u99127 commented on a change in pull request #4274: [µTVM] Enable AutoTVM for
ARM STM32F746XX Boards
URL: https://github.com/apache/incubator-tvm/pull/4274#discussion_r349797868
##########
File path: src/runtime/micro/micro_session.h
##########
@@ -174,29 +204,24 @@ class MicroSession : public ModuleNode {
/*! \brief array of memory allocators for each on-device section */
std::shared_ptr<MicroSectionAllocator>
section_allocators_[static_cast<size_t>(SectionKind::kNumKinds)];
- /*! \brief total number of bytes of usable device memory for this session */
- size_t memory_size_;
- /*! \brief uTVM runtime binary info */
- BinaryInfo runtime_bin_info_;
- /*! \brief path to uTVM runtime source code */
- std::string runtime_binary_path_;
- /*! \brief offset of the runtime entry function */
- DevBaseOffset utvm_main_symbol_;
- /*! \brief offset of the runtime exit breakpoint */
- DevBaseOffset utvm_done_symbol_;
+ /*! \brief number of bytes in a word on the target device */
+ size_t word_size_;
+ /*! \brief whether the target device requires a thumb-mode bit on function
addresses
+ *
+ * ARM and other manufacturers use the lowest bit of a function address to
determine
+ * whether it's a "thumb mode" function. The Thumb ISA is more restricted,
but
+ * results in more compact binaries.
+ */
+ bool thumb_mode_;
Review comment:
For Cortex-M based devices, the only variant of the AArch32 ISA supported is
Thumb (i.e. T16 / T32) so that's always default. Thus I don't think you need to
worry about this. It's only on older cores (i.e. CPUs implementing Armv6
(except Armv6T2), Armv4t, Armv5t*, or with cores from other profiles that you
need to worry between A32 and T16 / T32.)
----------------------------------------------------------------
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]
With regards,
Apache Git Services
