| 1 |
Hello all, |
| 2 |
|
| 3 |
This week I mainly focused on dev-libs/rocm-opencl-runtime. |
| 4 |
|
| 5 |
I bumped dev-libs/rocm-opencl-runtime to 5.1.3. That's relatively easy. |
| 6 |
The difficult part is enabling its tests. I came across a major problem, |
| 7 |
which is oclgl test requiring X server. I compiled using debug options |
| 8 |
and use gdb to dive into the code, but found there is no simple |
| 9 |
solution. Currently the test needs a X server where OpenGL vender is |
| 10 |
AMD. Xvfb only provides llvmpipe, not meeting the requirements. I |
| 11 |
consulted some friends, they said NVIDIA recommends using EGL when there |
| 12 |
is no X [1], but apparently ROCm can only get OpenGL from X [2]. So my |
| 13 |
workaround is to let user passing an X display into the ebuild, by |
| 14 |
reading the environment variable OCLGL_DISPLAY (DISPLAY variable will be |
| 15 |
wiped when calling emerge, while this can survive). If no display is |
| 16 |
detected, or glxinfo shows the OpenGL vendor is not AMD, then src_test |
| 17 |
dies, throwing indications about running an X server using amdgpu |
| 18 |
driver. |
| 19 |
|
| 20 |
I was also trapped by CRLF problem in src_test of |
| 21 |
dev-libs/rocm-opencl-runtime. Tests in oclperf.exclude should be skipped |
| 22 |
for oclperf test, but it did not. After numerous trials, I finally found |
| 23 |
that this file is using CRLF, not LF, which causes the exclusion failed |
| 24 |
:-( |
| 25 |
|
| 26 |
Nevertheless, rocm-opencl-runtime tests passed on Radeon RX 6700XT! A |
| 27 |
good thing, because I know many user in Gentoo rely on this package to |
| 28 |
provide opencl in their computation, and the correctness is vital. |
| 29 |
Before we does not have src_test enabled. The PR is now in [6]. |
| 30 |
|
| 31 |
Other works including starting wiki writing [3,4], refine rocm.eclass |
| 32 |
according to feedback (not much, see gentoo-dev mailing list), and found |
| 33 |
a bug of dev-util/hip -- FindHIP.cmake module is not in the correct |
| 34 |
place. Fix can be found in [5] but I need to further polish the patch |
| 35 |
before PR. |
| 36 |
|
| 37 |
If no further suggestions on rocm.eclass, I'll land rocm.eclass in |
| 38 |
::gentoo next week, and start bumping the sci-libs version already done |
| 39 |
locally. |
| 40 |
|
| 41 |
[1] https://developer.nvidia.com/blog/egl-eye-opengl-visualization-without-x-server/ |
| 42 |
[2] https://github.com/RadeonOpenCompute/ROCm-OpenCL-Runtime/blob/bbdc87e08b322d349f82bdd7575c8ce94d31d276/tests/ocltst/module/common/OCLGLCommonLinux.cpp |
| 43 |
[3] https://wiki.gentoo.org/wiki/ROCm |
| 44 |
[4] https://wiki.gentoo.org/wiki/HIP |
| 45 |
[5] https://github.com/littlewu2508/gentoo/tree/hip-correct-cmake |
| 46 |
[6] https://github.com/gentoo/gentoo/pull/26870 |
| 47 |
|
| 48 |
Best wishes, |
| 49 |
-- |
| 50 |
Yiyang Wu |
Archives