[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-webassembly-runtimes-2026-who-got-faster-zh":3,"article-related-webassembly-runtimes-2026-who-got-faster-zh":37,"series-industry-835183a4-b35d-4171-9689-c836901454c8":82},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":29,"views":33,"created_at":34,"published_at":35,"topic_cluster_id":36},"835183a4-b35d-4171-9689-c836901454c8","webassembly-runtimes-2026-who-got-faster-zh","2026 WebAssembly 執行器誰變快了","\u003Cp data-speakable=\"summary\">這篇比較 7 個 WebAssembly 執行器在 2026 年的 libsodium 加密基準，幫你判斷哪個最適合追求速度。\u003C\u002Fp>\u003Cp>看完這份\u003Ca href=\"\u002Fnews\u002F2016-star-awesome-harness-engineering-list-zh\">清單\u003C\u002Fa>，你可以直接判斷：你的 Wasm 工作負載該優先測哪個 runtime，以及是否值得為 wide_arithmetic 重新跑一次基準。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>2024\u003C\u002Fth>\u003Cth>2025\u003C\u002Fth>\u003Cth>2026\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Wasmtime\u003C\u002Ftd>\u003Ctd>2.67x native\u003C\u002Ftd>\u003Ctd>2.54x native\u003C\u002Ftd>\u003Ctd>2.41x native\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Node\u003C\u002Ftd>\u003Ctd>8.60x native\u003C\u002Ftd>\u003Ctd>7.95x native\u003C\u002Ftd>\u003Ctd>7.95x native\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>WAMR\u003C\u002Ftd>\u003Ctd>n\u002Fa\u003C\u002Ftd>\u003Ctd>1.59x native\u003C\u002Ftd>\u003Ctd>1.57x native\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>wasm2c\u003C\u002Ftd>\u003Ctd>2.83x native\u003C\u002Ftd>\u003Ctd>2.66x native\u003C\u002Ftd>\u003Ctd>2.49x native\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>WasmEdge\u003C\u002Ftd>\u003Ctd>2.20x native\u003C\u002Ftd>\u003Ctd>2.02x native\u003C\u002Ftd>\u003Ctd>1.74x native\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Wasmer\u003C\u002Ftd>\u003Ctd>2.30x native\u003C\u002Ftd>\u003Ctd>2.37x native\u003C\u002Ftd>\u003Ctd>2.08x native\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Bun\u003C\u002Ftd>\u003Ctd>12.45x native\u003C\u002Ftd>\u003Ctd>26.03x native\u003C\u002Ftd>\u003Ctd>8.77x native\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>1. Wasmtime：穩定變快，還吃得到新指令\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fwasmtime.dev\u002F\">Wasmtime\u003C\u002Fa> 是這組資料裡最像「穩定進步」的例子。基礎版本從 2024 的 2.67x native，走到 2026 的 2.41x native，速度提升不誇張，但趨勢很清楚。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782456472060-bs3x.png\" alt=\"2026 WebAssembly 執行器誰變快了\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>真正的轉折在 wide_arithmetic。2026 版一開這個特性，Wasmtime 直接降到 1.46x native，對 libsodium 這類算術密集工作是很實際的收益。\u003C\u002Fp>\u003Cul>\u003Cli>2026 基準：2.41x native\u003C\u002Fli>\u003Cli>wide_arithmetic：1.46x native\u003C\u002Fli>\u003Cli>適合：能利用新版 Wasm 指令的加密程式\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>2. Wasmer：開啟 wide_arithmetic 後最有爆發力\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fwasmer.io\u002F\">Wasmer\u003C\u002Fa> 的基準歷史不算一路順風，但 2026 版很關鍵，因為它終於明顯吃到 wide_arithmetic 的紅利。純基礎版是 2.08x native，屬於穩健但不特別突出。\u003C\u002Fp>\u003Cp>一旦啟用 wide_arithmetic，Wasmer 下降到 1.33x native，成為這組測試裡最快的完整現代結果。對 CPU-heavy 的 crypto 工作，這種差距足以改變部署選擇。\u003C\u002Fp>\u003Cul>\u003Cli>2026 基準：2.08x native\u003C\u002Fli>\u003Cli>wide_arithmetic：1.33x native\u003C\u002Fli>\u003Cli>重點：功能支援比單看基準更重要\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>3. WAMR：小而快，但不吃新算術指令\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fbytecodealliance\u002Fwasm-micro-runtime\">WAMR\u003C\u002Fa> 的故事比較安靜。在 AOT 模式下，2025 與 2026 版都維持在 1.6x native 左右，對可攜式 WebAssembly 來說已經很強。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782456472409-o3wr.png\" alt=\"2026 WebAssembly 執行器誰變快了\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>\u003Ca href=\"\u002Fnews\u002Fxiaomi-mimo-v2-5-pro-pricing-benchmarks-limits-zh\">限制\u003C\u002Fa>也很明確。WAMR 能跑基礎版和 SIMD-lime 變體，但 wide_arithmetic 會因為不支援的 opcode 直接失敗。如果你的程式需要最新算術指令，它不是目前最穩的選擇。\u003C\u002Fp>\u003Cul>\u003Cli>2025 基準：1.59x native\u003C\u002Fli>\u003Cli>2026 基準：1.57x native\u003C\u002Fli>\u003Cli>限制：這次測試不支援 wide_arithmetic\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>4. WasmEdge：模式選對，成績差很多\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fwasmedge.org\u002F\">WasmEdge\u003C\u002Fa> 的進步有感，但這份測試也提醒一件事：執行模式會嚴重影響結果。2026 年一開始誤用 interpreter mode，成績看起來很差，改成 AOT 後才回到合理水準。\u003C\u002Fp>\u003Cp>在 AOT 下，2026 基準落在 1.74x native，比前幾年更好，足以把它放進「速度不錯」的名單。前提是你得自己控制執行模式。\u003C\u002Fp>\u003Cul>\u003Cli>2024 基準：2.20x native\u003C\u002Fli>\u003Cli>2025 基準：2.02x native\u003C\u002Fli>\u003Cli>2026 AOT 基準：1.74x native\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>5. wasm2c：走編譯鏈，換來穩定可控\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FWebAssembly\u002Fwabt\u002Ftree\u002Fmain\u002Fwasm2c\">wasm2c\u003C\u002Fa> 是比較務實的路線：先把 Wasm 轉成 C，再交給主機編譯器處理。它不追 runtime 花招，但在這組測試裡一直維持不錯的表現。\u003C\u002Fp>\u003Cp>2026 基準是 2.49x native，雖然不是最快，但改善幅度穩定。若你的建置流程本來就偏向原生編譯，這條路很順。\u003C\u002Fp>\u003Ccode>wasm2c -> C -> zig cc -O3 -march=native\u003C\u002Fcode>\u003Ch2>6. Node：慢慢進步，但還離最快很遠\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fnodejs.org\u002F\">Node\u003C\u002Fa> 的改善幅度不大，但至少不是停滯。基準從 2024 的 8.60x native，走到 2025 的 7.95x native，2026 也大致維持在這個區間。\u003C\u002Fp>\u003Cp>它還暴露了一個實務問題：某些 libsodium 測試要先設好明確的最大線性記憶體，否則會失敗。也就是說，能不能跑完，有時比跑多快更重要。\u003C\u002Fp>\u003Cul>\u003Cli>2024 基準：8.60x native\u003C\u002Fli>\u003Cli>2026 基準：約 7.95x native\u003C\u002Fli>\u003Cli>注意：要設 explicit linear memory\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>7. Bun：一年內變化最大，但波動也最大\u003C\u002Fh2>\u003Cp>\u003Ca href=\"https:\u002F\u002Fbun.sh\u002F\">Bun\u003C\u002Fa> 是這份資料裡年際變化最戲劇化的項目。2024 和 2025 的結果都落後不少，但 2026 版把差距大幅縮小，基準成績來到 8.77x native。\u003C\u002Fp>\u003Cp>這仍然比 Node 慢，但改善幅度夠大，代表它不再是這類工作負載的明顯外圍選項。如果團隊已經在用 Bun，2026 數字值得重新\u003Ca href=\"\u002Fnews\u002Fmoneygram-solana-validator-infrastructure-not-marketing-zh\">驗證\u003C\u002Fa>自己的程式。\u003C\u002Fp>\u003Cul>\u003Cli>2024 基準：12.45x native\u003C\u002Fli>\u003Cli>2025 基準：26.03x native\u003C\u002Fli>\u003Cli>2026 基準：8.77x native\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>哪種適合你：先看功能，再看名次\u003C\u002Fh2>\u003Cp>如果你要的是 2026 年這組 crypto Wasm 測試裡的最快結果，先測 Wasmer 和 Wasmtime，再確認你的模組能不能用 wide_arithmetic。這兩個 runtime 吃到新指令後，排名會明顯翻動。\u003C\u002Fp>\u003Cp>如果你重視可攜性、AOT、或既有建置流程，WAMR、WasmEdge 和 wasm2c 更值得先試。Node 和 Bun 都在進步，但就這個工作負載來看，還是落後專用 runtime。最後的結論很簡單：跑你自己的模組，因為執行模式和功能支援，常常比名字更重要。\u003C\u002Fp>","7 個 WebAssembly 執行器在 2026 年的 libsodium 基準顯示明顯分化，wide_arithmetic 對加密效能的影響最大。","00f.net","https:\u002F\u002F00f.net\u002F2026\u002F06\u002F23\u002Fwebassembly-runtimes-2026\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782456472060-bs3x.png","industry","zh","f55523f3-f80c-42f0-a97e-b4bd2c527565",[17,18,19,20,21,22,23,24,25,26,27,28],"WebAssembly","runtime","libsodium","Wasmtime","Wasmer","WAMR","WasmEdge","wasm2c","Node","Bun","wide_arithmetic","AOT",[30,31,32],"wide_arithmetic 是 2026 年影響加密效能最大的變數。","Wasmer 和 Wasmtime 在啟用新指令後最接近最快結果。","WAMR、WasmEdge、wasm2c 適合重視部署方式與可攜性的團隊。",0,"2026-06-26T06:47:26.158778+00:00","2026-06-26T06:47:26.15+00:00","caa87b65-9bbc-46fe-bba8-4f4158dd2d8b",{"tags":38,"relatedLang":41,"relatedPosts":45},[39],{"name":17,"slug":40},"webassembly",{"id":15,"slug":42,"title":43,"language":44},"webassembly-runtimes-2026-who-got-faster-en","WebAssembly runtimes in 2026: who got faster","en",[46,52,58,64,70,76],{"id":47,"slug":48,"title":49,"cover_image":50,"image_url":50,"created_at":51,"category":13},"e5ddd81c-7a48-4394-8271-71863ee7034c","product-hunt-best-prompt-engineering-tools-2026-zh","5 款 Prompt 工具各自最適合的工作","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782507772922-ljrs.png","2026-06-26T21:02:19.140423+00:00",{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":13},"bc15b4e6-06b0-4237-869b-23ec02cc24e1","xcode-266-gemini-ai-coding-stack-zh","Xcode 26.6 把 Gemini 納入 Apple AI 編碼堆疊","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782497870523-0glb.png","2026-06-26T18:17:25.179841+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":13},"4edf30e4-8159-43f6-aa25-073d98a971cf","openai-anthropic-must-sell-efficiency-not-excess-zh","OpenAI 與 Anthropic 必須賣效率，不是賣耗用","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782496966108-3grw.png","2026-06-26T18:02:19.661289+00:00",{"id":65,"slug":66,"title":67,"cover_image":68,"image_url":68,"created_at":69,"category":13},"315693f2-be56-411d-a51f-4d1d7f94d29c","ai-code-review-tools-catch-issues-earlier-zh","10 款 AI 程式碼審查工具，提早抓出問題","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782491579498-09ad.png","2026-06-26T16:32:31.773823+00:00",{"id":71,"slug":72,"title":73,"cover_image":74,"image_url":74,"created_at":75,"category":13},"bfdb1e4f-5d0d-47b7-9d77-18ffbe8505fc","openai-ipo-delay-turns-hype-into-caution-zh","OpenAI 延後 IPO，把熱度變警惕","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782489796442-d5rz.png","2026-06-26T16:02:50.943965+00:00",{"id":77,"slug":78,"title":79,"cover_image":80,"image_url":80,"created_at":81,"category":13},"6060d642-55b6-444c-811d-60388b99d5d1","suno-launches-spark-indie-artists-zh","Suno 用 Spark 拉攏獨立音樂人","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782486181801-h39c.png","2026-06-26T15:02:31.668699+00:00",[83,88,93,98,103,108,113,118,123,128],{"id":84,"slug":85,"title":86,"created_at":87},"ee073da7-28b3-4752-a319-5a501459fb87","ai-in-2026-what-actually-matters-now-zh","2026 AI 真正重要的事","2026-03-26T07:09:12.008134+00:00",{"id":89,"slug":90,"title":91,"created_at":92},"83bd1795-8548-44c9-9a7e-de50a0923f71","trump-ai-framework-power-speech-state-preemption-zh","川普 AI 框架瞄準電力、言論與州權","2026-03-26T07:12:18.695466+00:00",{"id":94,"slug":95,"title":96,"created_at":97},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","2026-03-26T07:14:26.62638+00:00",{"id":99,"slug":100,"title":101,"created_at":102},"4bcec76f-4c36-4daa-909f-54cd702f7c93","claude-users-spreading-out-and-getting-better-zh","Claude 用戶更分散，也更會用","2026-03-26T07:22:52.325888+00:00",{"id":104,"slug":105,"title":106,"created_at":107},"bd903b15-2473-4178-9789-b7557816e535","openclaw-raises-hard-question-for-ai-models-zh","OpenClaw 逼問 AI 模型價值","2026-03-26T07:24:54.707486+00:00",{"id":109,"slug":110,"title":111,"created_at":112},"eeac6b9e-ad9d-4831-8eec-8bba3f9bca6a","gap-google-gemini-checkout-fashion-search-zh","Gap 把結帳搬進 Gemini","2026-03-26T07:28:23.937768+00:00",{"id":114,"slug":115,"title":116,"created_at":117},"0740e53f-605d-4d57-8601-c10beb126f3c","google-pushes-gemini-transition-to-march-2026-zh","Google 把 Gemini 轉換延到 2026 年 3…","2026-03-26T07:30:12.825269+00:00",{"id":119,"slug":120,"title":121,"created_at":122},"e660d801-2421-4529-8fa9-86b82b066990","metas-llama-4-benchmark-scandal-gets-worse-zh","Meta Llama 4 分數風波又擴大","2026-03-26T07:34:21.156421+00:00",{"id":124,"slug":125,"title":126,"created_at":127},"183f9e7c-e143-40bb-a6d5-67ba84a3a8bc","accenture-mistral-ai-sovereign-enterprise-deal-zh","Accenture 攜手 Mistral AI 賣主權 AI","2026-03-26T07:38:14.818906+00:00",{"id":129,"slug":130,"title":131,"created_at":132},"191d9b1b-768a-478c-978c-dd7431a38149","mistral-ai-faces-its-hardest-year-yet-zh","Mistral AI 迎來最硬的一年","2026-03-26T07:40:23.716374+00:00"]