Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

ggml-webgpu: add mulmat with overlapping src0/src1 (e.g., for minimax-01) (#27321)

Browse Source
This commit is contained in:
Masashi Yoshimura
2026-08-19 16:29:33 +09:00
committed by GitHub
parent 0adcc3bb57
commit 5112b9738b
10 changed files with 189 additions and 92 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_vec.wgsl
+14 -4
View File
@@ -7,6 +7,11 @@ enable f16;
requires packed_4x8_integer_dot_product;
#endif
#ifdef SRC_OVERLAP
#define SRC0 merged_src
#define SRC1 merged_src
#endif
#define DECLARE_BYTE_LOADERS_SRC0
#include "common_decls.tmpl"
@@ -35,17 +40,22 @@ struct MulMatParams {
broadcast3: u32
};
#if defined(MMVQ)
@group(0) @binding(0) var<storage, read_write> src0: array<SRC0_TYPE>;
#ifdef MMVQ
@group(0) @binding(1) var<storage, read_write> src1q: array<q8_1>;
#define DST_BINDING 2
#elif defined(SRC_OVERLAP)
@group(0) @binding(0) var<storage, read_write> merged_src: array<SRC0_TYPE>;
#define DST_BINDING 1
#else
@group(0) @binding(0) var<storage, read_write> src0: array<SRC0_TYPE>;
@group(0) @binding(1) var<storage, read_write> src1: array<SRC1_TYPE>;
#define DST_BINDING 2
#endif
@group(0) @binding(2) var<storage, read_write> dst: array<f32>;
@group(0) @binding(DST_BINDING) var<storage, read_write> dst: array<f32>;
// "mul_mat_vec_acc.tmpl" requires params.k, params.m, params.stride_01
@group(0) @binding(3) var<uniform> params: MulMatParams;
@group(0) @binding(DST_BINDING + 1) var<uniform> params: MulMatParams;
// Flattened as [row][thread] to keep each row's reduction contiguous in memory.
var<workgroup> partial_sums: array<f32, OUTPUTS_PER_WG * WG_SIZE>;