| ??? 09/03/08 11:22 Read: times |
#157925 - Is really MAC so nice for 8bit MCU? Responding to: ???'s previous message |
For 8-bit MCU, just MAC unit is not enough, like SiLabs C8051F12x.
On this chip, multi-bytes parameters and data are transferred to the MAC unit over the MCU core instructions, one byte by one. It takes more instruction cycles than the multiplication or barrel shift on the MAC. Then, the throughput is determined not by the MAC, but by the MCU core speed. I wonder why SiLabs didn't assign DMA to the MAC. Real DSP has dedicated bus for this purpose, other than the bus for the MCU core. Then, don't expect so much for the MAC unit on 8bit MCU, unless it has any speed-up mechanism for data loading/ unloading to the MAC. Tsuneo |
| Topic | Author | Date |
| Dale Grover's DSP Microcontroller Book | 01/01/70 00:00 | |
| I've read it a while ago ... | 01/01/70 00:00 | |
| MAC is nice, but good multiply often enough | 01/01/70 00:00 | |
| Sure but ... | 01/01/70 00:00 | |
| Not manufacturers but books. | 01/01/70 00:00 | |
| Is really MAC so nice for 8bit MCU? | 01/01/70 00:00 | |
| It\'s nicer than not having one, ... | 01/01/70 00:00 | |
| Yes, need for MAC depends on processor size | 01/01/70 00:00 | |
Yes, a MAC is nice for 8 Bit MCU | 01/01/70 00:00 | |
| Comb Filter | 01/01/70 00:00 | |
| Depends. | 01/01/70 00:00 | |
| Bucket Brigade Comb Filter | 01/01/70 00:00 | |
| Any further questions ? | 01/01/70 00:00 |