Skip to main content

Module coopmat

Module coopmat 

Structs§

CoopMatShape
Problem and tile descriptor for Cooperative Matrix / Tensor Core routing.

Constants§

MATMUL_TC_F16_16X16_ENTRY
Entry point of the 16x16x16 FP16 cooperative-matrix GEMM.
MATMUL_TC_I8_16X16X32_ENTRY
Entry point of the 16x16x32 INT8 cooperative-matrix GEMM.
MATMUL_TC_TILED_ENTRY
Entry point of the tiled cooperative-matrix GEMM (matmul_tc_wgsl_tiled).
TILE
Default portable cooperative matrix tile dimension (8x8x8).
WGPU_API_VERSION
Target Etalon / WGPU API version.
WGPU_COOPMAT_BLOCKED
Cooperative matrix blocked flag: false on Etalon because etalon-hal enables vulkan_memory_model_device_scope.

Functions§

matmul_tc_f16_16x16_wgsl
Emits a 16x16x16 mixed-precision FP16->F32 cooperative-matrix GEMM tile.
matmul_tc_i8_16x16x32_wgsl
Emits a 16x16x32 INT8 (I8 -> I32) cooperative-matrix GEMM tile using Etalon’s CooperativeSize::ThirtyTwo (coop_mat16x32 / coop_mat32x16 / coop_mat16x16<i32, C>).
matmul_tc_wgsl
Emits the cooperative-matrix 8x8 GEMM tile C = A * B, all-f32.
matmul_tc_wgsl_tiled
Emits a tiled cooperative-matrix GEMM C[m x n] = A[m x k] * B[k x n] (row-major, all-f32).
should_use_coopmat
Determines whether Cooperative Matrix / WMMA should be used for shape.