{"id":3410,"date":"2025-04-18T08:02:48","date_gmt":"2025-04-17T23:02:48","guid":{"rendered":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/?p=3410"},"modified":"2025-04-18T08:02:49","modified_gmt":"2025-04-17T23:02:49","slug":"productivity-performance-oneapi-2025-1","status":"publish","type":"post","link":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/2025\/04\/18\/productivity-performance-oneapi-2025-1\/","title":{"rendered":"\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI \u3068\u30a4\u30f3\u30c6\u30eb\u306e AI \u30c4\u30fc\u30eb 2025.1: \u751f\u6210 AI\u3001\u30d3\u30b8\u30e5\u30a2\u30eb AI\u3001\u30a8\u30c3\u30b8\u304b\u3089\u30af\u30e9\u30a6\u30c9 HPC \u307e\u3067\u751f\u7523\u6027\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a"},"content":{"rendered":"\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<p>\u3053\u306e\u8a18\u4e8b\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u306e\u30a6\u30a7\u30d6\u30b5\u30a4\u30c8\u3067\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u300c<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/productivity-performance-oneapi-2025-1.html\" target=\"_blank\">oneAPI and AI Tools 2025.1<\/a>\u300d\u306e\u65e5\u672c\u8a9e\u53c2\u8003\u8a33\u3067\u3059\u3002\u539f\u6587\u306f\u66f4\u65b0\u3055\u308c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308a\u307e\u3059\u3002\u539f\u6587\u3068\u7ffb\u8a33\u6587\u306e\u5185\u5bb9\u304c\u7570\u306a\u308b\u5834\u5408\u306f\u539f\u6587\u3092\u512a\u5148\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u306e\u6700\u65b0\u306e\u30a2\u30c3\u30d7\u30c7\u30fc\u30c8 2025.1 \u304c\u30ea\u30ea\u30fc\u30b9\u3055\u308c\u307e\u3057\u305f\u3002<\/p>\n<ul>\n<li><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/topic-technology\/artificial-intelligence\/frameworks-tools.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u306e AI \u30c4\u30fc\u30eb\u3068\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af<\/a> (\u82f1\u8a9e)<\/li>\n<li><a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/python\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30c7\u30a3\u30b9\u30c8\u30ea\u30d3\u30e5\u30fc\u30b7\u30e7\u30f3\u306e Python*<\/a><\/li>\n<li><a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/oneapi\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI \u30d9\u30fc\u30b9\u30fb\u30c4\u30fc\u30eb\u30ad\u30c3\u30c8<\/a><\/li>\n<li><a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/oneapi\/index.html?tab=1#features-hpc\">\u30a4\u30f3\u30c6\u30eb\u00ae HPC \u30c4\u30fc\u30eb\u30ad\u30c3\u30c8<\/a><\/li>\n<\/ul>\n<p>\u4eca\u56de\u306e\u30ea\u30ea\u30fc\u30b9\u3067\u306f\u3001CPU\u3001GPU\u3001NPU \u4e0a\u3067\u306e AI \u306e\u751f\u7523\u6027\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5411\u4e0a\u306b\u529b\u3092\u5165\u308c\u3066\u3044\u307e\u3059\u3002\u30d3\u30b8\u30e5\u30a2\u30eb AI \u63a8\u8ad6\u306e\u30b5\u30dd\u30fc\u30c8\u3092\u5f37\u5316\u3057\u3001\u3059\u3079\u3066\u306e\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u8005\u5411\u3051\u306b\u9ad8\u5ea6\u306a\u4e26\u5217\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u306e\u751f\u7523\u6027\u3068\u30b3\u30fc\u30c9\u54c1\u8cea\u4fdd\u8a3c\u306e\u5411\u4e0a\u3092\u500d\u5897\u3055\u305b\u307e\u3057\u305f\u3002\u30af\u30e9\u30a6\u30c9\u306b\u63a5\u7d9a\u3055\u308c\u305f\u30a8\u30c3\u30b8\u63a8\u8ad6\u30c7\u30d0\u30a4\u30b9\u3001\u30a8\u30af\u30b5\u30b9\u30b1\u30fc\u30eb\u30fb\u30b9\u30fc\u30d1\u30fc\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u30fc\u3001\u3042\u308b\u3044\u306f\u305d\u306e\u4e2d\u9593\u3092\u30bf\u30fc\u30b2\u30c3\u30c8\u3068\u3057\u3066\u3044\u308b\u5834\u5408\u3067\u3082\u3001\u672c\u30ea\u30ea\u30fc\u30b9\u306e\u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u3068\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u6a5f\u80fd\u306e\u66f4\u65b0\u306b\u3088\u308a\u3001\u4f5c\u696d\u304c\u5bb9\u6613\u306b\u306a\u308a\u3001\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n<p>\u3053\u306e\u8a18\u4e8b\u3067\u306f\u3001\u3053\u308c\u3089\u306e\u6539\u5584\u304c\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u306b\u3069\u306e\u3088\u3046\u306a\u30e1\u30ea\u30c3\u30c8\u3092\u3082\u305f\u3089\u3059\u304b\u306b\u3064\u3044\u3066\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n<p class=\"note\">\u2192 \u4e3b\u306a\u6a5f\u80fd\u6539\u5584\u306e\u6982\u8981\u306f\u3001<a href=\"https:\/\/www.isus.jp\/products\/oneapi\/oneapi-news-and-updates\/\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u6700\u65b0\u60c5\u5831<\/a>\u306e\u30c8\u30d4\u30c3\u30af\u300c<a href=\"https:\/\/www.isus.jp\/products\/oneapi\/oneapi-news-and-updates\/\" target=\"_blank\" rel=\"noopener\">\u3088\u308a\u9ad8\u901f\u306a AI\u3001\u30ea\u30a2\u30eb\u30bf\u30a4\u30e0\u30fb\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u3001\u3088\u308a\u30b9\u30de\u30fc\u30c8\u306a HPC: \u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb 2025.1 \u30ea\u30ea\u30fc\u30b9!<\/a>\u300d\u3067\u7d39\u4ecb\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u307e\u305a\u3001\u751f\u6210 AI (GenAI) \u3068\u30d3\u30b8\u30e5\u30a2\u30eb AI \u306e\u30e6\u30fc\u30b9\u30b1\u30fc\u30b9\u304b\u3089\u898b\u3066\u3044\u304d\u307e\u3057\u3087\u3046\u3002<\/p>\n<h2>\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9 AI \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u6d3b\u7528<\/h2>\n<p>\u30a4\u30f3\u30c6\u30eb\u306f\u3001\u696d\u754c\u6a19\u6e96\u306e\u30c4\u30fc\u30eb\u3084\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306b\u6ce8\u76ee\u3057\u3001\u6700\u9069\u5316\u306e\u30ce\u30a6\u30cf\u30a6\u3084\u7d4c\u9a13\u3092\u63d0\u4f9b\u3059\u308b\u3053\u3068\u3067 AI \u306b\u53d6\u308a\u7d44\u3093\u3067\u3044\u307e\u3059\u3002\u3053\u306e\u30a2\u30d7\u30ed\u30fc\u30c1\u306f\u3001\u65e2\u5b58\u306e\u958b\u767a\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3084 AI \u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u4f7f\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u3067\u3001\u304a\u5ba2\u69d8\u306b\u30e1\u30ea\u30c3\u30c8\u3092\u3082\u305f\u3089\u3057\u307e\u3059\u3002<\/p>\n<p>\u305d\u306e\u305f\u3081\u3001\u7279\u5b9a\u306e\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u30fb\u30bb\u30c3\u30c8\u306b\u4f9d\u5b58\u3057\u305f\u308a\u3001\u30ec\u30ac\u30b7\u30fc\u30fb\u30b3\u30fc\u30c9\u30d9\u30fc\u30b9\u3092\u653e\u68c4\u3059\u308b\u3053\u3068\u306a\u304f\u3001\u305d\u308c\u3089\u306e\u5229\u7528\u3092\u52b9\u7387\u5316\u3057\u3001\u30a4\u30f3\u30c6\u30eb\u306e\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u306e\u30e1\u30ea\u30c3\u30c8\u3092\u6d3b\u7528\u3067\u304d\u308b\u3088\u3046\u306b\u79fb\u884c\u3059\u308b\u3053\u3068\u304c\u53ef\u80fd\u3067\u3059\u3002<\/p>\n<p>TensorFlow \u3084 PyTorch\u3001\u3055\u3089\u306b\u306f JAX Python \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u306b\u3082\u3001\u30a4\u30f3\u30c6\u30eb\u306e\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u5411\u3051\u306e\u6700\u9069\u5316\u304c\u8ffd\u52a0\u3055\u308c\u3066\u3044\u307e\u3059\u3002JAX Python \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u306f\u30a4\u30f3\u30c6\u30eb\u306e AI \u30c4\u30fc\u30eb\u306b\u7d71\u5408\u3055\u308c\u3001\u30de\u30b7\u30f3\u30e9\u30fc\u30cb\u30f3\u30b0\u30fb\u30e2\u30c7\u30eb\u306e\u6570\u5024\u8a08\u7b97\u3084\u79d1\u5b66\u8a08\u7b97\u3001\u6700\u9069\u5316\u30bf\u30b9\u30af\u3092\u9ad8\u901f\u5316\u3059\u308b<a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/python\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30c7\u30a3\u30b9\u30c8\u30ea\u30d3\u30e5\u30fc\u30b7\u30e7\u30f3\u306e Python*<\/a> \u306e\u6700\u9069\u5316\u306e\u6069\u6075\u3092\u53d7\u3051\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<h3>PyTorch \u5411\u3051\u306e\u6700\u65b0\u306e\u6700\u9069\u5316<\/h3>\n<p><a href=\"https:\/\/developer.intel.com\/pytorch\" target=\"_blank\" rel=\"noopener\">PyTorch* \u5411\u3051\u30a4\u30f3\u30c6\u30eb\u00ae \u30a8\u30af\u30b9\u30c6\u30f3\u30b7\u30e7\u30f3<\/a> (\u82f1\u8a9e) \u3067\u5229\u7528\u53ef\u80fd\u3067\u3042\u308a\u3001PyTorch 2.6 \u306b\u3082\u30a2\u30c3\u30d7\u30b9\u30c8\u30ea\u30fc\u30e0\u3055\u308c\u3066\u3044\u308b\u6700\u65b0\u306e\u8ffd\u52a0\u6a5f\u80fd\u3068\u6700\u9069\u5316\u306f\u3001AI \u30bd\u30ea\u30e5\u30fc\u30b7\u30e7\u30f3\u304c\u30a4\u30f3\u30c6\u30eb\u306e CPU \u3068 GPU \u3092\u30b7\u30fc\u30e0\u30ec\u30b9\u306b\u30bf\u30fc\u30b2\u30c3\u30c8\u306b\u3067\u304d\u308b\u3053\u3068\u3092\u793a\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<p><a href=\"https:\/\/github.com\/intel\/intel-extension-for-deepspeed\" target=\"_blank\" rel=\"noopener\">DeepSpeed* \u5411\u3051\u30a4\u30f3\u30c6\u30eb\u00ae \u30a8\u30af\u30b9\u30c6\u30f3\u30b7\u30e7\u30f3<\/a> (\u82f1\u8a9e) \u304c PyTorch \u30ea\u30ea\u30fc\u30b9\u306b\u7d71\u5408\u3055\u308c\u307e\u3057\u305f\u3002\u307e\u305f\u3001\u30a4\u30f3\u30c6\u30eb\u00ae GPU \u30c7\u30d0\u30a4\u30b9\u4e0a\u306e SYCL \u30ab\u30fc\u30cd\u30eb\u30b5\u30dd\u30fc\u30c8\u3068\u540c\u69d8\u306e\u3082\u306e\u3092\u3001DeepSpeed \u306b\u3059\u3067\u306b\u30a2\u30c3\u30d7\u30b9\u30c8\u30ea\u30fc\u30e0\u3055\u308c\u3066\u3044\u308b Microsoft Research \u306e\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u6700\u9069\u5316\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u306b\u8ffd\u52a0\u3057\u3001\u5e45\u5e83\u3044\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb (LLM) \u306b\u8ffd\u52a0\u306e\u6700\u9069\u5316\u3092\u9069\u7528\u3057\u307e\u3057\u305f\u3002<\/p>\n<p>2024\u5e74\u304b\u3089\u306f\u3001torch.tensor \u30c7\u30d0\u30a4\u30b9\u306e\u540d\u524d\u3092\u5909\u66f4\u3059\u308b\u3060\u3051\u3067\u3001SYCL \u5bfe\u5fdc\u30c7\u30d0\u30a4\u30b9\u4e0a\u3067\u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30bf\u30fc\u3078\u306e\u30aa\u30d5\u30ed\u30fc\u30c9\u30b3\u30fc\u30c9\u3092\u5b9f\u884c\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3057\u305f<\/p>\n<pre><code class=\"language-python\"># CUDA Code\u00a0\ntensor = torch.tensor([1.0, 2.0]).to(\"cuda\")\u00a0\n\n\n# SYCL Code\ntensor = torch.tensor([1.0, 2.0]).to(\"xpu\")<\/code><\/pre>\n<p>FP32\u3001BF16\u3001FP16\u3001\u81ea\u52d5\u6df7\u5408\u7cbe\u5ea6 (AMP) \u306a\u3069\u306e\u30c7\u30fc\u30bf\u578b\u306e\u4e2d\u304b\u3089\u3001AI \u30e6\u30fc\u30b9\u30b1\u30fc\u30b9\u306b\u6700\u9069\u306a\u30c7\u30fc\u30bf\u578b\u3092\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002GenAI LLM \u306f INT4 \u306b\u6700\u9069\u5316\u3055\u308c\u3001\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30fb\u30a8\u30f3\u30b8\u30f3\u306f FP8 \u3092\u30b5\u30dd\u30fc\u30c8\u3057\u307e\u3059\u3002<\/p>\n<p>\u4f8b\u3048\u3070\u3001\u7e2e\u5c0f\u3055\u308c\u305f FP16 \u6d6e\u52d5\u5c0f\u6570\u70b9\u30c7\u30fc\u30bf\u578b\u3092\u3001<a href=\"https:\/\/www.intel.co.jp\/content\/www\/jp\/ja\/products\/details\/processors\/xeon\/xeon6-p-cores.html\" target=\"_blank\" rel=\"noopener\">P-cores \u642d\u8f09\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae 6 \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc<\/a>\u306a\u3069\u306e CPU \u4e0a\u3067 CPU \u30cd\u30a4\u30c6\u30a3\u30d6\u306e<a href=\"https:\/\/www.intel.co.jp\/content\/www\/jp\/ja\/products\/docs\/accelerator-engines\/advanced-matrix-extensions\/overview.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30a2\u30c9\u30d0\u30f3\u30b9\u30c8\u30fb\u30de\u30c8\u30ea\u30af\u30b9\u30fb\u30a8\u30af\u30b9\u30c6\u30f3\u30b7\u30e7\u30f3 (\u30a4\u30f3\u30c6\u30eb\u00ae AMX)<\/a> \u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30bf\u30fc\u3068\u7d44\u307f\u5408\u308f\u305b\u3066\u4f7f\u7528\u3059\u308b\u3068\u3001PyTorch \u306e Inductor \u30e2\u30fc\u30c9\u3068 Eager \u30e2\u30fc\u30c9\u306e\u4e21\u65b9\u3067\u63a8\u8ad6\u304c\u5927\u5e45\u306b\u30b9\u30d4\u30fc\u30c9\u30a2\u30c3\u30d7\u3057\u307e\u3059\u3002<\/p>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae Arc&#x2122;\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u306e\u5168\u30d5\u30a1\u30df\u30ea\u30fc (\u30a4\u30f3\u30c6\u30eb\u00ae Core&#x2122; Ultra \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc (\u30b7\u30ea\u30fc\u30ba 2)\u3001\u30a4\u30f3\u30c6\u30eb\u00ae Arc&#x2122; A \u30b7\u30ea\u30fc\u30ba\u30fb\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u3001\u30a4\u30f3\u30c6\u30eb\u00ae Arc&#x2122; B \u30b7\u30ea\u30fc\u30ba\u30fb\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u306a\u3069) \u3092\u542b\u3080\u30a4\u30f3\u30c6\u30eb\u00ae GPU \u5411\u3051\u306b\u6700\u9069\u5316\u304c\u62e1\u5f35\u3055\u308c\u307e\u3057\u305f\u3002Windows \u3067\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u00ae GPU \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u30fb\u30b9\u30bf\u30c3\u30af\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u304c\u7c21\u7d20\u5316\u3055\u308c\u3001torch-xpu \u30bf\u30fc\u30b2\u30c3\u30c8\u306e\u30b5\u30dd\u30fc\u30c8\u3068 torch core \u306e\u30d0\u30a4\u30ca\u30ea\u30fc\u30ea\u30ea\u30fc\u30b9\u3092\u30ef\u30f3\u30af\u30ea\u30c3\u30af\u3067\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3057\u305f\u3002<\/p>\n<h3>FlexAttention \u306b\u3088\u308b\u63a8\u8ad6\u306e\u9ad8\u901f\u5316<\/h3>\n<p>\u3082\u3046 1 \u3064\u306e\u8208\u5473\u6df1\u3044\u8ffd\u52a0\u6a5f\u80fd\u306f\u3001FlexAttention \u306e TorchInductor CPP \u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u6700\u9069\u5316\u306b\u304a\u3051\u308b x86 CPU \u30b5\u30dd\u30fc\u30c8\u3067\u3059\u3002\u3053\u308c\u306b\u3088\u308a\u3001\u9ad8\u5ea6\u306a\u91cd\u307f\u4ed8\u304d\u63a8\u8ad6\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n<p>\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u8a2d\u5b9a\u306f\u3001\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u5165\u529b\u30b7\u30fc\u30b1\u30f3\u30b9\u306e\u6700\u3082\u95a2\u9023\u6027\u306e\u9ad8\u3044\u90e8\u5206\u306b\u30d5\u30a9\u30fc\u30ab\u30b9\u3059\u308b\u305f\u3081\u3001\u30e2\u30c7\u30eb\u306b\u91cd\u307f\u3092\u5272\u308a\u5f53\u3066\u308b\u624b\u6bb5\u3067\u3059\u3002\u3057\u304b\u3057\u3001\u3053\u308c\u3089\u306e\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u91cd\u307f\u306b\u306f\u591a\u304f\u306e\u30d0\u30ea\u30a2\u30f3\u30c8\u304c\u5b58\u5728\u3057\u307e\u3059\u3002\u671b\u307e\u3057\u3044\u7d50\u679c\u3092\u5f97\u308b\u305f\u3081\u3001\u91cd\u307f\u306e\u7d44\u307f\u5408\u308f\u305b\u3092\u9ad8\u5ea6\u306b\u30ab\u30b9\u30bf\u30de\u30a4\u30ba\u3059\u308b\u5fc5\u8981\u304c\u983b\u7e41\u306b\u751f\u3058\u308b\u5834\u5408\u3001\u958b\u767a\u8005\u3084\u30c7\u30fc\u30bf\u30a2\u30ca\u30ea\u30b9\u30c8\u304c\u3001\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30e1\u30ab\u30cb\u30ba\u30e0\u304c\u9069\u5207\u306b\u8abf\u6574\u3055\u308c\u3066\u3044\u306a\u3044\u305f\u3081\u306b\u63a8\u8ad6\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u4f4e\u4e0b\u306b\u983b\u7e41\u306b\u906d\u9047\u3059\u308b\u306e\u306f\u5f53\u7136\u3067\u3057\u3087\u3046\u3002\u305d\u3053\u3067 PyTorch \u306e FlexAttenstion \u306e\u51fa\u756a\u3067\u3059\u3002FlexAttenstion \u306f\u3001\u6570\u884c\u306e PyTorch \u30b3\u30fc\u30c9\u3067\u518d\u5229\u7528\u3067\u304d\u308b\u6700\u9069\u5316\u6e08\u307f\u306e\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30e1\u30ab\u30cb\u30ba\u30e0\u306e\u30d0\u30ea\u30a2\u30f3\u30c8\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002\u5fc5\u8981\u306a\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30e1\u30ab\u30cb\u30ba\u30e0\u306e\u7d44\u307f\u5408\u308f\u305b\u306f\u3001torch.compile \u3092\u4f7f\u3063\u3066 FlexAttension \u30ab\u30fc\u30cd\u30eb\u306b\u878d\u5408\u3057\u3066\u6700\u9069\u5316\u3067\u304d\u307e\u3059\u3002\u624b\u66f8\u304d\u30b3\u30fc\u30c9\u306b\u5339\u6575\u3059\u308b\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5b9f\u73fe\u3059\u308b\u305f\u3081\u3001\u4f7f\u7528\u3055\u308c\u308b\u3059\u3079\u3066\u306e\u30a2\u30c6\u30f3\u30b7\u30e7\u30f3\u30fb\u30d0\u30ea\u30a2\u30f3\u30c8\u306b\u5305\u62ec\u7684\u306a\u6700\u9069\u5316\u3092\u9069\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n<p class=\"note\">\u2192 \u30d6\u30ed\u30b0\u8a18\u4e8b\u300c<a href=\"https:\/\/pytorch.org\/blog\/unlocking-pt-2-6-intel\/\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u5411\u3051\u306e PyTorch 2.6 \u306e\u6700\u65b0\u6a5f\u80fd\u3092\u6d3b\u7528\u3059\u308b<\/a>\u300d (\u82f1\u8a9e) \u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n<h3>\u30a4\u30f3\u30c6\u30eb\u00ae oneDNN \u3068 Python \u3092\u30d9\u30fc\u30b9\u306b\u69cb\u7bc9<\/h3>\n<p>\u30a4\u30f3\u30c6\u30eb\u306e PyTorch \u5411\u3051\u6700\u9069\u5316\u306f\u3001<a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/python\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30c7\u30a3\u30b9\u30c8\u30ea\u30d3\u30e5\u30fc\u30b7\u30e7\u30f3\u306e Python*<\/a> \u3068<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/tools\/oneapi\/onednn.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI \u30c7\u30a3\u30fc\u30d7\u30fb\u30cb\u30e5\u30fc\u30e9\u30eb\u30fb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u30fb\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc (\u30a4\u30f3\u30c6\u30eb\u00ae oneDNN)<\/a> (\u82f1\u8a9e) \u306e\u6700\u65b0\u306e\u6a5f\u80fd\u3082\u6d3b\u7528\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae \u30c7\u30a3\u30b9\u30c8\u30ea\u30d3\u30e5\u30fc\u30b7\u30e7\u30f3\u306e Python* \u306f\u30013 \u3064\u306e\u57fa\u672c\u30d1\u30c3\u30b1\u30fc\u30b8\u3067\u69cb\u6210\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<ul>\n<li>dpnp (<a href=\"https:\/\/numpy.org\/\" target=\"_blank\" rel=\"noopener\">NumPy<\/a> (\u82f1\u8a9e) \u5411\u3051\u30c7\u30fc\u30bf\u30fb\u30d1\u30e9\u30ec\u30eb\u30fb\u30a8\u30af\u30b9\u30c6\u30f3\u30b7\u30e7\u30f3)<\/li>\n<li>numba_dpex (<a href=\"https:\/\/numba.pydata.org\/\" target=\"_blank\" rel=\"noopener\">Numba<\/a> (\u82f1\u8a9e) \u5411\u3051\u30c7\u30fc\u30bf\u30fb\u30d1\u30e9\u30ec\u30eb\u30fb\u30a8\u30af\u30b9\u30c6\u30f3\u30b7\u30e7\u30f3)<\/li>\n<li>dpctl (\u30c7\u30d0\u30a4\u30b9\u9078\u629e\u3001\u30c7\u30d0\u30a4\u30b9\u4e0a\u306e\u30c7\u30fc\u30bf\u5272\u308a\u5f53\u3066\u3001\u30c6\u30f3\u30bd\u30eb\u30c7\u30fc\u30bf\u69cb\u9020\u306e\u30b5\u30dd\u30fc\u30c8\u3001Python \u5411\u3051\u30e6\u30fc\u30b6\u30fc\u5b9a\u7fa9\u30c7\u30fc\u30bf\u4e26\u5217\u62e1\u5f35\u7528\u306e\u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u30fc\u3092\u5099\u3048\u305f\u30c7\u30fc\u30bf\u30fb\u30d1\u30e9\u30ec\u30eb\u30fb\u30b3\u30f3\u30c8\u30ed\u30fc\u30eb\u30fb\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc)<\/li>\n<\/ul>\n<p>\u672c\u30ea\u30ea\u30fc\u30b9\u3067\u306f\u3001\u4ee5\u4e0b\u306b\u5bfe\u5fdc\u3057\u307e\u3057\u305f\u3002<\/p>\n<ul>\n<li>Python Array API \u4ed5\u69d8 (\u30ea\u30d3\u30b8\u30e7\u30f3\u300c2023.12\u300d) \u306b 100% \u6e96\u62e0<\/li>\n<li>NumPy 2.0 \u30b5\u30dd\u30fc\u30c8\u3068 NumPy 2.2.3 \u3068\u306e\u4e92\u63db\u6027<\/li>\n<li>CuPy \u3068\u306e 90% \u306e\u6a5f\u80fd\u4e92\u63db\u6027<\/li>\n<\/ul>\n<p>\u3053\u308c\u306b\u3088\u308a\u3001\u30a4\u30f3\u30c6\u30eb\u306e\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u3092\u30bf\u30fc\u30b2\u30c3\u30c8\u3068\u3059\u308b\u5834\u5408\u3001\u65e2\u5b58\u306e Python \u5b9f\u88c5\u3092\u5909\u66f4\u3059\u308b\u3053\u3068\u306a\u304f\u3001\u3055\u3089\u306b\u7c21\u5358\u306b\u4f7f\u7528\u3067\u304d\u307e\u3059\u3002<\/p>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae oneDNN \u306b\u3082\u591a\u5927\u306a\u52b4\u529b\u3068\u958b\u767a\u304c\u6295\u5165\u3055\u308c\u307e\u3057\u305f\u3002\u6700\u65b0\u306e\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u3068\u5e45\u5e83\u3044\u30a4\u30f3\u30c6\u30eb\u00ae Arc&#x2122; \u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u5411\u3051\u306e\u6700\u9069\u5316\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u30fb\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u304c\u5f37\u5316\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u958b\u767a\u8005\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u00ae AMX \u547d\u4ee4\u30bb\u30c3\u30c8\u5bfe\u5fdc\u306e\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae 6 \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u3092\u542b\u3080\u6700\u65b0\u306e\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u5411\u3051\u306e\u6700\u9069\u5316\u306b\u3088\u308a\u3001\u5f37\u5316\u3055\u308c\u305f\u884c\u5217\u4e57\u7b97\u3068\u7573\u307f\u8fbc\u307f\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6d3b\u7528\u3057\u3066\u3001\u30c7\u30fc\u30bf\u30bb\u30f3\u30bf\u30fc\u306e AI \u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u3092\u9ad8\u901f\u5316\u3067\u304d\u307e\u3059\u3002<\/p>\n<p>\u30af\u30e9\u30a4\u30a2\u30f3\u30c8 CPU \u3067\u306e AI \u63a8\u8ad6\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u00ae Arc&#x2122; \u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u4e0a\u3001\u7279\u306b\u30a4\u30f3\u30c6\u30eb\u00ae Core&#x2122; Ultra \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc (\u30b7\u30ea\u30fc\u30ba 2) \u642d\u8f09 AI PC \u3068\u30a4\u30f3\u30c6\u30eb\u00ae Arc&#x2122; B \u30b7\u30ea\u30fc\u30ba\u306e\u30c7\u30a3\u30b9\u30af\u30ea\u30fc\u30c8\u30fb\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u4e0a\u3067\u30a4\u30f3\u30c6\u30eb\u00ae oneDNN \u3092\u4f7f\u7528\u3059\u308b\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002\u30a4\u30f3\u30c6\u30eb\u00ae oneDNN \u306f\u3001\u6697\u9ed9\u7684\u306a\u30ab\u30b8\u30e5\u30a2\u30eb\u30de\u30b9\u30af\u3092\u4f7f\u7528\u3057\u305f\u30b2\u30fc\u30c8\u4ed8\u304d\u591a\u5c64\u30d1\u30fc\u30bb\u30d7\u30c8\u30ed\u30f3 (\u30b2\u30fc\u30c8\u4ed8\u304d MLP) \u3068 SPDA (Scaled Dot-Product Attention) \u306b\u3088\u308a\u3001AI \u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6700\u9069\u5316\u3057\u307e\u3059\u3002Graph API \u306b\u3088\u308b int8 \u307e\u305f\u306f int4 \u5727\u7e2e\u30ad\u30fc\u3068\u5024\u306e\u30b5\u30dd\u30fc\u30c8\u306f\u3001\u63a8\u8ad6\u306e\u901f\u5ea6\u3068\u52b9\u7387\u306e\u4e21\u65b9\u3092\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n<h3>\u30e2\u30c7\u30eb\u5727\u7e2e\u306b\u3088\u308b\u63a8\u8ad6\u306e\u9ad8\u901f\u5316<\/h3>\n<p><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/github.com\/intel\/neural-compressor\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30cb\u30e5\u30fc\u30e9\u30eb\u30fb\u30b3\u30f3\u30d7\u30ec\u30c3\u30b5\u30fc<\/a> (\u82f1\u8a9e) \u306e\u76ee\u7684\u306f\u3001\u63a8\u8ad6\u52b9\u7387\u306e\u5411\u4e0a\u3092\u652f\u63f4\u3059\u308b\u3053\u3068\u3067\u3059\u3002<\/p>\n<p>\u30e2\u30c7\u30eb\u6700\u9069\u5316\u306b\u3088\u308a\u30e2\u30c7\u30eb\u30b5\u30a4\u30ba\u3092\u7e2e\u5c0f\u3057\u3001CPU\u3001GPU\u3001\u307e\u305f\u306f\u30a4\u30f3\u30c6\u30eb\u00ae Gaudi\u00ae AI \u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30bf\u30fc\u4e0a\u3067\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u63a8\u8ad6\u306e\u901f\u5ea6\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002<\/p>\n<p>\u3053\u308c\u306f\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u306e Python \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u3067\u3042\u308a\u3001\u6700\u5148\u7aef\u306e <code>INT8\/FP8\/INT4\/FP4\/NF4<\/code> (\u6b63\u898f\u5316 Float 4) \u306b\u3088\u308b\u4f4e\u30d3\u30c3\u30c8 LLM \u91cf\u5b50\u5316\u3068\u30b9\u30d1\u30fc\u30b9\u6027\u3092\u5229\u7528\u3057\u3066\u3001\u8907\u6570\u306e\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u30fb\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306b\u304a\u3051\u308b\u91cf\u5b50\u5316\u3001\u30d7\u30eb\u30fc\u30cb\u30f3\u30b0\u3001\u77e5\u8b58\u84b8\u7559\u3068\u3044\u3063\u305f\u4e00\u822c\u7684\u306a\u30e2\u30c7\u30eb\u6700\u9069\u5316\u624b\u6cd5\u3092\u81ea\u52d5\u5316\u3057\u307e\u3059\u3002<\/p>\n<p>\u30a8\u30c3\u30b8\u304b\u3089\u30af\u30e9\u30a6\u30c9\u3078\u306e\u63a8\u8ad6\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u306e\u4e8b\u524d\u6e96\u5099\u3068\u3057\u3066\u3001\u5927\u898f\u6a21\u30e2\u30c7\u30eb\u304b\u3089\u77e5\u8b58\u3092\u84b8\u7559\u3057\u3001\u5c0f\u898f\u6a21\u30e2\u30c7\u30eb\u306e\u7cbe\u5ea6\u3092\u5411\u4e0a\u3059\u308b\u306e\u306b\u6700\u9069\u306a\u30c4\u30fc\u30eb\u3067\u3059\u3002<\/p>\n<p style=\"text-align: center\"><img decoding=\"async\" style=\"max-width: 100%\" src=\"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-content\/uploads\/sites\/5\/2025\/04\/20_neural-compressor-models-1024x574.png\" alt=\"\" width=\"1509\" height=\"auto\" \/><br \/><strong>\u56f3 1: \u30a4\u30f3\u30c6\u30eb\u00ae \u30cb\u30e5\u30fc\u30e9\u30eb\u30fb\u30b3\u30f3\u30d7\u30ec\u30c3\u30b5\u30fc\u306e\u6982\u8981<\/strong><\/p>\n<h3>\u30d3\u30b8\u30e5\u30a2\u30eb AI \u3068\u30ec\u30f3\u30c0\u30ea\u30f3\u30b0\u306e\u9ad8\u901f\u5316<\/h3>\n<p>\u65b0\u30ea\u30ea\u30fc\u30b9\u306e<a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/compilers\/dpcpp\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc<\/a>\u3068<a href=\"https:\/\/www.isus.jp\/dpc-compatibility-tool\/\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae DPC++ \u4e92\u63db\u6027\u30c4\u30fc\u30eb<\/a>\u3067\u306f\u3001Vulkan \u304a\u3088\u3073 DirectX 12 \u3068\u306e SYCL \u76f8\u4e92\u904b\u7528\u6027\u304c\u5f37\u5316\u3055\u308c\u3066\u3044\u307e\u3059\u3002GPU \u304b\u3089\u30a4\u30e1\u30fc\u30b8\u30fb\u30c6\u30af\u30b9\u30c1\u30e3\u30fc\u30fb\u30de\u30c3\u30d7\u30fb\u30c7\u30fc\u30bf\u3092\u76f4\u63a5\u5171\u6709\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u3001CPU \u3068 GPU \u9593\u306e\u4f59\u5206\u306a\u30a4\u30e1\u30fc\u30b8\u30b3\u30d4\u30fc\u304c\u6392\u9664\u3055\u308c\u3001\u30a4\u30e1\u30fc\u30b8\u51e6\u7406\u3068\u9ad8\u5ea6\u306a\u30ec\u30f3\u30c0\u30ea\u30f3\u30b0\u30fb\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3067\u30b7\u30fc\u30e0\u30ec\u30b9\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u304c\u78ba\u4fdd\u3055\u308c\u3001\u30b3\u30f3\u30c6\u30f3\u30c4\u4f5c\u6210\u306e\u751f\u7523\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n<p>\u3055\u3089\u306b\u3001\u7279\u306b\u52d5\u7684\u753b\u50cf\u914d\u5217\u3084\u3001\u30a2\u30af\u30bb\u30b5\u30fc\u3067\u306f\u306a\u304f\u30cf\u30f3\u30c9\u30eb\u3092\u4ecb\u3057\u305f\u30a4\u30e1\u30fc\u30b8\u30a2\u30af\u30bb\u30b9\u306b\u304a\u3044\u3066\u30a4\u30e1\u30fc\u30b8\u7ba1\u7406\u6a5f\u80fd\u3092\u5f37\u5316\u3059\u308b\u3001SYCL 2020 \u306e<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/sycl-bindless-images-migration.html\" target=\"_blank\" rel=\"noopener\">\u30d0\u30a4\u30f3\u30c9\u30ec\u30b9\u753b\u50cf\u62e1\u5f35<\/a> (\u82f1\u8a9e) \u304c\u63d0\u6848\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u3053\u308c\u3089\u306e\u6539\u5584\u306f\u3001SYCL \u5bfe\u5fdc GPU \u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30b7\u30e7\u30f3\u3092\u6d3b\u7528\u3057\u305f\u30b2\u30fc\u30e0\u3001\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u3001\u30d3\u30b8\u30e5\u30a2\u30eb AI\u3001\u304a\u3088\u3073\u8996\u899a\u7684\u306b\u9b45\u529b\u7684\u306a\u52d5\u7684\u30b3\u30f3\u30c6\u30f3\u30c4\u306e\u958b\u767a\u3092\u7c21\u7d20\u5316\u3059\u308b\u3053\u3068\u3092\u76ee\u7684\u3068\u3057\u3066\u3044\u307e\u3059\u3002\u3053\u308c\u306b\u3088\u308a\u5f97\u3089\u308c\u308b\u751f\u7523\u6027\u306e\u5411\u4e0a\u306f\u3001GenAI\u3001\u30d3\u30b8\u30e5\u30a2\u30eb AI\u3001\u305d\u3057\u3066\u9ad8\u5ea6\u3067\u8907\u96d1\u306a\u753b\u50cf\u30c7\u30fc\u30bf\u89e3\u6790\u306b\u30e1\u30ea\u30c3\u30c8\u3092\u3082\u305f\u3089\u3057\u307e\u3059\u3002<\/p>\n<h2>AI \u63a8\u8ad6\u306e\u30d7\u30ed\u30d5\u30a1\u30a4\u30eb\u3068\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0<\/h2>\n<p>\u30a8\u30c3\u30b8\u63a8\u8ad6\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3067\u306f\u3001\u6f5c\u5728\u7684\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u30fb\u30dc\u30c8\u30eb\u30cd\u30c3\u30af\u3092\u7406\u89e3\u3057\u3001\u30a4\u30f3\u30c6\u30eb\u00ae Core&#x2122; Ultra \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u642d\u8f09 AI PC \u4e0a\u3067\u5b9f\u884c\u3055\u308c\u308b\u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u3092\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u3053\u3068\u304c\u91cd\u8981\u3067\u3059\u3002\u3053\u306e\u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u306b\u306f\u3001CPU\u3001GPU\u3001\u30cb\u30e5\u30fc\u30e9\u30eb\u30fb\u30d7\u30ed\u30bb\u30c3\u30b7\u30f3\u30b0\u30fb\u30e6\u30cb\u30c3\u30c8 (NPU) \u306e 3 \u3064\u306e\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u30fb\u30a8\u30f3\u30b8\u30f3\u304c\u642d\u8f09\u3055\u308c\u3066\u3044\u307e\u3059\u3002\u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u306f\u3001DirectML \u307e\u305f\u306f WinML API \u3092\u547c\u3073\u51fa\u3059 AI \u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u30fb\u30dc\u30c8\u30eb\u30cd\u30c3\u30af\u3092\u7279\u5b9a\u3057\u3001Python 3.12 \u3067\u6700\u3082\u6642\u9593\u306e\u304b\u304b\u308b\u30b3\u30fc\u30c9\u9818\u57df\u3068\u30af\u30ea\u30c6\u30a3\u30ab\u30eb\u306a\u30b3\u30fc\u30c9\u30d1\u30b9\u3092\u7279\u5b9a\u3059\u308b\u3053\u3068\u3067\u3001\u30d3\u30b8\u30e5\u30a2\u30eb AI \u306e\u6700\u9069\u5316\u3092\u652f\u63f4\u3057\u307e\u3059\u3002<\/p>\n<p>AI PC \u306e\u5834\u5408\u3001NPU \u3092\u542b\u3080\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u3068\u30de\u30a4\u30af\u30ed\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u30fc\u304c\u8003\u616e\u3055\u308c\u307e\u3059\u3002NPU \u306f\u3001\u9577\u6642\u9593\u306b\u308f\u305f\u3063\u3066\u6301\u7d9a\u7684\u306b\u5b9f\u884c\u3055\u308c\u308b\u4f4e\u6d88\u8cbb\u96fb\u529b\u306e AI \u652f\u63f4\u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u306b\u6700\u9069\u3067\u3059\u3002\u4f8b\u3048\u3070\u3001\u80cc\u666f\u306e\u307c\u304b\u3057\u3084\u753b\u50cf\u306e\u30bb\u30b0\u30e1\u30f3\u30c6\u30fc\u30b7\u30e7\u30f3\u306a\u3069\u306e\u30a6\u30a7\u30d6\u30ab\u30e1\u30e9\u30fb\u30bf\u30b9\u30af\u3092\u5b9f\u884c\u3059\u308b\u5834\u5408\u3001NPU \u3067\u5b9f\u884c\u3059\u308b\u3053\u3068\u3067\u5168\u4f53\u7684\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n<p>AI \u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6700\u5927\u9650\u306b\u9ad8\u3081\u308b\u306b\u306f\u3001NPU \u306e\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u30fb\u30ea\u30bd\u30fc\u30b9\u3068\u30e1\u30e2\u30ea\u30fc\u30ea\u30bd\u30fc\u30b9\u3092\u6700\u5927\u9650\u306b\u6d3b\u7528\u3059\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001NPU \u3067\u5b9f\u884c\u3055\u308c\u3066\u3044\u308b\u30b3\u30fc\u30c9\u306e\u30dc\u30c8\u30eb\u30cd\u30c3\u30af\u3092\u7279\u5b9a\u3067\u304d\u307e\u3059\u3002\u4ee5\u4e0b\u306e\u3053\u3068\u304c\u53ef\u80fd\u3067\u3059\u3002<\/p>\n<ul>\n<li>NPU \u3068 DDR \u30e1\u30e2\u30ea\u30fc\u9593\u3067\u8ee2\u9001\u3055\u308c\u308b\u30c7\u30fc\u30bf\u91cf\u306e\u628a\u63e1<\/li>\n<li>NPU \u3067\u6700\u3082\u6642\u9593\u306e\u304b\u304b\u308b\u30bf\u30b9\u30af\u306e\u7279\u5b9a<\/li>\n<li>NPU \u306e\u4f7f\u7528\u7387\u3092\u6642\u9593\u7d4c\u904e\u3068\u3068\u3082\u306b\u8996\u899a\u5316<\/li>\n<li>NPU \u306e\u5e2f\u57df\u5e45\u4f7f\u7528\u7387\u306e\u628a\u63e1<\/li>\n<\/ul>\n<p class=\"note\">\u2192 AI PC \u306e\u63a8\u8ad6\u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u3092\u9ad8\u901f\u5316\u3059\u308b\u65b9\u6cd5\u3092\u7d39\u4ecb\u3057\u305f\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u300c<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/vtune-ai-profile-npu.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae Core&#x2122; Ultra \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u4e0a\u3067\u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u3092\u4f7f\u7528\u3057\u3066 NPU \u4f9d\u5b58\u306e AI PC \u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3092\u9ad8\u901f\u5316\u3059\u308b<\/a>\u300d (\u82f1\u8a9e) \u3092\u78ba\u8a8d\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n<p style=\"text-align: center\"><img decoding=\"async\" style=\"max-width: 100%\" src=\"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-content\/uploads\/sites\/5\/2025\/04\/20_figure_5-1024x572.png\" alt=\"\" width=\"1527\" height=\"auto\" \/><br \/><strong>\u56f3 2: \u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u306e NPU \u5168\u822c\u89e3\u6790\u306e\u30b9\u30ec\u30c3\u30c9\u30d3\u30e5\u30fc<\/strong><\/p>\n<p class=\"note\">\u2192 \u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u3092\u4f7f\u7528\u3057\u305f GenAI \u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u89e3\u6790\u3068\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306b\u95a2\u3059\u308b\u4e00\u822c\u7684\u306a\u30c1\u30e5\u30fc\u30c8\u30ea\u30a2\u30eb\u306f\u3001\u300c<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/enhance-genai-llm-performance-using-vtune-profiler.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u3092\u4f7f\u7528\u3057\u305f GenAI LLM \u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5411\u4e0a<\/a>\u300d (\u82f1\u8a9e) \u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n<h2>\u9ad8\u6027\u80fd\u30c7\u30fc\u30bf\u30fb\u30a2\u30ca\u30ea\u30c6\u30a3\u30af\u30b9<\/h2>\n<p>\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u306f\u3001\u30d5\u30a9\u30fc\u30e0\u30d5\u30a1\u30af\u30bf\u30fc\u3084\u6d88\u8cbb\u96fb\u529b\u304c\u5236\u9650\u3055\u308c\u305f\u74b0\u5883\u3067\u30a8\u30c3\u30b8\u63a8\u8ad6\u3084\u30a4\u30e1\u30fc\u30b8\u51e6\u7406\u3092\u30b5\u30dd\u30fc\u30c8\u3059\u308b\u3060\u3051\u3067\u306a\u304f\u3001\u5206\u6563\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u69cb\u6210\u3084\u30b9\u30fc\u30d1\u30fc\u30b3\u30f3\u30d4\u30e5\u30fc\u30bf\u30fc\u3060\u3051\u304c\u63d0\u4f9b\u3067\u304d\u308b\u6700\u9ad8\u30ec\u30d9\u30eb\u306e\u6a5f\u80fd\u3092\u5fc5\u8981\u3068\u3059\u308b\u30e6\u30fc\u30b9\u30b1\u30fc\u30b9\u306b\u3082\u3001\u540c\u3058\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u306e\u539f\u5247\u3068\u30c4\u30fc\u30eb\u3092\u9069\u7528\u3057\u62e1\u5f35\u3057\u307e\u3059\u3002<\/p>\n<h3>\u5206\u6563\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u5411\u3051\u306e\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0<\/h3>\n<p><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/tools\/oneapi\/oneccl.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI \u30b3\u30ec\u30af\u30c6\u30a3\u30d6\u30fb\u30b3\u30df\u30e5\u30cb\u30b1\u30fc\u30b7\u30e7\u30f3\u30fb\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc (\u30a4\u30f3\u30c6\u30eb\u00ae oneCCL)<\/a> (\u82f1\u8a9e) \u306f\u3001\u975e\u5e38\u306b\u5927\u898f\u6a21\u306a\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u63a8\u8ad6\u3068\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3092\u6700\u9069\u5316\u3057\u3001\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002<\/p>\n<p>\u4e16\u754c\u898f\u6a21\u306e\u74b0\u5883\u30c7\u30fc\u30bf\u8abf\u67fb\u3001\u5927\u898f\u6a21\u306a\u4eba\u53e3\u7d71\u8a08\u8abf\u67fb\u3001\u75ab\u5b66\u7814\u7a76\u3001\u91d1\u878d\u30ea\u30b9\u30af\u8a55\u4fa1\u306a\u3069\u306b AI \u3092\u9069\u7528\u3059\u308b\u3053\u3068\u3092\u60f3\u50cf\u3057\u3066\u307f\u3066\u304f\u3060\u3055\u3044\u3002\u9ad8\u6027\u80fd\u5206\u6563\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u306e\u539f\u5247\u3092 AI \u306b\u9069\u7528\u3059\u308b\u3053\u3068\u3067\u3001\u8ffd\u52a0\u306e\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u30fb\u30cb\u30fc\u30ba\u3068\u3001\u81a8\u5927\u306a\u91cf\u306e\u30a2\u30c9\u30ec\u30b9\u6307\u5b9a\u53ef\u80fd\u306a\u30e2\u30c7\u30eb\u30c7\u30fc\u30bf\u3078\u306e\u30a2\u30af\u30bb\u30b9\u306b\u5bfe\u5fdc\u3067\u304d\u307e\u3059\u3002\u3053\u308c\u306b\u3088\u308a\u3001\u5206\u6563\u30de\u30eb\u30c1\u30ce\u30fc\u30c9\u69cb\u6210\u3092\u4f7f\u7528\u3057\u3066\u3001\u8907\u96d1\u307e\u305f\u306f\u5927\u898f\u6a21\u306a\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u30fb\u30e2\u30c7\u30eb\u3092\u3088\u308a\u8fc5\u901f\u306b\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u3053\u3068\u304c\u53ef\u80fd\u3067\u3059\u3002\u3053\u308c\u306f\u4e00\u822c\u7684\u306a MPI \u304a\u3088\u3073 lib-fabrics \u6a19\u6e96\u306b\u57fa\u3065\u3044\u3066\u69cb\u7bc9\u3055\u308c\u3066\u304a\u308a\u3001Cornelis Networks\u3001InfiniBand\u3001\u30a4\u30fc\u30b5\u30cd\u30c3\u30c8\u306a\u3069\u306e\u3055\u307e\u3056\u307e\u306a\u76f8\u4e92\u63a5\u7d9a\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u5171\u901a\u6a19\u6e96\u3078\u306e\u6e96\u62e0\u306b\u3088\u308a\u3001\u65b0\u898f\u304a\u3088\u3073\u65e2\u5b58\u306e\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u30fb\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3078\u306e\u7d71\u5408\u304c\u5bb9\u6613\u306b\u306a\u308a\u307e\u3059\u30022025.1 \u3067\u306f\u3001\u4ee5\u4e0b\u306b\u5bfe\u5fdc\u3057\u307e\u3057\u305f\u3002<\/p>\n<ul>\n<li>\u96c6\u5408\u901a\u4fe1\u306e\u5236\u5fa1\u3092\u5f37\u5316\u3059\u308b\u305f\u3081\u3001\u96c6\u5408\u64cd\u4f5c\u3092\u30b5\u30dd\u30fc\u30c8\u3059\u308b Group API \u3092\u62e1\u5f35\u3057\u3001\u30b3\u30df\u30e5\u30cb\u30b1\u30fc\u30bf\u30fc\u3092\u5206\u5272\u3059\u308b\u65b0\u3057\u3044API \u3092\u8ffd\u52a0\u3057\u307e\u3057\u305f\u3002<\/li>\n<li>\u5404\u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u304c\u307b\u304b\u306e\u3059\u3079\u3066\u306e\u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u306b\u500b\u5225\u306e\u30e1\u30c3\u30bb\u30fc\u30b8\u3092\u9001\u4fe1\u3059\u308b\u3001\u5358\u4e00\u30d7\u30ed\u30b0\u30e9\u30e0\u306b\u3088\u308b\u8907\u6570\u30c7\u30fc\u30bf\u64cd\u4f5c\u3067\u3042\u308b all-to-all \u306e\u6700\u9069\u5316\u306b\u3088\u308a\u3001\u3088\u308a\u52b9\u679c\u7684\u306a\u30b9\u30b1\u30fc\u30eb\u30a2\u30c3\u30d7\u3092\u5b9f\u73fe\u3057\u307e\u3059\u3002<\/li>\n<\/ul>\n<p>\u5927\u898f\u6a21\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f7f\u7528\u3059\u308b AI \u306e\u8a08\u7b97\u9818\u57df\u306f\u3001\u5f93\u6765\u306e\u30cf\u30a4\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u30fb\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u3068 AI \u306e\u7d44\u307f\u5408\u308f\u305b\u3067\u3059\u3002\u5927\u898f\u6a21\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f7f\u7528\u3059\u308b\u5206\u6563\u578b\u306e\u53ef\u8996\u5316\u304a\u3088\u3073\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u30fb\u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u306f\u307e\u3059\u307e\u3059\u5897\u3048\u3066\u304a\u308a\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3001\u63a8\u8ad6\u3001\u975e\u7dda\u5f62\u65b9\u7a0b\u5f0f\u3084\u5fae\u5206\u65b9\u7a0b\u5f0f\u306e\u6570\u5024\u89e3\u6cd5\u3001\u30e2\u30f3\u30c6\u30ab\u30eb\u30ed\u30fb\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u306b\u304a\u3044\u3066 GPU \u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30b7\u30e7\u30f3\u30fb\u30aa\u30d5\u30ed\u30fc\u30c9\u30fb\u30ab\u30fc\u30cd\u30eb\u3092\u6d3b\u7528\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u3053\u306e\u3088\u3046\u306a\u30b1\u30fc\u30b9\u3067\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u00ae MPI \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u3068\u65b0\u305f\u306b\u30ea\u30ea\u30fc\u30b9\u3055\u308c\u305f<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/tools\/oneapi\/shmem.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae SHMEM \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc<\/a> (\u82f1\u8a9e) \u3092\u4f75\u7528\u3059\u308b\u3053\u3068\u3067\u3001\u5206\u6563\u74b0\u5883\u306b\u304a\u3051\u308b\u901a\u4fe1\u3068\u30ea\u30e2\u30fc\u30c8\u30fb\u30e1\u30e2\u30ea\u30fc\u30fb\u30a2\u30af\u30bb\u30b9 (RMA) \u306e\u52b9\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<p>\u30dd\u30a4\u30f3\u30c8\u30c4\u30fc\u30dd\u30a4\u30f3\u30c8\u306e\u30ea\u30e2\u30fc\u30c8\u30fb\u30e1\u30e2\u30ea\u30fc\u30fb\u30a2\u30af\u30bb\u30b9 (RMA)\u3001OpenSHMEM 1.6 \u30b9\u30c8\u30e9\u30a4\u30c9 RMA \u64cd\u4f5c\u3001\u30a2\u30c8\u30df\u30c3\u30af\u30e1\u30e2\u30ea\u30fc\u64cd\u4f5c (AMO) \u3001\u30b7\u30b0\u30ca\u30ea\u30f3\u30b0\u3001\u30e1\u30e2\u30ea\u30fc\u30aa\u30fc\u30c0\u30fc\u3001\u30c1\u30fc\u30e0\u3001\u96c6\u5408\u64cd\u4f5c\u3001\u540c\u671f\u306a\u3069\u3001OpenSHMEM 1.5 \u6e96\u62e0\u306e\u6a5f\u80fd\u3092\u5099\u3048\u305f\u30de\u30eb\u30c1\u30ce\u30fc\u30c9\u30fb\u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30bf\u30fc\u30fb\u30c7\u30d0\u30a4\u30b9\u3068\u30db\u30b9\u30c8\u3092\u30bf\u30fc\u30b2\u30c3\u30c8\u306b\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002\u5bfe\u79f0\u30d2\u30fc\u30d7 API \u3092\u4f7f\u7528\u3057\u305f\u30a4\u30f3\u30c6\u30eb\u00ae SHMEM \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u306e SYCL \u30ad\u30e5\u30fc\u9806\u5e8f\u4ed8\u304d RMA \u304a\u3088\u3073 SYCL \u30db\u30b9\u30c8 USM \u30a2\u30af\u30bb\u30b9\u306b\u3088\u308a\u3001\u5206\u6563\u30de\u30eb\u30c1\u30ce\u30fc\u30c9 SYCL \u30c7\u30d0\u30a4\u30b9\u30a2\u30af\u30bb\u30b9\u3092\u7c21\u7d20\u5316\u3057\u307e\u3059\u3002<\/p>\n<p class=\"note\">\u2192 <a href=\"https:\/\/github.com\/oneapi-src\/ishmem\/tree\/main\/examples\" target=\"_blank\" rel=\"noopener\">\u30b5\u30f3\u30d7\u30eb\u30b3\u30fc\u30c9<\/a> (\u82f1\u8a9e) \u3068<a href=\"https:\/\/oneapi-src.github.io\/ishmem\/supported_features.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae SHMEM \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8<\/a> (\u82f1\u8a9e) \u3092\u78ba\u8a8d\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n<p>\u3053\u308c\u3089\u3059\u3079\u3066\u3092\u30b5\u30dd\u30fc\u30c8\u3059\u308b\u305f\u3081\u3001MPI \u6a19\u6e96\u30d9\u30fc\u30b9\u306e\u5206\u6563\u30e1\u30c3\u30bb\u30fc\u30b8\u30d1\u30c3\u30b7\u30f3\u30b0\u306b\u3082\u5e45\u5e83\u3044\u6539\u5584\u304c\u52a0\u3048\u3089\u308c\u307e\u3057\u305f\u3002<\/p>\n<ul>\n<li>\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u306f\u3001P-cores \u642d\u8f09\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae 6\u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u5411\u3051\u306e\u65b0\u3057\u3044\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u30fb\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u6d3b\u7528\u3067\u304d\u307e\u3059\u3002\u3053\u308c\u306b\u306f\u3001\u30b9\u30b1\u30fc\u30eb\u30a2\u30c3\u30d7\u3092\u5411\u4e0a\u3059\u308b\u30b9\u30ec\u30c3\u30c9\u5206\u5272\u3068\u30cf\u30f3\u30c9\u30aa\u30d5\u3001\u30a4\u30f3\u30c6\u30eb\u00ae MPI\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u3092\u30d0\u30c3\u30af\u30a8\u30f3\u30c9\u3068\u3057\u3066\u4f7f\u7528\u3057\u305f\u30a4\u30f3\u30c6\u30eb\u00ae oneCCL \u5411\u3051\u306e CPU \u63a8\u8ad6\u306e\u6700\u9069\u5316\u3001\u30dd\u30a4\u30f3\u30c8\u30c4\u30fc\u30dd\u30a4\u30f3\u30c8\u5171\u6709\u30e1\u30e2\u30ea\u30fc\u64cd\u4f5c\u306e\u6700\u9069\u5316\u306a\u3069\u304c\u542b\u307e\u308c\u307e\u3059\u3002<\/li>\n<li>P-cores \u304a\u3088\u3073 E-cores \u642d\u8f09\u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u306e\u30ea\u30bd\u30fc\u30b9\u4f7f\u7528\u7387\u3092\u5411\u4e0a\u3055\u305b\u308b\u305f\u3081\u3001\u30d4\u30cb\u30f3\u30b0\u30fb\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u304c\u6539\u826f\u3055\u308c\u307e\u3057\u305f\u3002<\/li>\n<\/ul>\n<ul style=\"list-style-type: square\">\n<li>\u30a4\u30f3\u30c6\u30eb\u00ae MPI\u30e9\u30a4\u30d6\u30e9\u30ea\u30fc\u306f\u3001MPI \u6a19\u6e96\u306b\u5148\u99c6\u3051\u3066\u3001\u30b5\u30dd\u30fc\u30c8\u3055\u308c\u3066\u3044\u308b GPU \u4e0a\u3067\u30c7\u30d0\u30a4\u30b9\u958b\u59cb MPI-RMA \u6a5f\u80fd\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u307e\u3057\u305f\u3002<\/li>\n<\/ul>\n<p class=\"note\">\u2192 <a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/oneapi\/index.html?tab=1#features-hpc\">\u30a4\u30f3\u30c6\u30eb\u00ae HPC \u30c4\u30fc\u30eb\u30ad\u30c3\u30c8<\/a>\u3068\u30a4\u30f3\u30c6\u30eb\u306e AI \u30c4\u30fc\u30eb\u304a\u3088\u3073\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306b\u3088\u308a <a href=\"https:\/\/www.intel.co.jp\/content\/www\/jp\/ja\/products\/details\/processors\/xeon.html\" target=\"_blank\" rel=\"noopener\">E-cores \u304a\u3088\u3073 P-cores \u642d\u8f09\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae 6 \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc<\/a>\u306e\u30d1\u30ef\u30fc\u3092\u6700\u5927\u9650\u306b\u6d3b\u7528\u3059\u308b\u65b9\u6cd5\u3092\u5b66\u3076\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<h3>\u4fe1\u983c\u6027\u3001\u4fdd\u5b88\u6027\u3001\u751f\u7523\u6027<\/h3>\n<p>\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u30fb\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u304c\u6271\u3046\u554f\u984c\u3084\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u306e\u8907\u96d1\u3055\u306f\u7d76\u3048\u305a\u5909\u5316\u3057\u3066\u304a\u308a\u3001SYCL\u3001OpenCL\u3001OpenMP \u306a\u3069\u306e\u8a00\u8a9e\u306b\u3088\u308b\u6a19\u6e96\u5316\u3055\u308c\u305f\u30de\u30eb\u30c1\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u30fc\u30fb\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u30fb\u30ab\u30fc\u30cd\u30eb\u30fb\u30aa\u30d5\u30ed\u30fc\u30c9\u306e\u5fc5\u8981\u6027\u3060\u3051\u3067\u306a\u304f\u3001\u751f\u6210\u3055\u308c\u305f\u30b3\u30fc\u30c9\u3092\u52b9\u7387\u826f\u304f\u8a18\u8ff0\u3057\u3001\u4fdd\u5b88\u3057\u3001\u305d\u3057\u3066\u4fe1\u983c\u6027\u3092\u78ba\u4fdd\u3059\u308b\u3053\u3068\u3082\u6c42\u3081\u3089\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u672c\u30ea\u30ea\u30fc\u30b9\u3067 LLVM \u30d9\u30fc\u30b9\u306e\u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u30fb\u30c6\u30af\u30ce\u30ed\u30b8\u30fc\u306f\u3001\u3053\u306e\u70b9\u3092\u6539\u826f\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306b\u5c0e\u5165\u3055\u308c\u305f <a href=\"https:\/\/ccache.dev\/\" target=\"_blank\" rel=\"noopener\">Ccache<\/a> (\u82f1\u8a9e) \u306f\u3001\u8efd\u5fae\u306a\u5909\u66f4\u5f8c\u306e\u30b3\u30fc\u30c9\u306e\u518d\u30b3\u30f3\u30d1\u30a4\u30eb\u6642\u9593\u3092\u77ed\u7e2e\u3057\u3001\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u3092\u52a0\u901f\u3057\u307e\u3059\u3002\u4ee5\u524d\u306e\u30b3\u30f3\u30d1\u30a4\u30eb\u3092\u30ad\u30e3\u30c3\u30b7\u30e5\u3057\u3066\u518d\u5229\u7528\u3059\u308b\u3053\u3068\u3067\u3001\u53cd\u5fa9\u51e6\u7406\u304c\u9ad8\u901f\u5316\u3057\u3001\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u304c\u52b9\u7387\u5316\u3055\u308c\u307e\u3059\u3002\u958b\u767a\u8005\u306f\u3001\u30d3\u30eb\u30c9\u306b\u6642\u9593\u3092\u304b\u3051\u305a\u306b\u3001\u9ad8\u54c1\u8cea\u306e\u30b3\u30fc\u30c9\u306e\u4f5c\u6210\u306b\u96c6\u4e2d\u3067\u304d\u307e\u3059\u3002<\/p>\n<p>\u30b3\u30de\u30f3\u30c9\u30e9\u30a4\u30f3\u3067\u4f7f\u7528\u3059\u308b\u5834\u5408\u306f\u3001\u6b21\u306e\u3088\u3046\u306b\u3001ccache \u30e6\u30fc\u30c6\u30a3\u30ea\u30c6\u30a3\u30fc\u30fb\u30d7\u30ea\u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u3078\u306e\u547c\u3073\u51fa\u3057\u3092\u8ffd\u52a0\u3059\u308b\u3060\u3051\u3067\u3059\u3002<\/p>\n<pre><code>$ ccache icx -c test.c\n$ ccache icpx -fsycl -c sycl_test.cpp<\/code><\/pre>\n<p>CMake \u74b0\u5883\u3067\u306f\u3001-DCMAKE_CXX_COMPILER_LAUNCHER \u30aa\u30d7\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002<\/p>\n<pre><code>$ cmake -DCMAKE_CXX_COMPILER=icpx -DCMAKE_CXX_COMPILER_LAUNCHER=ccache ..<\/code><\/pre>\n<p>\u751f\u7523\u6027\u3092\u5411\u4e0a\u3055\u305b\u3001\u88fd\u54c1\u30ea\u30ea\u30fc\u30b9\u3092\u8fc5\u901f\u5316\u3059\u308b\u3082\u3046 1 \u3064\u306e\u65b9\u6cd5\u306f\u3001\u88fd\u54c1\u30ea\u30ea\u30fc\u30b9\u524d\u306e\u30c7\u30d0\u30c3\u30b0\u56de\u6570\u3084\u691c\u8a3c\u306b\u3088\u308b\u9045\u308c\u3092\u6e1b\u3089\u3059\u3053\u3068\u3067\u3059\u3002\u3053\u308c\u306f\u3001\u30b3\u30fc\u30c9\u8a2d\u8a08\u3068\u5b9f\u88c5\u306e\u521d\u671f\u6bb5\u968e\u3067\u6f5c\u5728\u7684\u306a\u554f\u984c\u3092\u691c\u51fa\u3059\u308b\u3053\u3068\u3067\u5b9f\u73fe\u3067\u304d\u307e\u3059\u3002<\/p>\n<p>\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u306f\u3001\u30b3\u30fc\u30c9\u5185\u306e\u671b\u307e\u3057\u304f\u306a\u3044\u52d5\u4f5c\u3084\u672a\u5b9a\u7fa9\u306e\u52d5\u4f5c\u3092\u7279\u5b9a\u3057\u3001\u30d4\u30f3\u30dd\u30a4\u30f3\u30c8\u3067\u691c\u51fa\u3059\u308b\u306e\u306b\u5f79\u7acb\u3061\u307e\u3059\u3002\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u306f\u3001\u30d7\u30ed\u30b0\u30e9\u30e0\u3092\u30a4\u30f3\u30b9\u30c8\u30eb\u30e1\u30f3\u30c8\u3059\u308b\u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u30fb\u30aa\u30d7\u30b7\u30e7\u30f3\u3067\u6709\u52b9\u306b\u306a\u308a\u3001\u30d0\u30a4\u30ca\u30ea\u30fc\u306b\u5b89\u5168\u30c1\u30a7\u30c3\u30af\u3092\u63d0\u4f9b\u3057\u307e\u3059\u3002\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u3092\u4f7f\u7528\u3059\u308b\u3068\u3001\u958b\u767a\u30d7\u30ed\u30bb\u30b9\u306e\u65e9\u3044\u6bb5\u968e\u3067\u52b9\u679c\u7684\u306b\u554f\u984c\u3092\u767a\u898b\u3057\u3001\u6642\u9593\u3092\u7bc0\u7d04\u3057\u3001\u88fd\u54c1\u30ec\u30d9\u30eb\u306e\u30b3\u30fc\u30c9\u3067\u30b3\u30b9\u30c8\u306e\u304b\u304b\u308b\u30a8\u30e9\u30fc\u304c\u767a\u751f\u3059\u308b\u53ef\u80fd\u6027\u3092\u8efd\u6e1b\u3067\u304d\u307e\u3059\u3002<\/p>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306f\u3001CPU \u304a\u3088\u3073 GPU \u3092\u30bf\u30fc\u30b2\u30c3\u30c8\u3068\u3059\u308b\u30b3\u30fc\u30c9\u5411\u3051\u306b\u3001\u8c4a\u5bcc\u306a\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u3092\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<ol>\n<li><strong>Address Sanitizer<\/strong> &#8211; \u30e1\u30e2\u30ea\u30fc\u306e\u5b89\u5168\u6027\u306e\u554f\u984c\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<li><strong>Undefined Behaviour Sanitizer<\/strong> &#8211; \u672a\u5b9a\u7fa9\u306e\u52d5\u4f5c\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<li><strong>Memory Sanitizer<\/strong> &#8211; \u521d\u671f\u5316\u3055\u308c\u3066\u3044\u306a\u3044\u30e1\u30e2\u30ea\u30fc\u306e\u4f7f\u7528\u306b\u95a2\u3059\u308b\u554f\u984c\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<li><strong>Thread Sanitizer<\/strong> &#8211; \u30c7\u30fc\u30bf\u7af6\u5408\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<li><strong>Numerical Stability Sanitizer<\/strong> &#8211; \u6570\u5024 (\u6d6e\u52d5\u5c0f\u6570\u70b9) \u306b\u95a2\u3059\u308b\u554f\u984c\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<li><strong>Device-Side Address Sanitizer<\/strong>\u2013 SYCL \u30c7\u30d0\u30a4\u30b9\u30b3\u30fc\u30c9\u306e\u30e1\u30e2\u30ea\u30fc\u306e\u5b89\u5168\u6027\u306e\u554f\u984c\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<li><strong>Device-Side Memory Sanitizer<\/strong> \u2013 \u521d\u671f\u5316\u3055\u308c\u3066\u3044\u306a\u3044\u30e1\u30e2\u30ea\u30fc\u306e\u4f7f\u7528\u306b\u95a2\u3059\u308b\u554f\u984c\u3092\u691c\u51fa\u3057\u307e\u3059\u3002<\/li>\n<\/ol>\n<p>\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc 2025.1 \u30ea\u30ea\u30fc\u30b9\u306e\u65b0\u6a5f\u80fd\u306f\u4ee5\u4e0b\u306e\u3068\u304a\u308a\u3067\u3059\u3002<\/p>\n<ul>\n<li>C++ \u30b3\u30fc\u30c9\u5411\u3051\u306e Numerical Stability Sanitizer (\u6570\u5024\u5b89\u5b9a\u6027\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc) \u306e\u5c0e\u5165<\/li>\n<li>SYCL \u304a\u3088\u3073 OpenMP \u30aa\u30d5\u30ed\u30fc\u30c9\u30fb\u30c7\u30d0\u30a4\u30b9\u30fb\u30b3\u30fc\u30c9\u5411\u3051\u306e\u30e1\u30e2\u30ea\u30fc\u304a\u3088\u3073\u30a2\u30c9\u30ec\u30b9\u30fb\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u30fb\u30b5\u30dd\u30fc\u30c8\u306e\u62e1\u5f35 (\u7121\u52b9\u306a\u30ab\u30fc\u30cd\u30eb\u5f15\u6570\u306e\u691c\u51fa\u3001NULL \u30dd\u30a4\u30f3\u30bf\u30fc\u306e\u691c\u51fa\u3001OpenMP \u30aa\u30d5\u30ed\u30fc\u30c9\u306e\u30d7\u30e9\u30a4\u30d9\u30fc\u30c8\u30fb\u30e1\u30e2\u30ea\u30fc\u30fb\u30b5\u30dd\u30fc\u30c8\u3092\u542b\u3080)<\/li>\n<\/ul>\n<p>OpenMP \u3068\u540c\u30ec\u30d9\u30eb\u306e\u30a2\u30c9\u30ec\u30b9\u30fb\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u30fb\u30b5\u30dd\u30fc\u30c8\u304c\u3001\u30a4\u30f3\u30c6\u30eb\u00ae Fortran \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306b\u3082\u8ffd\u52a0\u3055\u308c\u307e\u3057\u305f\u3002<\/p>\n<ul>\n<li>\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306f\u3001\u3059\u3067\u306b CPU \u3067\u30b5\u30dd\u30fc\u30c8\u3057\u3066\u3044\u308b\u30e1\u30e2\u30ea\u30fc\u30fb\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u306e\u6a5f\u80fd\u3092\u3001GPU \u3092\u542b\u3080\u30c7\u30d0\u30a4\u30b9\u5074\u306b\u62e1\u5f35\u3057\u307e\u3059\u3002\u3053\u306e\u62e1\u5f35\u306b\u3088\u308a\u3001\u958b\u767a\u8005\u306f CPU \u3068\u30c7\u30d0\u30a4\u30b9\u30b3\u30fc\u30c9\u306e\u4e21\u65b9\u3067\u554f\u984c\u3092\u7c21\u5358\u306b\u691c\u51fa\u3057\u3066\u30c8\u30e9\u30d6\u30eb\u30b7\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u3067\u304d\u308b\u305f\u3081\u3001\u3088\u308a\u4fe1\u983c\u6027\u304c\u9ad8\u3044\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u73fe\u3067\u304d\u307e\u3059\u3002\u3055\u3089\u306b\u3001\u6570\u5024\u5b89\u5b9a\u6027\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc (NSan) \u304c\u8ffd\u52a0\u3055\u308c\u3001\u6d6e\u52d5\u5c0f\u6570\u70b9\u6f14\u7b97\u306b\u95a2\u9023\u3059\u308b\u30b3\u30fc\u30c9\u306e\u554f\u984c\u3092\u8fc5\u901f\u306b\u7279\u5b9a\u3057\u3066\u89e3\u6c7a\u3059\u308b\u3053\u3068\u3067\u3001\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u306e\u54c1\u8cea\u3068\u5b89\u5b9a\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/li>\n<\/ul>\n<p class=\"greyHighlight\">\u2192 C\/C++ \u30b3\u30fc\u30c9\u5411\u3051\u306e\u6570\u5024\u5b89\u5b9a\u6027\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u306f <code>-fsanitize=numerical<\/code> \u30aa\u30d7\u30b7\u30e7\u30f3\u3067\u6709\u52b9\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n<ul>\n<li>\u30a4\u30f3\u30c6\u30eb\u00ae Fortran \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306f\u3001CPU MemorySanitizer \u306e\u30b5\u30dd\u30fc\u30c8\u3092 GPU \u3092\u542b\u3080\u30c7\u30d0\u30a4\u30b9\u306b\u62e1\u5f35\u3057\u307e\u3057\u305f\u3002\u3053\u308c\u306b\u3088\u308a\u3001Fortran \u958b\u767a\u8005\u306f CPU \u3068\u30c7\u30d0\u30a4\u30b9\u30b3\u30fc\u30c9\u306e\u4e21\u65b9\u3067\u554f\u984c\u3092\u5bb9\u6613\u306b\u691c\u51fa\u3057\u3001\u30c8\u30e9\u30d6\u30eb\u30b7\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u3067\u304d\u308b\u305f\u3081\u3001\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u306e\u4fe1\u983c\u6027\u3068\u5805\u7262\u6027\u3092\u9ad8\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/li>\n<\/ul>\n<p class=\"greyHighlight\">\u2192 \u30c7\u30d0\u30a4\u30b9\u30b3\u30fc\u30c9\u3067\u3053\u306e\u6a5f\u80fd\u3092\u6709\u52b9\u306b\u3059\u308b\u306b\u306f\u3001<code>-Xarch_device -fsanitize=memory<\/code> \u30aa\u30d7\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3057\u3001<code>UR_ENABLE_LAYERS=UR_LAYER_MSAN<\/code> \u30e9\u30f3\u30bf\u30a4\u30e0\u5909\u6570\u3092\u8a2d\u5b9a\u3057\u307e\u3059\u3002\u30db\u30b9\u30c8\u30b3\u30fc\u30c9\u3067\u30e1\u30e2\u30ea\u30fc\u30a2\u30af\u30bb\u30b9\u306e\u554f\u984c\u3092\u7279\u5b9a\u3059\u308b\u306b\u306f\u3001<code>-Xarch_host -fsanitize=address<\/code> \u30aa\u30d7\u30b7\u30e7\u30f3\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002<\/p>\n<p class=\"note\">\u2192 \u3059\u3079\u3066\u306e\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u6a5f\u80fd\u306e\u8a73\u7d30\u306b\u3064\u3044\u3066\u306f\u3001\u300c<a href=\"https:\/\/www.isus.jp\/products\/c-compilers\/find-bugs-quickly-using-sanitizers-with-oneapi-compiler\/\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u3067\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u3092\u4f7f\u7528\u3057\u3066\u30d0\u30b0\u3092\u7d20\u65e9\u304f\u691c\u51fa<\/a>\u300d\u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n<ul>\n<li>\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306e\u30b3\u30fc\u30c9\u30fb\u30ab\u30d0\u30ec\u30c3\u30b8\u30fb\u30c4\u30fc\u30eb\u306b\u306f\u3001C\/C++\u3001SYCL\u3001OpenMP \u3092\u4f7f\u7528\u3059\u308b\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u5411\u3051\u306e GPU \u30b5\u30dd\u30fc\u30c8\u3068\u62e1\u5f35 CPU \u30ab\u30d0\u30ec\u30c3\u30b8\u304c\u8ffd\u52a0\u3055\u308c\u307e\u3057\u305f\u3002\u30c6\u30b9\u30c8\u6e08\u307f\u304a\u3088\u3073\u672a\u30c6\u30b9\u30c8\u306e\u30b3\u30fc\u30c9\u9818\u57df\u3092\u8b58\u5225\u3059\u308b\u8a73\u7d30\u306a\u5206\u6790\u3068\u5305\u62ec\u7684\u306a HTML \u30ec\u30dd\u30fc\u30c8\u304c\u63d0\u4f9b\u3055\u308c\u308b\u305f\u3081\u3001\u30c6\u30b9\u30c8\u30ab\u30d0\u30ec\u30c3\u30b8\u3068\u30b3\u30fc\u30c9\u54c1\u8cea\u304c\u5411\u4e0a\u3057\u3001\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3078\u306e\u5bb9\u6613\u306a\u7d71\u5408\u304c\u4fdd\u8a3c\u3055\u308c\u307e\u3059\u3002<\/li>\n<\/ul>\n<h3>\u6700\u65b0\u306e\u30aa\u30fc\u30d7\u30f3\u6a19\u6e96\u898f\u683c\u306e\u30b5\u30dd\u30fc\u30c8<\/h3>\n<p>\u6700\u65b0\u306e OpenMP \u304a\u3088\u3073 SYCL \u6a5f\u80fd\u306b\u3088\u308b\u4e26\u5217\u30aa\u30d5\u30ed\u30fc\u30c9\u306e\u52b9\u7387\u6027\u3092\u5229\u7528\u3067\u304d\u307e\u3059\u3002\u5f15\u6570\u51e6\u7406\u3068\u30ea\u30b9\u30c8\u5909\u6570\u306b\u95a2\u3059\u308b\u6700\u65b0\u306e Fortran 23 \u6a5f\u80fd\u306b\u3088\u308a\u3001\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u306e\u67d4\u8edf\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/p>\n<ul>\n<li>\u30a4\u30f3\u30c6\u30eb\u00ae Fortran \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306f\u3001\u30b9\u30ec\u30c3\u30c9\u9593\u3067\u4f5c\u696d\u3092\u52b9\u7387\u826f\u304f\u5206\u6563\u3059\u308b <code>WORKDISTRIBUTE<\/code> \u69cb\u9020\u3068\u3001\u30eb\u30fc\u30d7\u30cd\u30b9\u30c8\u5185\u306e\u30eb\u30fc\u30d7\u9806\u5e8f\u3092\u5909\u66f4\u3059\u308b <code>INTERCHANGE<\/code> \u69cb\u9020\u3092\u5c0e\u5165\u3059\u308b\u3053\u3068\u3067\u3001OpenMP 6.0 \u6a19\u6e96\u306e\u30b5\u30dd\u30fc\u30c8\u3092\u62e1\u5f35\u3057\u3001\u4e26\u5217\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3068\u30b3\u30fc\u30c9\u6700\u9069\u5316\u3092\u5411\u4e0a\u3055\u305b\u307e\u3059\u3002<\/li>\n<li>\u30a4\u30f3\u30c6\u30eb\u00ae Fortran \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306f\u3001<code>SYSTEM_CLOCK<\/code> \u7d44\u8fbc\u307f\u95a2\u6570\u306e\u6574\u6570\u5f15\u6570\u306e\u7a2e\u5225\u306e\u4e00\u8cab\u6027\u3092\u78ba\u4fdd\u3057\u3001<code>PUBLIC NAMELIST<\/code> \u30b0\u30eb\u30fc\u30d7\u306b <code>PRIVATE<\/code> \u5909\u6570\u3092\u542b\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b\u3053\u3068\u3067\u3001Fortran 23 \u306e\u30b5\u30dd\u30fc\u30c8\u3092\u5f37\u5316\u3057\u307e\u3057\u305f\u3002\u3053\u308c\u306b\u3088\u308a\u3001Fortran 23 \u8a00\u8a9e\u6a19\u6e96\u3078\u306e\u6e96\u62e0\u304c\u5411\u4e0a\u3057\u3001\u30b3\u30fc\u30c9\u306e\u67d4\u8edf\u6027\u304c\u5411\u4e0a\u3057\u307e\u3059\u3002<\/li>\n<li>C++ with SYCL GPU \u30ab\u30fc\u30cd\u30eb\u306f\u3001<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/tools\/oneapi\/dpc-library.html#gs.g18gln\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI DPC++ \u30e9\u30a4\u30d6\u30e9\u30ea\u30fc (\u30a4\u30f3\u30c6\u30eb\u00ae oneDPL)<\/a> (\u82f1\u8a9e)\u3092\u4f7f\u7528\u3059\u308b\u3053\u3068\u3067\u3001<code><em>copy<\/em>\u3001<em>transform<\/em>\u3001<em>order-changing<\/em><\/code>\u3001<em><code>generation<\/code><\/em>\u3001\u304a\u3088\u3073 <code><em>set<\/em><\/code> \u64cd\u4f5c\u306a\u3069\u306e\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u306e\u5b9f\u884c\u901f\u5ea6\u304c\u6700\u5927 3 \u500d\u5411\u4e0a\u3057\u307e\u3059\u3002\u307e\u305f\u3001\u7bc4\u56f2\u30d9\u30fc\u30b9\u306e C++20 \u6a19\u6e96\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u306e <code><em>copy<\/em><\/code>\u3001<em><code>transform<\/code><\/em>\u3001\u304a\u3088\u3073 <em><code>merge<\/code><\/em> \u3092\u30b5\u30dd\u30fc\u30c8\u3057\u307e\u3057\u305f\u3002<\/li>\n<\/ul>\n<p class=\"note\">\u2192 <a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/devcloud\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae Tiber&#x2122; AI \u30af\u30e9\u30a6\u30c9<\/a>\u3067\u30db\u30b9\u30c8\u3055\u308c\u3066\u3044\u308b Jupyter Notebook \u3067\u6700\u65b0\u306e\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u3092\u7121\u6599\u3067\u8a66\u3057\u3066\u3001\u53ef\u80fd\u6027\u3092\u63a2\u3063\u3066\u307f\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n<h2>\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u306e\u751f\u7523\u6027\u3092\u9ad8\u3081\u307e\u3057\u3087\u3046<\/h2>\n<p>\u3059\u3079\u3066\u306e\u30c4\u30fc\u30eb\u306e 2025.1 \u30ea\u30ea\u30fc\u30b9\u306f\u3001\u4ee5\u4e0b\u304b\u3089\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u3067\u304d\u307e\u3059\u3002<\/p>\n<ul>\n<li><a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/oneapi\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI \u30d9\u30fc\u30b9\u30fb\u30c4\u30fc\u30eb\u30ad\u30c3\u30c8<\/a> &#8211; \u30aa\u30fc\u30d7\u30f3\u30fb\u30a2\u30af\u30bb\u30e9\u30ec\u30fc\u30c6\u30c3\u30c9\u30fb\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u5411\u3051\u306e\u30de\u30eb\u30c1\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u30fc\u5bfe\u5fdc\u306e C++ \u304a\u3088\u3073 Python \u958b\u767a\u30c4\u30fc\u30eb<\/li>\n<li><a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/oneapi\/index.html?tab=1#features-hpc\">\u30a4\u30f3\u30c6\u30eb\u00ae HPC \u30c4\u30fc\u30eb\u30ad\u30c3\u30c8<\/a> &#8211; \u30af\u30e9\u30b9\u30bf\u30fc\u5168\u4f53\u3067\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3059\u308b\u9ad8\u901f\u306a\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3092\u5b9f\u73fe<\/li>\n<li><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/tools\/oneapi\/ai-tools-selector.html\" target=\"_blank\" rel=\"noopener\">AI \u30c4\u30fc\u30eb\u30bb\u30ec\u30af\u30bf\u30fc<\/a> (\u82f1\u8a9e) &#8211; oneAPI \u306b\u3088\u308a AI \u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u3067\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5b9f\u73fe<\/li>\n<\/ul>\n<h3>\u5c0f\u3055\u306a\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u30fb\u30d1\u30c3\u30b1\u30fc\u30b8\u3082\u5229\u7528\u53ef\u80fd<\/h3>\n<p><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/tools\/oneapi\/base-toolkit-download.html?packages=cpp-essentials\" target=\"_blank\" rel=\"noopener\">\u30c4\u30fc\u30eb\u30ad\u30c3\u30c8\u30fb\u30bb\u30ec\u30af\u30bf\u30fc<\/a> (\u82f1\u8a9e) \u3092\u4f7f\u7528\u3057\u3066\u3001\u30d5\u30eb\u30ad\u30c3\u30c8\u307e\u305f\u306f\u7279\u5b9a\u306e\u30e6\u30fc\u30b9\u30b1\u30fc\u30b9\u306b\u5fc5\u8981\u306a\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u306e\u307f\u3092\u542b\u3080\u3088\u308a\u5c0f\u3055\u306a\u30b5\u30a4\u30ba\u306e\u65b0\u3057\u3044\u30b5\u30d6\u30d0\u30f3\u30c9\u30eb\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u3053\u3068\u3067\u3001\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u306e\u30bb\u30c3\u30c8\u30a2\u30c3\u30d7\u3092\u52b9\u7387\u5316\u3067\u304d\u307e\u3059\u3002\u30a4\u30f3\u30c6\u30eb\u00ae C++ \u30a8\u30c3\u30bb\u30f3\u30b7\u30e3\u30eb\u30ba\u3001\u30a4\u30f3\u30c6\u30eb\u00ae Fortan \u30a8\u30c3\u30bb\u30f3\u30b7\u30e3\u30eb\u30ba\u3001\u30a4\u30f3\u30c6\u30eb\u00ae \u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u30fb\u30a8\u30c3\u30bb\u30f3\u30b7\u30e3\u30eb\u30ba\u3092\u5229\u7528\u3057\u3066\u3001\u6642\u9593\u3092\u7bc0\u7d04\u3057\u3001\u624b\u9593\u3092\u6e1b\u3089\u3057\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u6700\u9069\u306a\u30c4\u30fc\u30eb\u3092\u5165\u624b\u3067\u304d\u307e\u3059\u3002<\/p>\n<h3>\u3059\u3050\u306b\u4f7f\u3048\u308b Jupyter Notebook<\/h3>\n<p><a href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/devcloud\/index.html\">\u30a4\u30f3\u30c6\u30eb\u00ae Tiber&#x2122; AI \u30af\u30e9\u30a6\u30c9<\/a>\u4e0a\u306e Jupyter Notebook \u3067\u3001\u65b0\u3057\u3044\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u3092\u7121\u6599\u3067\u304a\u8a66\u3057\u3044\u305f\u3060\u3051\u307e\u3059\u3002<\/p>\n<h2>\u95a2\u9023\u60c5\u5831<\/h2>\n<p>\u88fd\u54c1\u30cb\u30e5\u30fc\u30b9<\/p>\n<ul>\n<li><a href=\"https:\/\/www.isus.jp\/products\/oneapi\/oneapi-news-and-updates\/\" target=\"_blank\" rel=\"noopener\">\u3088\u308a\u9ad8\u901f\u306a AI\u3001\u30ea\u30a2\u30eb\u30bf\u30a4\u30e0\u30fb\u30b0\u30e9\u30d5\u30a3\u30c3\u30af\u30b9\u3001\u3088\u308a\u30b9\u30de\u30fc\u30c8\u306a HPC: \u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb 2025.1 \u30ea\u30ea\u30fc\u30b9!<\/a><\/li>\n<li><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/platform\/xeon.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae Xeon\u00ae \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u5411\u3051\u958b\u767a\u30c4\u30fc\u30eb<\/a> (\u82f1\u8a9e)<\/li>\n<\/ul>\n<p>\u30b3\u30fc\u30c9\u751f\u6210\u3068\u30b3\u30fc\u30c9\u306e\u6b63\u5f53\u6027<\/p>\n<ul>\n<li><a href=\"https:\/\/www.isus.jp\/products\/c-compilers\/find-bugs-quickly-using-sanitizers-with-oneapi-compiler\/\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u3067\u30b5\u30cb\u30bf\u30a4\u30b6\u30fc\u3092\u4f7f\u7528\u3057\u3066\u30d0\u30b0\u3092\u7d20\u65e9\u304f\u691c\u51fa<\/a><\/li>\n<li><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/sycl-bindless-images-migration.html\" target=\"_blank\" rel=\"noopener\">SYCL \u30d0\u30a4\u30f3\u30c9\u30ec\u30b9\u753b\u50cf\u62e1\u5f35: \u30b7\u30f3\u30d7\u30eb\u306a\u79fb\u884c\u30c7\u30e2<\/a> (\u82f1\u8a9e)<\/li>\n<\/ul>\n<p>AI \u30ef\u30fc\u30af\u30ed\u30fc\u30c9\u306e\u30d7\u30ed\u30d5\u30a1\u30a4\u30eb<\/p>\n<ul>\n<li><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/vtune-ai-profile-npu.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae Core&#x2122; Ultra \u30d7\u30ed\u30bb\u30c3\u30b5\u30fc\u4e0a\u3067\u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u3092\u4f7f\u7528\u3057\u3066 NPU \u4f9d\u5b58\u306e AI PC \u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u3092\u9ad8\u901f\u5316\u3059\u308b<\/a> (\u82f1\u8a9e)<\/li>\n<li><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/developer\/articles\/technical\/enhance-genai-llm-performance-using-vtune-profiler.html\" target=\"_blank\" rel=\"noopener\">\u30a4\u30f3\u30c6\u30eb\u00ae VTune&#x2122; \u30d7\u30ed\u30d5\u30a1\u30a4\u30e9\u30fc\u3092\u4f7f\u7528\u3057\u305f GenAI LLM \u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u5411\u4e0a<\/a> (\u82f1\u8a9e)<\/li>\n<\/ul>\n<hr \/>\n<h4>\u88fd\u54c1\u304a\u3088\u3073\u6027\u80fd\u306b\u95a2\u3059\u308b\u60c5\u5831<\/h4>\n<p class=\"disclaimer\"><sup>1<\/sup> \u6027\u80fd\u306f\u3001\u4f7f\u7528\u72b6\u6cc1\u3001\u69cb\u6210\u3001\u305d\u306e\u4ed6\u306e\u8981\u56e0\u306b\u3088\u3063\u3066\u7570\u306a\u308a\u307e\u3059\u3002\u8a73\u7d30\u306b\u3064\u3044\u3066\u306f\u3001<a title=\"Follow link\" href=\"http:\/\/www.intel.com\/PerformanceIndex\" target=\"_blank\" rel=\"noopener\">http:\/\/www.intel.com\/PerformanceIndex\/<\/a> (\u82f1\u8a9e) \u3092\u53c2\u7167\u3057\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n<div class=\"wp-block-group gray-bg\"><div class=\"wp-block-group__inner-container is-layout-constrained wp-block-group-is-layout-constrained\">\n<h3 class=\"wp-block-heading\">\u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb 2025.1 \u5bfe\u5fdc ! \u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u5411\u3051\u30b5\u30dd\u30fc\u30c8\u30b5\u30fc\u30d3\u30b9<\/h3>\n\n\n\n<p>\u30a8\u30af\u30bb\u30eb\u30bd\u30d5\u30c8\u304c\u63d0\u4f9b\u3059\u308b\u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u5411\u3051\u30b5\u30dd\u30fc\u30c8\u30b5\u30fc\u30d3\u30b9\u3067\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u00ae \u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u306e\u65e7\u30d0\u30fc\u30b8\u30e7\u30f3\u304b\u3089\u65b0\u30d0\u30fc\u30b8\u30e7\u30f3\u3078\u306e\u79fb\u884c\u3001CUDA \u304b\u3089 SYCL \u3078\u306e\u30b3\u30fc\u30c9\u79fb\u884c\u3001\u4ed6\u793e\u88fd GPU \u3068\u306e\u30b3\u30fc\u30c9\u4e92\u63db\u306a\u3069\u3001\u65b0\u3057\u3044\u74b0\u5883\u3067\u3053\u308c\u307e\u3067\u901a\u308a\u696d\u52d9\u3092\u9042\u884c\u3059\u308b\u305f\u3081\u306e\u79fb\u884c\u3092\u652f\u63f4\u3057\u307e\u3059\u3002\u88fd\u54c1\u306e\u79fb\u884c\u306b\u95a2\u3057\u3066\u304a\u60a9\u307f\u3001\u8cea\u554f\u306a\u3069\u304a\u6c17\u8efd\u306b\u304a\u554f\u3044\u5408\u308f\u305b\u304f\u3060\u3055\u3044\u3002<\/p>\n\n\n\n<h5 class=\"wp-block-heading\">\u6700\u65b0\u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u65e5\u672c\u8a9e\u53c2\u8003\u8a33<\/h5>\n\n\n\n<p>\u30a8\u30af\u30bb\u30eb\u30bd\u30d5\u30c8\u3067\u306f\u3001\u30d0\u30fc\u30b8\u30e7\u30f3 2025.1 \u306b\u5bfe\u5fdc\u3059\u308b\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306e\u65e5\u672c\u8a9e\u53c2\u8003\u8a33\u3092\u4e00\u822c\u516c\u958b\u3057\u3066\u3044\u307e\u3059\u3002\u73fe\u5728\u3001\u4ee5\u4e0b\u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306b\u95a2\u3059\u308b\u65e5\u672c\u8a9e\u53c2\u8003\u8a33\u3092\u3054\u53c2\u7167\u3044\u305f\u3060\u3051\u307e\u3059\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u30a4\u30f3\u30c6\u30eb<sup>\u00ae<\/sup>&nbsp;\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30c4\u30fc\u30eb\u306e\u30ea\u30ea\u30fc\u30b9\u30ce\u30fc\u30c8\u3001\u52d5\u4f5c\u74b0\u5883 (\u30d9\u30fc\u30b9\u30fb\u30c4\u30fc\u30eb\u30ad\u30c3\u30c8\u3001\u30d9\u30fc\u30b9 &amp; HPC \u30c4\u30fc\u30eb\u30ad\u30c3\u30c8)<\/li>\n\n\n\n<li>\u30a4\u30f3\u30c6\u30eb<sup>\u00ae<\/sup>&nbsp;oneAPI DPC++\/C++ \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306e\u30ea\u30ea\u30fc\u30b9\u30ce\u30fc\u30c8\u3001\u52d5\u4f5c\u74b0\u5883 <\/li>\n\n\n\n<li>\u30a4\u30f3\u30c6\u30eb<sup>\u00ae<\/sup>&nbsp;DPC++ \u4e92\u63db\u6027\u30c4\u30fc\u30eb\u306e\u30ea\u30ea\u30fc\u30b9\u30ce\u30fc\u30c8\u3001\u52d5\u4f5c\u74b0\u5883<\/li>\n\n\n\n<li>\u30a4\u30f3\u30c6\u30eb<sup>\u00ae<\/sup>&nbsp;Fortran \u30b3\u30f3\u30d1\u30a4\u30e9\u30fc\u306e\u30ea\u30ea\u30fc\u30b9\u30ce\u30fc\u30c8\u3001\u52d5\u4f5c\u74b0\u5883 <\/li>\n<\/ul>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button has-custom-font-size\" style=\"font-size:15px\"><a class=\"wp-block-button__link has-white-color has-text-color has-background wp-element-button\" href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/tech\/documents.html?tab=1#doc-oneapi\" style=\"border-radius:10px;background-color:#e64946\"><strong>\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u4e00\u89a7\u306f\u3053\u3061\u3089<\/strong><\/a><\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button has-custom-font-size\" style=\"font-size:15px\"><a class=\"wp-block-button__link has-white-color has-text-color has-background wp-element-button\" href=\"https:\/\/www.xlsoft.com\/jp\/products\/intel\/support\/oneapi.html\" style=\"border-radius:10px;background-color:#e64946\"><strong>\u30b5\u30dd\u30fc\u30c8\u30b5\u30fc\u30d3\u30b9\u306e\u8a73\u7d30\u306f\u3053\u3061\u3089<\/strong><\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div><\/div>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<div class=\"mh-excerpt\"><p>\u3053\u306e\u8a18\u4e8b\u306f\u3001\u30a4\u30f3\u30c6\u30eb\u306e\u30a6\u30a7\u30d6\u30b5\u30a4\u30c8\u3067\u516c\u958b\u3055\u308c\u3066\u3044\u308b <a class=\"mh-excerpt-more\" href=\"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/2025\/04\/18\/productivity-performance-oneapi-2025-1\/\" title=\"\u30a4\u30f3\u30c6\u30eb\u00ae oneAPI \u3068\u30a4\u30f3\u30c6\u30eb\u306e AI \u30c4\u30fc\u30eb 2025.1: \u751f\u6210 AI\u3001\u30d3\u30b8\u30e5\u30a2\u30eb AI\u3001\u30a8\u30c3\u30b8\u304b\u3089\u30af\u30e9\u30a6\u30c9 HPC \u307e\u3067\u751f\u7523\u6027\u3068\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u5411\u4e0a\">[&#8230;]<\/a><\/p>\n<\/div>","protected":false},"author":1,"featured_media":3423,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[12,38,14,120,113],"class_list":["post-3410","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-tech-tips","tag-ai","tag-hpc","tag-oneapi","tag-120","tag-113"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/posts\/3410","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/comments?post=3410"}],"version-history":[{"count":15,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/posts\/3410\/revisions"}],"predecessor-version":[{"id":3475,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/posts\/3410\/revisions\/3475"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/media\/3423"}],"wp:attachment":[{"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/media?parent=3410"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/categories?post=3410"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.xlsoft.com\/jp\/blog\/intel\/wp-json\/wp\/v2\/tags?post=3410"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}