Browse Prior Art Database

Self-Tuning Bimodal 64-bit Stack Storage Manager Disclosure Number: IPCOM000205111D
Publication Date: 2011-Mar-15
Document File: 2 page(s) / 18K

Publishing Venue

The Prior Art Database


Disclosed is a Self-Tuning Bimodal 64-bit Stack Storage Manager which uses a bimodal approach to address virtual and real storage consumption and performance issues related to stack management.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

Self-Tuning Bimodal 64-bit Stack Storage Manager

Allocating and freeing stack storage

                        (i.e., autodata) for a procedure is critical to program function and performance. The considerations for stack management performance center on the caching of segments. Caching can improve performance but can result in virtual and real storage constraints. The choices as solutions to the problem are to either eliminate caching at the expense of performance or cache without regard to real/virtual storage consumption.

Any hybrid approaches require arbitrary constants in

place to control cache growth. With such approaches, caching behavior remains constant and does not react to the operating environment. In addition, hybrid approaches typically result in the reduction of the cache size which ultimately harms cache performance.

Embodiments of the invention concern a Self-Tuning Bimodal 64-bit Stack Storage Manager. The embodiments address virtual/real storage consumption and performance aspects of stack management. The stack manager is bimodal such that it tunes for high performance when real storage is available but will tune for real storage conservation when real frames are detected to be in high use. To eliminate virtual storage constraints, stack is allocated from 64-bit virtual addresses. Segments are cached to ensure high performance for allocation and de-allocation. Since using a large virtual space can result in an equally large amount of real storage consumption, system paging is monitored so the stack manager can discard unused real frames.

The advantage of this bimodal approach is to maximize stack performance as dictated by the amount of real storage available. The stack manager does not reduce the cache size and performs optimally based on the operating environment rather than preset constants.

The Self-Tuning Bi-Modal 64-bit Stack Storage Manager operates with, for example, a large 64-bit virtual pool dedicated to stack segment allocation. Stack segments are allocated in large fixed virtual sizes so allocation and de-allocation need not search for segments based on size.

A cache may be established when the application starts with

stack segments being placed on a push-down free queue.

pop a segment from the queue and use as needed.

procedure will monitor and mark the high water mark used in a given stack segment.