Himax WEI int8 build not float32

Anyone else finding out that the WEI-I only works with “int8” not “float32” settings before building the final firmware?

Here is the error message I get with float32.

Creating job... OK (ID: 472426)

Writing templates OK

Scheduling job in cluster...
Job started
Compiling EON model...
Compiling EON model OK

Removing clutter...
Removing clutter OK

Copying output...
Copying output OK

Job started
Building firmware...
-- Configuring done
-- Generating done
-- Build files have been written to: /app/firmware/build
Scanning dependencies of target firmware.elf
[  1%] Building CXX object CMakeFiles/firmware.elf.dir/ei_run_impulse.cpp.o
[  2%] Building CXX object CMakeFiles/firmware.elf.dir/tflite-model/trained_model_compiled.cpp.o
[  3%] Linking CXX executable firmware.elf
[ldac] "../memory.lcf",line 72: Section .data (address range=0x20069700..202023ff)
|	falls outside its designated MEMORY region SYSTEM0(0x20020800..0x2010a2ef)
|	by 1016080 (0xf8110) bytes.
make[2]: *** [CMakeFiles/firmware.elf.dir/build.make:2383: firmware.elf] Error 1
make[1]: *** [CMakeFiles/Makefile2:123: CMakeFiles/firmware.elf.dir/all] Error 2
make: *** [Makefile:103: all] Error 2

Application exited with code 2 (Error)

Job failed (see above)

```.

@Rocksetta, the model does not fit in flash.

[ldac] "../memory.lcf",line 72: Section .data (address range=0x20069700..202023ff)
|	falls outside its designated MEMORY region SYSTEM0(0x20020800..0x2010a2ef)
|	by 1016080 (0xf8110) bytes.

Which makes sense, given that the float32 model probably is 4x larger.

1 Like