issues
search
c3sr
/
comm_scope
NUMA-aware multi-CPU multi-GPU data transfer benchmarks
https://github.com/c3sr/scope
Apache License 2.0
21
stars
3
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
`hipMemcpyHostToDevice` -> `hipMemcpyAsyncDeviceToDevice`
#47
cwpearson
opened
1 year ago
0
Remove `-Werror`
#46
cwpearson
closed
1 year ago
1
Let's remove `-Wfatal-errors` from the flags
#45
cwpearson
closed
1 year ago
1
added more combinations for gputogpu transfers
#44
ZaidQureshi
closed
2 years ago
0
PCIe A100 can deliver incorrect results when GPU is idle due to low clocks
#43
cwpearson
opened
3 years ago
0
ThetaGPU: error in src/cudaMemcpyPeerAsync_Duplex_GPUGPUPeer: a PTX JIT compilation failed
#42
cwpearson
opened
3 years ago
0
NUMA node boundary benchmarks?
#41
rlerdorf
closed
3 years ago
2
use cudaEvent to measure empty kernel time
#40
cwpearson
opened
4 years ago
0
add NVSHMEM benchmarks
#39
cwpearson
opened
4 years ago
0
Break benchmarks out into latency and bandwidth
#38
cwpearson
opened
4 years ago
0
How to fix the number of iterations in a given micro-benchmark?
#37
Palwisha-18
closed
4 years ago
5
Undefined when `USE_NUMA != 1`
#36
cwpearson
closed
4 years ago
0
Undefined when `USE_NUMA != 1`
#35
cwpearson
closed
4 years ago
0
Make sure data is random to foil compression
#34
cwpearson
opened
4 years ago
0
Allow comm_scope targets to build without being in scope
#33
cwpearson
closed
4 years ago
1
Unknown CMake command "sugar_include".
#32
Yiltan
closed
4 years ago
4
`nvcc fatal: Unknown option '-pthread' in CMake 3.17.2
#31
cwpearson
closed
1 year ago
1
How to give size as an argument for comm_scope?
#30
Palwisha-18
closed
4 years ago
2
Any lessons to be learned from EasyPerf?
#29
cwpearson
opened
5 years ago
1
Investigate using cudaLaunchHostFunc for getting wall time when a stream operation ends
#28
cwpearson
opened
5 years ago
0
sync should clobber memory
#27
cwpearson
closed
5 years ago
1
mfence should clobber memory
#26
cwpearson
closed
5 years ago
1
use cudaEventWaitStream for multi-device duplex transfers
#25
cwpearson
closed
5 years ago
3
create zero-copy H2D
#24
cwpearson
closed
5 years ago
1
prefetch-duplex CPU/GPU should probably occur in two different streams
#23
cwpearson
closed
5 years ago
1
prefetch-duplex GPU/GPU may be able to associate both streams with a single device
#22
cwpearson
opened
5 years ago
0
Anything we should learn from NVIDIA/multi-gpu-programming-models
#21
cwpearson
opened
5 years ago
0
rename UM_Coherence to UM_Demand
#20
cwpearson
closed
5 years ago
1
Create a rai_build.yml file for reproducibility.
#19
cwpearson
closed
4 years ago
1
Error if numa is not found
#18
cwpearson
closed
5 years ago
1
nvcc 7.5 has incompatibility with spdlog 0.16.3-p1
#17
cwpearson
opened
5 years ago
0
add multi-threaded explicit transfer benchmarks
#16
cwpearson
opened
5 years ago
1
Flush caches in unified memory host-to-gpu
#15
cwpearson
opened
5 years ago
0
stack-smashing error during do_after_inits
#14
cwpearson
closed
4 years ago
1
zero-copy GPU/GPU should enable bidirectional peer access and free memory
#13
cwpearson
closed
5 years ago
0
Add a CUDA 10 docker image
#12
cwpearson
closed
1 year ago
1
dcbf should clobber memory
#11
cwpearson
closed
5 years ago
1
Use clflush with "+m" output operand?
#10
cwpearson
closed
5 years ago
0
This numa/wr.cpp argument no longer exists
#9
cwpearson
closed
6 years ago
0
Use aligned_alloc in numa-memcpy/pinned_to_gpu
#8
cwpearson
closed
6 years ago
0
Use aligned_alloc in numa-memcpy/host_to_pinned
#7
cwpearson
closed
6 years ago
0
programatically register benchmarks based on available CUDA and NUMA devices
#6
cwpearson
closed
5 years ago
0
Clean up memcpy/gpu_to_gpu_nopeer
#5
cwpearson
closed
6 years ago
1
Clean up memcpy/pinned_to_gpu
#4
cwpearson
closed
6 years ago
1
Clean up memcpy/host_to_gpu
#3
cwpearson
closed
6 years ago
1
Causes scope to crash during benchmark registration on non-gpu systems
#2
cwpearson
closed
6 years ago
1
Graceful handling on non-NUMA systems
#1
cwpearson
closed
6 years ago
1