Hardware to support looping code in an image processing system

• US 8,892,853 B2
• Filed: 06/10/2010
• Issued: 11/18/2014
• Est. Priority Date: 06/10/2010
• Status: Active Grant

• Current Assignee: MobilEye Vision Technologies Ltd. (Intel Corporation)
• Original Assignee: Mobileye Technologies Limited (Intel Corporation)
• Inventors: Kreinin, Yosef, Dogon, Gil, Sixsou, Emmanuel, Arbeli, Yosi, Navon, Mois, Sajman, Roman
• Primary Examiner(s): LI, AIMEE J
• Application Number:
  US12/797,689
• Publication Number:
  US 20110307684A1
• Time in Patent Office: 1,622 Days
• Field of Search: 712/6, 712/7, 712/241, 711/214
• US Class Current: 712/241
• CPC Class Codes: G06F 9/30043 LOAD or STORE instructions; Clear instruction
• CPC Class Codes: G06F 9/3555 using scaling, e.g. multiplication of index
• CPC Class Codes: G06F 9/3853 of compound instructions
• CPC Class Codes: G06F 9/3889 controlled by multiple instructions, e.g. MIMD, decoupled access or execute
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining
• CPC Class Codes: G06V 10/955 using specific electronic processors
• CPC Class Codes: G06V 20/56 exterior to a vehicle by using sensors mounted on the vehicle

Abstract

An image processing system including a vector processor and a memory adapted for attaching to the vector processor. The memory is adapted to store multiple image frames. The vector processor includes an address generator operatively attached to the memory to access the memory. The address generator is adapted for calculating addresses of the memory over the multiple image frames. The addresses may be calculated over the image frames based upon an image parameter. The image parameter may specify which of the image frames are processed simultaneously. A scalar processor may be attached to the vector processor. The scalar processor provides the image parameter(s) to the address generator for address calculation over the multiple image frames. An input register may be attached to the vector processor. The input register may be adapted to receive a very long instruction word (VLIW) instruction. The VLIW instruction may be configured to transfer only: (i) parameters for image processing calculations over the image frames by the ALU units and (ii) a single bit to the address generator.

Claims

1. An image processing system comprising:

a vector processor (404); and

a memory (402) adapted for operatively attaching to said vector processor (404), wherein said memory (402) is adapted to store a plurality of image frames (15),

wherein said vector processor (404) includes an address generator (408) operatively attached to said memory (402) to access said memory (402),

wherein said address generator (408) is adapted for calculating addresses of said memory (402) over said image frames (15),

wherein said address generator (408) is programmed with parameters of an execution loop over the image frames (15),

wherein said parameters are selected from a group of horizontal (X) and vertical (Y) image parameters, the group consisting;

a horizontal increment value (Xstep), an initial horizontal value (XStart), a horizontal product of the horizontal increment value (Xstep) and a horizontal count iteration value (Xcount), a vertical increment value (YStep), an initial vertical value (Ystart), a vertical product of said vertical increment value (Ystep) and a vertical count iteration value (Ycount), a number of memory addresses (Width) and a memory size value of said memory (402) per image frame (Base);

wherein said address generator (408) includes;

a horizontal counter (72X) with a horizontal output of X=X+Xstep, wherein the a horizontal counter 72X has inputs from the horizontal increment value (Xstep) and the horizontal value (XStart);

an output of a comparator (74X) connected to another input of the horizontal counter (72X), wherein the comparator (74X) compares the horizontal output of X=X+Xstep with the horizontal product of the horizontal increment value (Xstep) and the horizontal count iteration value (Xcount);

a vertical counter (72Y) with a vertical output of Y=Y+Ystep, wherein the vertical counter 72Y has inputs from the vertical increment value (Ystep), the vertical value (YStart) and a signal (DONE) from the horizontal counter (72X);

an output of a comparator (74Y) connected to another input of the vertical counter (72Y), wherein the comparator (74Y) compares the vertical output of Y=Y+Ystep with the vertical product of the vertical increment value (Ystep) and the vertical count iteration value (Ycount);

a multiplier (76), wherein a multiplier output of the a multiplier (76) is the vertical output of Y=Y+Ystep multiplied with the number of memory addresses (Width);

an adder (78), wherein a first adder output of the adder (78) is the multiplier output added with the horizontal output of X=X+Xstep;

a second adder (79), wherein a second adder output (NextADDR) of the second adder (79) is the first adder output added with memory size value of said memory (402) per image frame (Base);

wherein said vector processor (404) includes hardware to support looping code during execution, and

wherein said address generator (408) is configured to generate a break signal to be received by said hardware to break execution of said looping code of the vector processor (404) over the image frames (15), wherein said break signal is a second output (DONE) of said vertical counter (72Y).

2. The image processing system according to claim 1, wherein said addresses are calculated over said image frames (15) based upon an image parameter.

3. The image processing system according to claim 2, wherein said image parameter specifies which of the image frames (15) are processed simultaneously, wherein said image frames (15) are processed in parallel for multiple driver assistance applications using a single hardware platform.

4. The image processing system according to claim 3, further comprising a scalar processor (418) operatively attached to said vector processor (404), wherein said scalar processor (418) provides said image parameter to said address generator (408) for said calculating over said image frames (15).

5. The image processing system according to claim 1, further comprising an input register (406) operatively attached to said vector processor (404), wherein said input register (406) is adapted to receive a very long instruction word (VLIW) instruction (422), wherein only one bit of said VLIW instruction (422) is provided to said address generator (408).

6. The image processing system according to claim 1, wherein said vector processor (404) includes:

at least one arithmetic and logic unit (ALU) (420); and

an input register (406) adapted for receiving a very long instruction word (VLIW) instruction (422), wherein said VLIW instruction (422) is configured to transfer only;

(i) operation codes for image processing calculations over said image frames (15) by said at least one ALU unit (420) and (ii) a single bit to said address generator (408).

7. The image processing system according to claim 1, wherein said addresses are calculated over said multiple image frames (15) based upon at most one image size parameter which specifies a dimension of said image frames (15).

8. The image processing system according to claim 1, wherein said addresses are calculated over said multiple image frames (15) based upon at most two image size parameters which specify a dimension of said image frames (15).

9. A method of processing images in a system including a vector processor (404) and a memory (402), wherein the vector processor (404) includes an address generator (408) operatively attached to the memory (402) to access the memory (402), wherein said address generator (408) is adapted for calculating addresses of said memory (402), wherein said vector processor (404) includes hardware to support looping code during execution, the method comprising:

storing a plurality of image frames (15) in the memory (402);

accessing the memory (402) by the address generator (408);

loading an image parameter, wherein said image parameter is selected from a group of horizontal (X) and vertical (Y) image parameters, the group consisting of;

a horizontal increment value (Xstep), an initial horizontal value (XStart), a horizontal product of the horizontal increment value (Xstep) and a horizontal count iteration value (Xcount), a vertical increment value (YStep), an initial vertical value (Ystart), a vertical product of said vertical increment value (Ystep) and a vertical count iteration value (Ycount), a number of memory addresses (Width) and a memory size value of said memory (402) per image frame (Base);

calculating addresses for said image frames (15) based upon said image parameter, wherein said calculating further comprises the steps;

counting with a horizontal counter (72X), wherein the horizontal counter (72X) includes a horizontal output of X=X+Xstep, wherein the horizontal counter (72X) has inputs from the horizontal increment value (Xstep) and the horizontal value XStart;

connecting an output of a comparator (74X) to another input of the horizontal counter (72X), wherein the comparator (74X) compares the horizontal output of X=X+Xstep with the horizontal roduct of the horizontal increment value (Xstep) and the horizontal count iteration value (Xcount);

counting with a vertical counter (72Y), wherein the vertical counter (72Y) includes a vertical output of Y=Y+Ystep, wherein the vertical counter (72Y) has inputs from the vertical increment value (Ystep), the vertical value (YStart) and a signal (DONE) from the horizontal counter (72X);

connecting an output of a comparator (74Y) to another input of the vertical counter (72Y) wherein the comparator (74Y) compares the vertical output of Y=Y+Ystep with the vertical product of the vertical increment value (Ystep) and the vertical count iteration value (Ycount);

multiplying with a multiplier (76), wherein a multiplier output of the a multiplier (76) is the vertical output of Y=Y+Ystep multiplied with the number of memory addresses (Width);

adding with an adder (78), wherein a first adder output of the adder (78) is the multiplier output added with the horizontal output of X=X+Xstep;

adding with a second adder (79), wherein a second adder output (NextADDR) of the second adder (79) is the first adder output added with memory size value of said memory (402) per image frame (Base); and

processing in parallel image data stored in said image frames (15),

based on said programming, said address generator (408) generating a break signal;

receiving by the hardware said break signal, thereby breaking execution of said looping code of the vector processor (404) over the image frames (15).

10. The method according to claim 9, wherein said loading is performed in a register operatively attached to the address generator (408).

11. The method according to claim 9, further comprising receiving a VLIW instruction (422) for said processing in parallel.

12. The method according to claim 11, wherein said processing in parallel is performed by the vector processor (404) according to said VLIW instruction (422), wherein said loading of said image parameter is performed externally to said VLIW instruction (422).

13. The method according to claim 12, further comprising receiving by said address generator (408) only a single bit of said VLIW instruction (422) to enable said calculating addresses.

14. The method according to claim 9, wherein said image parameter specifies at least two of said image frames (15) to be processed in parallel.

15. The method according to claim 9, wherein said calculating is based upon said image parameter including at most two image size parameters which specify the size of said image frames (15).

16. The method according to claim 9, wherein said calculating is based upon said image parameter including at most one image size parameter which specifics the size of said image frames (15).

17. A method of processing images in a system including a vector processor (404) and a memory (402), wherein the vector processor (404) includes an address generator (408) operatively attached to the memory (402) to access the memory (402), wherein said address generator (408) is adapted for calculating addresses of said memory (402), and wherein said vector processor (404) includes hardware to support looping code during execution, the method comprising:

storing an image frame in the memory (402);

loading and reading a window parameter which specifies multiple windows within said image frame, wherein said window parameter is selected from a group of horizontal (X) and vertical (Y) image parameters, the group consisting;

a horizontal increment value (Xstep), an initial horizontal value (XStart), a horizontal product of the horizontal increment value (Xstep) and a horizontal count iteration value (Xcount), a vertical increment value (YStep), an initial vertical value (Ystart), a vertical product of said vertical increment value (Ystep) and a vertical count iteration value (Ycount), a number of memory addresses (Width) and a memory size value of said memory (402) per image frame (Base);

calculating addresses for said multiple windows within said image frame based upon said image parameter, wherein said calculating comprises;

counting with a horizontal counter (72X), wherein the horizontal counter (72X) includes a horizontal output of X=X+Xstep, wherein the horizontal counter (72X) has inputs from the horizontal increment value (Xstep) and the horizontal value (XStart);

connecting an output of a comparator (74X) to another input of the horizontal counter (72X), wherein the comparator (74X) compares the horizontal output of X=X+Xstep with the horizontal roduct of the horizontal increment value (Xstep) and the horizontal count iteration value (Xcount);

counting with a vertical counter (72Y), wherein the vertical counter (72Y) includes a vertical output of Y=Y+Ystep, wherein the vertical counter (72Y) has inputs from the vertical increment value (Ystep), the vertical value (YStart) and a signal (DONE) from the horizontal counter (72X);

connecting an output of a comparator (74Y) to another input of the vertical counter (72Y), wherein the comparator (74Y) compares the vertical output of Y=Y+Ystep with the vertical product of the vertical increment value (Ystep) and the vertical count iteration value (Ycount);

multiplying with a multiplier (76), wherein a multiplier output of the a multiplier (76) is the vertical output of Y=Y+Ystep multiplied with the number of memory addresses (Width);

adding with an adder (78), wherein a first adder output of the adder (78) is the multiplier output added with the horizontal output of X=X+Xstep;

adding with a second adder (79), wherein a second adder output (NextADDR) of the second adder (79) is the first adder output added with memory size value of said memory (402) per image frame (Base); and

processing in parallel image data of said multiple windows stored at said addresses,

based on said programming, said address generator (408) generating a break signal;

receiving by the hardware said break signal, thereby breaking execution of said looping code of the vector processor (404) over the image frame.

18. The method according to claim 17, further comprising receiving a VLIW instruction (422) for said processing in parallel.

19. The method according to claim 18, wherein said processing in parallel is performed by the vector processor (404) according to said VLIW instruction (422), wherein said loading said window parameter is performed externally to said VLIW instruction (422).

20. The method according to claim 17, further comprising receiving by said address generator (408) only a single bit of said VLIW instruction (422) to enable said calculating.