background image

CVTPI2PD—Convert Packed Dword Integers to Packed Double-Precision FP Values

INSTRUCTION SET REFERENCE, A-L

Vol. 2A 3-243

CVTPI2PD—Convert Packed Dword Integers to Packed Double-Precision FP Values

Instruction Operand Encoding

Description

Converts two packed signed doubleword integers in the source operand (second operand) to two packed double-
precision floating-point values in the destination operand (first operand). 
The source operand can be an MMX technology register or a 64-bit memory location. The destination operand is an 
XMM register. In addition, depending on the operand configuration:

For operands xmm, mm: the instruction causes a transition from x87 FPU to MMX technology operation (that 
is, the x87 FPU top-of-stack pointer is set to 0 and the x87 FPU tag word is set to all 0s [valid]). If this 
instruction is executed while an x87 FPU floating-point exception is pending, the exception is handled before 
the CVTPI2PD instruction is executed.

For operands xmm, m64: the instruction does not cause a transition to MMX technology and does not take 
x87 FPU exceptions.

In 64-bit mode, use of the REX.R prefix permits this instruction to access additional registers (XMM8-XMM15).

Operation

DEST[63:0] ← Convert_Integer_To_Double_Precision_Floating_Point(SRC[31:0]);

DEST[127:64] ← Convert_Integer_To_Double_Precision_Floating_Point(SRC[63:32]);

Intel C/C++ Compiler Intrinsic Equivalent

CVTPI2PD:

__m128d _mm_cvtpi32_pd(__m64 a)

SIMD Floating-Point Exceptions

None

Other Exceptions

See Table 22-6, “Exception Conditions for Legacy SIMD/MMX Instructions with XMM and without FP Exception,” in 
the Intel® 64 and IA-32 Architectures Software Developer’s Manual, Volume 3B
.

Opcode/

Instruction

Op/ 

En

64-Bit 

Mode

Compat/

Leg Mode

Description

66 0F 2A /r
CVTPI2PD xmmmm/m64*

RM

Valid

Valid

Convert two packed signed doubleword 

integers from mm/mem64 to two packed 

double-precision floating-point values in xmm.

NOTES:

*Operation is different for different operand sets; see the Description section.

Op/En

Operand 1

Operand 2

Operand 3

Operand 4

RM

ModRM:reg (w)

ModRM:r/m (r)

NA

NA