u99127 commented on a change in pull request #6279:
URL: https://github.com/apache/incubator-tvm/pull/6279#discussion_r470713079
##########
File path: tests/python/contrib/test_arm_compute_lib/infrastructure.py
##########
@@ -25,15 +26,22 @@
from tvm.contrib import graph_runtime
from tvm.relay.op.contrib import arm_compute_lib
from tvm.contrib import util
+from tvm.autotvm.measure import request_remote
class Device:
- """Adjust the following settings to connect to and use a remote device for
tests."""
- use_remote = False
- target = "llvm -mtriple=aarch64-linux-gnu -mattr=+neon"
- # Enable cross compilation when connecting a remote device from a non-arm
platform.
- cross_compile = None
- # cross_compile = "aarch64-linux-gnu-g++"
+ """
+ Use the following environment variables to connect to and use a remote
device.
+ The RPC mechanism can be used in the case of compiling an ACL module on an
x86 machine
+ but running on an AArch64 machine.
+ """
+ rpc_host = os.environ.get("TVM_ARM_COMPUTE_LIB_RPC_HOST", "localhost")
+ rpc_port = int(os.environ.get("TVM_ARM_COMPUTE_LIB_RPC_PORT", 9090))
+ rpc_device_key = os.environ.get("TVM_ARM_COMPUTE_LIB_RPC_DEVICE_KEY", "")
+ target = os.environ.get("TVM_ARM_COMPUTE_LIB_TARGET", "llvm
-mtriple=aarch64-linux-gnu -mattr=+neon")
Review comment:
Yeah we'd need ways of skipping the test on devices that aren't
supported.
----------------------------------------------------------------
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]
