User Guide

128-Bit Media and Scientific Programming 107
24592—Rev. 3.15—November 2009 AMD64 Technology
the source operands. The result of the operation replaces the first source operand. There are also
instructions that operate on vectors of words, doublewords, or quadwords.
Figure 4-1. Parallel Operations on Vectors of Integer Elements
4.2.3 Floating-Point Vector Operations
There are almost as many 128-bit floating-point instructions as integer instructions. Figure 4-2 shows
an example of parallel operations on vectors containing four 32-bit single-precision floating-point
values. There are also instructions that operate on vectors containing two 64-bit double-precision
floating-point values.
Figure 4-2. Parallel Operations on Vectors of Floating-Point Elements
Integer and floating-point instructions can be freely intermixed in the same procedure. The floating-
point instructions allow media applications such as 3D graphics to accelerate geometry, clipping, and
lighting calculations. Pixel data are typically integer-based, although both integer and floating-point
operand 1
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
127 0
operand 2
127 0
operation
operation
513-163.eps
result
127 0
513-164.eps
. .. .
. .
operation
operation
result
operand 1
127 0
127 0
operand 2
127 0
FP single FP single FP single FP single FP single FP single FP single FP single
FP single FP single FP single FP single