|
|
@@ -64,7 +64,7 @@ package("mkl")
|
|
|
end
|
|
|
end
|
|
|
|
|
|
- add_configs("threading", {description = "Choose threading modal for mkl.", default = "tbb", type = "string", values = {"tbb", "openmp", "seq"}})
|
|
|
+ add_configs("threading", {description = "Choose threading modal for mkl.", default = "tbb", type = "string", values = {"tbb", "openmp", "gomp", "seq"}})
|
|
|
add_configs("interface", {description = "Choose index integer size for the interface.", default = 32, values = {32, 64}})
|
|
|
|
|
|
on_fetch("fetch")
|
|
|
@@ -73,6 +73,8 @@ package("mkl")
|
|
|
add_syslinks("pthread", "dl")
|
|
|
end
|
|
|
on_load("windows", "macosx", "linux", function (package)
|
|
|
+ -- Refer to [oneAPI Math Kernel Library Link Line Advisor](https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-link-line-advisor.html)
|
|
|
+ -- to get the link option for MKL library.
|
|
|
local suffix = (package:config("interface") == 32 and "lp64" or "ilp64")
|
|
|
if package:config("interface") == 64 then
|
|
|
package:add("defines", "MKL_ILP64")
|
|
|
@@ -91,6 +93,10 @@ package("mkl")
|
|
|
table.insert(flags, "-lmkl_sequential")
|
|
|
elseif threading == "openmp" then
|
|
|
table.insert(flags, "-lmkl_intel_thread")
|
|
|
+ table.insert(flags, "-lomp")
|
|
|
+ elseif threading == "gomp" then
|
|
|
+ table.insert(flags, "-lmkl_gnu_thread")
|
|
|
+ table.insert(flags, "-lgomp")
|
|
|
end
|
|
|
table.insert(flags, "-lmkl_core")
|
|
|
table.insert(flags, "-Wl,--end-group")
|
|
|
@@ -104,7 +110,9 @@ package("mkl")
|
|
|
elseif threading == "seq" then
|
|
|
package:add("links", "mkl_sequential")
|
|
|
elseif threading == "openmp" then
|
|
|
- package:add("links", "mkl_intel_thread")
|
|
|
+ package:add("links", "mkl_intel_thread", "omp")
|
|
|
+ elseif threading == "gomp" then
|
|
|
+ package:add("links", "mkl_gnu_thread", "gomp")
|
|
|
end
|
|
|
package:add("links", "mkl_core")
|
|
|
end
|
Chen Yufei