[GH-PAGES] Updated website

triton-lang · Feb 26, 2025 · 0dc92b6 · 0dc92b6
1 parent f942eaa
commit 0dc92b6
Show file tree

Hide file tree

Showing 65 changed files with 493 additions and 487 deletions.
diff --git a/main/.doctrees/dialects/TritonGPUOps.doctree b/main/.doctrees/dialects/TritonGPUOps.doctree
diff --git a/main/.doctrees/environment.pickle b/main/.doctrees/environment.pickle
diff --git a/main/.doctrees/getting-started/tutorials/01-vector-add.doctree b/main/.doctrees/getting-started/tutorials/01-vector-add.doctree
diff --git a/main/.doctrees/getting-started/tutorials/02-fused-softmax.doctree b/main/.doctrees/getting-started/tutorials/02-fused-softmax.doctree
diff --git a/main/.doctrees/getting-started/tutorials/03-matrix-multiplication.doctree b/main/.doctrees/getting-started/tutorials/03-matrix-multiplication.doctree
diff --git a/main/.doctrees/getting-started/tutorials/04-low-memory-dropout.doctree b/main/.doctrees/getting-started/tutorials/04-low-memory-dropout.doctree
diff --git a/main/.doctrees/getting-started/tutorials/05-layer-norm.doctree b/main/.doctrees/getting-started/tutorials/05-layer-norm.doctree
diff --git a/main/.doctrees/getting-started/tutorials/06-fused-attention.doctree b/main/.doctrees/getting-started/tutorials/06-fused-attention.doctree
diff --git a/main/.doctrees/getting-started/tutorials/07-extern-functions.doctree b/main/.doctrees/getting-started/tutorials/07-extern-functions.doctree
diff --git a/main/.doctrees/getting-started/tutorials/08-grouped-gemm.doctree b/main/.doctrees/getting-started/tutorials/08-grouped-gemm.doctree
diff --git a/main/.doctrees/getting-started/tutorials/09-persistent-matmul.doctree b/main/.doctrees/getting-started/tutorials/09-persistent-matmul.doctree
diff --git a/main/.doctrees/getting-started/tutorials/sg_execution_times.doctree b/main/.doctrees/getting-started/tutorials/sg_execution_times.doctree
diff --git a/main/.doctrees/sg_execution_times.doctree b/main/.doctrees/sg_execution_times.doctree
diff --git a/main/_downloads/032b2a144fc26b286cf422d1aecab3b6/05-layer-norm.zip b/main/_downloads/032b2a144fc26b286cf422d1aecab3b6/05-layer-norm.zip
diff --git a/main/_downloads/3790c08ba249d4a4a4e22801792cc0e1/10-block-scaled-matmul.zip b/main/_downloads/3790c08ba249d4a4a4e22801792cc0e1/10-block-scaled-matmul.zip
diff --git a/main/_downloads/3cf54e8aaddcfce69d180b77518fd544/07-extern-functions.zip b/main/_downloads/3cf54e8aaddcfce69d180b77518fd544/07-extern-functions.zip
diff --git a/main/_downloads/4e511f795844d864249b83f016d8ce09/01-vector-add.zip b/main/_downloads/4e511f795844d864249b83f016d8ce09/01-vector-add.zip
diff --git a/main/_downloads/509ee5615f08cb2e7336a73c0511fba2/06-fused-attention.zip b/main/_downloads/509ee5615f08cb2e7336a73c0511fba2/06-fused-attention.zip
diff --git a/main/_downloads/662999063954282841dc90b8945f85ce/tutorials_jupyter.zip b/main/_downloads/662999063954282841dc90b8945f85ce/tutorials_jupyter.zip
diff --git a/main/_downloads/763344228ae6bc253ed1a6cf586aa30d/tutorials_python.zip b/main/_downloads/763344228ae6bc253ed1a6cf586aa30d/tutorials_python.zip
diff --git a/main/_downloads/9241eab99db7582ceb6cd81f77524214/04-low-memory-dropout.zip b/main/_downloads/9241eab99db7582ceb6cd81f77524214/04-low-memory-dropout.zip
diff --git a/main/_downloads/a13fad8f9b954e9e27a17a2c27aee1b8/09-persistent-matmul.zip b/main/_downloads/a13fad8f9b954e9e27a17a2c27aee1b8/09-persistent-matmul.zip
diff --git a/main/_downloads/d74d81ab958b2efbd334253f48cdb202/03-matrix-multiplication.zip b/main/_downloads/d74d81ab958b2efbd334253f48cdb202/03-matrix-multiplication.zip
diff --git a/main/_downloads/f66de4fbee2c4ba20b6f7f3ae99f7de3/02-fused-softmax.zip b/main/_downloads/f66de4fbee2c4ba20b6f7f3ae99f7de3/02-fused-softmax.zip
diff --git a/main/_downloads/fecddac383ee03c4c47e2cf2ec91448a/08-grouped-gemm.zip b/main/_downloads/fecddac383ee03c4c47e2cf2ec91448a/08-grouped-gemm.zip
diff --git a/main/_images/sphx_glr_01-vector-add_001.png b/main/_images/sphx_glr_01-vector-add_001.png
diff --git a/main/_images/sphx_glr_01-vector-add_thumb.png b/main/_images/sphx_glr_01-vector-add_thumb.png
diff --git a/main/_images/sphx_glr_02-fused-softmax_001.png b/main/_images/sphx_glr_02-fused-softmax_001.png
diff --git a/main/_images/sphx_glr_02-fused-softmax_thumb.png b/main/_images/sphx_glr_02-fused-softmax_thumb.png
diff --git a/main/_images/sphx_glr_03-matrix-multiplication_001.png b/main/_images/sphx_glr_03-matrix-multiplication_001.png
diff --git a/main/_images/sphx_glr_03-matrix-multiplication_002.png b/main/_images/sphx_glr_03-matrix-multiplication_002.png
diff --git a/main/_images/sphx_glr_03-matrix-multiplication_thumb.png b/main/_images/sphx_glr_03-matrix-multiplication_thumb.png
diff --git a/main/_images/sphx_glr_05-layer-norm_001.png b/main/_images/sphx_glr_05-layer-norm_001.png
diff --git a/main/_images/sphx_glr_05-layer-norm_thumb.png b/main/_images/sphx_glr_05-layer-norm_thumb.png
diff --git a/main/_images/sphx_glr_06-fused-attention_001.png b/main/_images/sphx_glr_06-fused-attention_001.png
diff --git a/main/_images/sphx_glr_06-fused-attention_002.png b/main/_images/sphx_glr_06-fused-attention_002.png
diff --git a/main/_images/sphx_glr_06-fused-attention_003.png b/main/_images/sphx_glr_06-fused-attention_003.png
diff --git a/main/_images/sphx_glr_06-fused-attention_thumb.png b/main/_images/sphx_glr_06-fused-attention_thumb.png
diff --git a/main/_images/sphx_glr_08-grouped-gemm_001.png b/main/_images/sphx_glr_08-grouped-gemm_001.png
diff --git a/main/_images/sphx_glr_08-grouped-gemm_thumb.png b/main/_images/sphx_glr_08-grouped-gemm_thumb.png
diff --git a/main/_sources/dialects/TritonGPUOps.md.txt b/main/_sources/dialects/TritonGPUOps.md.txt
@@ -482,13 +482,15 @@ default {
 }
 partition0(%arg0: i32, %arg1: i32) num_warps(8) {
   some_async_dispatch(%arg0, %arg1)
+  ttg.warp_return
 }
 partition1(%arg0: i32, %arg1: i32) num_warps(1) {
   some_async_dispatch(%arg0, %arg1)
+  ttg.warp_return
 } : (i32, i32) -> i32
 ```
 
-Traits: `RecursiveMemoryEffects`, `RecursivelySpeculatableImplTrait`, `VerifyTensorLayoutsTrait`
+Traits: `AsyncRegions`, `RecursiveMemoryEffects`, `RecursivelySpeculatableImplTrait`, `VerifyTensorLayoutsTrait`
 
 Interfaces: `ConditionallySpeculatable`, `RegionBranchOpInterface`
 
@@ -497,6 +499,7 @@ Interfaces: `ConditionallySpeculatable`, `RegionBranchOpInterface`
 <table>
 <tr><th>Attribute</th><th>MLIR Type</th><th>Description</th></tr>
 <tr><td><code>partitionNumWarps</code></td><td>::mlir::DenseI32ArrayAttr</td><td>i32 dense array attribute</td></tr>
+<tr><td><code>warpGroupStartIds</code></td><td>::mlir::DenseI32ArrayAttr</td><td>i32 dense array attribute</td></tr>
 </table>
 
 #### Operands:
@@ -518,7 +521,7 @@ _Container op for `ttg.warp_specialize`_
 Because MLIR requires entire operations be isolated from above, this op
 contains the actual isolated from above regions of `ttg.warp_specialize`.
 
-Traits: `HasParent<WarpSpecializeOp>`, `IsolatedFromAbove`, `RecursiveMemoryEffects`, `RecursivelySpeculatableImplTrait`, `SingleBlockImplicitTerminator<WarpReturnOp>`, `SingleBlock`, `Terminator`, `VerifyTensorLayoutsTrait`
+Traits: `HasParent<WarpSpecializeOp>`, `IsolatedFromAbove`, `RecursiveMemoryEffects`, `RecursivelySpeculatableImplTrait`, `Terminator`, `VerifyTensorLayoutsTrait`
 
 Interfaces: `ConditionallySpeculatable`
 
@@ -543,7 +546,7 @@ Example:
 ttg.warp_yield %a, %b : i32, tensor<32xbf16, #blocked>
 ```
 
-Traits: `AlwaysSpeculatableImplTrait`, `HasParent<WarpSpecializeOp>`, `Terminator`, `VerifyTensorLayoutsTrait`
+Traits: `AlwaysSpeculatableImplTrait`, `HasParent<WarpSpecializeOp>`, `ReturnLike`, `Terminator`, `VerifyTensorLayoutsTrait`
 
 Interfaces: `ConditionallySpeculatable`, `NoMemoryEffect (MemoryEffectOpInterface)`, `RegionBranchTerminatorOpInterface`
 

diff --git a/main/_sources/getting-started/tutorials/01-vector-add.rst.txt b/main/_sources/getting-started/tutorials/01-vector-add.rst.txt
@@ -242,20 +242,20 @@ We can now run the decorated function above. Pass `print_data=True` to see the p
     7      524288.0   614.400016   614.400016
     8     1048576.0   819.200021   819.200021
     9     2097152.0  1023.999964  1023.999964
-    10    4194304.0  1260.307736  1260.307736
+    10    4194304.0  1260.307736  1228.800031
     11    8388608.0  1424.695621  1424.695621
     12   16777216.0  1560.380965  1560.380965
-    13   33554432.0  1624.859540  1624.859540
+    13   33554432.0  1631.601649  1624.859540
     14   67108864.0  1669.706983  1662.646960
-    15  134217728.0  1684.008546  1680.410210
+    15  134217728.0  1684.008546  1678.616907
 
 
 
 
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** (0 minutes 6.268 seconds)
+   **Total running time of the script:** (0 minutes 12.340 seconds)
 
 
 .. _sphx_glr_download_getting-started_tutorials_01-vector-add.py:

diff --git a/main/_sources/getting-started/tutorials/02-fused-softmax.rst.txt b/main/_sources/getting-started/tutorials/02-fused-softmax.rst.txt
@@ -321,104 +321,104 @@ We will then compare its performance against (1) :code:`torch.softmax` and (2) t
 
     softmax-performance:
               N       Triton        Torch
-    0     256.0   470.104985   690.281403
-    1     384.0   612.362382   812.550651
-    2     512.0   763.491036   918.700448
-    3     640.0   814.447857   954.721493
-    4     768.0   885.888117  1022.381372
-    5     896.0   942.848306  1064.142450
-    6    1024.0  1011.711306  1115.557064
-    7    1152.0  1107.077578   614.295965
-    8    1280.0  1152.577055   669.344330
-    9    1408.0  1167.526302   721.998479
-    10   1536.0  1201.721384   783.800793
-    11   1664.0  1218.754022   812.894195
-    12   1792.0  1244.226716   856.205481
-    13   1920.0  1248.015489   908.772535
-    14   2048.0  1272.270320   954.953708
-    15   2176.0  1238.031810   975.084352
-    16   2304.0  1253.581173  1008.397320
-    17   2432.0  1278.147710  1052.904453
-    18   2560.0  1284.657260  1085.608260
-    19   2688.0  1293.098879  1105.287008
-    20   2816.0  1303.949358  1128.452353
-    21   2944.0  1318.000673  1168.943749
-    22   3072.0  1331.887083  1182.483759
-    23   3200.0  1336.420411  1191.352660
-    24   3328.0  1339.755982  1224.091949
-    25   3456.0  1352.389611  1252.248056
-    26   3584.0  1359.466596  1257.676754
-    27   3712.0  1360.841709  1271.636771
-    28   3840.0  1378.556602  1301.933257
-    29   3968.0  1370.985594  1313.381834
-    30   4096.0  1383.261189  1326.226685
-    31   4224.0  1333.390160  1159.842644
-    32   4352.0  1336.014943  1176.527085
-    33   4480.0  1360.151812  1184.711118
-    34   4608.0  1361.846293  1197.956257
-    35   4736.0  1357.644105  1197.741940
-    36   4864.0  1378.771634  1221.352060
-    37   4992.0  1372.291192  1237.347392
-    38   5120.0  1379.591869  1252.228819
-    39   5248.0  1378.745632  1260.218099
-    40   5376.0  1373.685357  1291.018429
-    41   5504.0  1382.927346  1303.080541
-    42   5632.0  1393.408137  1316.063796
-    43   5760.0  1391.734831  1325.760982
-    44   5888.0  1387.845938  1342.315862
-    45   6016.0  1403.380174  1353.743928
-    46   6144.0  1406.296044  1373.403538
-    47   6272.0  1413.915813  1376.605064
-    48   6400.0  1420.386355  1388.776885
-    49   6528.0  1416.030037  1391.930540
-    50   6656.0  1419.273005  1406.550947
-    51   6784.0  1413.499047  1412.385093
-    52   6912.0  1432.862458  1426.738666
-    53   7040.0  1425.279737  1433.056175
-    54   7168.0  1424.313568  1433.093011
-    55   7296.0  1432.636525  1440.870595
-    56   7424.0  1427.727433  1448.754156
-    57   7552.0  1429.101604  1452.824232
-    58   7680.0  1434.593437  1462.035053
-    59   7808.0  1435.434822  1466.194826
-    60   7936.0  1433.926902  1468.153974
-    61   8064.0  1437.045556  1476.091091
-    62   8192.0  1440.641601  1485.736867
-    63   8320.0  1380.752488  1402.321298
-    64   8448.0  1375.650455  1407.151385
-    65   8576.0  1388.579711  1394.617460
-    66   8704.0  1386.627876  1400.752286
-    67   8832.0  1378.158574  1404.995037
-    68   8960.0  1395.450610  1412.721953
-    69   9088.0  1403.743469  1418.040275
-    70   9216.0  1396.971444  1424.372707
-    71   9344.0  1394.211472  1425.115372
-    72   9472.0  1391.432092  1431.222078
-    73   9600.0  1386.520902  1431.687475
-    74   9728.0  1394.817878  1441.334588
-    75   9856.0  1407.459228  1446.004223
-    76   9984.0  1395.751174  1450.250798
-    77  10112.0  1408.307508  1456.828408
-    78  10240.0  1416.706218  1467.569295
-    79  10368.0  1406.572420  1460.161186
-    80  10496.0  1410.425398  1468.508112
-    81  10624.0  1410.180487  1466.648627
-    82  10752.0  1398.942180  1474.278067
-    83  10880.0  1396.972988  1477.723918
-    84  11008.0  1416.958974  1477.673515
-    85  11136.0  1420.424056  1484.033260
-    86  11264.0  1424.310719  1487.364373
-    87  11392.0  1412.053047  1488.211948
-    88  11520.0  1418.055040  1497.987539
-    89  11648.0  1416.977769  1498.288067
-    90  11776.0  1427.057601  1501.545180
-    91  11904.0  1434.981281  1505.169316
-    92  12032.0  1421.369815  1507.671595
-    93  12160.0  1419.429800  1509.576661
-    94  12288.0  1431.906591  1392.188839
-    95  12416.0  1440.775729  1389.181018
-    96  12544.0  1441.661003  1390.394040
-    97  12672.0  1449.605979  1394.322508
+    0     256.0   466.653129   687.163315
+    1     384.0   607.174224   805.467444
+    2     512.0   759.233427   915.545158
+    3     640.0   816.907949   948.492846
+    4     768.0   892.093796  1012.737606
+    5     896.0   947.357658  1061.925921
+    6    1024.0  1006.725502  1109.807297
+    7    1152.0  1104.283245   611.824783
+    8    1280.0  1156.094471   667.443080
+    9    1408.0  1158.083439   722.430799
+    10   1536.0  1186.161698   780.061504
+    11   1664.0  1214.162317   815.011669
+    12   1792.0  1241.639183   863.680728
+    13   1920.0  1253.928740   910.738940
+    14   2048.0  1284.366161   959.903838
+    15   2176.0  1236.314838   976.703918
+    16   2304.0  1254.431170  1012.349902
+    17   2432.0  1270.022558  1057.588365
+    18   2560.0  1280.709616  1083.786331
+    19   2688.0  1287.735237  1099.916919
+    20   2816.0  1294.990187  1131.826596
+    21   2944.0  1315.164979  1168.428417
+    22   3072.0  1334.566019  1182.335406
+    23   3200.0  1332.914727  1197.247024
+    24   3328.0  1348.971413  1220.738084
+    25   3456.0  1353.903346  1252.850697
+    26   3584.0  1356.408509  1263.392878
+    27   3712.0  1367.322740  1268.345548
+    28   3840.0  1377.397255  1298.786564
+    29   3968.0  1372.063580  1313.919552
+    30   4096.0  1383.325900  1329.002425
+    31   4224.0  1338.175130  1157.823610
+    32   4352.0  1333.527466  1176.968477
+    33   4480.0  1356.379291  1181.746465
+    34   4608.0  1359.403881  1196.124952
+    35   4736.0  1359.105667  1197.596995
+    36   4864.0  1378.394559  1218.556791
+    37   4992.0  1372.498410  1237.658170
+    38   5120.0  1374.298993  1249.933010
+    39   5248.0  1374.565125  1256.543992
+    40   5376.0  1375.000741  1282.988345
+    41   5504.0  1383.061121  1304.223127
+    42   5632.0  1386.625030  1314.125315
+    43   5760.0  1394.630025  1328.179048
+    44   5888.0  1390.262590  1346.377287
+    45   6016.0  1399.947168  1356.576389
+    46   6144.0  1409.798735  1373.462579
+    47   6272.0  1412.657628  1376.270410
+    48   6400.0  1415.086686  1385.343832
+    49   6528.0  1411.922554  1391.314924
+    50   6656.0  1418.828518  1403.513862
+    51   6784.0  1414.505939  1412.491109
+    52   6912.0  1423.626828  1426.011035
+    53   7040.0  1418.103391  1432.886073
+    54   7168.0  1426.754462  1435.561785
+    55   7296.0  1434.272954  1440.618765
+    56   7424.0  1426.474240  1448.000637
+    57   7552.0  1425.746304  1456.983219
+    58   7680.0  1434.059305  1461.729784
+    59   7808.0  1436.480986  1466.430142
+    60   7936.0  1433.465977  1468.297207
+    61   8064.0  1436.916836  1473.042602
+    62   8192.0  1437.379090  1483.669877
+    63   8320.0  1382.868502  1400.927992
+    64   8448.0  1376.743800  1406.677530
+    65   8576.0  1388.943070  1395.674086
+    66   8704.0  1384.972246  1401.521452
+    67   8832.0  1377.646883  1404.894324
+    68   8960.0  1385.526431  1410.106704
+    69   9088.0  1402.766787  1419.185055
+    70   9216.0  1400.745140  1422.292112
+    71   9344.0  1391.479678  1423.078108
+    72   9472.0  1393.258844  1431.417554
+    73   9600.0  1391.929790  1434.810907
+    74   9728.0  1397.388438  1442.276910
+    75   9856.0  1408.521936  1441.911980
+    76   9984.0  1397.047501  1454.908745
+    77  10112.0  1408.846096  1457.304623
+    78  10240.0  1410.085540  1467.631420
+    79  10368.0  1415.178015  1461.542929
+    80  10496.0  1407.868140  1468.447799
+    81  10624.0  1405.938936  1468.760847
+    82  10752.0  1395.392816  1472.134690
+    83  10880.0  1392.289004  1479.920326
+    84  11008.0  1415.650982  1473.824756
+    85  11136.0  1419.900506  1486.753559
+    86  11264.0  1425.138421  1486.569650
+    87  11392.0  1413.175574  1490.166507
+    88  11520.0  1416.727179  1493.169083
+    89  11648.0  1418.707622  1497.968086
+    90  11776.0  1431.502521  1503.033613
+    91  11904.0  1436.744463  1506.083408
+    92  12032.0  1417.715170  1509.020402
+    93  12160.0  1413.315645  1512.943001
+    94  12288.0  1426.379877  1392.936739
+    95  12416.0  1445.511245  1390.635466
+    96  12544.0  1438.549138  1393.105350
+    97  12672.0  1439.769107  1393.185312
 
 
 
@@ -433,7 +433,7 @@ In the above plot, we can see that:
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** (0 minutes 23.287 seconds)
+   **Total running time of the script:** (0 minutes 23.176 seconds)
 
 
 .. _sphx_glr_download_getting-started_tutorials_02-fused-softmax.py: