ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

sycl: addressing non-contiguous src1 mul_mats (nc and batched) (#13343)

Browse source 
* sycl: fixed non-contiguous src1 mul_mats (nc and batched)

* Fixed wrong static_cast inside kernel
This commit is contained in:
Alberto Cabrera Pérez 2025-05-08 10:08:01 +01:00 • committed by GitHub
parent 814f795e06
commit 8733e0cf6e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 138 additions and 125 deletions
Download patch file Download diff file Expand all files Collapse all files

21
ggml/src/ggml-sycl/convert.hpp
View file

@ -1,6 +1,6 @@
//
// MIT license
// Copyright (C) 2024 Intel Corporation
// Copyright (C) 2025 Intel Corporation
// SPDX-License-Identifier: MIT
//
@ -16,12 +16,19 @@
#include "common.hpp"
template <typename T>
using to_t_sycl_t = void (*)(const void *__restrict__ x, T *__restrict__ y,
int64_t k, dpct::queue_ptr stream);
typedef to_t_sycl_t<float> to_fp32_sycl_t;
using to_t_sycl_t = void (*)(const void * __restrict__ x, T * __restrict__ y, int64_t k, dpct::queue_ptr stream);
typedef to_t_sycl_t<float> to_fp32_sycl_t;
typedef to_t_sycl_t<sycl::half> to_fp16_sycl_t;
to_fp16_sycl_t ggml_get_to_fp16_sycl(ggml_type type, ggml_tensor *dst);
to_fp32_sycl_t ggml_get_to_fp32_sycl(ggml_type type, ggml_tensor *dst);
to_fp16_sycl_t ggml_get_to_fp16_sycl(ggml_type type, ggml_tensor * dst);
to_fp32_sycl_t ggml_get_to_fp32_sycl(ggml_type type, ggml_tensor * dst);
#endif // GGML_SYCL_CONVERT_HPP
// Nc = Non-contiguous
template <typename T>
using to_t_nc_sycl_t = void (*)(const void * x, T * y, int64_t ne00, int64_t ne01, int64_t ne02, int64_t ne03,
int64_t s01, int64_t s02, int64_t s03, dpct::queue_ptr queue);
typedef to_t_nc_sycl_t<sycl::half> to_fp16_nc_sycl_t;
to_fp16_nc_sycl_t get_to_fp16_nc_sycl(ggml_type type);
#endif // GGML_SYCL_CONVERT_HPP