apanwariisc / custom Goto Github PK

This repository contains two simple c programs written to observe
TLB behavior and the impact of huge pages on x86 processors.

Both the programs start with allocating two buffers of 1GB each
but references them in different manners.

sequential.c: Iterate over the buffers touching 1 byte from each.
              An iteration picks one buffer first and traverses it
              touching 1 byte at 4KB jumps before moving to the
              second buffer. Successive memory references are expected
              to be 4KB apart generating TLB pressure.

random.c:     Iterate over the buffers touching one byte from each.
              Unlike sequential, traverses both the buffers at the
	      same time. As a result, successive memory references
              are expected to be 1GB apart.


How to compile:
	run "make" to compile.
	"make clean" to clean.

How to run:
	Both the programs expect two arguments from the command line -
	1. Access mode
	2. The number of iterations to be performed.

Access Modes:
	This argument influences how the buffers will be accessed. Following
	modes are supported -
	0 - READ_ONLY - self-explanatory
	1 - WRITE_ONLY - self-explanatory
	2 - READ_WRITE - Every byte being touched is read first and then written.
	3 - WRITE_READ - Every byte being touched is written first and then read.

	Example run-
	./sequential 0 100000
	./random     1 100000