Method and apparatus for staggering execution of a single packed data instruction using the same circuit

A method comprising:

• receiving a single macro instruction specifying a first and a second data operand in first and second registers, respectively, and specifying an operation to be performed on the data operands;
• retrieving the first and second data operands;
• splitting the first and second data operands into a first corresponding segment and a second corresponding segment, the first corresponding segment including a first portion of the first data operand and a first portion of the second data operand, and the second corresponding segment including a second portion of the first data operand and a second portion of the second data operand;
• performing the operation on the first corresponding segment;
• and
• upon completion of the operation on the first corresponding segment, performing the operation on the second corresponding segment.

A method comprising:

• at time T:
• retrieving first and second data operands from their respective registers;
• splitting the first and second data operands into first and second corresponding segments;
• and
• adding the first corresponding segments to generate a first result;
• at time T+1:
• adding the second corresponding segments to generate a second result;
• and
• storing the first and second results sin a third register;
• and
• retrieving the first and second data operands from their respective registers;
• splitting the first and second data operands into first and second corresponding segments;
• and
• multiplying the first corresponding segments to generate a third result;
• and
• at time T+2:
• multiplying the second corresponding segments to generate a fourth result;
• and
• storing the third and fourth results of the added corresponding segments in a fourth register.

A method comprising:

• executing an add instruction on full width first and second data operands stored in a first and second register, respectively, the executing being performed in a half width execution unit by:
• at time T, retrieving the first and second data operands from their respective registers;
• splitting the first and second data operands into first and second corresponding segments;
• adding the first corresponding segments to generate a first result;
• at time T+1, adding the second corresponding segments to generate a second result;
• and
• storing the first and second results sin a third register;
• and
• executing a multiply instruction on the full width first and second data operands stored in the first and second register, respectively, the executing being performed in a half width execution unit by:
• at time T+1, retrieving the first and second data operands from their respective registers;
• splitting the first and second data operands into first and second corresponding segments;
• multiplying the first corresponding segments to generate a third result;
• at time T+2, multiplying the second corresponding segments to generate a fourth result;
• and
• storing the third and fourth results of the added corresponding segments in a fourth register.

A method comprising:

• receiving a first macro instruction specifying a first operation, and specifying a first and a second data operand in first and second registers, respectively;
• receiving a second macro instruction specifying a second operation, and specifying the first and the second data operand in the first and second registers, respectively;
• at time T, performing the first operation on lower order data segments of the first and second data operands on a first set of execution units;
• at time T+1, performing the first operation on high order data segments of the first and second data operands on the first set of execution units;
• at time T+2, performing the second operation on lower order data segments of the first and second data operands on the first set of execution units;
• and
• at time T+3, performing the second operation on high order data segments of the first and second data operands on the first set of execution units.

A method comprising:

• receiving a first macro instruction specifying a first add operation, and specifying a first and a second data operand in first and second registers, respectively;
• receiving a second macro instruction specifying a second add operation, and specifying the first and the second data operand in the first and second registers, respectively;
• determining that a third macro instruction specifying a multiply operation, and specifying the first and the second data operand in the first and second registers, respectively, further down an instruction stream may be performed out-or-order;
• at time T, performing the first add operation on lower order data segments of the first and second data operands on a first set of execution units;
• at time T+1, performing the first add operation on high order data segments of the first and second data operands on the first set of execution units, and performing the multiply operation on the low order data segments of the first and second data operands on a second set of execution units;
• at time T+2, performing the multiply operation on high order data segments of the first and second data operands on the second set of execution units.