0_top = and2
bench0_act = ${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.act
bench0_verilog = ${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/and2/and2.v
[SCRIPT_PARAM_MIN_ROUTE_CHAN_WIDTH]
end_flow_with_test=

7
openfpga_flow/tasks/basic_tests/k4_series/k4n4_rstOnLut/config/pin_constraints_reset.xml Normal file
View File

@ -0,0 +1,7 @@
<pin_constraints>
<!-- For a given .blif file, we want to assign
- the reset signal to the op_reset[0] port of the FPGA fabric
-->
<set_io pin="op_reset[0]" net="rst"/>
</pin_constraints>

42
openfpga_flow/tasks/basic_tests/k4_series/k4n4_rstOnLut/config/task.conf Normal file
View File

@ -0,0 +1,42 @@
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# Configuration file for running experiments
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
# timeout_each_job : FPGA Task script splits fpga flow into multiple jobs
# Each job execute fpga_flow script on combination of architecture & benchmark
# timeout_each_job is timeout for each job
# = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
[GENERAL]
run_engine=openfpga_shell
power_tech_file = ${PATH:OPENFPGA_PATH}/openfpga_flow/tech/PTM_45nm/45nm.xml
power_analysis = false
spice_output=false
verilog_output=true
timeout_each_job = 3*60
fpga_flow=yosys_vpr
[OpenFPGA_SHELL]
openfpga_shell_template=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_shell_scripts/ignore_global_nets_on_pins_example_script.openfpga
openfpga_arch_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_arch/k4_frac_N4_fracff_40nm_cc_openfpga.xml
openfpga_sim_setting_file=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_simulation_settings/fixed_sim_openfpga.xml
[ARCHITECTURES]
arch0=${PATH:OPENFPGA_PATH}/openfpga_flow/vpr_arch/k4_frac_N4_tileable_fracff_rstOnLut_40nm.xml
[BENCHMARKS]
bench0=${PATH:OPENFPGA_PATH}/openfpga_flow/benchmarks/micro_benchmark/rst_on_lut/rst_on_lut.v
[SYNTHESIS_PARAM]
# Yosys script parameters
bench_yosys_cell_sim_verilog_common=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_yosys_techlib/openfpga_dff_sim.v
bench_yosys_dff_map_verilog_common=${PATH:OPENFPGA_PATH}/openfpga_flow/openfpga_yosys_techlib/openfpga_dff_map.v
bench_read_verilog_options_common = -nolatches
bench_yosys_common=${PATH:OPENFPGA_PATH}/openfpga_flow/misc/ys_tmpl_yosys_vpr_dff_flow.ys
bench_yosys_rewrite_common=${PATH:OPENFPGA_PATH}/openfpga_flow/misc/ys_tmpl_yosys_vpr_flow_with_rewrite.ys;${PATH:OPENFPGA_PATH}/openfpga_flow/misc/ys_tmpl_rewrite_flow.ys
bench0_top = rst_on_lut
bench0_openfpga_pin_constraints_file = ${PATH:TASK_DIR}/config/pin_constraints_reset.xml
[SCRIPT_PARAM_MIN_ROUTE_CHAN_WIDTH]
end_flow_with_test=
vpr_fpga_verilog_formal_verification_top_netlist=

1
openfpga_flow/vpr_arch/README.md
View File

@ -22,6 +22,7 @@ Please reveal the following architecture features in the names to help quickly s
- reduced\_io: If I/Os only appear a certain or multiple sides of FPGAs
- registerable\_io: If I/Os are registerable (can be either combinational or sequential)
- CustomIoLoc: Use OpenFPGA's extended custom I/O location syntax
- rstOnLut: The reset signal of CLB can feed LUT inputs through a local routing architecture
- <feature\_size>: The technology node which the delay numbers are extracted from.
- TileOrgz<Type>: How tile is organized.
* Top-left (Tl): the pins of a tile are placed on the top side and left side only

148
openfpga_flow/vpr_arch/k4_N4_tileable_customIoLoc_40nm.xml
View File

@ -43,7 +43,10 @@
<output name="inpad" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="custom">
<loc side="bottom">io_top[0:3].inpad io_top[0:7].outpad io_top[4:7].inpad</loc>
<loc side="top">io_top[0:1].inpad io_top[0:1].outpad</loc>
<loc side="right">io_top[3:2].inpad</loc>
<loc side="bottom">io_top[4:5].inpad io_top[2:4].outpad</loc>
<loc side="left">io_top[6:7].inpad io_top[5:7].outpad</loc>
</pinlocations>
</tile>
<tile name="io_bottom" capacity="6" area="0">
@ -54,18 +57,10 @@
<output name="inpad" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="custom">
<loc side="top">io_bottom[0:1].outpad io_bottom[0:3].inpad io_bottom[2:5].outpad io_bottom[4:5].inpad</loc>
</pinlocations>
</tile>
<tile name="io_left" capacity="4" area="0">
<equivalent_sites>
<site pb_type="io"/>
</equivalent_sites>
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="custom">
<loc side="right">io_left.inpad io_left.outpad</loc>
<loc side="top">io_bottom[0:1].outpad</loc>
<loc side="right">io_bottom[2:2].outpad io_bottom[0:3].inpad</loc>
<loc side="bottom">io_bottom[3:3].outpad</loc>
<loc side="left">io_bottom[4:5].outpad io_bottom[4:5].inpad</loc>
</pinlocations>
</tile>
<tile name="io_right" capacity="2" area="0">
@ -76,7 +71,10 @@
<output name="inpad" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="custom">
<loc side="left">io_right[1:1].inpad io_right[1:0].outpad io_right[0:0].inpad</loc>
<loc side="top">io_right[0:0].inpad</loc>
<loc side="right">io_right[1:1].outpad</loc>
<loc side="bottom">io_right[0:0].outpad</loc>
<loc side="left">io_right[1:1].inpad</loc>
</pinlocations>
</tile>
<tile name="clb" area="53894">
@ -94,61 +92,123 @@
<!-- Physical descriptions begin -->
<layout tileable="true">
<auto_layout aspect_ratio="1.0">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<row type="io_top" starty="H-1" priority="100"/>
<row type="io_bottom" starty="0" priority="100"/>
<col type="io_left" startx="0" priority="100"/>
<col type="io_right" startx="W-1" priority="100"/>
<!--Perimeter of 'EMPTY' blocks, I/Os are placed on the inner ring
Intend to have no I/Os on the left side, it is to check the correctness of I/O indexing in OpenFPGA
-->
<row type="io_top" starty="H-2" priority="90"/>
<row type="io_bottom" starty="1" priority="91"/>
<col type="io_right" startx="W-2" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</auto_layout>
<fixed_layout name="2x2" width="4" height="4">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<row type="io_top" starty="H-1" priority="100"/>
<row type="io_bottom" starty="0" priority="100"/>
<col type="io_left" startx="0" priority="100"/>
<col type="io_right" startx="W-1" priority="100"/>
<!--Perimeter of 'EMPTY' blocks, I/Os are placed on the inner ring -->
<row type="io_top" starty="H-2" priority="90"/>
<row type="io_bottom" starty="1" priority="91"/>
<col type="io_right" startx="W-2" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="4x4" width="6" height="6">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<row type="io_top" starty="H-1" priority="100"/>
<row type="io_bottom" starty="0" priority="100"/>
<col type="io_left" startx="0" priority="100"/>
<col type="io_right" startx="W-1" priority="100"/>
<!--Perimeter of 'EMPTY' blocks, I/Os are placed on the inner ring -->
<row type="io_top" starty="H-2" priority="90"/>
<row type="io_bottom" starty="1" priority="91"/>
<col type="io_right" startx="W-2" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="4x4_io_center" width="6" height="6">
<!--Perimeter of 'clb' blocks, I/Os are placed in the center-->
<row type="clb" starty="H-2" priority="90"/>
<row type="clb" starty="1" priority="91"/>
<col type="clb" startx="W-2" priority="93"/>
<col type="clb" startx="1" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="io_top" priority="10"/>
</fixed_layout>
<fixed_layout name="4x3_io_center" width="6" height="5">
<!--Perimeter of 'clb' blocks, I/Os are placed in the center-->
<row type="clb" starty="H-2" priority="90"/>
<row type="clb" starty="1" priority="91"/>
<col type="clb" startx="W-2" priority="93"/>
<col type="clb" startx="1" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="io_top" priority="10"/>
</fixed_layout>
<fixed_layout name="3x4_io_center" width="5" height="6">
<!--Perimeter of 'clb' blocks, I/Os are placed in the center-->
<row type="clb" starty="H-2" priority="90"/>
<row type="clb" starty="1" priority="91"/>
<col type="clb" startx="W-2" priority="93"/>
<col type="clb" startx="1" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="io_top" priority="10"/>
</fixed_layout>
<fixed_layout name="48x48" width="50" height="50">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<row type="io_top" starty="H-1" priority="100"/>
<row type="io_bottom" starty="0" priority="100"/>
<col type="io_left" startx="0" priority="100"/>
<col type="io_right" startx="W-1" priority="100"/>
<!--Perimeter of 'EMPTY' blocks, I/Os are placed on the inner ring -->
<row type="io_top" starty="H-2" priority="90"/>
<row type="io_bottom" starty="1" priority="91"/>
<col type="io_right" startx="W-2" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="72x72" width="74" height="74">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<row type="io_top" starty="H-1" priority="100"/>
<row type="io_bottom" starty="0" priority="100"/>
<col type="io_left" startx="0" priority="100"/>
<col type="io_right" startx="W-1" priority="100"/>
<!--Perimeter of 'EMPTY' blocks, I/Os are placed on the inner ring -->
<row type="io_top" starty="H-2" priority="90"/>
<row type="io_bottom" starty="1" priority="91"/>
<col type="io_right" startx="W-2" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="96x96" width="98" height="98">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<row type="io_top" starty="H-1" priority="100"/>
<row type="io_bottom" starty="0" priority="100"/>
<col type="io_left" startx="0" priority="100"/>
<col type="io_right" startx="W-1" priority="100"/>
<!--Perimeter of 'EMPTY' blocks, I/Os are placed on the inner ring -->
<row type="io_top" starty="H-2" priority="90"/>
<row type="io_bottom" starty="1" priority="91"/>
<col type="io_right" startx="W-2" priority="93"/>
<row type="EMPTY" starty="H-1" priority="101"/>
<row type="EMPTY" starty="0" priority="102"/>
<col type="EMPTY" startx="0" priority="103"/>
<col type="EMPTY" startx="W-1" priority="104"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>

627
openfpga_flow/vpr_arch/k4_frac_N4_tileable_fracff_rstOnLut_40nm.xml Normal file
View File

@ -0,0 +1,627 @@
<!--
Flagship Heterogeneous Architecture (No Carry Chains) for VTR 7.0.
- 40 nm technology
- General purpose logic block:
K = 4, N = 4, fracturable 4 LUTs (can operate as one 4-LUT or two 3-LUTs with all 3 inputs shared)
with optionally registered outputs
- Routing architecture: L = 4, fc_in = 0.15, Fc_out = 0.1
Authors: Xifan Tang
-->
<architecture>
<!--
ODIN II specific config begins
Describes the types of user-specified netlist blocks (in blif, this corresponds to
".model [type_of_block]") that this architecture supports.
Note: Basic LUTs, I/Os, and flip-flops are not included here as there are
already special structures in blif (.names, .input, .output, and .latch)
that describe them.
-->
<models>
<!-- A virtual model for I/O to be used in the physical mode of io block -->
<model name="io">
<input_ports>
<port name="outpad"/>
</input_ports>
<output_ports>
<port name="inpad"/>
</output_ports>
</model>
<!-- A virtual model for I/O to be used in the physical mode of io block -->
<model name="frac_lut4">
<input_ports>
<port name="in"/>
</input_ports>
<output_ports>
<port name="lut3_out"/>
<port name="lut4_out"/>
</output_ports>
</model>
<!-- A virtual model for scan-chain flip-flop to be used in the physical mode of FF -->
<model name="dff">
<input_ports>
<port name="D" clock="C"/>
<port name="C" is_clock="1"/>
</input_ports>
<output_ports>
<port name="Q" clock="C"/>
</output_ports>
</model>
<!-- A virtual model for scan-chain flip-flop to be used in the physical mode of FF -->
<model name="dffr">
<input_ports>
<port name="D" clock="C"/>
<port name="R" clock="C"/>
<port name="C" is_clock="1"/>
</input_ports>
<output_ports>
<port name="Q" clock="C"/>
</output_ports>
</model>
<!-- A virtual model for scan-chain flip-flop to be used in the physical mode of FF -->
<model name="dffrn">
<input_ports>
<port name="D" clock="C"/>
<port name="RN" clock="C"/>
<port name="C" is_clock="1"/>
</input_ports>
<output_ports>
<port name="Q" clock="C"/>
</output_ports>
</model>
</models>
<tiles>
<!-- Do NOT add clock pins to I/O here!!! VPR does not build clock network in the way that OpenFPGA can support
If you need to register the I/O, define clocks in the circuit models
These clocks can be handled in back-end
-->
<tile name="io" capacity="8" area="0">
<equivalent_sites>
<site pb_type="io"/>
</equivalent_sites>
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10"/>
<pinlocations pattern="custom">
<loc side="left">io.outpad io.inpad</loc>
<loc side="top">io.outpad io.inpad</loc>
<loc side="right">io.outpad io.inpad</loc>
<loc side="bottom">io.outpad io.inpad</loc>
</pinlocations>
</tile>
<tile name="clb" area="53894">
<equivalent_sites>
<site pb_type="clb"/>
</equivalent_sites>
<input name="I" num_pins="12" equivalent="full"/>
<input name="reset" num_pins="1" is_non_clock_global="true"/>
<output name="O" num_pins="8" equivalent="none"/>
<clock name="clk" num_pins="1"/>
<fc in_type="frac" in_val="0.15" out_type="frac" out_val="0.10">
<fc_override port_name="clk" fc_type="frac" fc_val="0"/>
<fc_override port_name="reset" fc_type="frac" fc_val="0"/>
</fc>
<pinlocations pattern="spread"/>
</tile>
</tiles>
<!-- ODIN II specific config ends -->
<!-- Physical descriptions begin -->
<layout tileable="true">
<auto_layout aspect_ratio="1.0">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</auto_layout>
<fixed_layout name="2x2" width="4" height="4">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="4x4" width="6" height="6">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
<fixed_layout name="48x48" width="50" height="50">
<!--Perimeter of 'io' blocks with 'EMPTY' blocks at corners-->
<perimeter type="io" priority="100"/>
<corners type="EMPTY" priority="101"/>
<!--Fill with 'clb'-->
<fill type="clb" priority="10"/>
</fixed_layout>
</layout>
<device>
<!-- VB & JL: Using Ian Kuon's transistor sizing and drive strength data for routing, at 40 nm. Ian used BPTM
models. We are modifying the delay values however, to include metal C and R, which allows more architecture
experimentation. We are also modifying the relative resistance of PMOS to be 1.8x that of NMOS
(vs. Ian's 3x) as 1.8x lines up with Jeff G's data from a 45 nm process (and is more typical of
45 nm in general). I'm upping the Rmin_nmos from Ian's just over 6k to nearly 9k, and dropping
RminW_pmos from 18k to 16k to hit this 1.8x ratio, while keeping the delays of buffers approximately
lined up with Stratix IV.
We are using Jeff G.'s capacitance data for 45 nm (in tech/ptm_45nm).
Jeff's tables list C in for transistors with widths in multiples of the minimum feature size (45 nm).
The minimum contactable transistor is 2.5 * 45 nm, so I need to multiply drive strength sizes in this file
by 2.5x when looking up in Jeff's tables.
The delay values are lined up with Stratix IV, which has an architecture similar to this
proposed FPGA, and which is also 40 nm
C_ipin_cblock: input capacitance of a track buffer, which VPR assumes is a single-stage
4x minimum drive strength buffer. -->
<sizing R_minW_nmos="8926" R_minW_pmos="16067"/>
<!-- The grid_logic_tile_area below will be used for all blocks that do not explicitly set their own (non-routing)
area; set to 0 since we explicitly set the area of all blocks currently in this architecture file.
-->
<area grid_logic_tile_area="0"/>
<chan_width_distr>
<x distr="uniform" peak="1.000000"/>
<y distr="uniform" peak="1.000000"/>
</chan_width_distr>
<switch_block type="wilton" fs="3" sub_type="subset" sub_fs="3"/>
<connection_block input_switch_name="ipin_cblock"/>
</device>
<switchlist>
<!-- VB: the mux_trans_size and buf_size data below is in minimum width transistor *areas*, assuming the purple
book area formula. This means the mux transistors are about 5x minimum drive strength.
We assume the first stage of the buffer is 3x min drive strength to be reasonable given the large
mux transistors, and this gives a reasonable stage ratio of a bit over 5x to the second stage. We assume
the n and p transistors in the first stage are equal-sized to lower the buffer trip point, since it's fed
by a pass transistor mux. We can then reverse engineer the buffer second stage to hit the specified
buf_size (really buffer area) - 16.2x minimum drive nmos and 1.8*16.2 = 29.2x minimum drive.
I then took the data from Jeff G.'s PTM modeling of 45 nm to get the Cin (gate of first stage) and Cout
(diff of second stage) listed below. Jeff's models are in tech/ptm_45nm, and are in min feature multiples.
The minimum contactable transistor is 2.5 * 45 nm, so I need to multiply the drive strength sizes above by
2.5x when looking up in Jeff's tables.
Finally, we choose a switch delay (58 ps) that leads to length 4 wires having a delay equal to that of SIV of 126 ps.
This also leads to the switch being 46% of the total wire delay, which is reasonable. -->
<switch type="mux" name="0" R="551" Cin=".77e-15" Cout="4e-15" Tdel="58e-12" mux_trans_size="2.630740" buf_size="27.645901"/>
<!--switch ipin_cblock resistance set to yeild for 4x minimum drive strength buffer-->
<switch type="mux" name="ipin_cblock" R="2231.5" Cout="0." Cin="1.47e-15" Tdel="7.247000e-11" mux_trans_size="1.222260" buf_size="auto"/>
</switchlist>
<segmentlist>
<!--- VB & JL: using ITRS metal stack data, 96 nm half pitch wires, which are intermediate metal width/space.
With the 96 nm half pitch, such wires would take 60 um of height, vs. a 90 nm high (approximated as square) Stratix IV tile so this seems
reasonable. Using a tile length of 90 nm, corresponding to the length of a Stratix IV tile if it were square. -->
<!-- GIVE a specific name for the segment! OpenFPGA appreciate that! -->
<segment name="L4" freq="1.000000" length="4" type="unidir" Rmetal="101" Cmetal="22.5e-15">
<mux name="0"/>
<sb type="pattern">1 1 1 1 1</sb>
<cb type="pattern">1 1 1 1</cb>
</segment>
</segmentlist>
<complexblocklist>
<!-- Define I/O pads begin -->
<!-- Capacity is a unique property of I/Os, it is the maximum number of I/Os that can be placed at the same (X,Y) location on the FPGA -->
<!-- Not sure of the area of an I/O (varies widely), and it's not relevant to the design of the FPGA core, so we're setting it to 0. -->
<pb_type name="io">
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
<!-- Do NOT add clock pins to I/O here!!! VPR does not build clock network in the way that OpenFPGA can support
If you need to register the I/O, define clocks in the circuit models
These clocks can be handled in back-end
-->
<!-- A mode denotes the physical implementation of an I/O
This mode will be not packable but is mainly used for fabric verilog generation
-->
<mode name="physical" disable_packing="true">
<pb_type name="iopad" blif_model=".subckt io" num_pb="1">
<input name="outpad" num_pins="1"/>
<output name="inpad" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="outpad" input="io.outpad" output="iopad.outpad">
<delay_constant max="1.394e-11" in_port="io.outpad" out_port="iopad.outpad"/>
</direct>
<direct name="inpad" input="iopad.inpad" output="io.inpad">
<delay_constant max="4.243e-11" in_port="iopad.inpad" out_port="io.inpad"/>
</direct>
</interconnect>
</mode>
<!-- IOs can operate as either inputs or outputs.
Delays below come from Ian Kuon. They are small, so they should be interpreted as
the delays to and from registers in the I/O (and generally I/Os are registered
today and that is when you timing analyze them.
-->
<mode name="inpad">
<pb_type name="inpad" blif_model=".input" num_pb="1">
<output name="inpad" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="inpad" input="inpad.inpad" output="io.inpad">
<delay_constant max="4.243e-11" in_port="inpad.inpad" out_port="io.inpad"/>
</direct>
</interconnect>
</mode>
<mode name="outpad">
<pb_type name="outpad" blif_model=".output" num_pb="1">
<input name="outpad" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="outpad" input="io.outpad" output="outpad.outpad">
<delay_constant max="1.394e-11" in_port="io.outpad" out_port="outpad.outpad"/>
</direct>
</interconnect>
</mode>
<!-- Every input pin is driven by 15% of the tracks in a channel, every output pin is driven by 10% of the tracks in a channel -->
<!-- IOs go on the periphery of the FPGA, for consistency,
make it physically equivalent on all sides so that only one definition of I/Os is needed.
If I do not make a physically equivalent definition, then I need to define 4 different I/Os, one for each side of the FPGA
-->
<!-- Place I/Os on the sides of the FPGA -->
<power method="ignore"/>
</pb_type>
<!-- Define I/O pads ends -->
<!-- Define general purpose logic block (CLB) begin -->
<!--- Area calculation: Total Stratix IV tile area is about 8100 um^2, and a minimum width transistor
area is 60 L^2 yields a tile area of 84375 MWTAs.
Routing at W=300 is 30481 MWTAs, leaving us with a total of 53000 MWTAs for logic block area
This means that only 37% of our area is in the general routing, and 63% is inside the logic
block. Note that the crossbar / local interconnect is considered part of the logic block
area in this analysis. That is a lower proportion of of routing area than most academics
assume, but note that the total routing area really includes the crossbar, which would push
routing area up significantly, we estimate into the ~70% range.
-->
<pb_type name="clb">
<input name="I" num_pins="12" equivalent="full"/>
<input name="reset" num_pins="1"/>
<output name="O" num_pins="8" equivalent="none"/>
<clock name="clk" num_pins="1"/>
<!-- Describe fracturable logic element.
Each fracturable logic element has a 6-LUT that can alternatively operate as two 5-LUTs with shared inputs.
The outputs of the fracturable logic element can be optionally registered
-->
<pb_type name="fle" num_pb="4">
<input name="in" num_pins="4"/>
<input name="reset" num_pins="1"/>
<output name="out" num_pins="2"/>
<clock name="clk" num_pins="1"/>
<!-- Physical mode definition begin (physical implementation of the fle) -->
<mode name="physical" disable_packing="true">
<pb_type name="fabric" num_pb="1">
<input name="in" num_pins="4"/>
<input name="reset" num_pins="1"/>
<output name="out" num_pins="2"/>
<clock name="clk" num_pins="1"/>
<pb_type name="frac_logic" num_pb="1">
<input name="in" num_pins="4"/>
<output name="out" num_pins="2"/>
<!-- Define LUT -->
<pb_type name="frac_lut4" blif_model=".subckt frac_lut4" num_pb="1">
<input name="in" num_pins="4"/>
<output name="lut3_out" num_pins="2"/>
<output name="lut4_out" num_pins="1"/>
</pb_type>
<interconnect>
<direct name="direct1" input="frac_logic.in" output="frac_lut4.in"/>
<direct name="direct2" input="frac_lut4.lut3_out[1]" output="frac_logic.out[1]"/>
<!-- Xifan Tang: I use out[0] because the output of lut6 in lut6 mode is wired to the out[0] -->
<mux name="mux1" input="frac_lut4.lut4_out frac_lut4.lut3_out[0]" output="frac_logic.out[0]"/>
</interconnect>
</pb_type>
<!-- Define flip-flop -->
<pb_type name="ff" blif_model=".subckt dffr" num_pb="2">
<input name="D" num_pins="1" port_class="D"/>
<input name="R" num_pins="1"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="ff.D" clock="C"/>
<T_setup value="66e-12" port="ff.R" clock="C"/>
<T_clock_to_Q max="124e-12" port="ff.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="fabric.in" output="frac_logic.in"/>
<direct name="direct2" input="frac_logic.out[1:0]" output="ff[1:0].D"/>
<complete name="direct3" input="fabric.clk" output="ff[1:0].C"/>
<complete name="direct4" input="fabric.reset" output="ff[1:0].R"/>
<mux name="mux1" input="ff[0].Q frac_logic.out[0]" output="fabric.out[0]">
<!-- LUT to output is faster than FF to output on a Stratix IV -->
<delay_constant max="25e-12" in_port="frac_logic.out[0]" out_port="fabric.out[0]"/>
<delay_constant max="45e-12" in_port="ff[0].Q" out_port="fabric.out[0]"/>
</mux>
<mux name="mux2" input="ff[1].Q frac_logic.out[1]" output="fabric.out[1]">
<!-- LUT to output is faster than FF to output on a Stratix IV -->
<delay_constant max="25e-12" in_port="frac_logic.out[1]" out_port="fabric.out[1]"/>
<delay_constant max="45e-12" in_port="ff[1].Q" out_port="fabric.out[1]"/>
</mux>
</interconnect>
</pb_type>
<interconnect>
<direct name="direct1" input="fle.in" output="fabric.in"/>
<direct name="direct2" input="fabric.out" output="fle.out"/>
<direct name="direct3" input="fle.clk" output="fabric.clk"/>
<direct name="direct4" input="fle.reset" output="fabric.reset"/>
</interconnect>
</mode>
<!-- Physical mode definition end (physical implementation of the fle) -->
<!-- Dual 3-LUT mode definition begin -->
<mode name="n2_lut3">
<pb_type name="lut3inter" num_pb="1">
<input name="in" num_pins="3"/>
<input name="reset" num_pins="1"/>
<output name="out" num_pins="2"/>
<clock name="clk" num_pins="1"/>
<pb_type name="ble3" num_pb="2">
<input name="in" num_pins="3"/>
<input name="reset" num_pins="1"/>
<output name="out" num_pins="1"/>
<clock name="clk" num_pins="1"/>
<!-- Define the LUT -->
<pb_type name="lut3" blif_model=".names" num_pb="1" class="lut">
<input name="in" num_pins="3" port_class="lut_in"/>
<output name="out" num_pins="1" port_class="lut_out"/>
<!-- LUT timing using delay matrix -->
<!-- These are the physical delay inputs on a Stratix IV LUT but because VPR cannot do LUT rebalancing,
we instead take the average of these numbers to get more stable results
82e-12
173e-12
261e-12
263e-12
398e-12
-->
<delay_matrix type="max" in_port="lut3.in" out_port="lut3.out">
235e-12
235e-12
235e-12
</delay_matrix>
</pb_type>
<!-- Define the flip-flop -->
<pb_type name="ff" num_pb="1">
<input name="D" num_pins="1"/>
<input name="R" num_pins="1"/>
<output name="Q" num_pins="1"/>
<clock name="C" num_pins="1"/>
<mode name="latch">
<pb_type name="latch" blif_model=".latch" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="clk" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="latch.D" clock="clk"/>
<T_clock_to_Q max="124e-12" port="latch.Q" clock="clk"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="latch.D"/>
<direct name="direct2" input="ff.C" output="latch.clk"/>
<direct name="direct3" input="latch.Q" output="ff.Q"/>
</interconnect>
</mode>
<mode name="dff">
<pb_type name="dff" blif_model=".subckt dff" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="dff.D" clock="C"/>
<T_clock_to_Q max="124e-12" port="dff.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="dff.D"/>
<direct name="direct2" input="ff.C" output="dff.C"/>
<direct name="direct3" input="dff.Q" output="ff.Q"/>
</interconnect>
</mode>
<mode name="dffr">
<pb_type name="dffr" blif_model=".subckt dffr" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<input name="R" num_pins="1"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="dffr.D" clock="C"/>
<T_setup value="66e-12" port="dffr.R" clock="C"/>
<T_clock_to_Q max="124e-12" port="dffr.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="dffr.D"/>
<direct name="direct2" input="ff.C" output="dffr.C"/>
<direct name="direct3" input="ff.R" output="dffr.R"/>
<direct name="direct4" input="dffr.Q" output="ff.Q"/>
</interconnect>
</mode>
<mode name="dffrn">
<pb_type name="dffrn" blif_model=".subckt dffrn" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<input name="RN" num_pins="1"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="dffrn.D" clock="C"/>
<T_setup value="66e-12" port="dffrn.RN" clock="C"/>
<T_clock_to_Q max="124e-12" port="dffrn.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="dffrn.D"/>
<direct name="direct2" input="ff.C" output="dffrn.C"/>
<direct name="direct3" input="ff.R" output="dffrn.RN"/>
<direct name="direct4" input="dffrn.Q" output="ff.Q"/>
</interconnect>
</mode>
</pb_type>
<interconnect>
<direct name="direct1" input="ble3.in[2:0]" output="lut3[0:0].in[2:0]"/>
<direct name="direct2" input="lut3[0:0].out" output="ff[0:0].D">
<!-- Advanced user option that tells CAD tool to find LUT+FF pairs in netlist -->
<pack_pattern name="ble3" in_port="lut3[0:0].out" out_port="ff[0:0].D"/>
</direct>
<direct name="direct3" input="ble3.clk" output="ff[0:0].C"/>
<direct name="direct4" input="ble3.reset" output="ff[0:0].R"/>
<mux name="mux1" input="ff[0:0].Q lut3.out[0:0]" output="ble3.out[0:0]">
<!-- LUT to output is faster than FF to output on a Stratix IV -->
<delay_constant max="25e-12" in_port="lut3.out[0:0]" out_port="ble3.out[0:0]"/>
<delay_constant max="45e-12" in_port="ff[0:0].Q" out_port="ble3.out[0:0]"/>
</mux>
</interconnect>
</pb_type>
<interconnect>
<direct name="direct1" input="lut3inter.in" output="ble3[0:0].in"/>
<direct name="direct2" input="lut3inter.in" output="ble3[1:1].in"/>
<direct name="direct3" input="ble3[1:0].out" output="lut3inter.out"/>
<complete name="complete1" input="lut3inter.clk" output="ble3[1:0].clk"/>
<complete name="complete2" input="lut3inter.reset" output="ble3[1:0].reset"/>
</interconnect>
</pb_type>
<interconnect>
<direct name="direct1" input="fle.in[2:0]" output="lut3inter.in"/>
<direct name="direct2" input="lut3inter.out" output="fle.out"/>
<direct name="direct3" input="fle.clk" output="lut3inter.clk"/>
<direct name="direct4" input="fle.reset" output="lut3inter.reset"/>
</interconnect>
</mode>
<!-- Dual 3-LUT mode definition end -->
<!-- 4-LUT mode definition begin -->
<mode name="n1_lut4">
<!-- Define 4-LUT mode -->
<pb_type name="ble4" num_pb="1">
<input name="in" num_pins="4"/>
<input name="reset" num_pins="1"/>
<output name="out" num_pins="1"/>
<clock name="clk" num_pins="1"/>
<!-- Define LUT -->
<pb_type name="lut4" blif_model=".names" num_pb="1" class="lut">
<input name="in" num_pins="4" port_class="lut_in"/>
<output name="out" num_pins="1" port_class="lut_out"/>
<!-- LUT timing using delay matrix -->
<!-- These are the physical delay inputs on a Stratix IV LUT but because VPR cannot do LUT rebalancing,
we instead take the average of these numbers to get more stable results
82e-12
173e-12
261e-12
263e-12
398e-12
397e-12
-->
<delay_matrix type="max" in_port="lut4.in" out_port="lut4.out">
261e-12
261e-12
261e-12
261e-12
</delay_matrix>
</pb_type>
<!-- Define the flip-flop -->
<pb_type name="ff" num_pb="1">
<input name="D" num_pins="1"/>
<input name="R" num_pins="1"/>
<output name="Q" num_pins="1"/>
<clock name="C" num_pins="1"/>
<mode name="latch">
<pb_type name="latch" blif_model=".latch" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="clk" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="latch.D" clock="clk"/>
<T_clock_to_Q max="124e-12" port="latch.Q" clock="clk"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="latch.D"/>
<direct name="direct2" input="ff.C" output="latch.clk"/>
<direct name="direct3" input="latch.Q" output="ff.Q"/>
</interconnect>
</mode>
<mode name="dff">
<pb_type name="dff" blif_model=".subckt dff" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="dff.D" clock="C"/>
<T_clock_to_Q max="124e-12" port="dff.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="dff.D"/>
<direct name="direct2" input="ff.C" output="dff.C"/>
<direct name="direct3" input="dff.Q" output="ff.Q"/>
</interconnect>
</mode>
<mode name="dffr">
<pb_type name="dffr" blif_model=".subckt dffr" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<input name="R" num_pins="1"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="dffr.D" clock="C"/>
<T_setup value="66e-12" port="dffr.R" clock="C"/>
<T_clock_to_Q max="124e-12" port="dffr.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="dffr.D"/>
<direct name="direct2" input="ff.C" output="dffr.C"/>
<direct name="direct3" input="ff.R" output="dffr.R"/>
<direct name="direct4" input="dffr.Q" output="ff.Q"/>
</interconnect>
</mode>
<mode name="dffrn">
<pb_type name="dffrn" blif_model=".subckt dffrn" num_pb="1">
<input name="D" num_pins="1" port_class="D"/>
<input name="RN" num_pins="1"/>
<output name="Q" num_pins="1" port_class="Q"/>
<clock name="C" num_pins="1" port_class="clock"/>
<T_setup value="66e-12" port="dffrn.D" clock="C"/>
<T_setup value="66e-12" port="dffrn.RN" clock="C"/>
<T_clock_to_Q max="124e-12" port="dffrn.Q" clock="C"/>
</pb_type>
<interconnect>
<direct name="direct1" input="ff.D" output="dffrn.D"/>
<direct name="direct2" input="ff.C" output="dffrn.C"/>
<direct name="direct3" input="ff.R" output="dffrn.RN"/>
<direct name="direct4" input="dffrn.Q" output="ff.Q"/>
</interconnect>
</mode>
</pb_type>
<interconnect>
<direct name="direct1" input="ble4.in" output="lut4[0:0].in"/>
<direct name="direct2" input="lut4.out" output="ff.D">
<!-- Advanced user option that tells CAD tool to find LUT+FF pairs in netlist -->
<pack_pattern name="ble4" in_port="lut4.out" out_port="ff.D"/>
</direct>
<direct name="direct3" input="ble4.clk" output="ff.C"/>
<direct name="direct4" input="ble4.reset" output="ff.R"/>
<mux name="mux1" input="ff.Q lut4.out" output="ble4.out">
<!-- LUT to output is faster than FF to output on a Stratix IV -->
<delay_constant max="25e-12" in_port="lut4.out" out_port="ble4.out"/>
<delay_constant max="45e-12" in_port="ff.Q" out_port="ble4.out"/>
</mux>
</interconnect>
</pb_type>
<interconnect>
<direct name="direct1" input="fle.in" output="ble4.in"/>
<direct name="direct2" input="ble4.out" output="fle.out[0:0]"/>
<direct name="direct3" input="fle.clk" output="ble4.clk"/>
<direct name="direct4" input="fle.reset" output="ble4.reset"/>
</interconnect>
</mode>
<!-- 6-LUT mode definition end -->
</pb_type>
<interconnect>
<!-- We use a full crossbar to get logical equivalence at inputs of CLB
The delays below come from Stratix IV. the delay through a connection block
input mux + the crossbar in Stratix IV is 167 ps. We already have a 72 ps
delay on the connection block input mux (modeled by Ian Kuon), so the remaining
delay within the crossbar is 95 ps.
The delays of cluster feedbacks in Stratix IV is 100 ps, when driven by a LUT.
Since all our outputs LUT outputs go to a BLE output, and have a delay of
25 ps to do so, we subtract 25 ps from the 100 ps delay of a feedback
to get the part that should be marked on the crossbar. -->
<complete name="crossbar" input="clb.I fle[3:0].out clb.reset" output="fle[3:0].in">
<delay_constant max="95e-12" in_port="clb.I clb.reset" out_port="fle[3:0].in"/>
<delay_constant max="75e-12" in_port="fle[3:0].out" out_port="fle[3:0].in"/>
</complete>
<complete name="clks" input="clb.clk" output="fle[3:0].clk">
</complete>
<complete name="resets" input="clb.reset" output="fle[3:0].reset">
</complete>
<!-- This way of specifying direct connection to clb outputs is important because this architecture uses automatic spreading of opins.
By grouping to output pins in this fashion, if a logic block is completely filled by 6-LUTs,
then the outputs those 6-LUTs take get evenly distributed across all four sides of the CLB instead of clumped on two sides (which is what happens with a more
naive specification).
-->
<direct name="clbouts1" input="fle[3:0].out[0:0]" output="clb.O[3:0]"/>
<direct name="clbouts2" input="fle[3:0].out[1:1]" output="clb.O[7:4]"/>
</interconnect>
<!-- Every input pin is driven by 15% of the tracks in a channel, every output pin is driven by 10% of the tracks in a channel -->
<!-- Place this general purpose logic block in any unspecified column -->
</pb_type>
<!-- Define general purpose logic block (CLB) ends -->
</complexblocklist>
</architecture>