In practice the memory associated with the relevant array slot is populated by the values. Given your code a small example shows what goes on. Please see comments inline. This is for a release build.
static void Main(string[] args) { Foo[] foos = new Foo[10]; foos[1] = new Foo(127, 255); Console.ReadLine(); }
The code above is JIT compiled as follows
// Method setup 00280050 55 push ebp 00280051 8bec mov ebp,esp 00280053 56 push esi // Create instance of Foo[] 00280054 b98a141d00 mov ecx,1D148Ah 00280059 ba0a000000 mov edx,0Ah 0028005e e8b121f4ff call CORINFO_HELP_NEWARR_1_VC (001c2214) 00280063 8bd0 mov edx,eax // Array range check 00280065 837a0401 cmp dword ptr [edx+4],1 00280069 7624 jbe // Assign foos[1] = new Foo(127, 255) 0028006b 8d4210 lea eax,[edx+10h] <-- load location of foos[1] in eax 0028006e ba7f000000 mov edx,7Fh <-- load 127 in edx 00280073 beff000000 mov esi,0FFh <-- load 255 in esi 00280078 8910 mov dword ptr [eax],edx <-- move the value 127 to foos[1] 0028007a 897004 mov dword ptr [eax+4],esi <-- move the value 255 to foos[1] + offset // This is just for the Console.ReadLine() part + rest of Main 0028007d e8d2436305 call mscorlib_ni!System.Console.get_In() (058b4454) 00280082 8bc8 mov ecx,eax 00280084 8b01 mov eax,dword ptr [ecx] 00280086 8b402c mov eax,dword ptr [eax+2Ch] 00280089 ff501c call dword ptr [eax+1Ch] // Epilog 0028008c 5e pop esi 0028008d 5d pop ebp 0028008e c3 ret //Exception handling 0028008f e8f05e7f70 call clr!JIT_RngChkFail (70a75f84) 00280094 cc int 3
So in short, the code loads the constants in registers and then copies values of these registers to the memory associated with the relevant part of the array instance.
structand when it is declared as aclass, theoretically, the size should benElements * IntPtr.Sizefor classes andnElements * sizeof(Foo)for structs; but I've never tried this in C# so I could be wrong.newwill actually override a Foo object directly, instead of creating and then copying.