new compute kernel design, less mem read delay

2026-04-07 01:04:05 +08:00
parent d702b3224c
commit 12a216d404
47 changed files with 225 additions and 145 deletions
@@ -138,7 +138,7 @@ extern __declspec(align(64)) sn_zero_kernel_u64 SN_ZERO_KERNEL;
                                                         int col_length, int row_length,       \
                                                         int col_index, int row_index)

-#if 0
+/* Declarations for AVX float variant 
 DECLARE_BUILD_PANEL(avx, float);
 DECLARE_BUILD_PANEL(512, float);

@@ -149,6 +149,6 @@ DECLARE_STORE_PANEL(512, float);
 DECLARE_BUILD_PANEL(512, double);

 DECLARE_STORE_PANEL(512, double);
-#endif
+/**/

-#endif /* SN_MATKERNEL */
+#endif /* SN_MATKERNEL */