[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-343239-105":59,"doc-detail-343239-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","alp-adaptive-lossless-floating-point-compression","ALP: Adaptive Lossless Floating-Point Compression","","IEEE 754 doubles introduce rounding errors during computation and when serializing to text, which prevents popular lightweight compression encodings like Delta and Frame Of Reference (FOR) from working reliably. Prior floating-point methods such as Gorilla, Chimp128, PseudoDecimals (PDE), Elf, and Patas improve compression but do not surpass general-purpose tools like Zstd and often suffer from slow decompression. ALP enhances speed and compression ratio via adaptive, vectorized lossless encoding.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":30,"@type":76,"position":81},"https://docshare.wps.com/document/technology/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/alp-adaptive-lossless-floating-point-compression/343239/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/alp-adaptive-lossless-floating-point-compression/343239.png","ImageObject",300,407,{"name":92,"@type":93},"Franzy","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-27","2026-09-22",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",11,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"Why do rounding errors make Delta and Frame Of Reference (FOR) unsuitable for raw doubles?","Question",{"text":113,"@type":114},"IEEE 754 doubles do not represent most real values exactly, so operations and serialization introduce rounding errors. Delta and FOR rely on arithmetic with fixed bases or previous values, so the rounding errors break their assumptions for lossless encoding.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What makes ALP faster and improve compression ratio compared with earlier schemes?",{"text":118,"@type":114},"ALP is designed for vectorized execution and uses a strongly enhanced PseudoDecimals approach to losslessly encode doubles as integers when they originated as decimals. Otherwise it applies vectorized compression to the doubles’ front bits, and its implementation uses auto-vectorizing scalar code plus an efficient two-stage algorithm.",{"name":120,"@type":111,"acceptedAnswer":121},"How does ALP compare to Zstd and other recent floating-point encodings?",{"text":122,"@type":114},"The evaluation reports that ALP is 1–2 orders of magnitude faster in decompression than competing schemes while maintaining an excellent compression ratio. Only Zstd achieves a similar compression ratio, but it is slow and block-based, while Elf is inferior across performance metrics.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},343239,1790686833,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":29,"category_name":30,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":145,"read_time":146},2336478945519,"https://ap-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45","ALP: Adaptive Lossless floating-Point Compression  \nAzim Afroozeh  \nCWI Amsterdam, The Netherlands  \nLeonardo Kuffó  \nCWI Amsterdam, The Netherlands  \nPeter Boncz  \nCWI Amsterdam, The Netherlands  \nABSTRACT  \nIEEE 754 doubles do not exactly represent most real values, introducing rounding errors in computations and [de]serialization to text. These rounding errors inhibit the use of existing lightweight compression schemes such as Delta and Frame Of Reference (FOR), but recently new schemes were proposed: Gorilla, Chimp128, PseudoDecimals (PDE), Elf and Patas. However, their compression ratios are not better than those of general-purpose compressors such as Zstd; while [de]compression is much slower than Delta and FOR.  \nWe propose and evaluate ALP, that significantly improves these previous schemes in both speed and compression ratio (Figure 1) . We created ALP after carefully studying the datasets used to evaluate the previous schemes. To obtain speed, ALP is designed to fit vectorized execution. This turned out to be key for also improving the compression ratio, as we found in-vector commonalities to create compression opportunities. ALP is an adaptive scheme that uses a strongly enhanced version of PseudoDecimals [31] to losslessly encode doubles as integers if they originated as decimals, and otherwise uses vectorized compression of the doubles’ front bits. Its high speeds stem from our implementation in scalar code that auto-vectorizes, using building blocks provided by our FastLanes library [6], and an efficient two-stage compression algorithm that first samples row-groups and then vectors.  \nKEYWORDS  \nlossless compression, floating point compression, lightweight compression, vectorized execution, columnar storage, big data formats  \nACM Reference Format:  \nAzim Afroozeh, Leonardo Kuffó, and Peter Boncz. 2024. ALP: Adaptive Lossless floating-Point Compression. In Proceedings of The 2024 International Conference on Management of Data (SIGMOD’24) . ACM, New York, NY, USA, 14 pages. [https://doi.org/XXXXXXX.XXXXXXX](https://doi.org/XXXXXXX.XXXXXXX)  \n1 INTRODUCTION  \nData analytics pipelines manipulate floating-point numbers (64-bit doubles) more frequently than classical enterprise database workloads, which typically rely on fixed-point decimals (systems often store these as 64-bit integers). Floating-point data is also a natural fit in scientific and sensor data; and can have a temporal component, yielding time series.  \nAnalytical data systems and big data formats have adopted columnar compressed storage [4, 12, 37, 41, 50, 51], where the compression in storage is either provided by general-purpose or  \nSIGMOD’24, June 09–15, 2024, Santiago, Chile © 2024 Association for Computing Machinery.  \nThis is the author’s version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in Proceedings of The 2024 International Conference on Management of Data (SIGMOD’24), [https:](https:)//[doi.org/XXXXXXX.XXXXXXX](doi.org/XXXXXXX.XXXXXXX).  \nDecom ression S eedas Tu les er CPU Cycle (Log Scale)  \n101  \n100  \n10−1  \n10−2  \n10−3 10−2 10−1  \nCompression S eed  \nas Tu les er CPU Cycle (Log Scale)  \nFigure 1: Compression performance for all schemes (on Intel Ice Lake). Each dot is one dataset. ALP is 1-2 orders of magnitude faster in [de]compression than all competing schemes, while providing an excellent compression ratio. The only one to achieve a compression ratio similar to ALP is Zstd, but it is slow and block-based (one cannot skip through compressed data). Elf is inferior to Zstd on all performance metrics. The evaluation framework is presented in Section 4.  \nlightweight compression. Lightweight methods, also called \"encodings\", exploit knowledge of the type and domain of a column. Examples are Frame Of Reference (FOR), Delta-, Dictionary-, and Run Length Encoding (RLE) [20, 44, 46] . The first two are used on high-cardinality columns and encode values as the","cbCaibgsbTJPKWmH","https://ap.wps.com/l/cbCaibgsbTJPKWmH","pdf",1117147,14,"English","# 1 Introduction\n## Overview of floating-point data in analytics pipelines\n# Compression background\n## Lightweight encodings and rounding-error limitations\n## General-purpose compressors in big data formats\n# 2 Related floating-point encodings\n## Gorilla, Chimp, Chimp128\n## PseudoDecimals (PDE), Patas, Elf\n# 3 ALP proposal and evaluation","[{\"question\":\"Why do rounding errors make Delta and Frame Of Reference (FOR) unsuitable for raw doubles?\",\"answer\":\"IEEE 754 doubles do not represent most real values exactly, so operations and serialization introduce rounding errors. Delta and FOR rely on arithmetic with fixed bases or previous values, so the rounding errors break their assumptions for lossless encoding.\"},{\"question\":\"What makes ALP faster and improve compression ratio compared with earlier schemes?\",\"answer\":\"ALP is designed for vectorized execution and uses a strongly enhanced PseudoDecimals approach to losslessly encode doubles as integers when they originated as decimals. Otherwise it applies vectorized compression to the doubles’ front bits, and its implementation uses auto-vectorizing scalar code plus an efficient two-stage algorithm.\"},{\"question\":\"How does ALP compare to Zstd and other recent floating-point encodings?\",\"answer\":\"The evaluation reports that ALP is 1–2 orders of magnitude faster in decompression than competing schemes while maintaining an excellent compression ratio. Only Zstd achieves a similar compression ratio, but it is slow and block-based, while Elf is inferior across performance metrics.\"}]","ALP: Adaptive Lossless Floating-Point Compression | PDF",1790050396,35]