Difference between revisions of "SoC 2009/ARM NEON Optimization"

From VideoLAN Wiki
Jump to navigation Jump to search
(New page: {{SoCProject|year=2009|student=David Conrad|mentor=Holger Lubitz}} == Abstract == Formerly used primarily in PDAs and cellphones, ARM is starting to target netbooks and laptops with its n...)
 
 
(6 intermediate revisions by the same user not shown)
Line 18: Line 18:
 
|5 || Chroma MC ||style="background: #ffff00"| In progress || Ported ffmpeg's and the general case works, but not the special cases
 
|5 || Chroma MC ||style="background: #ffff00"| In progress || Ported ffmpeg's and the general case works, but not the special cases
 
|-
 
|-
−
|6 || (I)DCT ||style="background: #ffff00"| In Progress || 4x4 (I)DCT done, 8x8 not started
+
|6 || (I)DCT ||style="background: #00ff00"| Done || 4x4 (I)DCT and 8x8 IDCT done, 8x8 DCT not started
 
|-
 
|-
 
|7 || (De)quant ||style="background: #ffff00"| In Progress || 4x4 quant done, rest not started
 
|7 || (De)quant ||style="background: #ffff00"| In Progress || 4x4 quant done, rest not started
 
|-
 
|-
−
|8 || SSIM ||style="background: #ffff00"| In Progress || core done, end in progress
+
|8 || SSIM ||style="background: #00ff00"| Done ||
 
|-
 
|-
 
|9 || Zigzag ||style="background: #ffff00"| In Progress || 4x4 scan/sub frame done, rest in progress
 
|9 || Zigzag ||style="background: #ffff00"| In Progress || 4x4 scan/sub frame done, rest in progress
 
|-
 
|-
−
|10 || Hadamard AC || Not Started ||
+
|10 || Hadamard AC || style="background: #00ff00"| Done  ||
 
|-
 
|-
−
|11 || Intra Prediction || Not Started ||
+
|11 || Intra Prediction || style="background: #ffff00"| In Progress || Simpler functions done
 
|-
 
|-
−
|12 || sa8d || Not Started ||
+
|12 || sa8d || style="background: #00ff00"| Done ||
 
|-
 
|-
−
|13 || pixel variance || Not Started ||
+
|13 || pixel variance || style="background: #00ff00"| Done ||
 
|-
 
|-
−
|14 || lowres_init || Not Started ||
+
|14 || lowres_init ||style="background: #00ff00"| Done ||
 
|-
 
|-
−
|15 || Loop Filter || Not Started ||
+
|15 || hpel_filter || Not Started ||
 
|-
 
|-
−
|16 || intra sad/satd/sa8d || Not Started ||
+
|16 || Loop Filter || Not Started ||
 
|-
 
|-
−
|17 || Integral init || Not Started ||
+
|17 || intra sad/satd/sa8d || Not Started ||
 
|-
 
|-
−
|18 || Denoise || Not Started ||
+
|18 || Integral init || Not Started ||
 
|-
 
|-
−
|19 || Decimate score || Not Started ||
+
|19 || Denoise || Not Started ||
 
|-
 
|-
−
|20 || Coeff last/level_run || Not Started ||
+
|20 || Decimate score || Not Started ||
 
|-
 
|-
−
|21 || cabac || Not Started ||
+
|21 || Coeff last/level_run || Not Started ||
 
|-
 
|-
−
|22 || ARMv6 optimizations ||style="background: #ffff00"| In progress || For smaller functions, write code using ARMv6's simd extensions. For instance, ARMv6 SAD width 4 is faster on Cortex-A8.
+
|22 || cabac || Not Started ||
 
|-
 
|-
−
|23 || Cortex-A9 optimizations ||style="background: #ffff00"| In progress || Primarily consists of running checkasm --bench on an A9, noting whether ARMv6 or NEON is faster, figuring out why, and ensuring that the faster function is selected.
+
|23 || ARMv6 optimizations ||style="background: #ffff00"| In progress || For smaller functions, write code using ARMv6's simd extensions. For instance, ARMv6 SAD width 4 is faster on Cortex-A8.
 
|-
 
|-
−
|24 || gcc optimization || Not Started || Investigate what gcc produces for important C functions and attempt to optimize it for ARM
+
|24 || Cortex-A9 optimizations ||style="background: #ffff00"| In progress || Primarily consists of running checkasm --bench on an A9, noting whether ARMv6 or NEON is faster, figuring out why, and ensuring that the faster function is selected.
 
|-
 
|-
−
|25 || Performance Counter Optimization || Not Started || Utilize the ARMv7 performance counters to optimize stalls, cache usage, NEON->ARM data hazards, and more
+
|25 || gcc optimization || Not Started || Investigate what gcc produces for important C functions and attempt to optimize it for ARM
 +
|-
 +
|26 || Performance Counter Optimization || Not Started || Utilize the ARMv7 performance counters to optimize stalls, cache usage, NEON->ARM data hazards, and more
 
|}
 
|}
 +
 +
== Repository ==
 +
I'm currently using gitorious to publish my work [http://gitorious.org/projects/x264-arm]
  
 
== Timeline ==
 
== Timeline ==
Line 68: Line 73:
 
''' May 23''' Program Starts
 
''' May 23''' Program Starts
  
−
''' May 23 - July 13 ''' Work on writing NEON asm for
+
''' May 23 - June 14''' Code
 +
 
 +
''' June 15 - June 21''' Vacation on Chincoteague
  
−
''' June 13 - July 13''' Code
+
''' June 22 - July 12''' Code
  
 
''' July 13''' Mid Term deadline. Specific goals pending, but I'd like to have none of the functions without asm equivalents to be more than about 2% of common profiles. In other words, all the major functions would have been written in NEON ASM
 
''' July 13''' Mid Term deadline. Specific goals pending, but I'd like to have none of the functions without asm equivalents to be more than about 2% of common profiles. In other words, all the major functions would have been written in NEON ASM

Latest revision as of 19:27, 1 July 2009

This project is part of Google Summer of Code 2009.
Student: David Conrad
Mentor: Holger Lubitz

Abstract

Formerly used primarily in PDAs and cellphones, ARM is starting to target netbooks and laptops with its new Cortex-A series of processors. One key feature of these new processors is the NEON SIMD unit, the use of which will massively boost the performance of many multimedia applications. This project will consist of writing NEON/ARMv6 SIMD assembly for all of the major DSP functions in x264, ideally speeding x264 up by over 4-5 times.

Goals

No. Task Description Status Comment
1 ARM framework Mostly Done Waiting on more info on CPUID emulation
2 SAD/SATD Done Qualifying task
3 SSD Done
4 Luma MC In progress get_ref_neon complete
5 Chroma MC In progress Ported ffmpeg's and the general case works, but not the special cases
6 (I)DCT Done 4x4 (I)DCT and 8x8 IDCT done, 8x8 DCT not started
7 (De)quant In Progress 4x4 quant done, rest not started
8 SSIM Done
9 Zigzag In Progress 4x4 scan/sub frame done, rest in progress
10 Hadamard AC Done
11 Intra Prediction In Progress Simpler functions done
12 sa8d Done
13 pixel variance Done
14 lowres_init Done
15 hpel_filter Not Started
16 Loop Filter Not Started
17 intra sad/satd/sa8d Not Started
18 Integral init Not Started
19 Denoise Not Started
20 Decimate score Not Started
21 Coeff last/level_run Not Started
22 cabac Not Started
23 ARMv6 optimizations In progress For smaller functions, write code using ARMv6's simd extensions. For instance, ARMv6 SAD width 4 is faster on Cortex-A8.
24 Cortex-A9 optimizations In progress Primarily consists of running checkasm --bench on an A9, noting whether ARMv6 or NEON is faster, figuring out why, and ensuring that the faster function is selected.
25 gcc optimization Not Started Investigate what gcc produces for important C functions and attempt to optimize it for ARM
26 Performance Counter Optimization Not Started Utilize the ARMv7 performance counters to optimize stalls, cache usage, NEON->ARM data hazards, and more

Repository

I'm currently using gitorious to publish my work [1]

Timeline

This is an estimation.

March-April Qualifying task, work on the more important (>5% in profiles) functions

May Final projects and exams

May 23 Program Starts

May 23 - June 14 Code

June 15 - June 21 Vacation on Chincoteague

June 22 - July 12 Code

July 13 Mid Term deadline. Specific goals pending, but I'd like to have none of the functions without asm equivalents to be more than about 2% of common profiles. In other words, all the major functions would have been written in NEON ASM

July 13 - August 17 Code

August 17 Program end

Retrieved from "https://wiki.videolan.org/index.php?title=SoC_2009/ARM_NEON_Optimization&oldid=14458"