Audio
There are no Audio documents available at this time. Stay tuned as new information is posted.
Communications
There are no Communication documents available at this time. Stay tuned as new information is posted.
DSP Kernels
There are no DSP Kernel documents available at this time. Stay tuned as new information is posted.
Graphics (2D)
JPEG Quantization process optimized for Pentium® II Processor
Current JPEG Quantization takes 64 DCT frequency components, divides them by a "Quantizer stepsize", and rounds
them to integers to form Quantified coefficients. This document shows optimizing techniques to gain substantial
performance improvement on the Quantization step running on a Pentium® II Processor.
JPEG Quantization process optimized for Pentium® II Processor
Also available in Adobe* Acrobat Format
Upsampling and color conversion are the final steps in JPEG decompression. These 2 steps can be merged together if the
chroma components are to be upsampled by simple replication i.e. box filtering. This appnote discusses techniques for
optimizing these steps using MMX technology.
Graphics (3D)
Mixed SW and HW 3D Rendering on the PC
This paper discusses the concept of mixed 3D rendering where some pixels on the screen are generated completely in software and others come
partly or completely from hardware renderers. SW-based rendering can provide neat cutting-edge advanced techniques that hardware is not
capable of, such as procedural texturing and geometric morphing. To date most PC applications have generated 3D images from software alone,
but recently hardware 3D acceleration has become fast and cheaper. Microsoft's DirectDraw provides primitives to allow some degree of mixed
rendering but could be made more mixed-friendly. THe paper explains how best to use DDraw for mixing, and how to acheive
concurrency/synchronization of HW/SW, with multithreading, semaphores, and various types of scene compositing.
Image Effects
There are no Image Effect documents available at this time. Stay tuned as new information is posted.
Miscellaneous
A Quick and Easy Way to Use CPUID at Run-time
Also available in Adobe* Acrobat Format
This paper explains a quick, easy, and effective method of using structured exception handling to determine the features of
a processor at run-time in order to execute code specifically for that processor. The paper gives an overview of
structured exception handling, explains the syntax, and includes code that sets flags when Pentium® II, Pentium® Pro,
Pentium®, and Pentium® processors with MMX technology are detected. The code should be easily modified to
detect features of future processors when they become available.
JPEG Inverse Discrete Cosine Transform and Dequantization Optimized for Pentium® II Processor
Also available in Adobe* Acrobat Format
This application note shows optimization techniques used to gain substantial performance improvement on the Inverse
Discrete Cosine Transformation (IDCT) and Dequantization in the JPEG decompression running on a Pentium® II
Processor.
A Quick and Easy Way to Use CPUID at Run-time
Also available in Adobe* Acrobat Format
This paper explains a quick, easy, and effective method of using structured exception handling to determine the features
of a processor at run-time in order to execute code specifically for that processor. The paper gives an overview of
structured exception handling, explains the syntax, and includes code that sets flags when Pentium® II, Pentium® Pro,
Pentium®, and Pentium® processors with MMX technology are detected. The code should be easily modified to
detect features of future processors when they become available.
Memory Ordering on Dynamic Execution (Pentium® Pro Family) Processors
Also available in Adobe* Acrobat Format
The order of memory access is very important, it is directly responsible for the performance. This collateral gives you an
overview of the cases where you can detect possible memory access stalls, serializations, etc. With the knowledge of the
ordering rules, you can make improvements to your code to avoid the performance hit, even though reordering the code
will offer very minimal performance gain.
Using MMX Technology Instructions in TASM
The current iammx.inc file is rejected by Borland's Turbo Assembler. This collertal lists some of the work arounds
available so that MMX technology instructions can be incorportated into projects using the Borland's Turbo Assembler.
16-bit Alpha Blend
The current procedure for 16-bit alpha blending is to use look up tables to avoid the costly multiplies and divides involved. The following appnote
demonstrates two algorithms which take in 16-bit data represented in 555 or 565 as a source, and destination and apply an 8-bit alpha mask to
it. The result is then placed back in the destination.
P6 Family Processor Performance Measurement Utility for Windows NT
The utility extends the Windows NT standard Performance Monitor (PerfMon) to collect and display the advanced CPU performance
information. With PerfMon this information can be displayed side-by-side with other system performance parameters.
Intel C/C++ Compiler Plug-in
The Intel C/C++ Compiler Plug-in version 2.4 can easily be
integrated into the Microsoft Developer Studio environment and
allows users to use Pentium® Pro processor and Pentium® II processor
inline assembly instructions which are currently not supported
with the latest version of Microsoft Visual C++, version 5.0.
Optimized String Search with MMX Technology
Also available in Adobe* Acrobat Format
This appnote presents an optimized version of a text-string search that uses MMX Technology to
enhance performance.
Using the RDTSC Instruction for Performance Monitoring
Also available in Adobe* Acrobat Format
This collateral will deal with the many issues involved in using the RDTSC instruction to measure cycles.
These issues include out-of-order execution, data preloading, instruction preloading,
register overwriting, and counter overflow. All of these topics will be covered in depth in this application note,
to give developers full knowledge of the counts that RDTSC will give in any situation. Software examples will
be provided.
Speech Recognition
There are no Speech Recognition documents available at this time. Stay tuned as new information is posted.
Video
There are no Video documents available at this time. Stay tuned as new information is posted.
Also see the Intel MMX™ Technology application notes
|