44#include " common.hpp"
55
66void ggml_sycl_op_repeat_back (ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
7+
78 GGML_ASSERT (dst->src [0 ]->type == GGML_TYPE_F32);
89 GGML_ASSERT (dst->type == GGML_TYPE_F32);
910
1011 const float * src0_dd = (const float *) dst->src [0 ]->data ;
1112 float * dst_dd = (float *) dst->data ;
1213
13- const int64_t ne0 = dst->ne [0 ];
14- const int64_t ne1 = dst->ne [1 ];
15- const int64_t ne2 = dst->ne [2 ];
16- const int64_t ne3 = dst->ne [3 ];
17- const int64_t ne00 = dst->src [0 ]->ne [0 ];
18- const int64_t ne01 = dst->src [0 ]->ne [1 ];
19- const int64_t ne02 = dst->src [0 ]->ne [2 ];
20- const int64_t ne03 = dst->src [0 ]->ne [3 ];
14+ const int64_t ne0 = dst->ne [0 ], ne1 = dst->ne [1 ], ne2 = dst->ne [2 ], ne3 = dst->ne [3 ];
15+ const int64_t ne00 = dst->src [0 ]->ne [0 ], ne01 = dst->src [0 ]->ne [1 ], ne02 = dst->src [0 ]->ne [2 ],
16+ ne03 = dst->src [0 ]->ne [3 ];
2117
2218 const int nr0 = (int ) (ne00 / ne0);
2319 const int nr1 = (int ) (ne01 / ne1);
@@ -29,7 +25,6 @@ void ggml_sycl_op_repeat_back(ggml_backend_sycl_context & ctx, ggml_tensor * dst
2925 const int num_blocks = (total + BLOCK_SIZE - 1 ) / BLOCK_SIZE;
3026
3127 queue_ptr stream = ctx.stream ();
32- stream->memset (dst_dd, 0 , total * sizeof (float ));
3328
3429 stream->parallel_for (
3530 sycl::nd_range<1 >(sycl::range<1 >(num_blocks * BLOCK_SIZE), sycl::range<1 >(BLOCK_SIZE)),
0 commit comments