0025264: Mesh very slow for Revol shape

(0050486)
git (administrator)
2016-02-08 15:52

Branch CR25264 has been created by oan.

SHA-1: 4bbba407a83264ae4b34853c1a88b2876120e119

Detailed log of new commits:

Author: oan
Date: Mon Feb 8 15:51:59 2016 +0300

    0025264: Mesh very slow for Revol shape

    Test case added

(0050487)
oan (developer)
2016-02-08 15:56

Dear Mikhail,

on current master meshing of the problematic shape takes about 2 seconds instead of 52 seconds mentioned in the bug. Difference between OCCT version 6.2 is due to changed meshing algorithm.

Test case has been pushed to branch CR25264. Please review.

(0050489)
msv (developer)
2016-02-08 16:10

Reviewed.

(0050503)
abv (manager)
2016-02-08 19:52

Oleg, the problem does exist: I have tried with OCCT 6.0, and it takes 0.2 sec, while on current master it works ~ 13 sec.

Note that it is not correct to rely on time of vdisplay command, as it may use different parameters in different versions.

Use incmesh command instead:

tclean result; chrono s reset; chrono s start; incmesh result 0.1; chrono s show

For me, it takes 0.05 sec (33 K triangles) in OCCT 6.0 and 2.5 sec (31 K triangles) in master.
For tolerance 0.001, it works 2.3 sec in OCCT 6.0, and 112 sec in master.

As we see, slow down is ~ 50 times regardless of deflection (and corresponding number of elements), and I believe that difference of the algorithm should not be blamed without in-depth analysis.

(0050513)
abv (manager)
2016-02-09 11:14

More timings, incmesh with deflection 0.1:

OCCT 6.0.0: 0.05 sec, 33030 triangles
OCCT 6.3.1: 14.4 sec, 32480 triangles
OCCT 6.5.0: 2.25 sec, 32624 triangles
OCCT 6.6.0: 2.37 sec, 31208 triangles
OCCT 6.7.0: 2.21 sec, 31208 triangles
OCCT 6.8.0: 1.9 sec, 31208 triangles
OCCT 6.9.0: 2.02 sec, 31090 triangles
OCCT 6.9.1: 2.01 sec, 31090 triangles
master : 2.02 sec, 31090 triangles

(0058773)
git (administrator)
2016-10-17 12:21

Branch CR25264_1 has been created by oan.

SHA-1: 9c651b4d32e64d4fcfef1e086018781035743c98

Detailed log of new commits:

Author: oan
Date: Mon Oct 17 12:21:25 2016 +0300

    0025264: Mesh very slow for Revol shape

    Test case added

(0058774)
oan (developer)
2016-10-17 12:27

Dear Bugmaster,

behaviour of BRepMesh component has been improved for case of thin cylindrical surfaces in context of #27442 issue.

The following timings used in test suites are produced on current master:
CPU user time: 0.0468003 sec for reference case
CPU user time: 0.0468003 sec for problematic case

Test case for current issue has been created and pushed to branch CR25264_1. Please test and close the issue.

Best regards.

(0059119)
git (administrator)
2016-10-25 20:02

Branch CR25264_1 has been updated by apn.

SHA-1: 9bc559e702a0a9e7025881c17b09a8de5b0ca574

Detailed log of new commits:

Author: apn
Date: Tue Oct 25 20:01:42 2016 +0300

Small corrections of test case

(0059120)
apn (administrator)
2016-10-25 20:09

Test case:
bugs mesh bug25264 - OK

(0059397)
git (administrator)
2016-10-28 21:37

Branch CR25264_1 has been deleted by kgv.

SHA-1: 9bc559e702a0a9e7025881c17b09a8de5b0ca574

(0059598)
git (administrator)
2016-10-28 21:51

Branch CR25264 has been deleted by kgv.

SHA-1: 4bbba407a83264ae4b34853c1a88b2876120e119

Related Changesets
occt: master f204ec4c Timestamp: 2016-10-27 14:18:18 Author: oan Committer: apn [ Details ] [ Diff ]	0025264: Mesh very slow for Revol shape Test case added Small corrections of test case
	add - tests/bugs/mesh/bug25264	[ Diff ] [ File ]

Issue History
Date Modified	Username	Field	Change
2014-09-23 14:35	eryar	New Issue
2014-09-23 14:35	eryar	Assigned To	=> oan
2014-09-23 14:35	eryar	File Added: mesh_revol.PNG
2014-09-23 14:35	eryar	File Added: wire.brep
2014-09-23 14:36	eryar	File Added: revol.tcl
2016-02-08 15:52	git	Note Added: 0050486
2016-02-08 15:55	oan	Assigned To	oan => msv
2016-02-08 15:55	oan	Status	new => resolved
2016-02-08 15:55	oan	Steps to Reproduce Updated	View Revisions
2016-02-08 15:55	oan	Steps to Reproduce Updated	View Revisions
2016-02-08 15:56	oan	Note Added: 0050487
2016-02-08 16:10	msv	Note Added: 0050489
2016-02-08 16:10	msv	Assigned To	msv => bugmaster
2016-02-08 16:10	msv	Status	resolved => reviewed
2016-02-08 17:05	apv	Assigned To	bugmaster => apv
2016-02-08 19:52	abv	Note Added: 0050503
2016-02-08 19:52	abv	Assigned To	apv => oan
2016-02-08 19:52	abv	Status	reviewed => assigned
2016-02-08 19:52	abv	Target Version	=> 7.1.0
2016-02-09 11:14	abv	Note Added: 0050513
2016-10-17 11:59	oan	Relationship added	duplicate of 0027442
2016-10-17 12:21	git	Note Added: 0058773
2016-10-17 12:27	oan	Note Added: 0058774
2016-10-17 12:28	oan	Assigned To	oan => bugmaster
2016-10-17 12:28	oan	Status	assigned => feedback
2016-10-25 20:02	git	Note Added: 0059119
2016-10-25 20:09	apn	Note Added: 0059120
2016-10-25 20:10	apn	Test case number	=> bugs mesh bug25264
2016-10-25 20:10	apn	Status	feedback => tested
2016-10-28 14:35	apn	Changeset attached	=> occt master f204ec4c
2016-10-28 14:35	apn	Assigned To	bugmaster => apn
2016-10-28 14:35	apn	Status	tested => verified
2016-10-28 14:35	apn	Resolution	open => fixed
2016-10-28 21:37	git	Note Added: 0059397
2016-10-28 21:51	git	Note Added: 0059598
2016-12-09 16:30	aiv	Status	verified => closed
2016-12-09 16:39	aiv	Fixed in Version	=> 7.1.0

Notes
(0050486) git (administrator) 2016-02-08 15:52	Branch CR25264 has been created by oan. SHA-1: 4bbba407a83264ae4b34853c1a88b2876120e119 Detailed log of new commits: Author: oan Date: Mon Feb 8 15:51:59 2016 +0300 0025264: Mesh very slow for Revol shape Test case added

(0050487) oan (developer) 2016-02-08 15:56	Dear Mikhail, on current master meshing of the problematic shape takes about 2 seconds instead of 52 seconds mentioned in the bug. Difference between OCCT version 6.2 is due to changed meshing algorithm. Test case has been pushed to branch CR25264. Please review.

(0050489) msv (developer) 2016-02-08 16:10	Reviewed.

(0050503) abv (manager) 2016-02-08 19:52	Oleg, the problem does exist: I have tried with OCCT 6.0, and it takes 0.2 sec, while on current master it works ~ 13 sec. Note that it is not correct to rely on time of vdisplay command, as it may use different parameters in different versions. Use incmesh command instead: tclean result; chrono s reset; chrono s start; incmesh result 0.1; chrono s show For me, it takes 0.05 sec (33 K triangles) in OCCT 6.0 and 2.5 sec (31 K triangles) in master. For tolerance 0.001, it works 2.3 sec in OCCT 6.0, and 112 sec in master. As we see, slow down is ~ 50 times regardless of deflection (and corresponding number of elements), and I believe that difference of the algorithm should not be blamed without in-depth analysis.

(0050513) abv (manager) 2016-02-09 11:14	More timings, incmesh with deflection 0.1: OCCT 6.0.0: 0.05 sec, 33030 triangles OCCT 6.3.1: 14.4 sec, 32480 triangles OCCT 6.5.0: 2.25 sec, 32624 triangles OCCT 6.6.0: 2.37 sec, 31208 triangles OCCT 6.7.0: 2.21 sec, 31208 triangles OCCT 6.8.0: 1.9 sec, 31208 triangles OCCT 6.9.0: 2.02 sec, 31090 triangles OCCT 6.9.1: 2.01 sec, 31090 triangles master : 2.02 sec, 31090 triangles

(0058773) git (administrator) 2016-10-17 12:21	Branch CR25264_1 has been created by oan. SHA-1: 9c651b4d32e64d4fcfef1e086018781035743c98 Detailed log of new commits: Author: oan Date: Mon Oct 17 12:21:25 2016 +0300 0025264: Mesh very slow for Revol shape Test case added

(0058774) oan (developer) 2016-10-17 12:27	Dear Bugmaster, behaviour of BRepMesh component has been improved for case of thin cylindrical surfaces in context of #27442 issue. The following timings used in test suites are produced on current master: CPU user time: 0.0468003 sec for reference case CPU user time: 0.0468003 sec for problematic case Test case for current issue has been created and pushed to branch CR25264_1. Please test and close the issue. Best regards.

(0059119) git (administrator) 2016-10-25 20:02	Branch CR25264_1 has been updated by apn. SHA-1: 9bc559e702a0a9e7025881c17b09a8de5b0ca574 Detailed log of new commits: Author: apn Date: Tue Oct 25 20:01:42 2016 +0300 Small corrections of test case

(0059120) apn (administrator) 2016-10-25 20:09	Test case: bugs mesh bug25264 - OK

(0059397) git (administrator) 2016-10-28 21:37	Branch CR25264_1 has been deleted by kgv. SHA-1: 9bc559e702a0a9e7025881c17b09a8de5b0ca574

(0059598) git (administrator) 2016-10-28 21:51	Branch CR25264 has been deleted by kgv. SHA-1: 4bbba407a83264ae4b34853c1a88b2876120e119

Relationships