Then there is a real problem because at some point that 3 per ms x 1000 ms/sec x 8 bytes x 10 bits each = 240000 baud that can fit through SPI (capable of like 1000000 Bytes/sec) is going to have to go through 115200 baud RX/TX serial and that ain't gonna happen no matter how many chips you dedicate!
IIRC (I've been helping about 4 or more people at a time, I try to not mix yinz up, honest!) there's a bluetooth link in this and it's not going to take all that.
When this thread started
This code sends a byte every 1 sec and prints the time taken between line A and C.
we didn't know that "prints the time taken between line A and C" happened 3x per ms.
Might I suggest that that time in usecs must be 333 or less does not need 8 bytes?
And if you only sent the difference between "how long", the value would likely fit in 1 char (signed byte)?
This would kind of be an engineering decision on your part. Some requirement(s) will have to change.
Either get rid of the serial bottleneck or send fewer bytes/message or send fewer messages than 3 per ms.
All 3 together at the same time just ain't happening!