https://jaehun.me/tags/programming-languages/Recent content in Programming Languages on Jaehun's BlogHugoko-krSat, 19 Sep 2026 01:44:51 +0900https://jaehun.me/posts/xgrammar-flexible-and-efficient-structured-generation-engine-for-large-language-models/Mon, 02 Jun 2025 00:00:00 +0900https://jaehun.me/posts/xgrammar-flexible-and-efficient-structured-generation-engine-for-large-language-models/<p><a href="https://arxiv.org/abs/2411.15100v3"target="_blank" class="inline-flex items-center gap-1" >논문 링크<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <h2 id="강점과-독창성-핵심-알고리즘-그리고-한계">강점과 독창성, 핵심 알고리즘, 그리고 한계<a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2> <blockquote class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1" > <div class="prose prose-sm max-w-none"> <p>물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘</p>https://jaehun.me/posts/mirage-a-multi-level-superoptimizer-for-tensor-programs/Sun, 13 Apr 2025 00:00:00 +0900https://jaehun.me/posts/mirage-a-multi-level-superoptimizer-for-tensor-programs/<p><a href="https://arxiv.org/abs/2405.05751v2"target="_blank" class="inline-flex items-center gap-1" >논문 링크<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <h2 id="강점과-독창성-핵심-알고리즘-그리고-한계">강점과 독창성, 핵심 알고리즘, 그리고 한계<a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2> <blockquote class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1" > <div class="prose prose-sm max-w-none"> <p>물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘</p>https://jaehun.me/posts/flex-attention-a-programming-model-for-generating-optimized-attention-kernels/Mon, 07 Apr 2025 00:00:00 +0900https://jaehun.me/posts/flex-attention-a-programming-model-for-generating-optimized-attention-kernels/<p><a href="https://arxiv.org/abs/2412.05496v1"target="_blank" class="inline-flex items-center gap-1" >논문 링크<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <h2 id="강점과-독창성-핵심-알고리즘-그리고-한계">강점과 독창성, 핵심 알고리즘, 그리고 한계<a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2> <blockquote class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1" > <div class="prose prose-sm max-w-none"> <p>물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘</p>https://jaehun.me/posts/scaling-deep-learning-training-with-mpmd-pipeline-parallelism/Mon, 10 Mar 2025 00:00:00 +0900https://jaehun.me/posts/scaling-deep-learning-training-with-mpmd-pipeline-parallelism/<p><a href="https://arxiv.org/abs/2412.14374"target="_blank" class="inline-flex items-center gap-1" >논문 링크<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <h2 id="강점과-독창성-핵심-알고리즘-그리고-한계">강점과 독창성, 핵심 알고리즘, 그리고 한계<a href="#%ea%b0%95%ec%a0%90%ea%b3%bc-%eb%8f%85%ec%b0%bd%ec%84%b1-%ed%95%b5%ec%8b%ac-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ea%b7%b8%eb%a6%ac%ea%b3%a0-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2> <blockquote class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1" > <div class="prose prose-sm max-w-none"> <p>물어본 것: 이 논문을 아주 자세하게 읽고 논문의 강점과 독창적인 지점을 설명해주고 핵심 알고리즘을 예시 입력을 들어서 전체적인 과정을 설명해줘 추가적으로 논문의 한계점에 대해서도 알려줘</p>https://jaehun.me/posts/hysynth-context-free-llm-approximation-for-guiding-program-synthesis/Sun, 10 Nov 2024 00:00:00 +0900https://jaehun.me/posts/hysynth-context-free-llm-approximation-for-guiding-program-synthesis/<p>논문 : <a href="https://arxiv.org/abs/2405.15880v2"target="_blank" class="inline-flex items-center gap-1" >https://arxiv.org/abs/2405.15880v2<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <p>아래글은 Scholar GPT와 대화를 나눈 내용입니다.</p> <h2 id="핵심-내용과-강점-알고리즘-한계">핵심 내용과 강점, 알고리즘, 한계<a href="#%ed%95%b5%ec%8b%ac-%eb%82%b4%ec%9a%a9%ea%b3%bc-%ea%b0%95%ec%a0%90-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2> <blockquote class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1" > <div class="prose prose-sm max-w-none"> <p>물어본 것: 논문의 핵심 내용과 강점, 알고리즘 설명, 그리고 한계점</p>https://jaehun.me/posts/meta-large-language-model-compiler-foundation-models-of-compiler-optimization/Thu, 07 Nov 2024 00:00:00 +0900https://jaehun.me/posts/meta-large-language-model-compiler-foundation-models-of-compiler-optimization/<p>논문 : <a href="https://arxiv.org/abs/2407.02524"target="_blank" class="inline-flex items-center gap-1" >https://arxiv.org/abs/2407.02524<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <p>아래글은 Scholar GPT와 대화를 나눈 내용입니다.</p> <h2 id="핵심-내용과-강점-알고리즘-한계">핵심 내용과 강점, 알고리즘, 한계<a href="#%ed%95%b5%ec%8b%ac-%eb%82%b4%ec%9a%a9%ea%b3%bc-%ea%b0%95%ec%a0%90-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ed%95%9c%ea%b3%84" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2> <blockquote class="border-primary bg-muted/30 hover:bg-muted/40 my-6 rounded-r-lg border-l-4 py-4 pr-4 pl-6 italic transition-all duration-200 ease-out hover:translate-x-1" > <div class="prose prose-sm max-w-none"> <p>물어본 것: 논문의 핵심 내용과 강점, 알고리즘 설명, 그리고 한계점</p>https://jaehun.me/posts/%EA%B0%84%EB%8B%A8%EB%85%BC%EB%AC%B8-%EC%A0%95%EB%A6%AC-fast-and-effective-orchestration-of-compiler-optimizationszhelong-panrudolf-eigenmannpurdue-university-cgo06/Fri, 12 Feb 2021 00:00:00 +0900https://jaehun.me/posts/%EA%B0%84%EB%8B%A8%EB%85%BC%EB%AC%B8-%EC%A0%95%EB%A6%AC-fast-and-effective-orchestration-of-compiler-optimizationszhelong-panrudolf-eigenmannpurdue-university-cgo06/<h2 id="제목">제목<a href="#%ec%a0%9c%eb%aa%a9" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2><p>Fast and Effective Orchestration of Compiler Optimizations</p> <h2 id="저자">저자<a href="#%ec%a0%80%ec%9e%90" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2><p>Zhelong Pan,Rudolf Eigenmann</p> <h2 id="motivation">Motivation<a href="#motivation" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2><p>compile-time optimizations 은 전반적으로 프로그램 성능을 향상시키지만 일부 기법은 성능 하락을 야기한다. 입력프로그램와 target architecture에 대한 불충분한 정보는 컴파일 시간에 정확도 향상을 향상 시키는 모델의 한계를 만든다.</p>https://jaehun.me/posts/%EA%B0%84%EB%8B%A8%EB%85%BC%EB%AC%B8-%EC%A0%95%EB%A6%AC-tvm-an-automated-end-to-end-optimizing-compiler-for-deep-learning-osdi-18/Fri, 12 Feb 2021 00:00:00 +0900https://jaehun.me/posts/%EA%B0%84%EB%8B%A8%EB%85%BC%EB%AC%B8-%EC%A0%95%EB%A6%AC-tvm-an-automated-end-to-end-optimizing-compiler-for-deep-learning-osdi-18/<p> <figure class="image-figure not-prose my-8" data-lightbox-enabled=" true " data-gallery-type="auto" data-image-width="0" data-image-height="0" data-image-src="https://jaehun.me/assets/images/tvm1.png" > <div class="image-container"> <img src="https://jaehun.me/assets/images/tvm1.png" alt="TVM 스택 도식. PyTorch·CoreML·TensorFlow·ONNX·Keras·MXNet 이 High-Level Differentiable IR 로 들어오고, Tensor Expression 과 최적화 탐색 공간을 거쳐 LLVM·CUDA·Metal 과 VTA 로 내려가며, AutoTVM 이 디바이스 플릿과 주고받으며 탐색한다" loading="lazy" decoding="async" data-gallery-src="https://jaehun.me/assets/images/tvm1.png" data-gallery-alt="TVM 스택 도식. PyTorch·CoreML·TensorFlow·ONNX·Keras·MXNet 이 High-Level Differentiable IR 로 들어오고, Tensor Expression 과 최적화 탐색 공간을 거쳐 LLVM·CUDA·Metal 과 VTA 로 내려가며, AutoTVM 이 디바이스 플릿과 주고받으며 탐색한다" data-gallery-title="" /></div> </figure> </p>https://jaehun.me/posts/%EB%85%BC%EB%AC%B8-%EC%A0%95%EB%A6%AC-neurovectorizer-end-to-end-vectorization-with-deep-reinforcement-learning-cgo-20/Fri, 12 Feb 2021 00:00:00 +0900https://jaehun.me/posts/%EB%85%BC%EB%AC%B8-%EC%A0%95%EB%A6%AC-neurovectorizer-end-to-end-vectorization-with-deep-reinforcement-learning-cgo-20/<p> <figure class="image-figure not-prose my-8" data-lightbox-enabled=" true " data-gallery-type="auto" data-image-width="0" data-image-height="0" data-image-src="https://jaehun.me/assets/images/nv1.png" > <div class="image-container"> <img src="https://jaehun.me/assets/images/nv1.png" alt="NeuroVectorizer 프레임워크 개요 도식의 축소판" loading="lazy" decoding="async" data-gallery-src="https://jaehun.me/assets/images/nv1.png" data-gallery-alt="NeuroVectorizer 프레임워크 개요 도식의 축소판" data-gallery-title="" /></div> </figure> </p> <h2 id="제목">제목<a href="#%ec%a0%9c%eb%aa%a9" class="heading-anchor" aria-label="이 섹션에 대한 링크"></a></h2><p>NeuroVectorizer: End-to-End Vectorization with Deep Reinforcement Learning</p>