README: describe the dataset as in the paper
Browse filesCo-authored-by: Shuaiwei Huang <112870732+ArtificialRay@users.noreply.github.com>
Co-authored-by: Allen Wu <110312634+wuallenwu@users.noreply.github.com>
Co-authored-by: Weihua Du <73473861+StigLidu@users.noreply.github.com>
README.md
CHANGED
|
@@ -28,9 +28,9 @@ configs:
|
|
| 28 |
|
| 29 |
# KernelShift
|
| 30 |
|
| 31 |
-
Benchmark data for
|
| 32 |
-
generations of vector and matrix extensions:
|
| 33 |
-
SVE2 (AWS Graviton4, 128-bit) and SME2 (Apple M4). Each kernel *definition*
|
| 34 |
comes with *workloads*, a scalar *reference* that serves as the correctness
|
| 35 |
check and the agent's starting point, and an *expert baseline*: a candidate
|
| 36 |
kernel is checked against the reference and timed against the baseline.
|
|
|
|
| 28 |
|
| 29 |
# KernelShift
|
| 30 |
|
| 31 |
+
Benchmark data for KernelShift: kernel optimization tasks for **Arm (AArch64)
|
| 32 |
+
CPUs** across four generations of vector and matrix extensions: Neon,
|
| 33 |
+
SVE (AWS Graviton3, 256-bit), SVE2 (AWS Graviton4, 128-bit) and SME2 (Apple M4). Each kernel *definition*
|
| 34 |
comes with *workloads*, a scalar *reference* that serves as the correctness
|
| 35 |
check and the agent's starting point, and an *expert baseline*: a candidate
|
| 36 |
kernel is checked against the reference and timed against the baseline.
|