search

enjoy-digital / litex

Build your hardware, easily!
Other
3.04k stars 573 forks source link

Endianness in WishboneDMAReader #1001

Open navaneeth-cirel opened 3 years ago

navaneeth-cirel commented 3 years ago

Hi,

Shouldn't this be opposite ? https://github.com/enjoy-digital/litex/blob/78c1751c4781ffe156128748810ee4af85fee058/litex/soc/cores/dma.py#L19-L20

Just to try this out I used the WishboneDMAReader in a module of my SoC (with vexriscv) simulation to read a word from SRAM and print it in my C code and it printed reversed. The complete system is little endian.

DMA Read from : 0x10003fac
Readback from DMA: 0xdf0feca
Actual value : 0xcafef00d

with the format_bytes reversed

 def format_bytes(s, endianness): 
     return {"little": s, "big": reverse_bytes(s)}[endianness]

the print is as expected

DMA Read from : 0x10003fac
Readback from DMA: 0xcafef00d
Actual value : 0xcafef00d
enjoy-digital commented 3 years ago

Hi @navaneeth-cirel,

this probably similar to https://github.com/enjoy-digital/litedram/issues/251. There is a endianness mismatch between the model and real hardware. I'll try to address this soon.

navaneeth-cirel commented 3 years ago

Hi @enjoy-digital ,

After brief look at the issue referenced it seems that there is a mismatch between the simulation and target hardware, but in my case I am seeing the mismatch in Simulation itself. Just to give more clarity , what I am doing is writing a 32bit value to a uint32_t variable in the C code and set its address to the DMA to read and transfer to a CSR status register which I am then reading and printing in the C code, where the mismatch is observed.