Processing vectors using wrapping minima and maxima instructions in the macroscalar architecture

• US 8,555,037 B2
• Filed: 09/24/2012
• Issued: 10/08/2013
• Est. Priority Date: 08/15/2008
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Gonion, Jeffry E.
• Primary Examiner(s): VICARY, KEITH E
• Application Number:
  US13/625,164
• Publication Number:
  US 20130036293A1
• Time in Patent Office: 379 Days
• Field of Search: None
• US Class Current: 712/222
• CPC Class Codes: G06F 8/4441 Reducing the execution time required by the program code
• CPC Class Codes: G06F 9/30014 with variable precision
• CPC Class Codes: G06F 9/30018 Bit or string instructions
• CPC Class Codes: G06F 9/30036 Instructions to perform operations on packed data, e.g. vector, tile or matrix operations
• CPC Class Codes: G06F 9/30072 to perform conditional operations, e.g. using predicates or guards
• CPC Class Codes: G06F 9/3838 Dependency mechanisms, e.g. register scoreboarding
• CPC Class Codes: G06F 9/3842 Speculative instruction execution

Abstract

Embodiments of a system and a method in which a processor may execute instructions that cause the processor to receive an input vector and a control vector are disclosed. The executed instructions may also cause the processor to perform a minima or maxima operation on another input vector dependent upon the input vector and the control vector.

Claims

1. A method, comprising:

performing, by a processor;

receiving an operand vector, a basis vector, a selection vector, and a control vector;

selecting a basis value from the basis vector dependent upon the selection vector; and

generating a result vector dependent upon the selected basis value;

wherein each of the operand vector, the basis vector, the selection vector, the control vector, and the result vector includes a respective plurality of elements that occupy N ordered element positions, wherein the N ordered element positions include an initial element position and a last element position;

wherein N is an integer greater than one;

wherein for each given one of the N ordered element positions, said generating comprises performing a minima or maxima operation dependent upon selected elements of the operand vector that are selected dependent upon the control vector and the basis value, such that for a particular one of the N ordered element positions, performing the minima or maxima operation for the particular element position is dependent upon one or more element positions prior to the particular element position.

2. The method of claim 1, wherein the selecting the basis value comprises:

selecting the basis value from the basis vector at an element position in the basis vector that corresponds to an initial active element position of the ordered element positions in the selection vector, wherein the initial active element position precedes an other active element positions in the selection vector.

3. The method of claim 1, wherein selecting the basis value comprises selecting the basis value from an element position of the basis vector corresponding to an element position immediately preceding an initial active element position of the selection vector, wherein the initial active element position precedes any other active element positions in the selection vector.

4. The method of claim 2, wherein selecting the basis value further comprises selecting the basis value from a last ordered element position of the basis vector in response to a determination that a first ordered element position of the selection vector is active.

5. A system, comprising:

one or more memories that, during operation of the system, store instructions; and

one or more processors that, during operation of the system, retrieve one or more instructions from the one or more memories and execute the one or more instructions to cause the system to perform operations comprising;

receiving an operand vector;

receiving a basis vector;

receiving a selection vector

receiving a control vector;

selecting a basis value from the basis vector dependent upon the selection vector; and

generating a result vector dependent upon the selected basis value;

wherein each of the operand vector, the basis vector, the selection vector, the control vector, and the result vector includes a respective plurality of elements that occupy N ordered element positions, wherein the N ordered element positions include an initial element position and a last element position;

wherein N is an integer greater than one

wherein generating the result vector comprises performing a minima or maxima operation dependent upon the basis value and selected elements of the operand vector that are selected dependent upon the control vector, such that for a particular one of the N ordered element positions, performing the minima or maxima operation for the particular element position is dependent upon one or more element positions prior to the particular element position.

6. The system of claim 5, wherein the minima or maxima operation comprises an operation to determine an arithmetic minimum or maximum value.

7. The system of claim 6, wherein selecting the basis value from the basis vector dependent upon the selection vector comprises selecting the basis value from the basis vector at an element position corresponding to an initial active element position of the ordered element positions in the selection vector, wherein the initial active element position precedes any other active element positions in the selection vector.

8. A processor, comprising:

one or more processor cores, wherein a given one of the one or more processor cores is configured to;

receive an operand vector, a selection vector, a basis vector, and a control vector;

select a basis value from the basis vector dependent upon the selection vector; and

generate a result vector dependent upon the selected basis value;

wherein each of the operand vector, the selection vector, the basis vector, the control vector, and the result vector includes a respective plurality of elements that occupy N ordered element positions, wherein the N ordered element positions include an initial element position and a last element position;

wherein N is an integer greater than one

wherein to generate the result vector, the given processor core is further configured to perform a minima or maxima operation dependent upon the basis value and selected elements of the operand vector that are selected dependent upon the control vector, such that to generate an element of the result vector at a particular one of the N ordered element positions of the result vector, the given processor core is configured to perform the minima or maxima operation for the particular element position dependent upon one or more element positions prior to the particular element position.

9. The processor of claim 8, wherein to select the basis value, the given one of the one or more processor cores is further configured to select the basis value from an element position of the basis vector corresponding to an element position immediately preceding an initial active element position of the selection vector, wherein the initial active element position precedes any other active element positions in the selection vector.

10. The processor of claim 8, wherein to select the basis value, the given one of the one or more processor cores is further configured to select the basis value from the last element position of the basis vector responsive to a determination that an element in the initial element position of the selection vector is active.

11. The processor of claim 10, wherein to generate the result vector, the given one of the one or more processor cores is further configured to set a given element position of the result vector to a value that is equivalent to performing a minima or maxima operation on a set of elements that includes the basis value and elements of the operand vector corresponding to active element positions of the control vector.

12. A method, comprising:

performing, by one or more processors;

receiving a basis vector, an operand vector, a selection vector, and a control vector;

selecting a basis value from the basis vector at a last element position in the basis vector responsive to a determination that an initial active element is in an initial element position of the selection vector; and

generating a result vector dependent upon the basis value;

wherein each of the basis vector, the operand vector, the selection vector, the control vector, and the result vector includes a respective plurality of elements that occupy N ordered element positions, wherein the N ordered element positions include an initial element position and a last element position;

wherein N is an integer greater than one

wherein for each given one of the N ordered element positions, said generating comprises performing a minima or maxima operation dependent upon the basis value and selected elements of the operand vector that are selected dependent upon the control vector.

13. The method of claim 12, wherein for a given element position indicated by the control vector, generating the result vector comprises setting the given element position of the result vector to a value that is equivalent to performing a minima or maxima operation on a set of elements that includes the basis value and elements of the operand vector corresponding to active element positions of the control vector prior to and including the given element position.

14. The method of claim 12, wherein for a given element position indicated by the control vector, generating the result vector comprises setting the given element position of the result vector to a value that is equivalent to performing a minima or maxima operation on a set of elements that includes the basis value and elements of the operand vector corresponding to active element positions of the control vector prior to and excluding the given element position.

15. A system, comprising:

one or more memories that, during operation of the system, store instructions; and

one or more processors that, during operation of the system, retrieve one or more instructions from the one or more memories and execute the one or more instructions to cause the system to perform operations comprising;

receiving a basis vector, an operand vector, a selection vector, and a control vector;

generating a result vector;

wherein each of the basis vector, the control vector, the operand vector, the selection vector and the result vector includes a respective plurality of elements that occupy N ordered element positions, wherein the N ordered element positions include an initial element position and a last element position;

wherein N is an integer greater than one

selecting a basis value from the last element position of the basis vector responsive to a determination that an initial active element of the selection vector is in the initial element position of the selection vector;

wherein generating the result vector comprises performing a minima or maxima operation dependent upon the basis value and selected elements of the operand vector that are selected dependent upon the control vector.

16. The system of claim 15, wherein selecting the basis value comprises selecting the basis value from an element position in the basis vector that corresponds to an initial active element of the ordered element positions in the selection vector, wherein the initial active element position precedes any other active element positions in the selection vector.

17. The system of claim 15, wherein for element positions prior to an initial active element position of the ordered element positions in the selection vector, wherein the initial active element position precedes any other active element position in the selection vector, generating the result vector further comprises copying a value from an element position of the basis vector into a corresponding inactive element position of the result vector, wherein the inactive element position of the result vector is indicated by the control vector.

18. A method, comprising:

performing, by a processor;

receiving an input vector, a basis vector, a selection vector, and a control vector;

selecting a basis value from the basis vector dependent upon the selection vector; and

generating a result vector dependent upon the selected basis value;

wherein each of the input vector, the basis vector, the selection vector, the control vector, and the result vector includes a respective plurality of elements that occupy N ordered element positions, wherein the N ordered element positions include an initial element position and a last element position;

wherein N is an integer greater than one

wherein for each given one of the N ordered element positions, said generating comprises performing a function dependent upon selected elements of the input vector that are selected dependent upon the control vector and the basis value, such that for a particular one of the N ordered element positions, performing the function for the particular element position is dependent upon one or more element positions prior to the particular element position.

19. The method of claim 18, wherein selecting the basis value further comprises selecting the basis value from an element position of the basis vector corresponding to an element position immediately preceding an initial active element position of the selection vector, wherein the initial active element position precedes any other element positions in the selection vector.

20. The method of claim 18, wherein selecting the basis value further comprises selecting the basis value from a last ordered element position of the basis vector in response to a determination that a first ordered element position of the selection vector is active.