leezu commented on a change in pull request #20266: URL: https://github.com/apache/incubator-mxnet/pull/20266#discussion_r632023242
########## File path: config/distribution/linux_aarch64_cpu.cmake ########## @@ -30,17 +30,13 @@ set(USE_F16C OFF CACHE BOOL "Build with x86 F16C instruction support") set(USE_DIST_KVSTORE OFF CACHE BOOL "Build with DIST_KVSTORE support") set(USE_MKLDNN ON CACHE BOOL "Build with MKL-DNN support") -# Uncomment the following line to build MKLDNN with APL support -# APL can be downloaded from https://developer.arm.com/tools-and-software/server-and-hpc/downloads/arm-performance-libraries -# APL needs to be added to LD_LIBRARY_PATH -## set(DNNL_BLAS_VENDOR “ARMPL” CACHE STRING “Build MKLDNN with Arm Performance Libraries as the BLAS library”) -# Uncomment the following lines to build MKLDNN with ACL support -# C++ 14 is requried to build with ACL and MKLDNN recommends building ACL from source rather than -# using the pre-built binaries from https://github.com/ARM-software/ComputeLibrary/releases -# If pre-built binaries are used anyways, make sure to copy and rename the appropriate binaries -# folder from <acl_root>/lib/<binaries_folder> to <acl_root>/build +# Pre-built binaries are available from https://github.com/ARM-software/ComputeLibrary/releases +# Make sure to copy and rename the appropriate binaries folder +# from <acl_root>/lib/<binaries_folder_for_your_arch> to <acl_root>/build Review comment: I see. Is there any reason for MKLDNN discouraging the use of the pre-built objects? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
