{"id":3743,"date":"2013-03-24T18:58:08","date_gmt":"2013-03-24T09:58:08","guid":{"rendered":"http:\/\/umezawa.dyndns.info\/wordpress\/?p=3743"},"modified":"2013-03-31T23:37:54","modified_gmt":"2013-03-31T14:37:54","slug":"x86-%e3%81%ae%e3%83%93%e3%83%83%e3%83%88%e3%82%b9%e3%82%ad%e3%83%a3%e3%83%b3%e5%91%bd%e4%bb%a4%e3%81%ae-simd-%e5%8c%96","status":"publish","type":"post","link":"http:\/\/umezawa.dyndns.info\/wordpress\/?p=3743","title":{"rendered":"x86 \u306e\u30d3\u30c3\u30c8\u30b9\u30ad\u30e3\u30f3\u547d\u4ee4\u306e SIMD \u5316"},"content":{"rendered":"<p>6\u6708\u306b Haswell \u304c\u51fa\u308b\u3068\u4e88\u60f3\u3055\u308c\u3066\u3044\u307e\u3059\u3002Haswell \u306b\u306f AVX2 \u304c\u642d\u8f09\u3055\u308c\u3001256bit \u30ec\u30b8\u30b9\u30bf\u3067\u6574\u6570 SIMD \u6f14\u7b97\u304c\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002\u4ed6\u306b\u3082\u8272\u3005\u3068\u65b0\u547d\u4ee4\u304c\u8ffd\u52a0\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u65b0\u547d\u4ee4\uff08\u5177\u4f53\u7684\u306b\u306f VPGATHERDD\uff09\u3092\u773a\u3081\u306a\u304c\u3089\u3001\u300c\u3053\u308c\u4f7f\u3063\u305f\u3089\u30cf\u30d5\u30de\u30f3\u30c7\u30b3\u30fc\u30c9\u306e\u4e26\u5217\u5b9f\u884c\u304c\u3067\u304d\u308b\u3093\u3058\u3083\u306d\uff1f\u300d\u3068\u601d\u3063\u3066 Ut Video Codec Suite \u306e\u30cf\u30d5\u30de\u30f3\u30c7\u30b3\u30fc\u30c9\u306e\u30a2\u30bb\u30f3\u30d6\u30e9\u30bd\u30fc\u30b9\u3092\u898b\u3066\u3044\u305f\u306e\u3067\u3059\u304c\u30011\u3064\u554f\u984c\u70b9\u304c\u3042\u308a\u307e\u3057\u305f\u3002BSR (Bit Scan Reverse) \u547d\u4ee4\u306e SIMD \u7248\u304c\u7121\u3044\u306e\u3067\u3059\u3002<\/p>\n<p><!--more--><\/p>\n<p>\u5143\u306e\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3092\u305d\u306e\u307e\u307e SIMD \u5316\u3057\u305f\u3044\u5834\u5408\u3001\u300c4\u3064\u306e\u30c0\u30d6\u30eb\u30ef\u30fc\u30c9\u305d\u308c\u305e\u308c\u306e BSR \u306e\u7d50\u679c\u30924\u3064\u306e\u30c0\u30d6\u30eb\u30ef\u30fc\u30c9\u306b\u683c\u7d0d\u3059\u308b\u300d\u3068\u3044\u3046\u3001PBSRD \u3068\u3067\u3082\u547d\u540d\u3055\u308c\u308b\u3088\u3046\u306a\u51e6\u7406\u304c\u5fc5\u8981\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n<h3>\u6c4e\u7528\u30ec\u30b8\u30b9\u30bf\u306b\u79fb\u52d5\u3057\u3066\u51e6\u7406\u3059\u308b\u5834\u5408<\/h3>\n<p>\u3082\u3063\u3068\u3082\u7c21\u5358\u306a\u306e\u306f\u3001\u4e00\u65e6\u30b9\u30ab\u30e9\u30fc\u306b\u3057\u3066\u51e6\u7406\u3059\u308b\u3068\u3044\u3046\u3082\u306e\u3067\u3059\u3002\u305f\u3068\u3048\u3070\u4ee5\u4e0b\u306e\u901a\u308a\u3002<\/p>\n<blockquote>\n<pre>    pextrd      eax, xmm0, 0\r\n    bsr         eax, eax\r\n    pinsrd      xmm1, eax, 0\r\n    pextrd      eax, xmm0, 1\r\n    bsr         eax, eax\r\n    pinsrd      xmm1, eax, 1\r\n    pextrd      eax, xmm0, 2\r\n    bsr         eax, eax\r\n    pinsrd      xmm1, eax, 2\r\n    pextrd      eax, xmm0, 3\r\n    bsr         eax, eax\r\n    pinsrd      xmm1, eax, 3\r\n<\/pre>\n<\/blockquote>\n<p>\u51e6\u7406\u306f\u81ea\u660e\u3060\u3057\u5206\u304b\u308a\u3084\u3059\u304f\u3066\u3044\u3044\u306e\u3067\u3059\u304c\u3001pextrd\/pinsrd \u306f\u305d\u308c\u307b\u3069\u901f\u304f\u306a\u3044\u306e\u304c\u554f\u984c\u3067\u3059\u3002ymm \u30ec\u30b8\u30b9\u30bf\u306b\u5bfe\u3057\u3066\u51e6\u7406\u3057\u3088\u3046\u3068\u3059\u308b\u3068\u3055\u3089\u306b\u500d\u66f8\u304f\u3053\u3068\u306b\u306a\u3063\u3066\u9045\u3055\u500d\u5897\u3067\u3059\u3057\u3001\u547d\u4ee4\u30c7\u30b3\u30fc\u30c0\u306b\u5bfe\u3059\u308b\u8ca0\u8377\u3082\u5fc3\u914d\u3067\u3059\u3002\u3067\u304d\u308c\u3070\u4ed6\u306e\u65b9\u6cd5\u3092\u53d6\u308a\u305f\u3044\u3068\u3053\u308d\u3067\u3059\u3002<\/p>\n<h3>\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306b\u5909\u63db\u3057\u3066\u51e6\u7406\u3059\u308b\u5834\u5408<\/h3>\n<p>\u5b9f\u306f\u3001BSR \u547d\u4ee4\u76f8\u5f53\u306e\u51e6\u7406\u306f\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306b\u5909\u63db\u3059\u308b\u3053\u3068\u3067\u9054\u6210\u3067\u304d\u308b\u3053\u3068\u304c\u77e5\u3089\u308c\u3066\u3044\u307e\u3059\u3002\uff08\u300c\u3088\u304f\u300d\u77e5\u3089\u308c\u3066\u3044\u308b\u304b\u3069\u3046\u304b\u306f\u5206\u304b\u308a\u307e\u305b\u3093\uff09<\/p>\n<p>\u305f\u3068\u3048\u3070\u30012\u9032\u6570\u3067 10010100<sub>2<\/sub> \u3092\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306b\u5909\u63db\u3059\u308b\u3068 1.0010100<sub>2<\/sub> \u00d7 2<sup>7<\/sup> \u306b\u306a\u308a\u307e\u3059\u304c\u3001\u6307\u6570\u90e8\u306b\u51fa\u3066\u304f\u308b 7 \u3053\u305d\u304c\u5143\u306e 10010100<sub>2<\/sub> \u306e BSR \u306e\u7d50\u679c\u3068\u306a\u308a\u307e\u3059\u3002<\/p>\n<p>\u3068\u3044\u3046\u308f\u3051\u3067\u3001\u4ee5\u4e0b\u306e\u30b3\u30fc\u30c9\u3067\u4e0a\u3068\u540c\u3058\u51e6\u7406\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<blockquote>\n<pre>    cvtdq2ps    xmm1, xmm0\r\n    psrld       xmm1, 23\r\n    psubd       xmm1, 0000007f0000007f0000007f0000007fh\r\n    ; \u5b9f\u969b\u306b\u306f\u4e0a\u306e\u3088\u3046\u306a\u5373\u5024\u306f\u4f7f\u3048\u306a\u3044\u306e\u3067\u3001\r\n    ; \u30e1\u30e2\u30ea\u306b\u7f6e\u3044\u3066\u304a\u304f\u304b\u3001\u305d\u308c\u3092\u30ec\u30b8\u30b9\u30bf\u306b\u30ed\u30fc\u30c9\u3057\u3066\u4f7f\u3046\r\n<\/pre>\n<\/blockquote>\n<p>cvtdq2ps \u30674\u3064\u306e\u30c0\u30d6\u30eb\u30ef\u30fc\u30c9\u30924\u3064\u306e\u5358\u7cbe\u5ea6\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306b\u5909\u63db\u3057\u3001psrld \u3067\u6307\u6570\u90e8\u3092 LSB \u307e\u3067\u79fb\u52d5\u3057\u3001psubd \u3067\u6307\u6570\u90e8\u306b\u5c65\u304b\u3055\u308c\u305f\u30b2\u30bf\u3092\u53d6\u308a\u9664\u304d\u307e\u3059\u3002\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u304c\u51fa\u3066\u304f\u308b\u306e\u3067\u901f\u5ea6\u304c\u6c17\u306b\u306a\u308b\u3068\u3053\u308d\u3067\u3059\u304c\u3001cvtdq2ps \u547d\u4ee4\u306e\u30ec\u30a4\u30c6\u30f3\u30b7\u306f\u305f\u3063\u305f\u306e3\u30af\u30ed\u30c3\u30af\u306a\u306e\u3067\u554f\u984c\u3067\u306f\u3042\u308a\u307e\u305b\u3093\u3002\u3053\u306e\u65b9\u6cd5\u3060\u3068 ymm \u30ec\u30b8\u30b9\u30bf\u306b\u306a\u3063\u3066\u3082\u5168\u304f\u540c\u3058\u306a\u306e\u3067\u9045\u304f\u306a\u308b\u5fc3\u914d\u3082\u3042\u308a\u307e\u305b\u3093\u3002<\/p>\n<p># \u3066\u3086\u30fc\u304b\u3001 CVTDQ2PS \u547d\u4ee4\u304c\u5b9f\u88c5\u3055\u308c\u3066\u3044\u308b\u3063\u3066\u3053\u3068\u306f\u3001<br \/>\n# \u56de\u8def\u7684\u306b\u306f PBSRD \u547d\u4ee4\u3068\u7b49\u4fa1\u306a\u51e6\u7406\u306f\u5b9f\u88c5\u3055\u308c\u3066\u3044\u308b\u308f\u3051\u3067\u3001<br \/>\n# \u305d\u308c\u304c\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u547d\u4ee4\u3068\u3057\u3066\u5b58\u5728\u3057\u3066\u3044\u306a\u3044\u306e\u304c\u6b8b\u5ff5\u3002<\/p>\n<p>\u305f\u3063\u305f3\u547d\u4ee4\u3067\u5b9f\u73fe\u3067\u304d\u3066\u4e07\u3005\u6b73\u2026\u3068\u8a00\u3044\u305f\u3044\u3068\u3053\u308d\u3067\u3059\u304c\u3001\u4e0a\u306e\u30b3\u30fc\u30c9\u3060\u30682\u3064\u306e\u70b9\u3067\u3046\u307e\u304f\u52d5\u304d\u307e\u305b\u3093\u3002<\/p>\n<ul>\n<li>\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\uff08BSR \u306e\u7d50\u679c\u304c 31 \u306b\u306a\u308b\u5834\u5408\uff09\u306b\u6b63\u3057\u3044\u7d50\u679c\u306b\u306a\u3089\u306a\u3044\u3002<\/li>\n<li>\u5165\u529b\u306b\u7acb\u3063\u3066\u3044\u308b\u30d3\u30c3\u30c8\u304c\u591a\u3044\u5834\u5408\u306b\u6b63\u3057\u3044\u7d50\u679c\u306b\u306a\u3089\u306a\u3044\u3053\u3068\u304c\u3042\u308b\u3002<\/li>\n<\/ul>\n<h4>\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306e\u30b1\u30a2<\/h4>\n<p>\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306b\u6b63\u3057\u3044\u7d50\u679c\u306b\u306a\u3089\u306a\u3044\u306e\u306f\u3001cvtdq2ps \u547d\u4ee4\u306e\u5165\u529b\u306f4\u3064\u306e\u300c\u7b26\u53f7\u4ed8\u304d\u300d\u30c0\u30d6\u30eb\u30ef\u30fc\u30c9\u3067\u3042\u308b\u305f\u3081\u306b\u30de\u30a4\u30ca\u30b9\u306e\u6570\u3068\u3057\u3066\u6271\u308f\u308c\u3066\u3057\u307e\u3046\u304b\u3089\u3067\u3059\u3002<\/p>\n<p>\u3068\u3044\u3046\u8a33\u3067\u3001\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306b\u7121\u7406\u3084\u308a 31 \u306b\u306a\u308b\u3088\u3046\u306b\u3057\u306a\u3051\u308c\u3070\u306a\u308a\u307e\u305b\u3093\u3002\u4ee5\u4e0b\u306e\u3088\u3046\u306a\u30b3\u30fc\u30c9\u3067\u3057\u3087\u3046\u304b\u3002<\/p>\n<blockquote>\n<pre>    cvtdq2ps    xmm1, xmm0\r\n    psrld       xmm1, 23\r\n    psubd       xmm1, 0000007f0000007f0000007f0000007fh\r\n\r\n    psrad       xmm0, 31\r\n    por         xmm1, xmm0\r\n    pand        xmm1, 0000001f0000001f0000001f0000001fh\r\n<\/pre>\n<\/blockquote>\n<p>SSE4.1 \u3060\u3068 PBLENDVB \u304c\u4f7f\u3048\u308b\u306e\u3067\u3001\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u3061\u3087\u3063\u3068\u77ed\u304f\u3067\u304d\u307e\u3059\uff08\u901f\u3044\u304b\u3069\u3046\u304b\u306f\u8b0e\uff09\u3002<\/p>\n<blockquote>\n<pre>    cvtdq2ps    xmm1, xmm0\r\n    psrld       xmm1, 23\r\n    psubd       xmm1, 0000007f0000007f0000007f0000007fh\r\n\r\n    psrad       xmm0, 31\r\n    pblendvb    xmm1, 0000001f0000001f0000001f0000001fh, xmm0\r\n<\/pre>\n<\/blockquote>\n<p>\u3061\u306a\u307f\u306b\u3001PBLENDVB \u306f\u30de\u30b9\u30af\u30ec\u30b8\u30b9\u30bf\u3068\u3057\u3066 xmm0 \u3057\u304b\u53d6\u308c\u305a\u82e5\u5e72\u4f7f\u3044\u3065\u3089\u3044\u306e\u3067\u3059\u304c\u3001AVX \u306e VPBLENDVB \u3060\u3068\u4efb\u610f\u306e xmm\/ymm \u30ec\u30b8\u30b9\u30bf\u3092\u6307\u5b9a\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u3063\u3066\u4f7f\u3044\u52dd\u624b\u304c\u5411\u4e0a\u3057\u3066\u3044\u307e\u3059\u3002<\/p>\n<h4>\u30d3\u30c3\u30c8\u304c\u305f\u304f\u3055\u3093\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306e\u30b1\u30a2<\/h4>\n<p>\u305f\u3068\u3048\u3070\u3001 0x7ffffff0 \u3092\u5165\u529b\u3057\u305f\u5834\u5408\u3001\u305d\u306e\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u8868\u73fe\u306f 1.11111111111111111111111111<sub>2<\/sub> \u00d7 2<sup>30<\/sup> \uff08\u5c0f\u6570\u70b9\u306e\u5f8c\u308d\u306b 1 \u304c 26 \u500b\uff09\u3067\u3059\u304c\u3001\u5358\u7cbe\u5ea6\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306f\u4eee\u6570\u90e8\u304c\uff08\u5c0f\u6570\u70b9\u4ee5\u4e0b\uff0923\u30d3\u30c3\u30c8\u3057\u304b\u306a\u3044\u306e\u3067\u3001\u4e38\u3081\u3066 1.0<sub>2<\/sub> \u00d7 2<sup>31<\/sup> \u306b\u5909\u63db\u3055\u308c\u307e\u3059\u3002\u7d50\u679c\u3068\u3057\u3066\u6b63\u3057\u3044\u7d50\u679c\u3088\u308a 1 \u5927\u304d\u304f\u306a\u308a\u307e\u3059\u3002<\/p>\n<p>\u4e38\u3081\u3092\u9632\u6b62\u3059\u308b\u306b\u306f\u4e0b\u306e\u65b9\u306e\u30d3\u30c3\u30c8\u3092\u843d\u3068\u3057\u3066\u3057\u307e\u3048\u3070\u3044\u3044\u306e\u3067\u3059\u304c\u3001\u3069\u3053\u307e\u3067\u304c\u843d\u3068\u3057\u3066\u3044\u3044\u30d3\u30c3\u30c8\u306a\u306e\u304b\u306f BSR \u306e\u7d50\u679c\u3092\u77e5\u3089\u306a\u3051\u308c\u3070\u5224\u65ad\u3067\u304d\u306a\u3044\u305f\u3081\u3001\u5225\u306e\u65b9\u6cd5\u304c\u5fc5\u8981\u3067\u3059\u3002\u7d50\u679c\u304c1\u5927\u304d\u304f\u306a\u3063\u3066\u3044\u308b\uff08\u7d50\u679c\u3067\u793a\u3055\u308c\u308b\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\uff09\u304b\u3069\u3046\u304b\u3067\u5224\u65ad\u3059\u308b\u307b\u3046\u304c\u697d\u3067\u3057\u3087\u3046\u3002<\/p>\n<blockquote>\n<pre>    cvtdq2ps    xmm1, xmm0\r\n    psrld       xmm1, 23\r\n    psubd       xmm1, 0000007f0000007f0000007f0000007fh\r\n\r\n    vpsllvd     xmm2, 00000001000000010000000100000001h, xmm1 ; AVX2 only\r\n    pand        xmm2, xmm0\r\n    pcmpeqd     xmm2, 00000000000000000000000000000000h\r\n    paddd       xmm1, xmm2\r\n<\/pre>\n<\/blockquote>\n<p><ins>3\/26\u8ffd\u8a18: \u305f\u3076\u3093\u4ee5\u4e0b\u306e\u65b9\u304c1\u30af\u30ed\u30c3\u30af\u901f\u3044\u3002<\/ins><\/p>\n<blockquote>\n<pre>    cvtdq2ps    xmm1, xmm0\r\n    psrld       xmm1, 23\r\n    psubd       xmm1, 0000007f0000007f0000007f0000007fh\r\n\r\n    vpsrlvd     xmm2, xmm0, xmm1 ; AVX2 only\r\n    pcmpeqd     xmm2, 00000000000000000000000000000000h\r\n    paddd       xmm1, xmm2\r\n<\/pre>\n<\/blockquote>\n<p><ins>3\/27\u8ffd\u8a18: \u3069\u3046\u305b VEX \u30d7\u30ec\u30d5\u30a3\u30c3\u30af\u30b9\u4ed8\u304d SIMD \u547d\u4ee4\u3092\u4f7f\u308f\u306a\u3051\u308c\u3070\u3044\u3051\u306a\u3044\u306e\u3060\u304b\u3089\u3001\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u547d\u4ee4\u306e\u4f9d\u5b58\u95a2\u4fc2\u3092\u6539\u5584\u3059\u308c\u3070\u305f\u3076\u3093\u3055\u3089\u306b1\u30af\u30ed\u30c3\u30af\u901f\u3044\u3002<\/ins><\/p>\n<blockquote>\n<pre>    vcvtdq2ps   xmm1, xmm0\r\n    vpsrld      xmm1, xmm1, 23\r\n    vpsubd      xmm3, xmm1, 00000080000000800000008000000080h\r\n    vpsubd      xmm1, xmm1, 0000007f0000007f0000007f0000007fh\r\n\r\n    vpsrlvd     xmm2, xmm0, xmm1 ; AVX2 only\r\n    vpaddd      xmm1, xmm3, xmm2\r\n<\/pre>\n<\/blockquote>\n<p>\u5404\u30c0\u30d6\u30eb\u30ef\u30fc\u30c9\u3054\u3068\u306b\u7570\u306a\u308b\u30d3\u30c3\u30c8\u6570\u306e\u30b7\u30d5\u30c8\u3092\u3059\u308b\u305f\u3081\u306b\u3001AVX2 \u3067\u8ffd\u52a0\u3055\u308c\u305f VPSLLVD\/VPSRLVD \u547d\u4ee4\u304c\u5fc5\u8981\u306b\u306a\u3063\u3066\u3057\u307e\u3044\u307e\u3057\u305f\u3002\u306a\u304a\u3001\u4e0a\u306e\u30b3\u30fc\u30c9\u306f\u30c6\u30b9\u30c8\u3057\u3066\u3044\u307e\u305b\u3093\uff08AVX2 \u30a8\u30df\u30e5\u30ec\u30fc\u30bf\u3092\u4f7f\u3048\u3070\u30c6\u30b9\u30c8\u3067\u304d\u307e\u3059\u304c\uff09\u3002<\/p>\n<p>\u5165\u529b\u304c 24bit \u4ee5\u4e0b\u3067\u3042\u308b\u3053\u3068\u304c\u4fdd\u8a3c\u3055\u308c\u3066\u3044\u308c\u3070\u4e38\u3081\u306f\u8d77\u304d\u306a\u3044\u306e\u3067\u3001\u305d\u3061\u3089\u306e\u65b9\u5411\u3067\u5bfe\u51e6\u3057\u305f\u65b9\u304c\u3044\u3044\u304b\u3082\u3057\u308c\u307e\u305b\u3093\u3002\u3064\u3044\u3067\u306b\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306e\u30b1\u30a2\u3082\u4e0d\u8981\u3067\u3059\u3002<\/p>\n<h4>\u500d\u7cbe\u5ea6\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306b\u5909\u63db\u3057\u3066\u51e6\u7406\u3059\u308b\u5834\u5408<\/h4>\n<p>\u4e38\u3081\u3092\u56de\u907f\u3059\u308b\u306b\u306f\u500d\u7cbe\u5ea6\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u3092\u4f7f\u3046\u65b9\u6cd5\u3082\u3042\u308a\u307e\u3059\u3002\u3053\u3061\u3089\u306f\u4eee\u6570\u90e8\u304c52\u30d3\u30c3\u30c8\u3042\u308b\u306e\u3067\u4e38\u3081\u306f\u8d77\u304d\u307e\u305b\u3093\u3002CVTDQ2PD \u306e\u5165\u529b\u306f CVTDQ2PS \u3068\u540c\u3058\u304f\u7b26\u53f7\u4ed8\u304d\u30c0\u30d6\u30eb\u30ef\u30fc\u30c9\u306a\u306e\u3067\u3001\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306e\u30b1\u30a2\u306f\u76f8\u5909\u308f\u3089\u305a\u5fc5\u8981\u3067\u3059\u3002<\/p>\n<blockquote>\n<pre>    pshufd      xmm1, xmm0, 0dh\r\n    pshufd      xmm0, xmm0, 08h\r\n\r\n    cvtdq2pd    xmm0, xmm0\r\n    psrld       xmm0, 52\r\n    psubd       xmm0, 00000000000003ff00000000000003ffh\r\n\r\n    cvtdq2pd    xmm1, xmm1\r\n    psrld       xmm1, 52\r\n    psubd       xmm1, 00000000000003ff00000000000003ffh\r\n\r\n    pslld       xmm1, 32\r\n    por         xmm1, xmm0\r\n<\/pre>\n<\/blockquote>\n<p>\u3067\u3082\u3053\u308c\u306f\u9577\u3044\u3067\u3059\u306d\u3002<\/p>\n<h3>\u901f\u3044\u306e\uff1f<\/h3>\n<p>\u3067\u3001\u5f53\u7136\u300c\u3053\u308c\u306f\u901f\u3044\u306e\u304b\uff1f\u300d\u3068\u3044\u3046\u7591\u554f\u304c\u51fa\u3066\u304f\u308b\u308f\u3051\u3067\u3059\u304c\u2026 \u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306f\u5225\u306e\u8a18\u4e8b\u3067\u3002<\/p>\n<p># \u305d\u3082\u305d\u3082\u3069\u3046\u3084\u3063\u3066\u8a08\u6e2c\u3057\u3088\u3046\u304b<\/p>\n<h3>BSF \u306e\u5834\u5408<\/h3>\n<p>\u4ee5\u4e0a\u306f BSR (Bit Scan Reverse) \u306e\u5834\u5408\u3067\u3059\u304c\u3001\u4e0b\u4f4d\u30d3\u30c3\u30c8\u304b\u3089\u30b9\u30ad\u30e3\u30f3\u3059\u308b BSF (Bit Scan Forward) \u547d\u4ee4\u306e\u5834\u5408\u306f\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306b\u5909\u63db\u3059\u308b\u65b9\u6cd5\u306f\u76f4\u63a5\u7684\u306b\u306f\u4f7f\u3048\u307e\u305b\u3093\u3002<\/p>\n<p>\u6700\u3082\u4e0b\u4f4d\u306e1\u306e\u30d3\u30c3\u30c8\u3060\u3051\u3092\u62bd\u51fa\u3057\u3066\u3057\u307e\u3048\u3070 BSF \u3082 BSR \u3082\u7d50\u679c\u306f\u540c\u3058\u306b\u306a\u308b\u306e\u3067\u3001\u305d\u306e\u65b9\u91dd\u3067\u884c\u304f\u3068\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002<\/p>\n<blockquote>\n<pre>    movdqa      xmm1, xmm0\r\n    psubd       xmm0, 00000001000000010000000100000001h\r\n    pandn       xmm0, xmm1\r\n\r\n    cvtdq2ps    xmm1, xmm0\r\n    psrld       xmm1, 23\r\n    psubd       xmm1, 0000007f0000007f0000007f0000007fh\r\n<\/pre>\n<\/blockquote>\n<p>\u3053\u306e\u5834\u5408\u3001\u6700\u4e0a\u4f4d\u30d3\u30c3\u30c8\u304c\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306e\u30b1\u30a2\u306f\uff08\u5834\u5408\u306b\u3088\u3063\u3066\u306f\uff09\u5fc5\u8981\u3067\u3059\u304c\u3001\u30d3\u30c3\u30c8\u304c\u305f\u304f\u3055\u3093\u7acb\u3063\u3066\u3044\u308b\u5834\u5408\u306e\u30b1\u30a2\u306f\u4e0d\u8981\u3067\u3059\u3002<\/p>\n<p>\u4ed6\u306e\u65b9\u6cd5\u306f\u3042\u308a\u307e\u3059\u304b\u306d\uff1f<\/p>\n","protected":false},"excerpt":{"rendered":"<p>6\u6708\u306b Haswell \u304c\u51fa\u308b\u3068\u4e88\u60f3\u3055\u308c\u3066\u3044\u307e\u3059\u3002Haswell \u306b\u306f AVX2 \u304c\u642d\u8f09\u3055\u308c\u3001256bit \u30ec\u30b8\u30b9\u30bf\u3067\u6574\u6570 SIMD \u6f14\u7b97\u304c\u3067\u304d\u308b\u3088\u3046\u306b\u306a\u308a\u307e\u3059\u3002\u4ed6\u306b\u3082\u8272\u3005\u3068\u65b0\u547d\u4ee4\u304c\u8ffd\u52a0\u3055\u308c\u3066\u3044\u307e\u3059\u3002 \u65b0\u547d\u4ee4\uff08\u5177\u4f53\u7684\u306b\u306f [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[27],"tags":[],"class_list":["post-3743","post","type-post","status-publish","format-standard","hentry","category-technology"],"_links":{"self":[{"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3743","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=3743"}],"version-history":[{"count":31,"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3743\/revisions"}],"predecessor-version":[{"id":3783,"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3743\/revisions\/3783"}],"wp:attachment":[{"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=3743"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=3743"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/umezawa.dyndns.info\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=3743"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}