> However, supporting 24 bit packed *can* be inefficient - it all depends on how the hardware supports 24bit. > > Which means, that the driver has to sit there and copy the 24bit packed data to 24bit unpacked DMA buffers - going through each sample - which would not be necessary if the application was not required to use 24bit packed. > > Although I'm not sure how this would really affect other processing, but I'm fairly certain that any code (user or kernel mode) that has to process 24bit packed data is going to be less efficient because of the lack of 24 bit math instructions on the CPUs. < But you are not doing "math", you are just reorganizing the data. Think of it as a fast codec. Peter ****************** WDMAUDIODEV addresses: Post message: mailto:wdmaudiodev@xxxxxxxxxxxxx Subscribe: mailto:wdmaudiodev-request@xxxxxxxxxxxxx?subject=subscribe Unsubscribe: mailto:wdmaudiodev-request@xxxxxxxxxxxxx?subject=unsubscribe Moderator: mailto:wdmaudiodev-moderators@xxxxxxxxxxxxx URL to WDMAUDIODEV page: http://www.wdmaudiodev.de/