onnxruntime/js/web/lib/wasm/jsep/webgpu/ops
Jiajia Qin fffefb1c22
[js/webgpu] Optimize matmul (#16969)
### Description
Changes in this PR:
1) use the optimized version `makeMatMulPacked[Vec4]Source` to support
matmul.
2) enable the conv2dByMatMul path.
3) support broadcast
4) use IndicesHelper.

MatMul with M = 512, K = 512, N = 512 becomes 2ms from 15ms when
enabling profilingMode on my ADL.
2023-08-29 12:40:57 -07:00
..
3rd-party [js/webgpu] Optimize matmul (#16969) 2023-08-29 12:40:57 -07:00
argminmax.ts [js/web] [webgpu] new incides helper (#16957) 2023-08-11 11:36:59 -07:00
binary-op.ts [js/webgpu] add kernel Not and Equal (#17306) 2023-08-27 19:50:17 -07:00
common.ts [js/webgpu] Optimize matmul (#16969) 2023-08-29 12:40:57 -07:00
concat.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
conv-grouped.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
conv-transpose.ts [Web/JS] Add ConvTranspose support (#16433) 2023-07-08 11:10:50 -07:00
conv.ts [js/webgpu] Optimize matmul (#16969) 2023-08-29 12:40:57 -07:00
conv2d-mm.ts
expand.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
fuse-utils.ts
gather-elements.ts [JS/WebGPU] Support GatherElements kernel (#17243) 2023-08-28 09:55:25 -07:00
gather.ts [js/web] JSEP node assignment optimization (#17128) 2023-08-15 18:58:05 -07:00
gemm.ts
instance-norm.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
layer-norm.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
matmul.ts [js/webgpu] Optimize matmul (#16969) 2023-08-29 12:40:57 -07:00
pool.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
reduce.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
resize.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
skip-layer-norm.ts [js/webgpu] fix 2 build breaks introduced in merge (#17273) 2023-08-23 18:09:50 -07:00
slice.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
softmax.ts js/webgpu: argmax,argmin,softmax support (#16882) 2023-08-02 18:16:19 -07:00
split.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
tile.ts [js/webgpu] fix 2 build breaks introduced in merge (#17273) 2023-08-23 18:09:50 -07:00
transpose.ts [js/webgpu] make IndicesHelper implementation implicit (#17193) 2023-08-23 14:41:35 -07:00
unary-op.ts [js/webgpu] add kernel Not and Equal (#17306) 2023-08-27 19:50:17 -07:00