System for converting packed integer data into packed floating point data in reduced time

A computer system, comprising:

• a storage device having stored therein one or more instructions;
• and
• a processor coupled to the storage device, the processor, in response to the one or more instructions, to,
• create a first intermediate data item by moving a first integer data element of a packed integer data item to a first data element of a first floating-point register and extending a sign of the first integer data element into a second data element of the first floating-point register,
• create a second intermediate data item by moving a second integer data element of the packed integer data item to a first data element of a second floating-point register and extending a sign of the second integer data element into a second data element of the second floating-point register,
• convert the first and second intermediate data items to a first and second floating-point data items, and
• pack the first and second floating-point data items from the first and second floating-point registers directly to a single floating-point register.

A processor, comprising:

• a decoder to decode instructions;
• and
• a circuit coupled to said decoder, said circuit in response to one or more decoded instructions to,
• create a first intermediate data item by moving a first integer data element of a packed integer data item to a first data element of a first floating-point register and extending a sign of the first integer data element into a second data element of the first floating-point register,
• create a second intermediate data item by moving a second integer data element of the packed integer data item to a first data element of a second floating-point register and extending a sign of the second integer data element into a second data element of the second floating-point register,
• convert the first and second intermediate data items to a first and second floating-point data items, and
• pack the first and second floating-point data items from the first and second floating-point registers directly to a single floating-point register.

A computer implemented method of converting a packed integer data to a packed floating-point data comprising:

• storing in a first floating-point register the result of,
• extending a sign of a first integer data element of the packed integer data, and
• concatenating the first integer data element with the extended sign of the first integer data element;
• storing in a second floating-point register the result of,
• extending a sign of a second integer data element of the packed integer data, and
• concatenating the second integer data element with the extended sign of the second integer data element;
• converting the integer data in the first-floating point register to floating-point data;
• converting the integer data in the second-floating point register to floating-point data;
• and
• packing the floating-point data of the first and second floating-point register directly into a single floating-point register.