LLM inference in C/C++
C++ ggml
vulkan: fix compile error (#30495)
Ruben Ortlam14310190cd0a
diff --git a/ggml/src/ggml-vulkan/ggml-vulkan.cpp b/ggml/src/ggml-vulkan/ggml-vulkan.cppindex 1496dd47a8..8357f01fe2 100644--- a/ggml/src/ggml-vulkan/ggml-vulkan.cpp+++ b/ggml/src/ggml-vulkan/ggml-vulkan.cpp@@ -6581,7 +6581,7 @@ static void ggml_vk_mul_mat_q_f16(ggml_backend_vk_context * ctx, vk_context& sub}// column stride in bytes of the B buffer: raw src1 keeps its own stride, staging buffers are tightly packed- const uint64_t b_col_bytes = (y_non_contig || qy_needs_dequant || quantize_y)+ const uint64_t b_col_bytes = (y_needs_reformat || qy_needs_dequant || quantize_y)? (quantize_y ? (uint64_t)(ne10 / ggml_blck_size(GGML_TYPE_Q8_1)) * ggml_type_size(GGML_TYPE_Q8_1) : (uint64_t)ne10 * sizeof(ggml_fp16_t)): (uint64_t)src1->nb[1];