| { |
| "op": "ai.onnx.Div", |
| "tunableSpace": { "WORKGROUP_SIZE": [64, 128, 256] }, |
| "cases": [ |
| { |
| "name": "div-f32-4m", |
| "preset": "smoke", |
| "vars": { "dtype": "float32", "count": 4194304 }, |
| "inputs": { |
| "a": { "shape": [4194304], "dtype": "float32", "dist": "normal", "seed": 402, "scale": 2 }, |
| "b": { "shape": [4194304], "dtype": "float32", "dist": "uniform", "seed": 403, "scale": 0.25, "offset": 1 } |
| }, |
| "outputs": { "c": { "shape": [4194304], "dtype": "float32" } }, |
| "bench": { |
| "primary": true, |
| "metrics": [{ "type": "bandwidth", "value": "args.count * dtypeBytes(args.dtype) * 3" }] |
| } |
| }, |
| { |
| "name": "div-f32-scalar-rhs-4m", |
| "preset": "smoke", |
| "vars": { "dtype": "float32", "count": 4194304 }, |
| "inputs": { |
| "a": { "shape": [4194304], "dtype": "float32", "dist": "normal", "seed": 511, "scale": 2 }, |
| "b": { "shape": [1], "dtype": "float32", "data": { "kind": "values", "values": [1.25] } } |
| }, |
| "outputs": { "c": { "shape": [4194304], "dtype": "float32" } }, |
| "bench": { "metrics": [{ "type": "bandwidth", "value": "(4194304 + 1 + 4194304) * dtypeBytes(args.dtype)" }] } |
| }, |
| { |
| "name": "div-f32-rank4-broadcast", |
| "preset": "smoke", |
| "vars": { "dtype": "float32", "count": 16777216 }, |
| "inputs": { |
| "a": { "shape": [8, 1, 512, 512], "dtype": "float32", "dist": "normal", "seed": 521, "scale": 2 }, |
| "b": { |
| "shape": [8, 8, 512, 512], |
| "dtype": "float32", |
| "dist": "uniform", |
| "seed": 522, |
| "scale": 0.25, |
| "offset": 1 |
| } |
| }, |
| "outputs": { "c": { "shape": [8, 8, 512, 512], "dtype": "float32" } }, |
| "bench": { |
| "metrics": [{ "type": "bandwidth", "value": "(2097152 + 16777216 + 16777216) * dtypeBytes(args.dtype)" }] |
| } |
| }, |
| { |
| "name": "div-f16-4m", |
| "preset": "smoke", |
| "vars": { "dtype": "float16", "count": 4194304 }, |
| "inputs": { |
| "a": { "shape": [4194304], "dtype": "float16", "dist": "normal", "seed": 531, "scale": 2 }, |
| "b": { "shape": [4194304], "dtype": "float16", "dist": "uniform", "seed": 532, "scale": 0.25, "offset": 1 } |
| }, |
| "outputs": { "c": { "shape": [4194304], "dtype": "float16" } }, |
| "bench": { "metrics": [{ "type": "bandwidth", "value": "args.count * dtypeBytes(args.dtype) * 3" }] } |
| }, |
| { |
| "name": "div-f32-broadcast-scalar-fallback-4m", |
| "preset": "stress", |
| "vars": { "dtype": "float32", "count": 4190208 }, |
| "inputs": { |
| "a": { "shape": [4096, 1023], "dtype": "float32", "dist": "uniform", "seed": 561, "scale": 10, "offset": 5 }, |
| "b": { "shape": [1, 1023], "dtype": "float32", "dist": "uniform", "seed": 562, "scale": 0.5, "offset": 1 } |
| }, |
| "outputs": { "c": { "shape": [4096, 1023], "dtype": "float32", "dist": "empty" } }, |
| "bench": { "metrics": [{ "type": "bandwidth", "value": "args.count * dtypeBytes(args.dtype) * 2" }] } |
| } |
| ] |
| } |
|
|