How to correctly use the -coordinates in write_spef?

[shinkuan]

Hello, I'm currently using ORFS and I want to get coordinates of the nodes in the 6_final.spef.

Using platform nangate45 with design aes that comes with ORFS.

I've read #3390 and #3434

I've tried to use the -coordinates in write_spef, but the result spef file still doesn't contain anything like
*N node_name *C x_coord y_coord

I modified the
https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/blob/89a2d1fc68e8df18a658d07d868c4bbbb8ad124c/flow/scripts/final_report.tcl#L31
to

  log_cmd write_spef -coordinates $::env(RESULTS_DIR)/6_final.spef

And confirmed the command was executed successfully in the logs:

write_spef -coordinates ./results/nangate45/aes/base/6_final.spef
[INFO RCX-0443] 16589 nets finished
[INFO PSM-0040] All shapes on net VDD are connected.
########## IR report #################
Net              : VDD
Corner           : default
Supply voltage   : 1.10e+00 V
Worstcase voltage: 1.04e+00 V
Average voltage  : 1.07e+00 V
Average IR drop  : 2.71e-02 V
Worstcase IR drop: 5.97e-02 V
Percentage drop  : 5.43 %
######################################
[INFO PSM-0040] All shapes on net VSS are connected.
########## IR report #################
Net              : VSS
Corner           : default
Supply voltage   : 0.00e+00 V
Worstcase voltage: 5.27e-02 V
Average voltage  : 2.40e-02 V
Average IR drop  : 2.40e-02 V
Worstcase IR drop: 5.27e-02 V
Percentage drop  : 4.79 %
######################################
Cell type report:                       Count       Area
  Fill cell                             16779   30636.82
  Tap cell                                411     109.33
  Clock buffer                            124     152.68
  Timing Repair Buffer                   2850    3051.82
  Inverter                                526     351.65
  Clock inverter                           63      59.05
  Sequential cell                         530    2453.58
  Multi-Input combinational cell        11488   15970.11
  Total                                 32771   52785.04
Report metrics stage 6, finish...

I've investigated the source code and found that:

OpenROAD/src/rcx/src/OpenRCX.tcl

Lines 153 to 174 in 118204e

	proc write_spef { args } {
	sta::parse_key_args "write_spef" args \
	keys { -net_id -nets } \
	flags { -coordinates }
	sta::check_argc_eq1 "write_spef" $args
	set spef_file $args
	set nets ""
	if { [info exists keys(-nets)] } {
	set nets $keys(-nets)
	}
	set net_id 0
	if { [info exists keys(-net_id)] } {
	set net_id $keys(-net_id)
	}
	set coordinates [info exists flags(-coordinates)]
	rcx::write_spef $spef_file $nets $net_id $coordinates
	}

Will call the below:

OpenROAD/src/rcx/src/ext.cpp

Lines 255 to 296 in 118204e

	void Ext::write_spef(const SpefOptions& options)
	{
	_ext->setBlockFromChip();
	if (options.end) {
	_ext->writeSPEF(true);
	return;
	}
	const char* name = options.ext_corner_name;
	uint netId = options.net_id;
	if (netId > 0) {
	_ext->writeSPEF(netId,
	options.single_pi,
	options.debug,
	options.corner,
	name,
	spef_version_);
	return;
	}
	_ext->writeSPEF((char*) options.file,
	(char*) options.nets,
	options.no_name_map,
	(char*) options.N,
	options.term_junction_xy,
	options.cap_units,
	options.res_units,
	options.gz,
	options.stop_after_map,
	options.w_clock,
	options.w_conn,
	options.w_cap,
	options.w_cc_cap,
	options.w_res,
	options.no_c_num,
	false,
	options.single_pi,
	options.no_backslash,
	options.corner,
	name,
	spef_version_,
	options.parallel);
	}

Finally, at:

OpenROAD/src/rcx/src/extSpef.cpp

Lines 1043 to 1073 in 118204e

	void extSpef::writeNodeCoords(const uint netId, dbSet<dbRSeg>& rSet)
	{
	const int dbunit = _block->getDbUnitsPerMicron();
	const double db2nm = 1.0 / ((double) dbunit);
	//*N *1:2 *C 3.06500 125.815
	//*N *1:3 *C 3.03000 120.555
	//*N *2:4 *C 3.07000 120.190
	//*N *2:5 *C 3.07000 120.190
	for (dbRSeg* rc : rSet) {
	const uint shapeId = rc->getShapeId();
	if (!_foreign && shapeId == 0) {
	continue;
	}
	dbCapNode* capNode = dbCapNode::getCapNode(_block, rc->getTargetNode());
	if (capNode->isITerm() || capNode->isBTerm()) {
	continue;
	}
	fprintf(_outFP, "*N ");
	writeCapNode(rc->getTargetNode(), netId);
	int x1, y1;
	rc->getCoords(x1, y1);
	fprintf(_outFP, "*C %f %f\n", db2nm * x1, db2nm * y1);
	}
	}

Will write out the *N node_name *C x_coord y_coord

But

OpenROAD/src/rcx/src/extSpef.cpp

Lines 1499 to 1502 in 118204e

	_writingNodeCoords = C_NONE;
	if (nodeCoord && nodeCoord[0] != '\0') {
	_writingNodeCoords = C_ON;
	}

OpenROAD/src/rcx/src/extSpef.cpp

Lines 1151 to 1153 in 118204e

	if (_writingNodeCoords == C_ON) {
	writeNodeCoords(netId, rcSet);
	}

OpenROAD/src/rcx/src/extSpef.cpp

Lines 1182 to 1184 in 118204e

	if (_writingNodeCoords == C_ON) {
	writeNodeCoords(netId, rcSet);
	}

OpenROAD/src/rcx/src/extSpef.cpp

Lines 1030 to 1032 in 3f01e9f

	if (!_foreign && shapeId == 0) {
	continue;
	}

I think one of them is preventing the RCX executing writeNodeCoords, either (nodeCoord && nodeCoord[0] != '\0') is false or !_foreign && shapeId == 0.

Thanks in advance for your time and assistance!

[shinkuan]

Don't know if this might be the cause of issue, but I found that:

OpenROAD/src/rcx/src/ext.i

Lines 79 to 96 in 118204e

	void
	write_spef(const char* file,
	const char* nets,
	int net_id,
	bool coordinates)
	{
	Ext* ext = getOpenRCX();
	SpefOptions opts;
	opts.file = file;
	// opts.corner= corner;
	opts.nets = nets;
	opts.net_id = net_id;
	opts.coordinates= coordinates;
	ext->write_spef(opts);
	}

sets opts.coordinates= coordinates;

But

OpenROAD/src/rcx/src/ext.cpp

Lines 278 to 299 in 3f01e9f

	_ext->writeSPEF((char*) options.file,
	(char*) options.nets,
	options.no_name_map,
	(char*) options.N,
	options.term_junction_xy,
	options.cap_units,
	options.res_units,
	options.gz,
	options.stop_after_map,
	options.w_clock,
	options.w_conn,
	options.w_cap,
	options.w_cc_cap,
	options.w_res,
	options.no_c_num,
	false,
	options.single_pi,
	options.no_backslash,
	options.corner,
	name,
	spef_version_,
	options.parallel);

doesn't use opts.coordinates at all

[shinkuan]

Figured it out, it was indeed ext.i's problem.
I've submited a PR to fix this. #8932