new compute kernel design, less mem read delay

StgSharp.Native/sn-head/StgSharpNative.h

@@ -12,30 +12,13 @@
 #include "../lib/stbi/stb_image.h"
 
 #include "gl.h"
-#include "wgl.h"
+// #include "wgl.h"
+#define GLFWAPI INTERNAL
 #include "glfw\glfw3.h"
 #include <immintrin.h>
 #include <assert.h>
 #include <glfw_function.h>
 
-#ifndef SN_APIDEF
-
-#ifdef _WIN32
-#define SN_API __declspec(dllexport)
-#elif defined _LINUX_
-#define SN_API __attribute__((dllexport))
-#elif defined _APPLE_
-#define SN_API __attribute__((visibility("default")))
-#else
-#define SN_API
-#endif
-
-#endif // ! SN_APIDEF
-
-#if __cplusplus
-extern "C" {
-#endif
-
 #define byte char
 #define cs_char uint16_t
 #define sn_null_assert(ptr) assert(ptr != NULL)
@@ -65,21 +48,12 @@ SN_API void SN_DECL unloadImageData(Image *out);
 SN_API char *SN_DECL readLog(void);
 SN_API void SN_DECL load_glfw_functions(GLFWFunctionTable *table);
 SN_API void SN_DECL load_intrinsic_function(void *intrinsic_context, uint64_t id);
-
-#pragma region numa
-SN_API SIMDID SN_DECL sn_get_unite_simd(SIMDID left, SIMDID right);
-SN_API int SN_DECL sn_compare_simd(SIMDID left, SIMDID right);
-SN_API void SN_DECL sn_set_thread_affinity(int core_id);
-SN_API int SN_DECL sn_get_numa_overall(int *out_numa_count, int *count_of_each_numa);
-SN_API int SN_DECL sn_get_numa_each(int numa_id, uint64_t *in_numa_cpuid, uint64_t *global_cpuid);
+SN_API SIMDID SN_DECL sn_get_simd_level_global();
 /* Per-core variant: always re-executes CPUID, never caches.
  * Must be called on the target core (after affinity is set). */
 SN_API SIMDID SN_DECL sn_get_simd_level_local();
-SN_API SIMDID SN_DECL sn_get_simd_level_global();
-
-#pragma endregion
-#ifdef __cplusplus
-}
-#endif
+SN_API SIMDID SN_DECL sn_get_unite_simd(SIMDID left, SIMDID right);
+SN_API int SN_DECL sn_compare_simd(SIMDID left, SIMDID right);
+SN_API void SN_DECL sn_set_thread_affinity(int core_id);
 
 #endif