Differences

This shows you the differences between two versions of the page.

seminars:seminar_03_28_14 [2014/03/13 18:05]
kevincha created
seminars:seminar_03_28_14 [2014/03/13 18:06] (current)
kevincha
Line 6: Line 6:
===Abstract=== ===Abstract===
Cache read misses stall the processor if Cache read misses stall the processor if
-there are no independent instructions to exe- +there are no independent instructions to execute. In contrast, most cache write misses are
-cute. In contrast, most cache write misses are+
off the critical path of execution, since writes off the critical path of execution, since writes
-can be buffered in the cache or the store buff- +can be buffered in the cache or the store buffer. With few exceptions, cache lines that serve
-er. With few exceptions, cache lines that serve+
loads are more critical for performance than loads are more critical for performance than
-cache lines that serve only stores. Unfortu- +cache lines that serve only stores. Unfortunately, traditional cache management mecha-
-nately, traditional cache management mecha-+
nisms do not take into account this disparity nisms do not take into account this disparity
-between read-write criticality. The key contri- +between read-write criticality. The key contribution of this work is the new idea of distin-
-bution of this work is the new idea of distin-+
guishing between lines that are reused by guishing between lines that are reused by
reads versus those that are reused only by reads versus those that are reused only by
writes to focus cache management policies on writes to focus cache management policies on
the more critical read lines. We propose a the more critical read lines. We propose a
-Read-Write Partitioning (RWP) policy that min- +Read-Write Partitioning (RWP) policy that minimizes read misses by dynamically partitioning
-imizes read misses by dynamically partitioning+
the cache into clean and dirty partitions, where the cache into clean and dirty partitions, where
partitions grow in size if they are more likely to partitions grow in size if they are more likely to
receive future read requests. We show that receive future read requests. We show that
-exploiting the differences in read-write criticali- +exploiting the differences in read-write criticality provides better performance over prior
-ty provides better performance over prior+
cache management mechanisms. cache management mechanisms.
===Bio=== ===Bio===
-Samira Khan is a Post Doctoral Re- +Samira Khan is a Post Doctoral Researcher at CMU working with Profes-
-searcher at CMU working with Profes-+
sor Onur Multu, She is also associated sor Onur Multu, She is also associated
with Intel Labs, Hillsboro. Her research with Intel Labs, Hillsboro. Her research
focuses on improving the performance, focuses on improving the performance,
-efficiency, and reliability of memory hi- +efficiency, and reliability of memory hierarchy of modern processors. She re- 
-erarchy of modern processors. She re- +ceived her PhD from the Department of Computer Science in the University of
-ceived her PhD from the Department of +
-Computer Science in the University of+
Texas at San Antonio. Texas at San Antonio.