Loading values from a value vector into subregisters of a single instruction multiple data register

• US 9,792,117 B2
• Filed: 09/10/2013
• Issued: 10/17/2017
• Est. Priority Date: 12/08/2011
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Ganesh, Amit, Chavan, Shasank K., Marwah, Vineet, Kamp, Jesse, Patthak, Anindya C., Gleeson, Michael J., Holloway, Allison L., Macnicol, Roger
• Primary Examiner(s): Giroux, George
• Application Number:
  US14/023,064
• Publication Number:
  US 20140013076A1
• Time in Patent Office: 1,498 Days
• Field of Search: None
• CPC Class Codes: G06F 16/221 Column-oriented storage; Management thereof
• CPC Class Codes: G06F 9/30018 Bit or string instructions
• CPC Class Codes: G06F 9/30032 Movement instructions, e.g. MOVE, SHIFT, ROTATE, SHUFFLE
• CPC Class Codes: G06F 9/30036 Instructions to perform operations on packed data, e.g. vector, tile or matrix operations
• CPC Class Codes: G06F 9/30038 using a mask

Abstract

A method and apparatus for efficiently processing data in various formats in a single instruction multiple data (“SIMD”) architecture is presented. Specifically, a method to unpack a fixed-width bit values in a bit stream to a fixed width byte stream in a SIMD architecture is presented. A method to unpack variable-length byte packed values in a byte stream in a SIMD architecture is presented. A method to decompress a run length encoded compressed bit-vector in a SIMD architecture is presented. A method to return the offset of each bit set to one in a bit-vector in a SIMD architecture is presented. A method to fetch bits from a bit-vector at specified offsets relative to a base in a SIMD architecture is presented. A method to compare values stored in two SIMD registers is presented.

Claims

1. A processor that, within the processor, loads values from a vector of values into a series of subregisters of a SIMD register:

wherein values within the vector of values are contiguous;

wherein each value in the vector of values is represented by a fixed number of bits;

wherein the fixed number of bits used to represent every value in the vector of values is the same;

wherein the processor is configured to respond to one or more instructions by;

based on the one or more instructions, subdividing the SIMD register into the series of subregisters;

wherein each subregister in the series of subregisters holds a greater number of bits than the fixed number of bits;

loading each subregister, of the series of subregisters, with a corresponding portion of the vector of values;

wherein loading each subregister includes loading the same portion of the vector of values into both a first subregister and a second subregister of the series of subregisters;

wherein the portion of the vector of values that is loaded into both the first subregister and the second subregister includes both a first value and a second value;

after loading each subregister with the corresponding portion of the vector of values, performing a plurality of subregister shifting operations;

wherein the plurality of subregister shifting operations includes a first subregister shifting operation that shifts, by a first amount, the first subregister of the series of subregisters;

wherein the plurality of subregister shifting operations includes a second subregister shifting operation that shifts, by a second amount, the second subregister of the series of subregisters;

wherein the first amount is different than the second amount;

wherein the plurality of subregister shifting operations causes one of;

the first value to be aligned, in the first subregister, with a left boundary of the first subregister and the second value to be aligned, in the second subregister, with a left boundary of the second subregister,

the first value to be aligned, in the first subregister, with a right boundary of the first subregister and the second value to be aligned, in the second subregister, with a right boundary of the second subregister, or

the first value and the second value to be byte-aligned; and

setting to zero all bits in the first subregister other than bits storing the first value; and

setting to zero all bits in the second subregister other than bits storing the second value.

2. The processor of claim 1, wherein the plurality of subregister shifting operations causes each value loaded into the series of subregisters to be byte aligned within the subregister.

3. The processor of claim 1, wherein each subregister, of the series of subregisters, is eight bits.

4. The processor of claim 1, wherein the one or more instructions are a single instruction.

5. The processor of claim 4, wherein the single instruction is performed in one cycle.

6. The processor of claim 1, wherein the one or more instructions designate how many bytes to use for each subregister of the series of subregisters.

7. The processor of claim 1, wherein each value in the vector of values is loaded into each subregister in the series of subregisters.

8. The processor of claim 1, wherein the plurality of subregister shifting operations causes each value loaded into the series of subregisters to be aligned with the left boundary of the subregister into which the value was loaded.

9. A processor configured to:

load each subregister of a first series of subregisters in a first SIMD register with a corresponding length-representing-value from a vector of length-representing-values;

wherein the first SIMD register resides within the processor;

wherein each length-representing-value in the vector of length-representing-values corresponds to a value in a vector of values, and indicates length of the corresponding value in the vector of values;

load values from the vector of values into a second series of subregisters of a second SIMD register that resides within the processor;

wherein, prior to being loaded into the second series of subregisters, values within the vector of values are contiguous;

wherein the values in the vector of values vary in length relative to each other; and

wherein loading values from the vector of values into the second series of subregisters includes;

based on one or more instructions, subdividing the second SIMD register into the second series of subregisters, wherein each subregister in the second series of subregisters is the same size; and

for each value in the vector of values, performing the steps of;

determining which length-representing-value, in the first series of subregisters, corresponds to the value;

based on the length-representing-value that corresponds to the value, determining which portion of the vector of values corresponds to the value; and

loading the portion of the vector of values that corresponds to the value from the vector of values into a corresponding subregister of the second series of subregisters.

10. The processor of claim 9 wherein the processor is configured to set all bytes, in each subregister of the second series of subregisters, other than bytes storing values from the vector of values, to zero.

11. The processor of claim 9 wherein the processor is configured to clear all bits in the second series of subregisters other than bytes storing values from the vector of values.

12. The processor of claim 9, wherein the one or more instructions are a single instruction.

13. The processor of claim 12, wherein the single instruction is performed in one cycle.

14. The processor of claim 9, wherein the one or more instructions designate how many bytes to use for each subregister of the second series of subregisters.