-
Hi,
I am using libfabric on a Cray with Xeon Phis and gnixu_get_pe_from_ip fails. I am not trying to run libfabric on the Phis. The problem is on the host cpu.
On the host, the IP address is assigne…
-
```
What steps will reproduce the problem?
1. create attached group groups/phi/TLB.txt
2. build
What is the expected output? What do you see instead?
build warning:
===> GENERATE GROUP HEADERS
Use …
-
I noticed with Isaac, that Cuda code with `__constant__` variables cannot be directly ported to Alpaka. Passing a const parameter to a kernel is slower than directly using `__constant__` variables.
S…
-
Hi all and thanks for such a great FS!
I saw 1eeb4562a72ab29345572609e1e4315ecd26c5a1 the other day and was intrigued. Unfortunately, my NAS only supports SSSE3, so I backported the algorithm. The mo…
-
I see #12 - but this topic did not lay out OpenCL support - Ben Sander in comments @ http://gpuopen.com/rocm-do-you-speaka-my-language/ says OpenCL is soon, but this was May 5th (4 months ago from thi…
-
```
The concatenation of features in the src/cpuid.c could lead to broken strings.
The trivial patch is enclosed.
What steps will reproduce the problem?
1. running valgrind with likwid.
What versio…
-
```
Port likwid tools to Intel Xeon Phi processor
```
Original issue reported on code.google.com by `jan.trei...@gmail.com` on 16 Nov 2012 at 9:20
-
```
What steps will reproduce the problem?
1. create attached group groups/phi/TLB.txt
2. build
What is the expected output? What do you see instead?
build warning:
===> GENERATE GROUP HEADERS
Use …
-
I'm getting around 100 GFLOPS using the dgemm.py example for 4096x4096 matrices vs ~300 GFLOPS reported in http://www.dlr.de/sc/Portaldata/15/Resources/dokumente/pyhpc2014/submissions/pyhpc2014_submis…
-
```
The concatenation of features in the src/cpuid.c could lead to broken strings.
The trivial patch is enclosed.
What steps will reproduce the problem?
1. running valgrind with likwid.
What versio…