Vol. 1 10-7

PROGRAMMING WITH INTEL® STREAMING SIMD EXTENSIONS (INTEL® SSE)

The scalar single-precision floating-point instructions operate on the low (least significant) doublewords of the two 
source operands (X0 and Y0); see Figure 10-6. The three most significant doublewords (X1, X2, and X3) of the first 
source operand are passed through to the destination. The scalar operations are similar to the floating-point oper-
ations performed in the x87 FPU data registers with the precision control field in the x87 FPU control word set for 
single precision (24-bit significand), except that x87 stack operations use a 15-bit exponent range for the result, 
while SSE operations use an 8-bit exponent range.

10.4.1.1   SSE Data Movement Instructions

SSE data movement instructions move single-precision floating-point data between XMM registers and between an 
XMM register and memory.
The MOVAPS (move aligned packed single-precision floating-point values) instruction transfers a double quadword 
operand containing four packed single-precision floating-point values from memory to an XMM register and vice 
versa, or between XMM registers. The memory address must be aligned to a 16-byte boundary; otherwise, a 
general-protection exception (#GP) is generated.
The MOVUPS (move unaligned packed single-precision, floating-point) instruction performs the same operations as 
the MOVAPS instruction, except that 16-byte alignment of a memory address is not required.
The MOVSS (move scalar single-precision floating-point) instruction transfers a 32-bit single-precision floating-
point operand from memory to the low doubleword of an XMM register and vice versa, or between XMM registers.
The MOVLPS (move low packed single-precision floating-point) instruction moves two packed single-precision 
floating-point values from memory to the low quadword of an XMM register and vice versa. The high quadword of 
the register is left unchanged.

Figure 10-5.  Packed Single-Precision Floating-Point Operation

Figure 10-6.  Scalar Single-Precision Floating-Point Operation

X3

X2

X1

X0

Y3

Y2

Y1

Y0

X3 OP Y3

X2 OP Y2

X1 OP Y1

X0 OP Y0

OP

OP

OP

OP

X3

X2

X1

X0

Y3

Y2

Y1

Y0

X3

X2

X1

X0 OP Y0

OP