---

NOTE:

1. this page uses JavaScript to display the diagrams generated by drawio; Fortunately the embedded js on this page is LibreJs-checked.
2. the lincensing of this article inherits whatever Bochs uses; All attributions/credits go to authors of Bochs¹ and other references^{2 3}.
3. to get the source files of the diagrams, click the diagram, then click the “edit (pen shape)” button to open the diagram in drawio online editor, then you can save/export the source file and/or do whatever you want.
4. I use a special patched version Bochs but it won’t hurt.

---

Table of Content

• black macro magic
• base CPU cycle2
• x86_64 long mode Memory access flow

# black macro magic

cpu/cpu.h

#if BX_USE_CPU_SMF == 0
// normal member functions.  This can ONLY be used within BX_CPU_C classes.
// Anyone on the outside should use the BX_CPU macro (defined in bochs.h)
// instead.
#  define BX_CPU_THIS_PTR  this->
#  define BX_CPU_THIS      this
#  define BX_SMF
#  define BX_CPU_C_PREFIX  BX_CPU_C::
// with normal member functions, calling a member fn pointer looks like
// object->*(fnptr)(arg, ...);
// Since this is different from when SMF=1, encapsulate it in a macro.
#  define BX_CPU_CALL_METHOD(func, args) \
            (this->*((BxExecutePtr_tR) (func))) args
#  define BX_CPU_CALL_METHODR(func, args) \
            (this->*((BxResolvePtr_tR) (func))) args
#else
// static member functions.  With SMF, there is only one CPU by definition.
#  define BX_CPU_THIS_PTR  BX_CPU(0)->
#  define BX_CPU_THIS      BX_CPU(0)
#  define BX_SMF           static
#  define BX_CPU_C_PREFIX
#  define BX_CPU_CALL_METHOD(func, args) \
            ((BxExecutePtr_tR) (func)) args
#  define BX_CPU_CALL_METHODR(func, args) \
            ((BxResolvePtr_tR) (func)) args
#endif

why: with BX_USE_CPU_SMF == 1 most methods can ba made static because this-> pointer is not needed (since there is only one instance of e.g. CPU).

i.e. with uni-processor configuration the whole CPU object can be made static (incl. its members and functions). A single CPU object is declared as bx_cpu.

w/ multiprocessor configuration

bx_cpu_array = new BX_CPU_C_PTR[BX_SMP_PROCESSORS];

HACK	UP	SMP	NOTE
BX_CPU_C	bx_cpu_c	bx_cpu_c	脱裤子放屁吗?
CPU object declare	bx_cpu_c bx_cpu	bx_cpu_c **bx_cpu_array
BX_CPU(x)	&bx_cpu	bx_cpu_array[x]	always return a ponter
BX_CPU_THIS_PTR	BX_CPU(0)->	this->
BX_CPU_THIS	BX_CPU(0)	this

debugging: BX_{DEBUG, INFO, ERROR, PANIC}(<fmtstr>) : each can be switched on/off at compile time.

• BOCHSAPI: I guess this is only for documentation purposes. (and win32 DLL building? to have __declspec(dllexport) on variables/functions or classes that the plugins can access.)

Multiple cpu object pointers as array bx_cpu_array[]. Access through BX_CPU(x) macro.

# base CPU cycle²

cpu/instr.h::bxInstruction_c

stores info of a instruction. dispatch func pointer, opcode, src/dest registers, immediate …

Each instruction is emulated by one or two functions, funcptr in bxInstruction_c::execute and bxInstruction_c::execute2. E.g. LOAD+OP uses two handler functions²

All instructions restartable from the register state + bxInstruction_c. Never possible to generate exception /after/ changing the visible state except for RIP (and in some cases RSP) registers CPU loops sets up return point using setjmp(), exception will call longjmp() and appear at the beginning of CPU loop²

// based of: Figure 1: Basic CPU emulation loop ²

# x86_64 long mode Memory access flow

The CPU can access memory through:

• direct host access
• access_{read,write}_physical()

Note that the TLB should have TLB_NoHostPtr bit set in the lpf[11] when direct access through host pointer is NOT allowed for the page. A memory operation asking for a direct access through host pointer does not have TLB_NoHostPtr bit set in its lpf (which is 4K always aligned) and thus will get TLB miss result when the direct access is not allowed. A memory operation that do not care for a direct access to the memory through host pointer should implicitly clear lpf[11] bit in the TLB before comparison with memory access lpf ²

the comment system on this blog works via email. The button
below will generate a mailto: link based on this page's url 
and invoke your email client - please edit the comment there!

[optional] even better, encrypt the email with my public key

- don't modify the subject field
- specify a nickname, otherwise your comment will be shown as   
  anonymous
- your email address will not be disclosed
- you agree that the comment is to be made public.
- to take down a comment, send the request via email.