Page 1 of 1
Does ESP32 have instruction set to accelerated math computing?
Posted: Thu Nov 15, 2018 8:12 am
by yanbin
Hi,
My project need to do more mathematical operations (e.g. multiplication, division, exponentiation ..).
Does ESP32 have instruction set to accelerated math computing?
Thanks.
Re: Does ESP32 have instruction set to accelerated math computing?
Posted: Thu Nov 15, 2018 1:24 pm
by ESP_Sprite
The ESP32 has integer and (single-precision) floating point multiplication and division; no need to do anything special for those as GCC will automatically spit out the right instructions for that. There's also an 16-bit MAC instruction, but you may need to write some assembly for that.
Re: Does ESP32 have instruction set to accelerated math computing?
Posted: Fri Nov 16, 2018 2:38 am
by yanbin
ESP_Sprite wrote: ↑Thu Nov 15, 2018 1:24 pm
The ESP32 has integer and (single-precision) floating point multiplication and division; no need to do anything special for those as GCC will automatically spit out the right instructions for that. There's also an 16-bit MAC instruction, but you may need to write some assembly for that.
Thank you @ESP_Sprite.
Does ESP32 support some features likes SIMD / NEON ?
Thanks.
Re: Does ESP32 have instruction set to accelerated math computing?
Posted: Fri Nov 16, 2018 3:58 am
by ESP_Sprite
No / hard to say. NEON is ARM-specific and SIMD is kinda vague. I know the MAC instruction is SIMD-ish, though.
Re: Does ESP32 have instruction set to accelerated math computing?
Posted: Fri Nov 16, 2018 8:24 am
by yanbin
ESP_Sprite wrote: ↑Fri Nov 16, 2018 3:58 am
No / hard to say. NEON is ARM-specific and SIMD is kinda vague. I know the MAC instruction is SIMD-ish, though.
Thank you @ESP_Sprite.
Now because of the SIMD/NEON, my applicatio could run fast on the arm platform.
On ESP32, it was a bit slow.
Would you please share any experience about accelerated math computing on ESP32 ?
Thanks.