search

viennacl / viennacl-dev

Developer repository for ViennaCL. Visit http://viennacl.sourceforge.net/ for the latest releases.
Other
282 stars 91 forks source link

Several test failures #105

Closed d-meiser closed 9 years ago

d-meiser commented 10 years ago

Hi, I've encountered several test failures with current master:

make test
Running tests...
...
The following tests FAILED:
     34 - matrix_product_float-opencl (Failed)
     35 - matrix_product_double-opencl (Failed)
     36 - blas3_solve-opencl (Failed)
     41 - matrix_vector-opencl (Failed)
     54 - sparse-opencl (OTHER_FAULT)
     65 - libviennacl-blas3 (Failed)
Errors while running CTest
make: *** [test] Error 8
test of viennacl-ser completed with result = 2.

The test logs are here: test logs

Here is the output from viennacl-info:

[dmeiser@ivy ser]$ ./examples/tutorial/viennacl-info 
# =========================================
#         Platform Information             
# =========================================
#
# Vendor and version: NVIDIA Corporation: OpenCL 1.1 CUDA 6.5.14
#
# ViennaCL uses this OpenCL platform by default.
# 
# Available Devices: 
# 

  -----------------------------------------
Address Bits:                  32
Available:                     1
Compiler Available:            1
Endian Little:                 1
Error Correction Support:      0
Execution Capabilities:        CL_EXEC_KERNEL 
Extensions:                    cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll  cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 
Global Mem Cache Size:         212992 Bytes
Global Mem Cache Type:         CL_READ_WRITE_CACHE 
Global Mem Cacheline Size:     128 Bytes
Global Mem Size:               5368512512 Bytes
Host Unified Memory:           0
Image Support:                 1
Image2D Max Height:            32768
Image2D Max Width:             32768
Image3D Max Depth:             4096
Image3D Max Height:            4096
Image3D Max Width:             4096
Local Mem Size:                49151 Bytes
Local Mem Type:                CL_LOCAL 
Max Clock Frequency:           705 MHz
Max Compute Units:             13
Max Constant Args:             9
Max Constant Buffer Size:      65536 Bytes
Max Mem Alloc Size:            1342128128 Bytes
Max Parameter Size:            4352 Bytes
Max Read Image Args:           256
Max Samplers:                  32
Max Work Group Size:           1024
Max Work Item Dimensions:      3
Max Work Item Sizes:           1024 1024 64 
Max Write Image Args:          16
Mem Base Addr Align:           4096
Min Data Type Align Size:      128 Bytes
Name:                          Tesla K20c
Native Vector Width char:      1
Native Vector Width short:     1
Native Vector Width int:       1
Native Vector Width long:      1
Native Vector Width float:     1
Native Vector Width double:    1
Native Vector Width half:      0
OpenCL C Version:              OpenCL C 1.1 
Platform:                      0x9a0a90
Preferred Vector Width char:   1
Preferred Vector Width short:  1
Preferred Vector Width int:    1
Preferred Vector Width long:   1
Preferred Vector Width float:  1
Preferred Vector Width double: 1
Preferred Vector Width half:   0
Profile:                       FULL_PROFILE
Profiling Timer Resolution:    1000 ns
Queue Properties:              CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE CL_QUEUE_PROFILING_ENABLE 
Single FP Config:              CL_FP_DENORM CL_FP_INF_NAN CL_FP_ROUND_TO_NEAREST CL_FP_ROUND_TO_ZERO CL_FP_ROUND_TO_INF CL_FP_FMA 
Type:                          GPU 
Vendor:                        NVIDIA Corporation
Vendor ID:                     4318
Version:                       OpenCL 1.1 CUDA
Driver Version:                340.29
  -----------------------------------------

###########################################
ptillet commented 10 years ago

Yap, this is known and was introduced in this commit: https://github.com/viennacl/viennacl-dev/commit/84dfe7a847f9b2dc0802071f3ffbde202a4a7893 We're trying to fix this...

d-meiser commented 10 years ago

Cool, thanks.

ptillet commented 10 years ago

The GEMM issues should be fixed in: f9b5e24bb58. I also have some failures with sparse-opencl on my nvidia laptop. I've had this bug for a very long time, so I don't close the issue until it is resolved!

d-meiser commented 10 years ago

Great. Thanks for the update.

karlrupp commented 10 years ago

Hi Dominic, these tests should now all succeed. Could you please let me know if this works for you? Thanks!

d-meiser commented 10 years ago

Hi Karl,

With commit 65224e30ba929 the qr_method-opencl test is failing (all others pass). This is on the same machine as above. I recompiled to run this test on the CPU device and got the same result. The log is here:

50/65 Testing: qr_method-opencl
50/65 Test: qr_method-opencl
Command: "/scr_ivyamd/dmeiser/viennacl-dev/build/tests/qr_method-test-opencl"
Directory: /scr_ivyamd/dmeiser/viennacl-dev/build/tests
"qr_method-opencl" start time: Nov 04 11:32 EST
Output:
----------------------------------------------------------
Reading...
Testing row-major matrix of size 9-by-9
Calculation...
Verification...
[[OK]] [9x9] ../examples/testdata/eigen/symm5.example time = 1.2655
tridiagonal = 1, hessenberg = 1 prod-diff = 0.000001 eigen-diff = 0.000002

Reading...
Testing column-major matrix of size 9-by-9
Calculation...
Verification...
[[OK]] [9x9] ../examples/testdata/eigen/symm5.example time = 0.0032
tridiagonal = 1, hessenberg = 1 prod-diff = 0.000001 eigen-diff = 0.000002

Reading...
Testing row-major matrix of size 10-by-10
Calculation...
Verification...
[FAIL] [10x10]  ../examples/testdata/eigen/nsm2.example time = 0.2015
tridiagonal = 1, hessenberg = 1 prod-diff = 1.659658 eigen-diff = 0.000000

<end of output>
Test time =   3.10 sec
----------------------------------------------------------
Test Failed.
"qr_method-opencl" end time: Nov 04 11:32 EST
"qr_method-opencl" time elapsed: 00:00:03
----------------------------------------------------------
karlrupp commented 10 years ago

Thanks, Dominic. I've got one more machine over here which shows exactly this same (one) failure, whereas all the others work fine. So at least I've got confirmation that this is something systematic - cheers!

d-meiser commented 10 years ago

Ok. Let me know if there is anything I can do to help. Cheers, Dominic

karlrupp commented 9 years ago

Summary:

Since there are no other problems in the test suite left (see #108), I close this issue.