Jenkins CI Failure - Vision Test -- OCL

[kiritigowda]

GPU device -gfx1030 Backend - OCL OS - Ubuntu 20.04/CentOS 8

Test Case - /opt/rocm/bin/runvx -frames:1000 -affinity:GPU -dump-profile node org.khronos.openvx.warp_perspective image:1920,1080,U008 "matrix:FLOAT32,3,3:INIT,{2;0.1;0;2;1.9;0;-1200;-360;1}" !NEAREST_NEIGHBOR image:1920,1080,U008

Running OpenVX Node - WarpPerspective_U8_U8_Nearest
runvx 1.0.0
OK: using AMD OpenVX 1.3.0
OK: OpenVX using GPU device#0 (gfx1030) [OpenCL 2.0 ] [SvmCaps 0 0]
csv,HEADER ,STATUS, COUNT,cur-ms,avg-ms,min-ms,clenqueue-ms,clwait-ms,clwrite-ms,clread-ms
Memory access fault by GPU node-8 (Agent handle: 0xca0ca0) on address 0x7f23bf2de000. Reason: Page not present or supervisor privilege.
Aborted (core dumped)

[kiritigowda]

@asalmanp -- need to try the GDB on this. Check CI -- OpenCL Ubuntu 20.04 for the log

[kiritigowda]

@asalmanp to run this test case do the following

$ set +H
$ /opt/rocm/bin/runvx -frames:1000 -affinity:CPU -dump-profile node org.khronos.openvx.warp_perspective image:1920,1080,U008 "matrix:FLOAT32,3,3:INIT,{2;0.1;0;2;1.9;0;-1200;-360;1}" !NEAREST_NEIGHBOR image:1920,1080,U008

Output for CPU

runvx 1.0.0
OK: using AMD OpenVX 1.3.0
csv,HEADER ,STATUS, COUNT,cur-ms,avg-ms,min-ms,clenqueue-ms,clwait-ms,clwrite-ms,clread-ms
csv,OVERALL,  PASS,  1000,      ,  2.81,  2.69,  0.00,  0.00,  0.00,  0.00 (median 2.764)
> total elapsed time:   2.81 sec
> graph profile: 
 COUNT,tmp(ms),avg(ms),min(ms),max(ms),DEV,KERNEL
  1000,  2.852,  2.810,  2.691,  3.631,CPU,GRAPH
  1000,  2.852,  2.810,  2.691,  3.629,CPU,com.amd.openvx.WarpPerspective_U8_U8_Nearest

Output for GPU -- HIP

runvx 1.0.0
OK: using AMD OpenVX 1.3.0
OK: OpenVX using GPU device#0 AMD Radeon PRO W6800 (gfx1030) (with 30 CUs) on PCI bus 0b:00.0
csv,HEADER ,STATUS, COUNT,cur-ms,avg-ms,min-ms,clenqueue-ms,clwait-ms,clwrite-ms,clread-ms
csv,OVERALL,  PASS,  1000,      ,  0.03,  0.03,  0.00,  0.00,  0.00,  0.00 (median 0.029)
> total elapsed time:   0.03 sec
> graph profile: 
 COUNT,tmp(ms),avg(ms),min(ms),max(ms),DEV,KERNEL
  1000,  0.029,  0.031,  0.027,  1.814,GPU,GRAPH
  1000,  0.029,  0.031,  0.027,  1.812,GPU,com.amd.openvx.WarpPerspective_U8_U8_Nearest
OK: HIP buffer usage: 4189988, 3/3

[kiritigowda]

PR #893 -- Test case disable till fixed for GFX 1030 -- @paveltc

[kiritigowda]

Failure reproducible on GFX 906

Running OpenVX Node - WarpPerspective_U8_U8_Nearest
runvx 1.0.0
OK: using AMD OpenVX 1.3.0
OK: OpenVX using GPU device - 0: gfx906:sramecc+:xnack- [OpenCL 2.0 ] [CL_DEVICE_SVM_CAPABILITIES 0 0]
csv,HEADER ,STATUS, COUNT,cur-ms,avg-ms,min-ms,clenqueue-ms,clwait-ms,clwrite-ms,clread-ms
Memory access fault by GPU node-2 (Agent handle: 0x1e599a0) on address 0x7f2076d18000. Reason: Page not present or supervisor privilege.

[kiritigowda]

Running OpenVX Node - Remap_S16_nearest
runvx 1.0.0
OK: using AMD OpenVX 1.3.0
OK: OpenVX using GPU device - 0: gfx1030 [OpenCL 2.0 ] [CL_DEVICE_SVM_CAPABILITIES 0 0]
csv,HEADER ,STATUS, COUNT,cur-ms,avg-ms,min-ms,clenqueue-ms,clwait-ms,clwrite-ms,clread-ms
Memory access fault by GPU node-5 (Agent handle: 0x1e61520) on address 0x7ff093830000. Reason: Page not present or supervisor privilege.
Aborted (core dumped)

[kiritigowda]

Issue no longer observed on TOT Master.