Fixed issue with dgemm performance regression

Using block size KC=256 for all sizes on zen3 architecture

Change-Id: I9fb571e014d152e9156e6549b4d2c0407f6733bb
AMD-Internal: [CPUPL-1401]

Author: dzambare

frame/3/gemm/bli_gemm_front.c

@@ -172,13 +172,6 @@ void bli_gemm_front
 	  rntm
 	);
 
-#ifdef BLIS_CONFIG_ZEN3
-	// On AMD Zen3 platform we support changing
-	// the blocksize at runtime to get optimial
-	// performance
-	BLIS_GEMM_DYNAMIC_BLOCK_SIZE_UPDATE(cntx, rntm, c_local);
-#endif
-
 	obj_t* cp    = &c_local;
 	obj_t* betap = beta;