You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+9-6Lines changed: 9 additions & 6 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -34,31 +34,34 @@ For more info on build options/flags when using the install script, use `./insta
34
34
RCCL build & installation helper script
35
35
Options:
36
36
--address-sanitizer Build with address sanitizer enabled
37
-
-d|--dependencies Install RCCL depdencencies
37
+
-c|--enable-code-coverage Enable code coverage
38
+
-d|--dependencies Install RCCL dependencies
38
39
--debug Build debug library
39
40
--enable_backtrace Build with custom backtrace support
40
41
--disable-colltrace Build without collective trace
41
42
--disable-msccl-kernel Build without MSCCL kernels
42
43
--disable-mscclpp Build without MSCCL++ support
44
+
--enable-mscclpp-clip Build MSCCL++ with clip wrapper on bfloat16 and half addition routines
45
+
--disable-roctx Build without ROCTX logging
43
46
-f|--fast Quick-build RCCL (local gpu arch only, no backtrace, and collective trace support)
44
47
-h|--help Prints this help message
45
48
-i|--install Install RCCL library (see --prefix argument below)
46
49
-j|--jobs Specify how many parallel compilation jobs to run ($nproc by default)
47
50
-l|--local_gpu_only Only compile forlocal GPU architecture
48
-
--amdgpu_targets Only compile for specified GPU architecture(s). For multiple targets, seperate by ';' (builds for all supported GPU architectures by default)
51
+
--amdgpu_targets Only compile for specified GPU architecture(s). For multiple targets, separate by ';' (builds for all supported GPU architectures by default)
49
52
--no_clean Don't delete files if they already exist
50
53
--npkit-enable Compile with npkit enabled
54
+
--log-trace Build with log trace enabled (i.e. NCCL_DEBUG=TRACE)
51
55
--openmp-test-enable Enable OpenMP in rccl unit tests
52
-
--roctx-enable Compile with roctx enabled (example usage: rocprof --roctx-trace ./rccl-program)
53
56
-p|--package_build Build RCCL package
54
57
--prefix Specify custom directory to install RCCL to (default: `/opt/rocm`)
55
-
--rm-legacy-include-dir Remove legacy include dir Packaging added for file/folder reorg backward compatibility
56
58
--run_tests_all Run all rccl unit tests (must be built already)
57
59
-r|--run_tests_quick Run small subset of rccl unit tests (must be built already)
58
60
--static Build RCCL as a static library instead of shared library
59
61
-t|--tests_build Build rccl unit tests, but do not run
60
62
--time-trace Plot the build time of RCCL (requires `ninja-build` package installed on the system)
61
63
--verbose Show compile commands
64
+
--force-reduce-pipeline Force reduce_copy sw pipeline to be used for every reduce-based collectives and datatypes
62
65
```
63
66
64
67
By default, RCCL builds for all GPU targets defined in `DEFAULT_GPUS` in `CMakeLists.txt`. To target specific GPU(s), and potentially reduce build time, use `--amdgpu_targets` as a `;` separated string listing GPU(s) to target.
0 commit comments