Add support for INT8 classification models - #38
Jithendar11 wants to merge 2 commits into
Conversation
Jithendar11
commented
Sep 28, 2026
Jithendar11
commented
- Add CIL library for INT8 model pre-processing and post-processing
- Add INT8 model_pte header file
- Add INT8 model report file
- Add CIL library for INT8 model pre-processing and post-processing - Add INT8 model_pte header file - Add INT8 model report file
baf5ddb to
d895e77
Compare
| // <i> Defines size of RTSS HP application memory region. | ||
| // <i> Default: 0x00200000 | ||
| #define APP_MRAM_HP_SIZE 0x00200000 | ||
| #define APP_MRAM_HP_SIZE 0x00400000 |
There was a problem hiding this comment.
Setting 0x00400000 would exceed the available MRAM size as MRAM size is 5.5 MB.
Since 0x00200000 is reserved for HE core it leaves the maximum of 0x00380000 for the HP core.
There was a problem hiding this comment.
Thanks for pointing this out. We increased this because some models have larger generated model/weight data and the previous HP MRAM allocation was not sufficient.
In this application we are not using the HE core, but I agree that with the current memory map APP_MRAM_HP_BASE starts after the HE-reserved 0x00200000 region, so setting HP size to 0x00400000 would exceed the available 5.5 MB MRAM range.
Updated APP_MRAM_HP_SIZE to 0x00380000, which is the maximum valid HP size with the current base address and HE reservation.
2 participants
