The instructions for the third question in Assignment 2 have been revised to improve clarity and readability, as follows.

Consider the nested loops in both test programs (Lines 22-26 in test1.c and Lines 22-27 in test2.c). Analyze all four combinations of test progra m and cache configuration:
test1.c with cache1.cfg, test1.c with cache2.cfg, test2.c with cache1.cfg, test2.c with cache2.cfg.

For each combination, calculate the cache miss ratio for the accesses to array A that are used to update the variable sum. The miss ratio is defined as the number of cache misses divided by the total number of accesses to array A.

To help with the calculation, think about the access pattern to array A: when does an access cause a cache miss, which elements of A are then loaded into the cache, and where are they placed? Because of spatial locality, some subsequent accesses may hit in the cache until another element that is not currently cached is accessed and causes the next miss. Try to identify the repeating pattern of misses and hits. For example, a pattern such as M H H H M H H H ... corresponds to one miss every four accesses, giving a miss ratio of 1/4 = 0.25.

Assume that the cache is empty when the nested loop starts and that an int occupies 4 bytes.

Finally, identify which of the four combinations gives the lowest miss ratio and which gives the highest miss ratio.

The following note has been added to Lab 3 (If you have already completed Lab 3, there is no need to redo the experiment. This update does not substantially affect the experimental results or conclusions and is mainly intended to provide additional clarification.):
Note. Please set -fetch:ifqsize in your configuration file to a sufficiently large value, such as 128, to ensure that enough instructions are available for the subsequent pipeline stages.
Note. You may want to write a script to iterate over the parameters. However, please be careful, as there appear to be some unresolved issues. The results obtained from scripted simulations may sometimes be unexpected and may differ from those obtained by running the simulations manually in the terminal. This issue has not yet been resolved. Therefore, please verify that the results produced by your script are consistent with those obtained from simulations run directly in the terminal.

Please submit the report for Lab 5 via Lisam. Only one report is required per group.