{ "type": "excalidraw", "version": 2, "source": "generated", "elements": [ { "id": "el0001", "type": "text", "x": 20, "y": 10, "width": 859.1, "height": 27.5, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1390851129, "version": 1, "versionNonce": 647892280, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "K2: one CTA per Q tile, online softmax over KV tiles, 4 threads per row", "fontSize": 22, "fontFamily": 1, "textAlign": "left", "verticalAlign": "top", "containerId": null, "originalText": "K2: one CTA per Q tile, online softmax over KV tiles, 4 threads per row", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0002", "type": "text", "x": 20, "y": 50, "width": 24.750000000000004, "height": 18.75, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1695753999, "version": 1, "versionNonce": 207388625, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "HBM", "fontSize": 15, "fontFamily": 1, "textAlign": "left", "verticalAlign": "top", "containerId": null, "originalText": "HBM", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0003", "type": "rectangle", "x": 20, "y": 75, "width": 80, "height": 60, "angle": 0, "strokeColor": "#1971c2", "backgroundColor": "#a5d8ff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 311111476, "version": 1, "versionNonce": 404285458, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0004", "type": "text", "x": 33.05, "y": 87.5, "width": 53.900000000000006, "height": 35.0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1570621945, "version": 1, "versionNonce": 249103478, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "Q tile\n(64, d)", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "Q tile\n(64, d)", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0005", "type": "rectangle", "x": 20, "y": 160, "width": 80, "height": 180, "angle": 0, "strokeColor": "#1971c2", "backgroundColor": "#a5d8ff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 922121677, "version": 1, "versionNonce": 161042649, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0006", "type": "text", "x": 36.9, "y": 168, "width": 46.2, "height": 35.0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 369140571, "version": 1, "versionNonce": 1862494043, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "K\n(N, d)", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "K\n(N, d)", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0007", "type": "rectangle", "x": 110, "y": 160, "width": 80, "height": 180, "angle": 0, "strokeColor": "#1971c2", "backgroundColor": "#a5d8ff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1796035740, "version": 1, "versionNonce": 300026768, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0008", "type": "text", "x": 126.9, "y": 168, "width": 46.2, "height": 35.0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1033639717, "version": 1, "versionNonce": 389609434, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "V\n(N, d)", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "V\n(N, d)", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0009", "type": "rectangle", "x": 24, "y": 250, "width": 72, "height": 30, "angle": 0, "strokeColor": "#e03131", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "dashed", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1823296039, "version": 1, "versionNonce": 253877687, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0010", "type": "rectangle", "x": 114, "y": 250, "width": 72, "height": 30, "angle": 0, "strokeColor": "#e03131", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "dashed", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 531725348, "version": 1, "versionNonce": 958804058, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0011", "type": "text", "x": 40.2, "y": 300, "width": 39.6, "height": 15.0, "angle": 0, "strokeColor": "#e03131", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 265695474, "version": 1, "versionNonce": 1703729685, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "tile j", "fontSize": 12, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "tile j", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0012", "type": "text", "x": 130.2, "y": 300, "width": 39.6, "height": 15.0, "angle": 0, "strokeColor": "#e03131", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 212984477, "version": 1, "versionNonce": 949539217, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "tile j", "fontSize": 12, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "tile j", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0013", "type": "arrow", "x": 105, "y": 105, "width": 170, "height": 0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 1.5, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 2 }, "seed": 200071089, "version": 1, "versionNonce": 571981486, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "points": [ [ 0, 0 ], [ 170, 0 ] ], "lastCommittedPoint": null, "startBinding": null, "endBinding": null, "startArrowhead": null, "endArrowhead": "arrow", "elbowed": false }, { "id": "el0014", "type": "text", "x": 143.8, "y": 85.5, "width": 92.4, "height": 17.5, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1243862423, "version": 1, "versionNonce": 1800188483, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "once per CTA", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "once per CTA", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0015", "type": "arrow", "x": 192, "y": 265, "width": 83, "height": 0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 1.5, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 2 }, "seed": 619570853, "version": 1, "versionNonce": 505913793, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "points": [ [ 0, 0 ], [ 83, 0 ] ], "lastCommittedPoint": null, "startBinding": null, "endBinding": null, "startArrowhead": null, "endArrowhead": "arrow", "elbowed": false }, { "id": "el0016", "type": "text", "x": 200.0, "y": 243, "width": 66.0, "height": 15.0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1324919353, "version": 1, "versionNonce": 776213900, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "per iter j", "fontSize": 12, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "per iter j", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0017", "type": "rectangle", "x": 280, "y": 50, "width": 300, "height": 300, "angle": 0, "strokeColor": "#495057", "backgroundColor": "#f8f9fa", "fillStyle": "solid", "strokeWidth": 1.5, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 442620899, "version": 1, "versionNonce": 806899910, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0018", "type": "text", "x": 310.65, "y": 56, "width": 238.70000000000002, "height": 17.5, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1599435268, "version": 1, "versionNonce": 418461139, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "shared memory, 48 KB at d = 128", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "shared memory, 48 KB at d = 128", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0019", "type": "rectangle", "x": 300, "y": 80, "width": 260, "height": 50, "angle": 0, "strokeColor": "#f08c00", "backgroundColor": "#ffec99", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 269676600, "version": 1, "versionNonce": 255985077, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0020", "type": "text", "x": 368.4, "y": 96.25, "width": 123.20000000000002, "height": 17.5, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 884585952, "version": 1, "versionNonce": 2132084005, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "sQ (64, d) BF16", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "sQ (64, d) BF16", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0021", "type": "rectangle", "x": 300, "y": 150, "width": 260, "height": 50, "angle": 0, "strokeColor": "#f08c00", "backgroundColor": "#ffec99", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1836494975, "version": 1, "versionNonce": 1349251824, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0022", "type": "text", "x": 310.65, "y": 166.25, "width": 238.70000000000002, "height": 17.5, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1999744785, "version": 1, "versionNonce": 1946412081, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "sKV (64, d) BF16: K_j, then V_j", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "sKV (64, d) BF16: K_j, then V_j", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0023", "type": "rectangle", "x": 300, "y": 220, "width": 260, "height": 50, "angle": 0, "strokeColor": "#f08c00", "backgroundColor": "#ffec99", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1552984409, "version": 1, "versionNonce": 1287489454, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0024", "type": "text", "x": 306.79999999999995, "y": 236.25, "width": 246.40000000000003, "height": 17.5, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1066984056, "version": 1, "versionNonce": 772092315, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "sS (64, 64) FP32: S_j, then P_j", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "sS (64, 64) FP32: S_j, then P_j", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0025", "type": "text", "x": 297.725, "y": 285, "width": 264.55, "height": 32.5, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1048386556, "version": 1, "versionNonce": 351564608, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "3 scalar loops per iteration:\nS = Q K^T, P = exp(S - m), O += P V", "fontSize": 13, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "3 scalar loops per iteration:\nS = Q K^T, P = exp(S - m), O += P V", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0026", "type": "rectangle", "x": 610, "y": 50, "width": 240, "height": 300, "angle": 0, "strokeColor": "#495057", "backgroundColor": "#f8f9fa", "fillStyle": "solid", "strokeWidth": 1.5, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1289560150, "version": 1, "versionNonce": 2126508551, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0027", "type": "text", "x": 622.2, "y": 56, "width": 215.60000000000002, "height": 17.5, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1475216846, "version": 1, "versionNonce": 1927728187, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "registers, row r (4 threads)", "fontSize": 14, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "registers, row r (4 threads)", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0028", "type": "rectangle", "x": 625, "y": 85, "width": 48, "height": 40, "angle": 0, "strokeColor": "#6741d9", "backgroundColor": "#d0bfff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1236683273, "version": 1, "versionNonce": 314395343, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0029", "type": "text", "x": 641.85, "y": 96.875, "width": 14.3, "height": 16.25, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 507088657, "version": 1, "versionNonce": 1795823849, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "t0", "fontSize": 13, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "t0", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0030", "type": "rectangle", "x": 680, "y": 85, "width": 48, "height": 40, "angle": 0, "strokeColor": "#6741d9", "backgroundColor": "#d0bfff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 708506837, "version": 1, "versionNonce": 1469118511, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0031", "type": "text", "x": 696.85, "y": 96.875, "width": 14.3, "height": 16.25, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 652768598, "version": 1, "versionNonce": 2100080515, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "t1", "fontSize": 13, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "t1", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0032", "type": "rectangle", "x": 735, "y": 85, "width": 48, "height": 40, "angle": 0, "strokeColor": "#6741d9", "backgroundColor": "#d0bfff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1811180650, "version": 1, "versionNonce": 168393880, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0033", "type": "text", "x": 751.85, "y": 96.875, "width": 14.3, "height": 16.25, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 333377415, "version": 1, "versionNonce": 1347535309, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "t2", "fontSize": 13, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "t2", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0034", "type": "rectangle", "x": 790, "y": 85, "width": 48, "height": 40, "angle": 0, "strokeColor": "#6741d9", "backgroundColor": "#d0bfff", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1460814403, "version": 1, "versionNonce": 1504004732, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0035", "type": "text", "x": 806.85, "y": 96.875, "width": 14.3, "height": 16.25, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 2133201996, "version": 1, "versionNonce": 1959386987, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "t3", "fontSize": 13, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "t3", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0036", "type": "text", "x": 611.2, "y": 132, "width": 237.60000000000002, "height": 15.0, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 295334610, "version": 1, "versionNonce": 401991736, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "each holds d/4 columns of O_r (FP32)", "fontSize": 12, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "each holds d/4 columns of O_r (FP32)", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0037", "type": "rectangle", "x": 625, "y": 160, "width": 210, "height": 34, "angle": 0, "strokeColor": "#2f9e44", "backgroundColor": "#b2f2bb", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 3 }, "seed": 1159380354, "version": 1, "versionNonce": 2036236842, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false }, { "id": "el0038", "type": "text", "x": 626.325, "y": 168.875, "width": 207.35000000000002, "height": 16.25, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 279172787, "version": 1, "versionNonce": 260573196, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "m_r, l_r replicated in all 4", "fontSize": 13, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "m_r, l_r replicated in all 4", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0039", "type": "text", "x": 631.0, "y": 205, "width": 198.00000000000003, "height": 60.0, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1329753548, "version": 1, "versionNonce": 1914012529, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "row max and row sum:\neach thread scans 16 of the 64\ncolumns of sS[r, :], then a\n2-step shuffle across t0..t3", "fontSize": 12, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "row max and row sum:\neach thread scans 16 of the 64\ncolumns of sS[r, :], then a\n2-step shuffle across t0..t3", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0040", "type": "arrow", "x": 585, "y": 245, "width": 30, "height": 0, "angle": 0, "strokeColor": "#1e1e1e", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 1.5, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": { "type": 2 }, "seed": 1222328496, "version": 1, "versionNonce": 1656961616, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "points": [ [ 0, 0 ], [ 30, 0 ] ], "lastCommittedPoint": null, "startBinding": null, "endBinding": null, "startArrowhead": null, "endArrowhead": "arrow", "elbowed": false }, { "id": "el0041", "type": "text", "x": 624.4, "y": 285, "width": 211.20000000000002, "height": 15.0, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1490376254, "version": 1, "versionNonce": 96907016, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "end: O_r = O_r / l_r, store BF16", "fontSize": 12, "fontFamily": 1, "textAlign": "center", "verticalAlign": "top", "containerId": null, "originalText": "end: O_r = O_r / l_r, store BF16", "autoResize": true, "lineHeight": 1.25 }, { "id": "el0042", "type": "text", "x": 20, "y": 370, "width": 1001.0000000000001, "height": 32.5, "angle": 0, "strokeColor": "#495057", "backgroundColor": "transparent", "fillStyle": "solid", "strokeWidth": 2, "strokeStyle": "solid", "roughness": 1, "opacity": 100, "groupIds": [], "frameId": null, "index": null, "roundness": null, "seed": 1982966163, "version": 1, "versionNonce": 1526706730, "isDeleted": false, "boundElements": null, "updated": 1, "link": null, "locked": false, "text": "Grid: (N / 64) x B CTAs, 256 threads each. N / 64 KV iterations per CTA (fewer under the causal mask: tiles above the diagonal are skipped).\nThe N x N score matrix never exists; only one 64 x 64 tile of it at a time.", "fontSize": 13, "fontFamily": 1, "textAlign": "left", "verticalAlign": "top", "containerId": null, "originalText": "Grid: (N / 64) x B CTAs, 256 threads each. N / 64 KV iterations per CTA (fewer under the causal mask: tiles above the diagonal are skipped).\nThe N x N score matrix never exists; only one 64 x 64 tile of it at a time.", "autoResize": true, "lineHeight": 1.25 } ], "appState": { "gridSize": 20, "gridStep": 5, "gridModeEnabled": false, "viewBackgroundColor": "#fdf8f6" }, "files": {} }