Intel ITANIUM ARCHITECTURE, Manual

Page: 476 / 604

Volume 4: IA-32 SSE Instruction Reference 4:469
sources ( Figure 4-6 ). When unpacking from a memory operand, the full 128-bit
operand is accessed from memory but only the high order 64 bits are utilized by the
instruction.
The UNPCKLPS (Unpacked low packed single-precision floating-point) instruction
performs an interleaved unpack of the low-order data elements of first and second
packed single-precision floating-point operands. It ignores the higher half part of the
sources (
Figure 4-7 ). When unpacking from a memory operand, the full 128-bit
operand is accessed from memory but only the low order 64 bits are utilized by the
instruction.
4.6.1.5 Conversion Instructions
These instructions support packed and scalar conversions between 128-bit SSE
registers and either 64-bit integer MMX technology registers or 32-bit integer IA-32
registers. The packed versions behave identically to original MMX technology
instructions, in the presence of x87-FP instructions, including:
• Transition from x87-FP to MMX technology (TOS=0, FP valid bits set to all valid).
• MMX technology instructions write ones (1’s) to the exponent part of the
corresponding x87-FP register.
• Use of EMMS for transition from MMX technology to x87-FP.
Figure 4-6. Unpack High Operation
Figure 4-7. Unpack Low Operation
X4 X3 X2 X1
Y4 Y3 Y2 Y1
Y4 X4 Y3 X3
X4 X3 X2 X1
Y4 Y3 Y2 Y1
Y2 X2 Y1 X1