nverke commented on code in PR #13721:
URL: https://github.com/apache/tvm/pull/13721#discussion_r1110184718
##########
include/tvm/meta_schedule/schedule_rule.h:
##########
@@ -210,6 +210,31 @@ class ScheduleRule : public runtime::ObjectRef {
Optional<Array<Integer>> vector_load_lens, Optional<Map<String,
ObjectRef>> reuse_read,
Optional<Map<String, ObjectRef>> reuse_write, bool
use_software_pipeline);
+ /*!
+ * \brief Extension of MultiLevelTiling for auto-tensorization with multiple
groups of candidate
+ * tensor core intrinsics
+ * \param intrin_groups A list of groups of tensor core intrinsics. The map
should contain key
+ * "compute" which represents the tensor intrin for computation. The value
of the map should be
+ * names of tensor intrinsics, must be registered via
+ * TensorIntrin.register(...) beforehand
+ * \param structure The tiling structure. Recommended:
+ * - 'SRSRS' on Hexagon
+ * \param tile_binds For each level of tiles, which thread axis it is bound
to. These are not
+ * supported on hexagon.
+ * \param max_innermost_factor The maximum size of the innermost factor.
NullOpt means no limit
+ * \param vector_load_lens The length of vector lane in vectorized
cooperative fetching.
Review Comment:
These have two separate uses, vector_load_lens is used for vector loads
outside of the tiling loops where as max_innermost_factor is for loop tiling.
As for the naming and usage I am not 100% sure.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
