diff --git a/gbhv/ept.c b/gbhv/ept.c
index 693cd88..6059ce9 100644
--- a/gbhv/ept.c
+++ b/gbhv/ept.c
@@ -173,7 +173,7 @@ PVMM_EPT_PAGE_TABLE HvEptAllocateAndCreateIdentityPageTable(PVMM_CONTEXT GlobalC
 	SIZE_T EntryIndex;
 	
 	/* Allocate all paging structures as 4KB aligned pages */
-	PageTable = OsAllocateContiguousAlignedPages(sizeof(VMM_EPT_PAGE_TABLE) / PAGE_SIZE);
+	PageTable = OsAllocateContiguousAlignedPages(ALIGN_UP_BY(sizeof(VMM_EPT_PAGE_TABLE), PAGE_SIZE) / PAGE_SIZE);
 
 	if(PageTable == NULL)
 	{
@@ -393,7 +393,7 @@ BOOL HvEptSplitLargePage(PVMM_PROCESSOR_CONTEXT ProcessorContext, SIZE_T Physica
 	* alternative was to use the contiguous aligned pages allocator because, in my testing, it resulted in only 4KB virtual
 	* allocations. This allocator also utilizes nonpaged pool frames, so it is more-or-less the same as the other allocator.
 	*/
-	NewSplit = (PVMM_EPT_DYNAMIC_SPLIT) OsAllocateContiguousAlignedPages(sizeof(VMM_EPT_DYNAMIC_SPLIT)/PAGE_SIZE);
+	NewSplit = (PVMM_EPT_DYNAMIC_SPLIT) OsAllocateContiguousAlignedPages(ALIGN_UP_BY(sizeof(VMM_EPT_DYNAMIC_SPLIT), PAGE_SIZE) / PAGE_SIZE);
 	if(!NewSplit)
 	{
 		HvUtilLogError("HvEptSplitLargePage: Failed to allocate dynamic split memory.\n");
diff --git a/gbhv/os.h b/gbhv/os.h
index 06632aa..f03494d 100644
--- a/gbhv/os.h
+++ b/gbhv/os.h
@@ -23,4 +23,4 @@ PVOID OsPhysicalToVirtual(PPHYSVOID PhysicalAddress);
 
 VOID OsZeroMemory(PVOID VirtualAddress, SIZE_T Length);
 
-VOID OsCaptureContext(PREGISTER_CONTEXT ContextRecord);
\ No newline at end of file
+VOID OsCaptureContext(PREGISTER_CONTEXT ContextRecord);
diff --git a/gbhv/os_nt.c b/gbhv/os_nt.c
index bb1390e..061305d 100644
--- a/gbhv/os_nt.c
+++ b/gbhv/os_nt.c
@@ -25,6 +25,7 @@ SIZE_T OsGetCurrentProcessorNumber()
 
 /*
  * Allocate a number of page-aligned, contiguous pages of memory and return a pointer to the region.
+ * When your data size is divided by PAGE_SIZE, you should be sure that no remainder is discarded.
  * 
  * Returns NULL if the pages could not be allocated.
  */
diff --git a/gbhv/vmm.h b/gbhv/vmm.h
index cd4399d..2b90195 100644
--- a/gbhv/vmm.h
+++ b/gbhv/vmm.h
@@ -39,7 +39,10 @@ typedef struct _VMM_HOST_STACK_REGION
 	/*
 	 * Top of the host stack must always have a pointer to the global context.
 	 * This allows the exit handler to access the global context after the host area is loaded.
+	 * In order to follow the x64 calling convention, must ensure that the end of rsp is 8 after entering HOST.
+	 * So we need a 16-byte aligned member in front of GlobalContext.
 	 */
+	DECLSPEC_ALIGN(16) UINT64 Alignment;
 	PVMM_CONTEXT GlobalContext;
 
 } VMM_HOST_STACK_REGION, *PVMM_HOST_STACK_REGION;
diff --git a/gbhv/vmxdefs.asm b/gbhv/vmxdefs.asm
index 975df26..816be6c 100644
--- a/gbhv/vmxdefs.asm
+++ b/gbhv/vmxdefs.asm
@@ -21,7 +21,7 @@ PushGeneralPurposeRegisterContext MACRO
 	push rdi
 	push rsi
 	push rbp
-	sub rsp, 8 ; placeholder for rsp
+	push rbp  ;rsp
 	push rbx
 	push rdx
 	push rcx
@@ -32,22 +32,22 @@ ENDM
 ; RSP is read from VMCS, so there's a placeholder.
 ; https://github.com/asamy/ksm/blob/e7e24931c9df26c33d6e2a0ea9a44c78d3ced7a6/vmx.asm#L62
 PopGeneralPurposeRegisterContext MACRO
-	pop	rax
-	pop	rcx
-	pop	rdx
-	pop	rbx
-	add	rsp, 8
-	pop	rbp
-	pop	rsi
-	pop	rdi
-	pop	r8
-	pop	r9
-	pop	r10
-	pop	r11
-	pop	r12
-	pop	r13
-	pop	r14
-	pop	r15
+	pop rax
+	pop rcx
+	pop rdx
+	pop rbx
+	pop rbp  ;rsp
+	pop rbp
+	pop rsi
+	pop rdi
+	pop r8
+	pop r9
+	pop r10
+	pop r11
+	pop r12
+	pop r13
+	pop r14
+	pop r15
 ENDM
 
 __invept PROC
@@ -111,6 +111,8 @@ HvBeginInitializeLogicalProcessor ENDP
 ; and we have to handle the VMRESUME failure.
 ; Interrupts are automatically disabled for us at this point.
 HvEnterFromGuest PROC
+	; rsp == 0x...............8, qword[rsp] == PVMM_GLOBAL_CONTEXT
+	
 	; Macro to push all GP registers
 	PushGeneralPurposeRegisterContext
 
@@ -123,15 +125,10 @@ HvEnterFromGuest PROC
 	; Second argument (RDX) is stack pointer, which is also the location of the general purpose registers
 	mov rdx, rsp
 
-	; Save XMM registers. The stack must be 16-byte aligned
-	; or a #GP exception is generated.
-	sub rsp, 68h
-	movaps xmmword ptr [rsp], xmm0
-	movaps xmmword ptr [rsp+10h], xmm1
-	movaps xmmword ptr [rsp+20h], xmm2
-	movaps xmmword ptr [rsp+30h], xmm3
-	movaps xmmword ptr [rsp+40h], xmm4
-	movaps xmmword ptr [rsp+50h], xmm5
+	; Save the x87 FPU, MMX, XMM, and MXCSR register state to m512byte.
+	; The stack must be 16-byte aligned, or a #GP exception is generated.
+	sub rsp, 208h
+	fxsave [rsp]
 
 	; Call HvHandleVmExit to actually handle the
 	; incoming exit
@@ -139,14 +136,9 @@ HvEnterFromGuest PROC
 	call HvHandleVmExit
 	add rsp, 20h
 
-	; Restore XMM registers
-	movaps xmm0, xmmword ptr [rsp]
-	movaps xmm1, xmmword ptr [rsp+10h]
-	movaps xmm2, xmmword ptr [rsp+20h]
-	movaps xmm3, xmmword ptr [rsp+30h]
-	movaps xmm4, xmmword ptr [rsp+40h]
-	movaps xmm5, xmmword ptr [rsp+50h]
-	add rsp, 68h
+	; Restore the x87 FPU, MMX, XMM, and MXCSR register state from m512byte.
+	fxrstor [rsp]
+	add rsp, 208h
 
 	; If it's not successful, we need to stop and figure out why
 	test al, al