Sciweavers

331 search results - page 5 / 67
» Limits on Multiple Instruction Issue
Sort
View
ISCA
1995
IEEE
110views Hardware» more  ISCA 1995»
14 years 11 days ago
Optimization of Instruction Fetch Mechanisms for High Issue Rates
Recent superscalar processors issue four instructions per cycle. These processors are also powered by highly-parallel superscalar cores. The potential performance can only be expl...
Thomas M. Conte, Kishore N. Menezes, Patrick M. Mi...
ICS
2004
Tsinghua U.
14 years 2 months ago
Scaling the issue window with look-ahead latency prediction
In contemporary out-of-order superscalar design, high IPC is mainly achieved by exposing high instruction level parallelism (ILP). Scaling issue window size can certainly provide ...
Yongxiang Liu, Anahita Shayesteh, Gokhan Memik, Gl...
ICCD
2000
IEEE
93views Hardware» more  ICCD 2000»
14 years 5 months ago
Cheap Out-of-Order Execution Using Delayed Issue
In superscalar architectures, out-of-order issue mechanisms increase performance by dynamically rescheduling instructions that cannot be statically reordered by the compiler. Whil...
J. P. Grossman
IEEEPACT
2000
IEEE
14 years 1 months ago
On Some Implementation Issues for Value Prediction on Wide-Issue ILP Processors
In this paper, we look at two issues which could affect the performance of value prediction on wide-issue ILP processors. One is the large number of accesses to the value predicti...
Sang Jeong Lee, Pen-Chung Yew
HPCA
1999
IEEE
14 years 1 months ago
Instruction Recycling on a Multiple-Path Processor
Processors that can simultaneously execute multiple paths of execution will only exacerbate the fetch bandwidth problem already plaguing conventional processors. On a multiple-pat...
Steven Wallace, Dean M. Tullsen, Brad Calder