@@ -2127,21 +2127,18 @@ inline void ggml_sycl_op_mul_mat_sycl(
21272127 const sycl::half *src1_ptr = src1->type == GGML_TYPE_F16
21282128 ? (const sycl::half *)src1->data + src1_padded_row_size
21292129 : src1_as_f16.get ();
2130- ggml_sycl_pool_alloc<sycl::half> dst_f16 (ctx.pool (), row_diff * src1_ncols);
21312130
21322131#if GGML_SYCL_DNNL
21332132 if (!g_ggml_sycl_disable_dnn) {
21342133 DnnlGemmWrapper::row_gemm (ctx, src1_ncols, row_diff, ne10, src1_ptr,
21352134 DnnlGemmWrapper::to_dt<sycl::half>(), src0_ptr, DnnlGemmWrapper::to_dt<sycl::half>(),
2136- dst_f16.get (), DnnlGemmWrapper::to_dt<sycl::half>(), stream);
2137- scope_op_debug_print scope_dbg_print (__func__, " /to_fp32_sycl" , dst, /* num_src=*/ 2 ,
2138- " : converting dst to fp32" );
2139- const to_fp32_sycl_t to_fp32_sycl = ggml_get_to_fp32_sycl (GGML_TYPE_F16, dst);
2140- to_fp32_sycl (dst_f16.get (), dst_dd_i, row_diff* src1_ncols, stream);
2135+ dst_dd_i, DnnlGemmWrapper::to_dt<float >(), stream);
21412136 }
21422137 else
21432138#endif
21442139 {
2140+ ggml_sycl_pool_alloc<sycl::half> dst_f16 (ctx.pool (), row_diff * src1_ncols);
2141+
21452142 const sycl::half alpha_f16 = 1 .0f ;
21462143 const sycl::half beta_f16 = 0 .0f ;
21472144 SYCL_CHECK (CHECK_TRY_ERROR (dpct::gemm (
0 commit comments