Masashi Yoshimura aa42b48312 ggml-webgpu: Improve prefill speeds for k-quants + refactor matmul for Q4/Q5/Q8 and k-quants (llama/24225) ...

* ggml-webgpu: Improve prefill speeds + refactor matmul for quants

* Fixes for editroconfig checker

2026-06-15 10:33:53 +03:00

..

add_id.wgsl

ggml-webgpu: Enables running gpt-oss-20b (llama/22906)

2026-05-14 21:26:48 +03:00

argmax.wgsl

ggml webgpu: support for backend sampling (llama/18880)

2026-01-30 15:56:40 +02:00

argsort_merge.wgsl

ggml webgpu: support for backend sampling (llama/18880)

2026-01-30 15:56:40 +02:00

argsort.wgsl

ggml webgpu: support for backend sampling (llama/18880)

2026-01-30 15:56:40 +02:00

binary.wgsl

Implement 2D workgroups for scale, binary, and unary ops (llama/24044)

2026-06-15 10:33:53 +03:00

common_decls.tmpl

ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594)

2026-05-29 09:47:30 +03:00

concat.wgsl

Handle buffer overlap / buffer aliasing for concat operator (llama/24000)

2026-06-15 10:33:53 +03:00

conv2d.wgsl

ggml-webgpu(shader): support conv2d kernels. (llama/21964)

2026-04-30 11:29:16 +03:00

cpy.wgsl

revert to using global_invocation_id for cpy shader (llama/23955)

2026-06-08 14:36:36 +03:00

cumsum.wgsl

ggml webgpu: support for backend sampling (llama/18880)

2026-01-30 15:56:40 +02:00

embed_wgsl.py

ggml-webgpu: port all AOT operators to JIT (llama/20728)

2026-04-30 11:28:58 +03:00

flash_attn_quant_staging.tmpl

ggml-webgpu: FlashAttention refactor + standardize quantization support (llama/23834)

2026-06-08 14:36:36 +03:00

flash_attn_tile.wgsl

ggml-webgpu: FlashAttention refactor + standardize quantization support (llama/23834)

2026-06-08 14:36:36 +03:00

flash_attn_vec_blk.wgsl

ggml-webgpu: enable FLASH_ATTN_EXT on browser without subgroup matrix (llama/22199)

2026-04-30 11:29:18 +03:00

flash_attn_vec_reduce.wgsl

ggml-webgpu: address precision issues for multimodal (llama/22808)

2026-05-14 21:26:48 +03:00

flash_attn_vec_split.wgsl

ggml-webgpu: FlashAttention refactor + standardize quantization support (llama/23834)

2026-06-08 14:36:36 +03:00

flash_attn.wgsl

ggml-webgpu: FlashAttention refactor + standardize quantization support (llama/23834)

2026-06-08 14:36:36 +03:00

gated_delta_net.wgsl

ggml-webgpu : extend GDN for K>1 (llama/23299)

2026-05-25 12:26:07 +03:00

get_rows.wgsl

ggml-webgpu: Enables running gpt-oss-20b (llama/22906)

2026-05-14 21:26:48 +03:00

glu.wgsl

ggml-webgpu: port all AOT operators to JIT (llama/20728)

2026-04-30 11:28:58 +03:00

im2col.wgsl

ggml-webgpu: add support for im2col (llama/22259)

2026-04-30 11:29:17 +03:00

memset.wgsl

ggml WebGPU: add support for quantization types (llama/15440)

2025-09-20 13:42:39 +03:00

mul_mat_decls.tmpl

ggml-webgpu: Improve prefill speeds for k-quants + refactor matmul for Q4/Q5/Q8 and k-quants (llama/24225)

2026-06-15 10:33:53 +03:00

mul_mat_id_gather.wgsl

ggml-webgpu: Fix how to dispatch WG to some ops (llama/23750)

2026-05-29 09:47:30 +03:00

mul_mat_id_vec.wgsl

ggml-webgpu: Improve performance of mat-vec and mat-mat for MUL_MAT_ID (llama/22464)

2026-05-01 13:07:35 +03:00

mul_mat_id.wgsl

ggml-webgpu: Fix dequantization helpers to not pass in pointers (llama/21872)

2026-04-30 11:29:10 +03:00

mul_mat_reg_tile.wgsl

ggml-webgpu: Fix dequantization helpers to not pass in pointers (llama/21872)

2026-04-30 11:29:10 +03:00

mul_mat_subgroup_matrix.wgsl

ggml-webgpu: Fix dequantization helpers to not pass in pointers (llama/21872)

2026-04-30 11:29:10 +03:00

mul_mat_vec_acc.tmpl

ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594)

2026-05-29 09:47:30 +03:00

mul_mat_vec_q_acc.tmpl

ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594)

2026-05-29 09:47:30 +03:00

mul_mat_vec.wgsl

ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594)

2026-05-29 09:47:30 +03:00

pad.wgsl

ggml webgpu: support for backend sampling (llama/18880)

2026-01-30 15:56:40 +02:00

quant_inner_loops.tmpl

ggml-webgpu: FlashAttention refactor + standardize quantization support (llama/23834)

2026-06-08 14:36:36 +03:00

quantize_q8.wgsl

ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594)

2026-05-29 09:47:30 +03:00

repeat.wgsl

ggml-webgpu: Add supports for GGML_OP_REPEAT (llama/20230)

2026-03-16 13:10:15 +02:00

rms_norm_mul.wgsl

ggml-webgpu: fix buffer aliasing for ssm_scan and refactor aliasing logic (llama/22456)

2026-04-30 11:29:22 +03:00

rope.wgsl

ggml-webgpu: port all AOT operators to JIT (llama/20728)

2026-04-30 11:28:58 +03:00

row_norm.wgsl

ggml-webgpu: add layer norm ops (llama/22406)

2026-05-14 21:26:48 +03:00

scale.wgsl

Implement 2D workgroups for scale, binary, and unary ops (llama/24044)

2026-06-15 10:33:53 +03:00

set_rows_quant.wgsl

ggml-webgpu: add q4_0/q8_0 SET_ROWS (llama/23760)

2026-06-08 14:36:36 +03:00

set_rows.wgsl

ggml-webgpu: add q4_0/q8_0 SET_ROWS (llama/23760)

2026-06-08 14:36:36 +03:00

set.wgsl

ggml webgpu: ops support for qwen3.5 (SET, TRI_SOLVE, SSM_CONV, GATED_DELTA_NET) + GET_ROWS optimization (llama/20687)

2026-03-29 15:04:36 +03:00

soft_max.wgsl

ggml-webgpu: port all AOT operators to JIT (llama/20728)

2026-04-30 11:28:58 +03:00

solve_tri.wgsl

ggml webgpu: ops support for qwen3.5 (SET, TRI_SOLVE, SSM_CONV, GATED_DELTA_NET) + GET_ROWS optimization (llama/20687)

2026-03-29 15:04:36 +03:00

ssm_conv.wgsl

ggml webgpu: ops support for qwen3.5 (SET, TRI_SOLVE, SSM_CONV, GATED_DELTA_NET) + GET_ROWS optimization (llama/20687)

2026-03-29 15:04:36 +03:00

ssm_scan.wgsl

ggml-webgpu: fix buffer aliasing for ssm_scan and refactor aliasing logic (llama/22456)

2026-04-30 11:29:22 +03:00

sum_rows.wgsl

ggml webgpu: support for backend sampling (llama/18880)

2026-01-30 15:56:40 +02:00

unary.wgsl

Implement 2D workgroups for scale, binary, and unary ops (llama/24044)

2026-06-15 10:33:53 +03:00

upscale.wgsl

ggml-webgpu: add the upscale shader (llama/22419)

2026-05-01 13:07:36 +03:00