anijain2305 commented on a change in pull request #5980:
URL: https://github.com/apache/incubator-tvm/pull/5980#discussion_r454011323
##########
File path: topi/python/topi/arm_cpu/tensor_intrin.py
##########
@@ -451,3 +451,55 @@ def _instr(index):
return te.decl_tensor_intrin(
C.op, _intrin_func, binds={data:a_buffer, kernel:b_buffer},
default_buffer_params=buffer_params)
+
+def _qmuls_arm(op):
+ """
+ Implementation of qmuls through arm intrinsics sqrdmulh and srshl
+ when q == 31.
+
+ Please note that this is introducing a small round-up error for
+ some corner cases. This is because we are rounding twice instead
+ than only once. I.e.:
+
+ * original qmuls: round(x*y*2^-s)
+ * arm qmuls: round(round(x*y)*2^-s)
+ """
+ x = op.args[0]
+ y = op.args[1]
+ q = op.args[2]
+ s = op.args[3]
+
+ # Don't use this intrinsic if we don't have a int32x4 vector
+ # and if we are not multiplying q31 numbers
+ if x.dtype != "int32x4" and q == 31:
Review comment:
Can you please double check the condition? Should there be `or` here?
----------------------------------------------------------------
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:
us...@infra.apache.org
