The info and logs even on a successful build would be pretty heavy. I used
output to the standard out to debug the SolverFactory in its current form.
I've recently updated them in the PR to be more intuitive what you should see
on a successful build.
mvn clean install -Pviennacl -Phadoop2 -DskipTests && cd Spark && mvn test
should show info (something like):
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
So having a few fallbacks to OpenMP is expected, since the test cluster is set
up to run a spark local context with 3 threads by default, and you only have
one card on your machine.
Since the goal is to have seamless integration, if there are several partitions
of a DRM operand and you're preforming %*%, the SolverFactory will attempt
first to run on the GPU, which should be successful (unless its still running
from a previous MMul. but for the second partition, the second thread will
fail to start a context since the GPU device is busy, and it will attempt to
run on the CPU using OpenMP.
This should always run on the CPU and not need to dump off to the JVM. (with a
few exceptions, eg in the case of an all zero sparse matrix)
.. except in the case of H2o, which cant make use of any of VCL Features.
I am noticing several of the type of failures that you're seeing since I
rebased last week. I am not sure whats going on and exactly what changed.
I'll try to figure it out...
________________________________
From: Andrew Musselman <andrew.mussel...@gmail.com>
Sent: Sunday, October 16, 2016 7:59:55 PM
To: dev@mahout.apache.org
Subject: Trying out PR 261
I'm not sure my setup is correct; on Ubuntu 16.04 I'm getting a firehose of
infos and warns about not finding the GPUMMul and OMPMMul classes. I've
installed libviennacl-dev 1.7.1 through apt-get. Any ideas?
Here's repro:
$ git fetch origin pull/261/head:viennacl-opmmul-a
$ mvn clean install -Pviennacl -Phadoop2
Infos and warns (these are littered consistently throughout test logs):
%*% B mapBlock {}
- C = A %*% B incompatible B keys
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Unable to create class OMPMMul: falling back to java version
- Spark-specific C = At %*% B , join
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Unable to create class OMPMMul: falling back to java version
- C = At %*% B , join, String-keyed
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Unable to create class OMPMMul: falling back to java version
- C = At %*% B , zippable, String-keyed
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Unable to create class OMPMMul: falling back to java version
- C = A %*% B.t
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Unable to create class OMPMMul: falling back to java version
{
0 => {0:26.0,1:35.0,2:46.0,3:51.0}
1 => {0:50.0,1:69.0,2:92.0,3:105.0}
2 => {0:62.0,1:86.0,2:115.0,3:132.0}
3 => {0:74.0,1:103.0,2:138.0,3:159.0}
}
- C = A %*% inCoreB
[INFO] Creating org.apache.mahout.viennacl.opencl.GPUMMul solver
[WARN] Unable to create class GPUMMul: attempting OpenMP version
[INFO] Creating org.apache.mahout.viennacl.openmp.OMPMMul solver
[INFO] Unable to create class OMPMMul: falling back to java version
{
0 => {0:26.0,1:35.0,2:46.0,3:51.0}
1 => {0:50.0,1:69.0,2:92.0,3:105.0}
2 => {0:62.0,1:86.0,2:115.0,3:132.0}
3 => {0:74.0,1:103.0,2:138.0,3:159.0}
}
Output of `clinfo`:
$ clinfo
Number of platforms 1
Platform Name NVIDIA CUDA
Platform Vendor NVIDIA Corporation
Platform Version OpenCL 1.2 CUDA 8.0.0
Platform Profile FULL_PROFILE
Platform Extensions
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 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_nv_copy_opts
Platform Extensions function suffix NV
Platform Name NVIDIA CUDA
Number of devices 1
Device Name GeForce GTX 750 Ti
Device Vendor NVIDIA Corporation
Device Vendor ID 0x10de
Device Version OpenCL 1.2 CUDA
Driver Version 367.44
Device OpenCL C Version OpenCL C 1.2
Device Type GPU
Device Profile FULL_PROFILE
Device Topology (NV) PCI-E, 01:00.0
Max compute units 5
Max clock frequency 1150MHz
Compute Capability (NV) 5.0
Device Partition (core)
Max number of sub-devices 1
Supported partition types None
Max work item dimensions 3
Max work item sizes 1024x1024x64
Max work group size 1024
Preferred work group size multiple 32
Warp size (NV) 32
Preferred / native vector sizes
char 1 / 1
short 1 / 1
int 1 / 1
long 1 / 1
half 0 / 0 (n/a)
float 1 / 1
double 1 / 1
(cl_khr_fp64)
Half-precision Floating-point support (n/a)
Single-precision Floating-point support (core)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Correctly-rounded divide and sqrt operations Yes
Double-precision Floating-point support (cl_khr_fp64)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Correctly-rounded divide and sqrt operations No
Address bits 64, Little-Endian
Global memory size 2095841280 (1.952GiB)
Error Correction support No
Max memory allocation 523960320 (499.7MiB)
Unified memory for Host and Device No
Integrated memory (NV) No
Minimum alignment for any data type 128 bytes
Alignment of base address 4096 bits (512 bytes)
Global Memory cache type Read/Write
Global Memory cache size 81920
Global Memory cache line 128 bytes
Image support Yes
Max number of samplers per kernel 32
Max size for 1D images from buffer 134217728 pixels
Max 1D or 2D image array size 2048 images
Max 2D image size 16384x16384 pixels
Max 3D image size 4096x4096x4096 pixels
Max number of read image args 256
Max number of write image args 16
Local memory type Local
Local memory size 49152 (48KiB)
Registers per block (NV) 65536
Max constant buffer size 65536 (64KiB)
Max number of constant args 9
Max size of kernel argument 4352 (4.25KiB)
Queue properties
Out-of-order execution Yes
Profiling Yes
Prefer user sync for interop No
Profiling timer resolution 1000ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels No
Kernel execution timeout (NV) Yes
Concurrent copy and kernel execution (NV) Yes
Number of async copy engines 1
printf() buffer size 1048576 (1024KiB)
Built-in kernels
Device Available Yes
Compiler Available Yes
Linker Available Yes
Device Extensions
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 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_nv_copy_opts
NULL platform behavior
clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...) NVIDIA CUDA
clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...) Success [NV]
clCreateContext(NULL, ...) [default] Success [NV]
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU) No devices found in
platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU) No platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR) No devices
found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM) No devices found in
platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL) No platform
ICD loader properties
ICD loader Name OpenCL ICD Loader
ICD loader Vendor OCL Icd free software
ICD loader Version 2.2.8
ICD loader Profile OpenCL 1.2
NOTE: your OpenCL library declares to support OpenCL 1.2,
but it seems to support up to OpenCL 2.1 too.
