Compare commits

...

7 Commits

Author SHA1 Message Date
David Galiffi
22572a9857 Add TransferBench and hipBLAS-common 2024-12-09 18:37:20 -05:00
David Galiffi
794b34f40e Update default.xml
Fixed merge conflicts
2024-12-09 11:18:39 -05:00
David Galiffi
25ef417b31 Merge branch 'develop' into JeniferC99-patch-1 2024-12-09 11:17:17 -05:00
Peter Park
78f9adc6ec fix rccl hip streams section in workload tuning guide (#4140) 2024-12-09 11:06:12 -05:00
David Galiffi
4abcae54a8 Update default.xml (#4136)
Add rocJPEG
Rename omniperf to rocprofiler-compute
Rename omnitrace to rocprofiler-compute
2024-12-09 10:56:07 -05:00
JeniferC99
2690506e64 Update default.xml
SWDEV-502858
Rename  Omnitrace and Omniperf
2024-12-06 23:01:56 -06:00
darren-amd
3dffe1998a External CI: add aomp dependency to rocprofiler-sdk (#4135)
External CI: Add aomp dependency for rocprofiler-sdk
2024-12-06 16:20:09 -05:00
3 changed files with 10 additions and 8 deletions

View File

@@ -41,6 +41,7 @@ parameters:
- ROCR-Runtime
- rocprofiler-register
- roctracer
- aomp
jobs:
- job: rocprofilersdk

View File

@@ -8,10 +8,7 @@
<!--list of projects for ROCm-->
<project name="ROCK-Kernel-Driver" />
<project name="ROCR-Runtime" />
<project name="ROCT-Thunk-Interface" />
<project name="amdsmi" />
<project name="omniperf" />
<project name="omnitrace" />
<project name="rdc" />
<project name="rocm_bandwidth_test" />
<project name="rocm_smi_lib" />
@@ -21,6 +18,8 @@
<project name="rocprofiler" />
<project name="rocprofiler-register" />
<project name="rocprofiler-sdk" />
<project name="rocprofiler-compute" />
<project name="rocprofiler-systems" />
<project name="roctracer" />
<!--HIP Projects-->
<project name="HIP" />
@@ -42,6 +41,7 @@
<project groups="mathlibs" name="ROCmValidationSuite" />
<project groups="mathlibs" name="Tensile" />
<project groups="mathlibs" name="composable_kernel" />
<project groups="mathlibs" name="hipBLAS-common" />
<project groups="mathlibs" name="hipBLAS" />
<project groups="mathlibs" name="hipBLASLt" />
<project groups="mathlibs" name="hipCUB" />
@@ -57,6 +57,7 @@
<project groups="mathlibs" name="rocALUTION" />
<project groups="mathlibs" name="rocBLAS" />
<project groups="mathlibs" name="rocDecode" />
<project groups="mathlibs" name="rocJPEG" />
<project groups="mathlibs" name="rocPyDecode" />
<project groups="mathlibs" name="rocFFT" />
<project groups="mathlibs" name="rocPRIM" />
@@ -67,6 +68,7 @@
<project groups="mathlibs" name="rocWMMA" />
<project groups="mathlibs" name="rocm-cmake" />
<project groups="mathlibs" name="rpp" />
<project groups="mathlibs" name="TransferBench" />
<!-- Projects for OpenMP-Extras -->
<project name="aomp" path="openmp-extras/aomp" />
<project name="aomp-extras" path="openmp-extras/aomp-extras" />

View File

@@ -2062,11 +2062,10 @@ collectives.
Multi-node FSDP and RCCL settings
---------------------------------
It's recommended to use high-priority HIP streams with RCCL.
The simplest way to enable this is by using the nightly PyTorch wheels, as the required changes from
`PR #122830 <https://github.com/pytorch/pytorch/pull/122830>`_ were not included in the PyTorch 2.3
release but are available in the nightly builds.
When using PyTorch's FSDP (Full Sharded Data Parallel) feature, the HIP
streams used by RCCL and HIP streams used for compute kernels do not
always overlap well. As a workaround, it's recommended to use
high-priority HIP streams with RCCL.
To configure high-priority streams: