Hybrid Data Prefetchers
=======================
Data Prefetching Championship Simulation Library in `lib/`.
Traces in `traces/`.
Sources in `src/`.
Documentation in `doc/`. Terminal Output is also documented in `doc/results_terminal.txt`.
Binaries and Executable Scripts in `bin/`.
Results are reproducible via the below commands (With working directory: Makefile Root)
Note: All scripts and binaries in `bin/` directory must be set executable if they are not already.
Commands
--------
To build both prefetchers:
make
To run all traces with all four configuration options on both prefetchers:
make run
Results:
--------
VLDP-AMPM hybrid, default setting
GemsFDTD 3.450102
gcc 0.331036
lbm 1.984225
leslie3d 1.299102
libquantum 3.279885
mcf 0.375427
milc 1.203386
omnetpp 2.251103
VLDP-AMPM hybrid, small_llc settings
GemsFDTD 3.450102
gcc 0.334651
lbm 1.768515
leslie3d 1.196737
libquantum 3.279885
mcf 0.236937
milc 1.117191
omnetpp 1.535260
VLDP-AMPM hybrid, low_bandwidth setting
GemsFDTD 3.411516
gcc 0.182784
lbm 0.987513
leslie3d 0.689992
libquantum 3.180119
mcf 0.224346
milc 0.615020
omnetpp 1.749399
VLDP-AMPM hybrid, scramble_loads setting
GemsFDTD 3.426760
gcc 0.331000
lbm 1.979308
leslie3d 1.260684
libquantum 3.278289
mcf 0.378711
milc 1.191201
omnetpp 2.254144
DCPT-VLDP hybrid, default setting
GemsFDTD 3.448920
gcc 0.328462
lbm 1.966279
leslie3d 1.243556
libquantum 3.273719
mcf 0.371777
milc 1.200224
omnetpp 2.245062
DCPT-VLDP hybrid, small_llc settings
GemsFDTD 3.448920
gcc 0.330768
lbm 1.768936
leslie3d 1.149879
libquantum 3.273719
mcf 0.237254
milc 1.114823
omnetpp 1.528507
DCPT-VLDP hybrid, low_bandwidth setting
GemsFDTD 3.411539
gcc 0.183068
lbm 0.986207
leslie3d 0.678101
libquantum 3.204195
mcf 0.223630
milc 0.613283
omnetpp 1.738227
DCPT-VLDP hybrid, scramble_loads setting
GemsFDTD 3.423804
gcc 0.328068
lbm 1.982051
leslie3d 1.207869
libquantum 3.274490
mcf 0.373476
milc 1.187882
omnetpp 2.245905
References
----------
1. Yasuo Ishii, Mary Inaba, and Kei Hiraki, “Access Map Pattern Matching Prefetch: Optimization Friendly Method”
2. Vinson Young, Ajit Krisshna, “Towards Bandwidth-Efficient Prefetching with Slim AMPM”
3. Manjunath Shevgoor, Sahil Koladiya, Rajeev Balasubramonian, Zeshan Chishti, “Efficiently Prefetching Complex Address Patterns”
aneesh-neelam/Hybrid-Data-Prefetchers
A hybrid data prefetcher for the Data Prefetching Championship
CNOASSERTION