5 ms·
How about a proprietary lower-resolution FP16 floating point instruction set that might speed up your AI inference?
by fxtentacle 7y ago
How about a proprietary lower-resolution FP16 floating point instruction set that might speed up your AI inference?
- _msw_ 7y agoThe FP16 instruction implementation comes from Arm Neoverse N1 cores, and is standard Armv8.2-FP16. https://en.wikichip.org/wiki/arm/armv8#ARMv8_Extensions_and_Processor_Features https://en.wikichip.org/wiki/arm/armv8#ARMv8_Extensions_and_...
- borramakot 7y agoIs this a criticism of the ARM chips, or of the TPU's bfloat format?