MerryMage 080b4b3aff Squashed 'externals/xbyak/' changes from 671fc805..4a6fac8a

4a6fac8a update version to 5.77
801cf3fd cosmetic change of getNumCores
d397e824 fix number of cores that share LLC cache
a669e092 support non-intel-cpu visual studio
af5f422e Merge branch 'fenghaitao-guard_x86' into develop
9b98dc17 Guard x86 specific codes with "#if defined(__i386__) || defined(__x86_64__)"
dd4173e1 move some member variables input private
f72646a7 update version
4612528f format change
4b95e862 Merge branch 'shelleygoel-master'
4c262fa6 add functionality to get num of cores using x2APIC ID
bc70e7e1 recover Xbyak::CastTo
d09a230f unlink Label when LabelManager is destroyed
973e8597 update version
afdb9fe9 Xbyak::CastTo is removed
b011aca4 add RegRip +/- int
acae93cd increase max temp regs for StackFrame
ea4e3562 util::StackFrame uses push/pop instead of mov
42462ef9 use evex encoding for vpslld/vpslldq/vpsraw/...(reg, mem, imm);
da9117a9 update version of readme.md
d35f4fb7 fix the encoding of vinsertps for disp8N
1de435ed bf uses Label class
613922bd add Label L() for convenience
43e15583 fix typo
93579ee6 add protect-re.cpp
60004b5c fix url of protect-re.cpp
348b2709 fix typo of doc
f34f6ed5 update manual
232110be update test
82b78bf0 add setProtectMode
dd8b290f put warning message if pageSize != 4096
64775ca2 a little refactoring
7c3e7b85 fix wrong VSIB encoding with idx >= 16

git-subtree-dir: externals/xbyak
git-subtree-split: 4a6fac8ade404f667b94170f713367fe7da2a852

2020-04-22 20:59:14 +01:00

19 KiB

Raw Blame History

Xbyak 5.77 ; JIT assembler for x86(IA32), x64(AMD64, x86-64) by C++

Abstract

This is a header file which enables dynamically to assemble x86(IA32), x64(AMD64, x86-64) mnemonic.

Feature

header file only
Intel/MASM like syntax
fully support AVX-512

Note: Xbyak uses and(), or(), xor(), not() functions, so -fno-operator-names option is necessary for gcc/clang.

Or define XBYAK_NO_OP_NAMES before including xbyak.h and use and_(), or_(), xor_(), not_() instead of them.

and_(), or_(), xor_(), not_() are always available.

XBYAK_NO_OP_NAMES will be defined in the feature version.

Supported OS

Windows Xp, Vista, Windows 7, Windows 10(32bit, 64bit)
Linux(32bit, 64bit)
Intel macOS

Supported Compilers

Almost C++03 or later compilers for x86/x64 such as Visual Studio, g++, clang++, Intel C++ compiler and g++ on mingw/cygwin.

Install

The following files are necessary. Please add the path to your compile directory.

xbyak.h
xbyak_mnemonic.h
xbyak_util.h

Linux:

make install

These files are copied into /usr/local/include/xbyak.

How to use it

Inherit Xbyak::CodeGenerator class and make the class method.

#define XBYAK_NO_OP_NAMES
#include <xbyak/xbyak.h>

struct Code : Xbyak::CodeGenerator {
    Code(int x)
    {
        mov(eax, x);
        ret();
    }
};

Make an instance of the class and get the function pointer by calling getCode() and call it.

Code c(5);
int (*f)() = c.getCode<int (*)()>();
printf("ret=%d\n", f()); // ret = 5

Syntax

Similar to MASM/NASM syntax with parentheses.

NASM              Xbyak
mov eax, ebx  --> mov(eax, ebx);
inc ecx           inc(ecx);
ret           --> ret();

Addressing

Use qword, dword, word and byte if it is necessary to specify the size of memory, otherwise use ptr.

(ptr|qword|dword|word|byte) [base + index * (1|2|4|8) + displacement]
                            [rip + 32bit disp] ; x64 only

NASM                   Xbyak
mov eax, [ebx+ecx] --> mov(eax, ptr [ebx+ecx]);
mov al, [ebx+ecx]  --> mov(al, ptr [ebx + ecx]);
test byte [esp], 4 --> test(byte [esp], 4);
inc qword [rax]    --> inc(qword [rax]);

Note: qword, ... are member variables, then don't use dword as unsigned int type.

How to use Selector (Segment Register)

mov eax, [fs:eax] --> putSeg(fs);
                      mov(eax, ptr [eax]);
mov ax, cs        --> mov(ax, cs);

Note: Segment class is not derived from Operand.

AVX

vaddps(xmm1, xmm2, xmm3); // xmm1 <- xmm2 + xmm3
vaddps(xmm2, xmm3, ptr [rax]); // use ptr to access memory
vgatherdpd(xmm1, ptr [ebp + 256 + xmm2*4], xmm3);

Note: If XBYAK_ENABLE_OMITTED_OPERAND is defined, then you can use two operand version for backward compatibility. But the newer version will not support it.

vaddps(xmm2, xmm3); // xmm2 <- xmm2 + xmm3

AVX-512

vaddpd zmm2, zmm5, zmm30                --> vaddpd(zmm2, zmm5, zmm30);
vaddpd xmm30, xmm20, [rax]              --> vaddpd(xmm30, xmm20, ptr [rax]);
vaddps xmm30, xmm20, [rax]              --> vaddps(xmm30, xmm20, ptr [rax]);
vaddpd zmm2{k5}, zmm4, zmm2             --> vaddpd(zmm2 | k5, zmm4, zmm2);
vaddpd zmm2{k5}{z}, zmm4, zmm2          --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2);
vaddpd zmm2{k5}{z}, zmm4, zmm2,{rd-sae} --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2 | T_rd_sae);
                                            vaddpd(zmm2 | k5 | T_z | T_rd_sae, zmm4, zmm2); // the position of `|` is arbitrary.
vcmppd k4{k3}, zmm1, zmm2, {sae}, 5     --> vcmppd(k4 | k3, zmm1, zmm2 | T_sae, 5);

vaddpd xmm1, xmm2, [rax+256]            --> vaddpd(xmm1, xmm2, ptr [rax+256]);
vaddpd xmm1, xmm2, [rax+256]{1to2}      --> vaddpd(xmm1, xmm2, ptr_b [rax+256]);
vaddpd ymm1, ymm2, [rax+256]{1to4}      --> vaddpd(ymm1, ymm2, ptr_b [rax+256]);
vaddpd zmm1, zmm2, [rax+256]{1to8}      --> vaddpd(zmm1, zmm2, ptr_b [rax+256]);
vaddps zmm1, zmm2, [rax+rcx*8+8]{1to16} --> vaddps(zmm1, zmm2, ptr_b [rax+rcx*8+8]);
vmovsd [rax]{k1}, xmm4                  --> vmovsd(ptr [rax] | k1, xmm4);

vcvtpd2dq xmm16, oword [eax+33]         --> vcvtpd2dq(xmm16, xword [eax+33]); // use xword for m128 instead of oword
                                            vcvtpd2dq(xmm16, ptr [eax+33]); // default xword
vcvtpd2dq xmm21, [eax+32]{1to2}         --> vcvtpd2dq(xmm21, ptr_b [eax+32]);
vcvtpd2dq xmm0, yword [eax+33]          --> vcvtpd2dq(xmm0, yword [eax+33]); // use yword for m256
vcvtpd2dq xmm19, [eax+32]{1to4}         --> vcvtpd2dq(xmm19, yword_b [eax+32]); // use yword_b to broadcast

vfpclassps k5{k3}, zword [rax+64], 5    --> vfpclassps(k5|k3, zword [rax+64], 5); // specify m512
vfpclasspd k5{k3}, [rax+64]{1to2}, 5    --> vfpclasspd(k5|k3, xword_b [rax+64], 5); // broadcast 64-bit to 128-bit
vfpclassps k5{k3}, [rax+64]{1to4}, 5    --> vfpclassps(k5|k3, yword_b [rax+64], 5); // broadcast 64-bit to 256-bit

Remark

k1, ..., k7 are opmask registers.
use | T_z, | T_sae, | T_rn_sae, | T_rd_sae, | T_ru_sae, | T_rz_sae instead of ,{z}, ,{sae}, ,{rn-sae}, ,{rd-sae}, ,{ru-sae}, ,{rz-sae} respectively.
k4 | k3 is different from k3 | k4.
use ptr_b for broadcast {1toX}. X is automatically determined.
specify xword/yword/zword(_b) for m128/m256/m512 if necessary.

Label

Two kinds of Label are supported. (String literal and Label class).

String literal

L("L1");
  jmp("L1");

  jmp("L2");
  ...
  a few mnemonics (8-bit displacement jmp)
  ...
L("L2");

  jmp("L3", T_NEAR);
  ...
  a lot of mnemonics (32-bit displacement jmp)
  ...
L("L3");

Call hasUndefinedLabel() to verify your code has no undefined label.
you can use a label for immediate value of mov like as mov(eax, "L2").

Support `@@`, `@f`, `@b` like MASM

L("@@"); // <A>
  jmp("@b"); // jmp to <A>
  jmp("@f"); // jmp to <B>
L("@@"); // <B>
  jmp("@b"); // jmp to <B>
  mov(eax, "@b");
  jmp(eax); // jmp to <B>

Local label

Label symbols beginning with a period between inLocalLabel() and outLocalLabel() are treated as a local label. inLocalLabel() and outLocalLabel() can be nested.

void func1()
{
    inLocalLabel();
  L(".lp"); // <A> ; local label
    ...
    jmp(".lp"); // jmp to <A>
  L("aaa"); // global label <C>
    outLocalLabel();

    inLocalLabel();
  L(".lp"); // <B> ; local label
    func1();
    jmp(".lp"); // jmp to <B>
    inLocalLabel();
    jmp("aaa"); // jmp to <C>
}

Label class

L() and jxx() support Label class.

  Xbyak::Label label1, label2;
L(label1);
  ...
  jmp(label1);
  ...
  jmp(label2);
  ...
L(label2);

Use putL for jmp table

    Label labelTbl, L0, L1, L2;
    mov(rax, labelTbl);
    // rdx is an index of jump table
    jmp(ptr [rax + rdx * sizeof(void*)]);
L(labelTbl);
    putL(L0);
    putL(L1);
    putL(L2);
L(L0);
    ....
L(L1);
    ....

assignL(dstLabel, srcLabel) binds dstLabel with srcLabel.

  Label label2;
  Label label1 = L(); // make label1 ; same to Label label1; L(label1);
  ...
  jmp(label2); // label2 is not determined here
  ...
  assignL(label2, label1); // label2 <- label1

The jmp in the above code jumps to label1 assigned by assignL.

Note:

srcLabel must be used in L().
dstLabel must not be used in L().

Label::getAddress() returns the address specified by the label instance and 0 if not specified.

// not AutoGrow mode
Label  label;
assert(label.getAddress() == 0);
L(label);
assert(label.getAddress() == getCurr());

Rip ; relative addressing

Label label;
mov(eax, ptr [rip + label]); // eax = 4
...

L(label);
dd(4);

int x;
...
  mov(eax, ptr[rip + &x]); // throw exception if the difference between &x and current position is larger than 2GiB

Code size

The default max code size is 4096 bytes. Specify the size in constructor of CodeGenerator() if necessary.

class Quantize : public Xbyak::CodeGenerator {
public:
  Quantize()
    : CodeGenerator(8192)
  {
  }
  ...
};

User allocated memory

You can make jit code on prepaired memory.

Call setProtectModeRE yourself to change memory mode if using the prepaired memory.

uint8_t alignas(4096) buf[8192]; // C++11 or later

struct Code : Xbyak::CodeGenerator {
    Code() : Xbyak::CodeGenerator(sizeof(buf), buf)
    {
        mov(rax, 123);
        ret();
    }
};

int main()
{
    Code c;
    c.setProtectModeRE(); // set memory to Read/Exec
    printf("%d\n", c.getCode<int(*)()>()());
}

Note: See sample/test0.cpp.

AutoGrow

The memory region for jit is automatically extended if necessary when AutoGrow is specified in a constructor of CodeGenerator.

Call ready() or readyRE() before calling getCode() to fix jump address.

struct Code : Xbyak::CodeGenerator {
  Code()
    : Xbyak::CodeGenerator(<default memory size>, Xbyak::AutoGrow)
  {
     ...
  }
};
Code c;
// generate code for jit
c.ready(); // mode = Read/Write/Exec