NumSharp vs NumPy Performance Baseline: NumPy (N=10M elements)
Ratio = NumSharp ÷ NumPy → Lower is better for NumSharp
Status
Ratio
Meaning
✅
Faster
<1.0
NumSharp beats NumPy
🟡
Close
1-2x
Acceptable parity
🟠
Slower
2-5x
Optimization target
🔴
Slow
>5x
Priority fix
⚪
Pending
-
C# benchmark not run
Summary: 64 ops | ✅ 61 | 🟡 3 | 🟠 0 | 🔴 0 | ⚪ 0
🏆 Top 15 Best (NumSharp closest to NumPy)
Operation
Type
NumPy
NumSharp
Ratio
✅
a % 7 (literal) (float64)
float64
267.0
39.8
0.1x
✅
a % b (element-wise) (float64)
float64
188.5
45.7
0.2x
✅
a % 7 (literal) (int32)
int32
49.7
13.8
0.3x
✅
a % b (element-wise) (int32)
int32
46.9
13.9
0.3x
✅
a % 7 (literal) (int64)
int64
49.2
23.9
0.5x
✅
a % b (element-wise) (int64)
int64
48.1
24.2
0.5x
✅
a % b (element-wise) (float32)
float32
155.0
84.3
0.5x
✅
a % 7 (literal) (float32)
float32
174.6
96.7
0.6x
✅
scalar - a (int32)
int32
11.4
7.2
0.6x
✅
a / b (element-wise) (int32)
int32
21.7
14.1
0.7x
✅
a - b (element-wise) (int32)
int32
11.1
7.4
0.7x
✅
a * a (square) (int32)
int32
9.7
6.7
0.7x
✅
a * b (element-wise) (int32)
int32
10.3
7.4
0.7x
✅
a + b (element-wise) (int64)
int64
20.3
14.8
0.7x
✅
np.add(a, b) (float64)
float64
20.3
14.9
0.7x
🔻 Top 15 Worst (Optimization priorities)
Operation
Type
NumPy
NumSharp
Ratio
🟡
a / scalar (int64)
int64
18.7
21.7
1.2x
🟡
a * scalar (float32)
float32
7.7
8.7
1.1x
🟡
a * 2 (literal) (float64)
float64
16.9
18.7
1.1x
✅
a * 2 (literal) (int32)
int32
9.0
8.9
1.0x
✅
scalar - a (int64)
int64
15.0
14.0
0.9x
✅
a * 2 (literal) (float32)
float32
7.7
6.9
0.9x
✅
a * b (element-wise) (int64)
int64
17.0
15.1
0.9x
✅
a + scalar (float32)
float32
7.8
6.9
0.9x
✅
a + b (element-wise) (float32)
float32
8.5
7.5
0.9x
✅
a * 2 (literal) (int64)
int64
15.0
13.2
0.9x
✅
a / b (element-wise) (float32)
float32
8.9
7.7
0.9x
✅
a - b (element-wise) (float32)
float32
8.6
7.5
0.9x
✅
a - scalar (int64)
int64
15.4
13.5
0.9x
✅
a * b (element-wise) (float32)
float32
8.6
7.4
0.9x
✅
a * a (square) (int64)
int64
15.8
13.6
0.9x
Operation
Type
NumPy
NumSharp
Ratio
✅
a + b (element-wise) (int32)
int32
9.5
7.5
0.8x
✅
np.add(a, b) (int32)
int32
9.6
7.7
0.8x
✅
a + scalar (int32)
int32
8.7
6.9
0.8x
✅
a + 5 (literal) (int32)
int32
9.1
6.9
0.8x
✅
a - b (element-wise) (int32)
int32
11.1
7.4
0.7x
✅
a - scalar (int32)
int32
9.0
6.8
0.8x
✅
scalar - a (int32)
int32
11.4
7.2
0.6x
✅
a * b (element-wise) (int32)
int32
10.3
7.4
0.7x
✅
a * a (square) (int32)
int32
9.7
6.7
0.7x
✅
a * scalar (int32)
int32
8.7
7.3
0.8x
✅
a * 2 (literal) (int32)
int32
9.0
8.9
1.0x
✅
a / b (element-wise) (int32)
int32
21.7
14.1
0.7x
✅
a / scalar (int32)
int32
17.5
14.4
0.8x
✅
scalar / a (int32)
int32
18.1
13.7
0.8x
✅
a % b (element-wise) (int32)
int32
46.9
13.9
0.3x
✅
a % 7 (literal) (int32)
int32
49.7
13.8
0.3x
✅
a + b (element-wise) (int64)
int64
20.3
14.8
0.7x
✅
np.add(a, b) (int64)
int64
18.8
14.9
0.8x
✅
a + scalar (int64)
int64
16.1
13.7
0.8x
✅
a + 5 (literal) (int64)
int64
16.5
13.9
0.8x
✅
a - b (element-wise) (int64)
int64
17.5
14.8
0.8x
✅
a - scalar (int64)
int64
15.4
13.5
0.9x
✅
scalar - a (int64)
int64
15.0
14.0
0.9x
✅
a * b (element-wise) (int64)
int64
17.0
15.1
0.9x
✅
a * a (square) (int64)
int64
15.8
13.6
0.9x
✅
a * scalar (int64)
int64
15.9
13.0
0.8x
✅
a * 2 (literal) (int64)
int64
15.0
13.2
0.9x
✅
a / b (element-wise) (int64)
int64
24.0
19.2
0.8x
🟡
a / scalar (int64)
int64
18.7
21.7
1.2x
✅
scalar / a (int64)
int64
18.8
15.4
0.8x
✅
a % b (element-wise) (int64)
int64
48.1
24.2
0.5x
✅
a % 7 (literal) (int64)
int64
49.2
23.9
0.5x
✅
a + b (element-wise) (float32)
float32
8.5
7.5
0.9x
✅
np.add(a, b) (float32)
float32
8.9
7.6
0.8x
✅
a + scalar (float32)
float32
7.8
6.9
0.9x
✅
a + 5 (literal) (float32)
float32
8.2
6.9
0.8x
✅
a - b (element-wise) (float32)
float32
8.6
7.5
0.9x
✅
a - scalar (float32)
float32
8.1
6.8
0.8x
✅
scalar - a (float32)
float32
8.4
7.1
0.8x
✅
a * b (element-wise) (float32)
float32
8.6
7.4
0.9x
✅
a * a (square) (float32)
float32
8.4
6.5
0.8x
🟡
a * scalar (float32)
float32
7.7
8.7
1.1x
✅
a * 2 (literal) (float32)
float32
7.7
6.9
0.9x
✅
a / b (element-wise) (float32)
float32
8.9
7.7
0.9x
✅
a / scalar (float32)
float32
8.9
6.6
0.7x
✅
scalar / a (float32)
float32
8.3
6.8
0.8x
✅
a % b (element-wise) (float32)
float32
155.0
84.3
0.5x
✅
a % 7 (literal) (float32)
float32
174.6
96.7
0.6x
✅
a + b (element-wise) (float64)
float64
20.1
15.0
0.8x
✅
np.add(a, b) (float64)
float64
20.3
14.9
0.7x
✅
a + scalar (float64)
float64
17.9
13.7
0.8x
✅
a + 5 (literal) (float64)
float64
17.9
13.6
0.8x
✅
a - b (element-wise) (float64)
float64
18.3
15.2
0.8x
✅
a - scalar (float64)
float64
17.2
13.5
0.8x
✅
scalar - a (float64)
float64
16.7
14.1
0.8x
✅
a * b (element-wise) (float64)
float64
18.6
14.7
0.8x
✅
a * a (square) (float64)
float64
17.0
13.1
0.8x
✅
a * scalar (float64)
float64
17.3
13.8
0.8x
🟡
a * 2 (literal) (float64)
float64
16.9
18.7
1.1x
✅
a / b (element-wise) (float64)
float64
18.6
15.2
0.8x
✅
a / scalar (float64)
float64
17.0
13.7
0.8x
✅
scalar / a (float64)
float64
17.9
13.5
0.8x
✅
a % b (element-wise) (float64)
float64
188.5
45.7
0.2x
✅
a % 7 (literal) (float64)
float64
267.0
39.8
0.1x