zhuwenxi commented on issue #7246:
URL: https://github.com/apache/tvm/issues/7246#issuecomment-784712542
The function looks pretty much like the fix I proposed. In my proposal
"reallocate the stack in parallel for loop", the function looks like this:
<pre>
fn myfunc() {
stack_tcode = @tir.tvm_stack_alloca("arg_tcode", 8)
stack_value = @tir.tvm_stack_alloca("arg_value", 8)
for i in range(10):
stack_tcode = @tir.tvm_stack_alloca("arg_tcode", 8)
// Do reallocation if current loop is parallel
stack_value = @tir.tvm_stack_alloca("arg_value", 8)
// Do reallocation if current loop is parallel
tir.tvm_call_packed_lowered("tvm.contrib.cblas.matmul", stack_1)
}
</pre>
So from this point of view, is it true that the only difference between
"packed_arg_alloca" and "tvm_stack_alloca" is the former one uses thread-local
storage allocation?
----------------------------------------------------------------
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]
