search

microsoft / OpenCLOn12

The OpenCL-on-D3D12 mapping layer
MIT License
108 stars 15 forks source link

FAHBench running error.. #23

Closed oscarbg closed 2 years ago

oscarbg commented 3 years ago

Hi, trying to run FAHBench (from Folding@Home) (https://github.com/pandegroup/fahbench/releases/download/2.3.1/FAHBench-2.3.1-win64.zip)

FAHBench has a mode for benchmarking double precision support.. so a motivation to support cl_khr_fp64 if possible.. at least my Titan V should support fast double precision ops even in D3D12..

anyway running it fails for no reason (full log below):

fahbench-cmd --platform-id 2
Something went wrong:
Force RMSE error of 282.431 with threshold of 5

running with disable accuracy check: fahbench-cmd --platform-id 2 --disable-accuracy-check

Starting Benchmark
Progress:   9.0%  estimate:   2.56 ns/day
Something went wrong:
Particle coordinate is nan

log:

fahbench-cmd --platform-id 2
FAHBench Simulation
-------------------
Plugin directory: "E:/a\FAHBench-2.3.1-win64\openmm"
Work unit: dhfr
WU Name: Dihydrofolate reductase
WU Description: A common system for benchmarking molecular dynamics
System XML: E:/a\FAHBench-2.3.1-win64\workunits\dhfr\system.xml
Integrator XML: E:/a\FAHBench-2.3.1-win64\workunits\dhfr\integrator.xml
State XML: E:/a\FAHBench-2.3.1-win64\workunits\dhfr\state.xml
Step chunk: 40
Device ID 0; Platform OpenCL; Platform ID 2
Run length: 60s

Loading plugins from plugin directory
Number of registered plugins: 3
Deserializing input files: system
Deserializing input files: state
Deserializing input files: integrator
Creating context (may take several minutes)
Checking accuracy against reference code
Creating reference context (may take several minutes)
Comparing forces and energy

Something went wrong:
Force RMSE error of 282.431 with threshold of 5
fahbench-cmd --platform-id 2 --disable-accuracy-check
FAHBench Simulation
-------------------
Plugin directory: "E:/a\FAHBench-2.3.1-win64\openmm"
Work unit: dhfr
WU Name: Dihydrofolate reductase
WU Description: A common system for benchmarking molecular dynamics
System XML: E:/a\FAHBench-2.3.1-win64\workunits\dhfr\system.xml
Integrator XML: E:/a\FAHBench-2.3.1-win64\workunits\dhfr\integrator.xml
State XML: E:/a\FAHBench-2.3.1-win64\workunits\dhfr\state.xml
Step chunk: 40
Device ID 0; Platform OpenCL; Platform ID 2
Run length: 60s

Loading plugins from plugin directory
Number of registered plugins: 3
Deserializing input files: system
Deserializing input files: state
Deserializing input files: integrator
Creating context (may take several minutes)
Starting Benchmark
Progress:   9.0%  estimate:   2.56 ns/day
Something went wrong:
Particle coordinate is nan
jenatali commented 2 years ago

This is fixed at https://github.com/microsoft/OpenCLOn12/commit/ddedec09aa6d403f14f4dee3a234993923327784, the first time I tried to run it this is what I hit, after fixing this it's passing for me, so I suspect it was actually fixed by a different change, perhaps https://github.com/microsoft/OpenCLOn12/commit/a7ac03579cc180ac8204d1d50f95a2ef361ef6c9.