mirror of
				https://github.com/ggml-org/llama.cpp.git
				synced 2025-11-04 09:32:00 +00:00 
			
		
		
		
	@@ -1407,11 +1407,11 @@ static inline bool ggml_can_repeat_rows(const struct ggml_tensor * t0, const str
 | 
			
		||||
////////////////////////////////////////////////////////////////////////////////
 | 
			
		||||
 | 
			
		||||
struct ggml_context * ggml_init(struct ggml_init_params params) {
 | 
			
		||||
    static bool is_first_call = false;
 | 
			
		||||
    static bool is_first_call = true;
 | 
			
		||||
 | 
			
		||||
    ggml_critical_section_start();
 | 
			
		||||
 | 
			
		||||
    if (!is_first_call) {
 | 
			
		||||
    if (is_first_call) {
 | 
			
		||||
        // initialize time system (required on Windows)
 | 
			
		||||
        ggml_time_init();
 | 
			
		||||
 | 
			
		||||
@@ -1422,7 +1422,8 @@ struct ggml_context * ggml_init(struct ggml_init_params params) {
 | 
			
		||||
            } u = {i};
 | 
			
		||||
            ggml_table_f32_f16[i] = GGML_COMPUTE_FP16_TO_FP32(u.fp16);
 | 
			
		||||
        }
 | 
			
		||||
        is_first_call = true;
 | 
			
		||||
 | 
			
		||||
        is_first_call = false;
 | 
			
		||||
    }
 | 
			
		||||
 | 
			
		||||
    ggml_critical_section_end();
 | 
			
		||||
 
 | 
			
		||||
		Reference in New Issue
	
	Block a user