EE 345S Real-Time Digital Signal Processing Laboratory - Lecture 23
Guest Lecture by
Mr. Ravi Bhargarva
- Lecture on Introduction to Native Signal Processing in
- Lecture on Evaluating MMX Technology Using DSP and Multimedia
This is a short version of the talk presented at the 1998 ACM/IEEE
International Symposium on Computer Architecture.
- MMX (1996) is an example of adding native signal processing operations
to the instruction set architecture of a general-purpose processor
- This is a recent trend among general-purpose processors, which started
with the Sun UltraSPARC Visual Instruction Set (VIS) in 1995
- MMX and VIS uses a single-instruction multiple-data architecture which
performs a single operation on a 64-bit register
- MMX has a multiply-accumulate instruction and saturation arithmetic,
whereas VIS does not
- Metrowerks CodeWarrior
Pro5 C++ compiler and Intel's C compiler do in fact
generate MMX code.
However, other C/C++ compilers do not generate MMX or VIS code;
instead, assembly code must be included in the C/C++ code.
- Speedups for digital signal processing applications are modest
(1.5:1 to 2:1) but speedups for graphics operations are higher
(4:1 to 6:1)
Last updated 12/22/04.
Send comments to