ARM-software / ComputeLibrary

The Compute Library is a set of computer vision and machine learning functions optimised for both Arm CPUs and GPUs using SIMD technologies.
MIT License
2.81k stars 774 forks source link

CPUInfo::get().has_fp16() returns true on RaspberryPi 4 #1096

Closed alvoron closed 6 months ago

alvoron commented 7 months ago

Platform: Raspberry Pi4 ACL version: 24.02 ACL build command: scons arch=armv7a neon=1 os=linux opencl=0 build=native -j 4 Werror=false fixed_format_kernels=1 multi_isa=0 openmp=0 cppthreads=1

CPUInfo::get().has_fp16() returns true on RaspberryPi 4 (Cortex-A72). I believe, it should be false.

alvoron commented 7 months ago

UPD: The issue is reproducible on 32-bit Debian 11 and not reproducible on 64-bit Debian 11.

morgolock commented 7 months ago

Hi @alvoron

Please see the following patch fixing the problem: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11242

Hope this helps

alvoron commented 6 months ago

@morgolock I implemented the same approach in our code :) As soon as the fix is included in ACL release I'll upgrade ACL version and remove our custom code.

Thank you.