Skip to content

Commit 4971b1a

Browse files
Update BULK.md
Updates to Gelu example in BULK.MD
1 parent ea0561b commit 4971b1a

1 file changed

Lines changed: 3 additions & 6 deletions

File tree

parser-ng/BULK.md

Lines changed: 3 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -261,7 +261,7 @@ evaluator.applyBulk(inputs, output);
261261

262262
```
263263

264-
### Matrix operation (Gelu-200x200) (3.8ns per element on a single core-2.5GHz)
264+
### Matrix operation (Gelu-200x200) (2.67ns per element on a single core-2.5GHz)
265265
```
266266
267267
@Test
@@ -275,15 +275,12 @@ evaluator.applyBulk(inputs, output);
275275
FlatMatrixF in1 = new FlatMatrixF(sz, sz);
276276
FlatMatrixF.randomFill(in1);
277277
278-
FlatMatrixF in2 = new FlatMatrixF(sz, sz);
279-
FlatMatrixF.randomFill(in2);
280-
281278
FlatMatrixF out = new FlatMatrixF(sz, sz);
282279
283280
double n = 10000;
284281
double t = System.nanoTime();
285282
for (int i = 0; i < n; i++) {
286-
evaluator.applyMatrixKernel(new FlatMatrixF[]{in1, in2}, out, "gelu");
283+
evaluator.applyMatrixKernel(new FlatMatrixF[]{in1}, out, "gelu");
287284
}
288285
289286
double t1 = System.nanoTime() - t;
@@ -295,7 +292,7 @@ evaluator.applyBulk(inputs, output);
295292
296293
}
297294
```
298-
### Matrix operation (Swiglu-200x200) (13.3ns per element on a single core-2.5GHz)
295+
### Matrix operation (Swiglu-200x200) (1.82ns per element on a single core-2.5GHz)
299296

300297

301298
```

0 commit comments

Comments
 (0)