B. Goeman, H. V. Dierendonck, and K. DeBosschere. Di#erential fcm: Increasing value prediction accuracy by improving table usage e#ciency. In HPCA-7, Jan. 2001. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Static Load Classification for Improving the Value.. - Burtscher, Diwan.. (2002) (3 citations) (Correct)
....during the repeated traversal of dynamic data structures. Note that this predictor can also predict alternating sequences and sequences exhibiting stride behavior as long as the sequence repeats and its length does not exceed the table size. The di#erential finite context method predictor dfcm [16] improves on fcm by retaining strides instead of absolute values. This approach reduces the chance of detrimental aliasing in the second level table, often increases the predictor s capacity, and enables it to predict values it has never before seen. Thus, dfcm combines the strengths of fcm and ....
B. Goeman, H. V. Dierendonck, and K. DeBosschere. Di#erential fcm: Increasing value prediction accuracy by improving table usage e#ciency. In HPCA-7, Jan. 2001.
Latency and Energy Aware Value Prediction for High-Frequency.. - Bhargava, John (2002) (1 citation) (Correct)
.... explored a more e#cient hybrid value predictor that exploits redundant predictions (e.g. last value predictor, stride predictor, and context predictor all unnecessarily do last value prediction) A very high performing context value predictor is the differential FCM predictor by Goeman et al. [2]. This predictor uses the FCM scheme for context prediction [24] to predict strides (di#erence in values) instead of the actual values. This strategy for prediction increases the e#ciency of the tables, especially the second level table. Tullsen et al. present a method of storageless value ....
H. V. B. Goeman and K. D. Bosschere. Di#erential FCM: Increasing value prediction accuracy by improving table usage e#ciency. In 7th International Symposium on High Performance Computer Architecture, Jan 2001.
Value Prediction Design for High-Frequency Microprocessors - Bhargava, John (2002) (Correct)
....tables are modeled as two read write ports plus two read and write port pairs. Eight bytes is the minimum allowable block size in the Cacti 2.0 model. Equations used to determine total energy for each predictor strategy: 1] At Fetch Hybrid E = VP Read hybrid E) VPUpdate hybrid E) [2] At Fetch Stride E = VP Read 4kTable E ) VP Update 4kTable E ) 3] Decoupled Hybrid E = VP Update hybrid E ) TR Read DVPPVCE ) TR Build DVPPVCE ) 4] Decoupled GatedE = VP Update 4kTable E ) TR Read DVPPVCE ) TR Build DVPPVCE ) 5] Post Decode Stride Ld VPE = ....
.... e#cient hybrid value predictor which reduces state due to redundant predictions (e.g. last value predictor, stride predictor, and context predictor all unnecessarily do last value prediction) 18] A very high performing context value predictor is the di#erential FCM predictor by Goeman et al. [2]. This predictor uses the FCM scheme for context prediction [24] to predict strides (di#erence in values) instead of the actual values. This strategy for prediction increases the e#ciency of the tables, especially the second level table. Tullsen et al. present a method of storageless value ....
H. V. B. Goeman and K. D. Bosschere. Di#erential FCM: Increasing value prediction accuracy by improving table usage e#ciency. In 7th International Symposium on High Performance Computer Architecture, Jan 2001.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC