Project plan
Hello, this blog post is to document my progress on the SPO600 project, I have selected a potential package to optimize the LZMA SDK found at https://www.7-zip.org/sdk.html.
At first I was unsure if I wanted to try working with the package because of how the directory's and make files were setup. But there is a folder called ASM containing assembler optimizations for both arm and x86 systems, however is only one optimization for arm32 bit systems used for crc calculations while there are many for x86. If it was worth while to write an optimization for arm32 it may be beneficial to arm64 systems to have that code ported.
So my initial plan for this project will be to attempt to port the code in the file 7zCrcOpt.asm to run on arm64 bit systems, I will first benchmark a use case that world normally use this 32bit optimization on a 64bit system to get a idea of how the performance currently would be without it. Then I will attempt to port the code and re run the same series of tests with the same code but this time with the optimization, This will show us weather or not there is additional benefit to having one more optimization.
Since the optimization is for arm 64 only I will not test on other systems as it will only have impact on the target system. I should also add I am not yet fully committed to this as a project and might choose to work on another package in witch case I will post a new plan. But for the time being this is what I intend to do,
At first I was unsure if I wanted to try working with the package because of how the directory's and make files were setup. But there is a folder called ASM containing assembler optimizations for both arm and x86 systems, however is only one optimization for arm32 bit systems used for crc calculations while there are many for x86. If it was worth while to write an optimization for arm32 it may be beneficial to arm64 systems to have that code ported.
So my initial plan for this project will be to attempt to port the code in the file 7zCrcOpt.asm to run on arm64 bit systems, I will first benchmark a use case that world normally use this 32bit optimization on a 64bit system to get a idea of how the performance currently would be without it. Then I will attempt to port the code and re run the same series of tests with the same code but this time with the optimization, This will show us weather or not there is additional benefit to having one more optimization.
Since the optimization is for arm 64 only I will not test on other systems as it will only have impact on the target system. I should also add I am not yet fully committed to this as a project and might choose to work on another package in witch case I will post a new plan. But for the time being this is what I intend to do,
Comments
Post a Comment