source-engine/tier0/InterlockedCompareExchange128.masm
FluorescentCIAAfricanAmerican 3bf9df6b27 1
2020-04-22 12:56:21 -04:00

79 lines
2.6 KiB
Plaintext

; call cpuid with args in eax, ecx
; store eax, ebx, ecx, edx to p
PUBLIC _InterlockedCompareExchange128
.CODE
_InterlockedCompareExchange128 PROC FRAME
; extern "C" unsigned char _InterlockedCompareExchange128( int64 volatile * Destination, int64 ExchangeHigh, int64 ExchangeLow, int64 * ComparandResult );
.endprolog
; fastcall conventions:
; RCX = Destination
; RDX = Exchange High
; R8 = Exchange Lo
; R9 = ComparandResult
; CMPXCHG16B refernece:
; http://download.intel.com/design/processor/manuals/253666.pdf
; Stash RBX in R11
mov r11, rbx
; Destination ptr to r10
mov r10, rcx
; RCX:RBX Exchange
mov rcx, rdx
mov rbx, r8
; RDX:RAX Comparand
mov rax, [r9]
mov rdx, [r9+8]
; Do the atomic operation
lock cmpxchg16b [r10]
; RDX:RAX now contains the value of the destination, before the atomic operation.
; (Either it already matched and was not changed, or it did not match, in which case
; the value has been loaded into RDX:RAX.) The _InterlockedCompareExchange128 intrinsic
; semantics specify that this is always written out to CompareResult, so give it
; back to the caller.
mov [r9], rax
mov [r9+8], rdx
; Return value is in AL, set it equal to the zero flag
setz al
; Restore RBX and get out
mov rbx, r11
ret
_InterlockedCompareExchange128 ENDP
; For reference, here's what VC2010 generated
;
; __declspec(noinline) unsigned char Test_InterlockedCompareExchange128( int64 volatile * Destination, int64 ExchangeHigh, int64 ExchangeLow, int64 * ComparandResult )
; {
; return _InterlockedCompareExchange128( Destination, ExchangeHigh, ExchangeLow, ComparandResult );
; }
;
;
;?Test_InterlockedCompareExchange128@GCSDK@@YAEPEC_J_J1PEA_J@Z (unsigned char __cdecl GCSDK::Test_InterlockedCompareExchange128(__int64 volatile *,__int64,__int64,__int64 *)):
; 0000000000000000: 48 89 5C 24 08 mov qword ptr [rsp+8],rbx
; 0000000000000005: 49 8B 01 mov rax,qword ptr [r9]
; 0000000000000008: 4C 8B D1 mov r10,rcx
; 000000000000000B: 48 8B CA mov rcx,rdx
; 000000000000000E: 49 8B 51 08 mov rdx,qword ptr [r9+8]
; 0000000000000012: 49 8B D8 mov rbx,r8
; 0000000000000015: F0 49 0F C7 0A lock cmpxchg16b oword ptr [r10]
; 000000000000001A: 48 8B 5C 24 08 mov rbx,qword ptr [rsp+8]
; 000000000000001F: 49 89 01 mov qword ptr [r9],rax
; 0000000000000022: 49 89 51 08 mov qword ptr [r9+8],rdx
; 0000000000000026: 0F 94 C0 sete al
; 0000000000000029: C3 ret
_TEXT ENDS
END