Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
composable_kernel
Commits
59dbb01f
Unverified
Commit
59dbb01f
authored
Oct 02, 2023
by
Illia Silin
Committed by
GitHub
Oct 02, 2023
Browse files
get rid of gfx900/906, set rocm5.7 as default (#958)
parent
9d58c421
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
4 additions
and
4 deletions
+4
-4
CMakeLists.txt
CMakeLists.txt
+2
-2
Dockerfile
Dockerfile
+1
-1
example/60_gemm_multi_ABD/gemm_multi_ABD_xdl_fp16.cpp
example/60_gemm_multi_ABD/gemm_multi_ABD_xdl_fp16.cpp
+1
-1
No files found.
CMakeLists.txt
View file @
59dbb01f
...
...
@@ -106,7 +106,7 @@ message("checking which targets are supported")
#Setting GPU_TARGETS on command line will override this list
if
(
NOT PROFILER_ONLY
)
rocm_check_target_ids
(
DEFAULT_GPU_TARGETS
TARGETS
"
gfx900;gfx906;
gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102"
)
TARGETS
"gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102"
)
else
()
add_definitions
(
-DPROFILER_ONLY
)
set
(
GPU_TARGETS
""
CACHE STRING
""
FORCE
)
...
...
@@ -114,7 +114,7 @@ else()
message
(
FATAL_ERROR
"For PROFILE_ONLY build, please do not set GPU_TARGETS, use GPU_ARCH = gfx90, gfx94, gfx10, or gfx11"
)
endif
()
if
(
GPU_ARCH MATCHES
"gfx90"
)
rocm_check_target_ids
(
DEFAULT_GPU_TARGETS TARGETS
"
gfx900;gfx906;
gfx908;gfx90a"
)
rocm_check_target_ids
(
DEFAULT_GPU_TARGETS TARGETS
"gfx908;gfx90a"
)
elseif
(
GPU_ARCH MATCHES
"gfx94"
)
rocm_check_target_ids
(
DEFAULT_GPU_TARGETS TARGETS
"gfx940;gfx941;gfx942"
)
elseif
(
GPU_ARCH MATCHES
"gfx10"
)
...
...
Dockerfile
View file @
59dbb01f
FROM
ubuntu:20.04
ARG
DEBIAN_FRONTEND=noninteractive
ARG
ROCMVERSION=5.
6
ARG
ROCMVERSION=5.
7
ARG
compiler_version=""
ARG
compiler_commit=""
...
...
example/60_gemm_multi_ABD/gemm_multi_ABD_xdl_fp16.cpp
View file @
59dbb01f
...
...
@@ -74,7 +74,7 @@ struct AddScale
a
=
scale
*
(
a0
+
a1
);
}
//this attribute will force copy_function applying element_wise with vector_type
//
this attribute will force copy_function applying element_wise with vector_type
static
constexpr
ck
::
index_t
vec_len
=
4
;
float
scale
=
1.0
;
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment