This sample application would launch a kernel on every available GPU, so
it's possible to monitor activities on every GPU. This could be used to help
verify roctracer is working properly on a multi-GPU scenario.
Temporary workaround for 2.10 release. RCCL, compiler, or firmware
must be corrected and this code reverted before another ASIC release.
Change-Id: I27851353289b93df9acb72d28b8c6ccb9f7f7d7a
We should be using bin/clang, not the build/lightining/bin/clang since
build/ is the project's internal build directory. This patch corrects
this where possible.
However, lightning does not install all it's end user files under out/
so some headers can not be found anywhere in out/ in an incremental build.
This header (opencl-c.h) if fetched out of the lightning source tree if
necessary.
Change-Id: I083d8b27bb39dd615fba3bb0711a789318f95e77
v2: Remove useless AcquireMem packet after the dispatch
v3: Minimum poll_interval is 4
Change-Id: I352eb21c781ed9e03d62c0febd532da6a9854afa
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
v2: optimize_ace_offload_mode=1, recommended by firmware team
Change-Id: Ia54e37242b4eaaf631c35e61a59f03ee0f85ca35
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
This patch adds the support for gfx90c apu. So far we treat it as "dgpu" and
gfx900. Will update hsa gfxip table while the isa/llvm is implemented on gfx90c.
Change-Id: I6ef164bf3e751fe6dd6287cac212a500dce84b1a
Signed-off-by: Huang Rui <ray.huang@amd.com>
The test actually tested all available SDMA queues, so change the name
to reflect the fact.
Change-Id: Ia23df3e5ac79b692b0b60194b05603ba8dd897a4
Signed-off-by: Yong Zhao <Yong.Zhao@amd.com>
New version of script generates all expected fields for KFD API. I checked all structs and made sure nested structs and unions, named or anonymous are handled properly.
If NUMA system no available memory on node 0, mbind will fail on node
0, so set flag NoNUMABind=1 to bypass mbind for all test cases which use
node 0 and allocate system memory.
Change-Id: I7962938ad2bed5a293ca5e6a8500c7f7e15ff453
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
On NUMA system, allocate queue ctx save restore area on the closest NUMA
node to the GPU which the queue is going to run. This will improve
performance on NUMA system generally by reducing schedule latency and
fix the multi-node rccl-tests unstable performance issue.
If the closest NUMA node has no memory available, set flags NoNUMABind=1
to bypass mbind, to use default NUMA memory policy to allocate system
memory.
Change-Id: Ic62bfa5bb2efbf4f6ae79ff403e9610ddf18d45c
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Evgeny
eshcherb
Wen-Heng (Jack) Chung
Sean Keely
Felix Kuehling
Yong Zhao
Huang Rui
Philip Cox
Rachida Kebichi
rkebichi
Laurent Morichetti
Chris Freehill
Philip Yang
Ramesh Errabolu