Apple Shake 4 - Manual (Page 350)

350 Chapter 13 Image Caching
The processing cache is mainly used to store image tiles (tiles are portions of the
complete image) generated by nodes that need surrounding pixels to perform their
image processing operations during a render. Example nodes include the Blur,
Transform, Warp, PinCushion, and Twirl nodes. The processing cache also provides
secondary functionality for caching rendering buffers (in particular for the QuickPaint
node that utilizes a full-frame rendering buffer), color look-up tables, and
transformation matricies. The global plugs used to control the processing cache are as
follows:
• cache.cacheMemory
• cache.cacheMemoryLimit (Shake version 3.5 and later)
Limits to Shake RAM Usage
Shake is currently compiled as a 32-bit application, which can theoretically address up
to 4 GB of virtual RAM (2^32). However, because of the RAM needs and constraints
imposed by the operating system, and competition for RAM from other running
applications, most 32-bit applications have a practical limit of approximately 2 GB of
addressable RAM per process.
Because of this, even if you install 4 GB or more of RAM on your workstation, each
Shake process can only take advantage of about 2 GB of that RAM. The good news is
that, if you launch a Flipbook while running Shake on a system with 4 GB of RAM, the
Flipbook (as a separate process) is able to take advantage of the additional 2 GB of
RAM and is less likely to swap to disk, which is slower.
Mac OS X v10.3 and above (a 64-bit operating system) running on a PowerPC G5
computer configured with 8 GB of RAM partly addresses this issue. A 32-bit application
running natively on a 64-bit OS is still limited to approximately 2 GB of addressable
RAM. However, a Macintosh G5 computer configured with 8 GB of RAM running
Panther can keep a larger number of applications in physical RAM without swapping
out any one application’s memory to disk. As a result, Shake is able to allocate larger
contiguous segments of physical RAM, allowing large Shake scripts to be edited and
rendered in less time.
The Image Cache
The main purpose of the image cache is to improve interactivity while you’re working
on a Shake script in the interface. Shake accomplishes this by attempting to output
image data from nodes in the compositing tree at, and near, the portion of the
compositing tree being edited or viewed. All nodes are capable of having their image
data cached.