YaConv: Convolution with Low Cache Footprint

نویسندگان

چکیده

This article introduces YaConv , a new algorithm to compute convolution using GEMM microkernels from Basic Linear Algebra Subprograms library that is efficient for multiple CPU architectures. Previous approaches either create copy of each image element filter or reload these elements into cache call, leading redundant instances the in cache. Instead, loads once and maximizes reuse elements. The output computed by scattering results microkernel calls correct locations image. main advantage this algorithm—which leads better performance comparison existing im2col approach on several architectures—is more use memory hierarchy. experimental evaluation convolutional layers PyTorch, along with parameterized study, indicates an average 24% speedup over convolution. Increased comes as result 3× reduction L3 accesses 2× fewer branch instructions.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Low Carbon Footprint Routes for Bird Watching

Bird watching is one of many recreational activities popular in ecotourism. Its popularity, therefore, prompts the need for studies on energy conservation. One such environmentally friendly approach toward minimizing bird watching’s ecological impact is ensuring a reduced carbon footprint by using an economic travel itinerary comprising a series of connected routes between tourist attractions t...

متن کامل

Based Low Footprint Security Chip for Portable

Elliptic Curve Cryptography is one of the most preferred public key Cryptography algorithms known for its security strength and reduced key size. This makes ECC most suitable for implementing the hardware security chip for providing security services in portable embedded devices. This paper presents an area efficient Security Chip housing ECC algorithm for data encryption over binary field and ...

متن کامل

Efficient Way-based Cache Partitioning for Low-Associativity Cache

Cache Partitioning is well-known technique to reduce destructive interference among co-running applications in a shared last-level cache (SLLC). Way-based cache partitioning is a popular partitioning scheme due to its simplicity, but it can dramatically reduce associativity of each partition. Also, most SLLC have limited associativity because the higher associativity causes the higher cache acc...

متن کامل

Distributed computing for carbon footprint reduction by exploiting low-footprint energy availability

Low carbon footprint energy sources such as solar and wind power typically suffer from unpredictable or limited availability. By globally distributing a number of these renewable sources, these effects can largely be compensated for. We look at the feasibility of this approach for powering already distributed data centers in order to operate at a reduced total carbon footprint. From our study w...

متن کامل

A Low Energy Set-Associative I-Cache with Extended BTB

This paper proposes a low-energy instruction-cache architecture, called history-based tag-comparison (HBTC) cache. The HBTC cache attempts to re-use tag-comparison results for avoiding unnecessary way activation in setassociative caches. The cache records tag-comparison results in an extended BTB, and re-uses them for directly selecting only the hit-way which includes the target instruction. In...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: ACM Transactions on Architecture and Code Optimization

سال: 2023

ISSN: ['1544-3973', '1544-3566']

DOI: https://doi.org/10.1145/3570305