inline assembly.zip_嵌入式_linux_其他

<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/626451364c65f41259cae270/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/626451364c65f41259cae270/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">© Copyright IBM Corporation 2001<span class="_ _0"> </span>T<span class="_ _1"></span>rademarks</div><div class="t m0 x1 h2 y2 ff1 fs0 fc1 sc0 ls0 ws0">Inline assembly for x86 in Linux<span class="_ _2"> </span>Page 1 of 9</div><div class="t m0 x1 h3 y3 ff2 fs1 fc2 sc0 ls0 ws0">Inline assembly for x86 in Linux</div><div class="t m0 x1 h4 y4 ff2 fs2 fc2 sc0 ls0 ws0">Putting the pieces together</div><div class="t m0 x1 h5 y5 ff1 fs3 fc0 sc0 ls0 ws0">Bharata Rao<span class="fc1"> (</span>rbharata@in.ibm.com<span class="fc1">)</span></div><div class="t m0 x1 h5 y6 ff1 fs3 fc1 sc0 ls0 ws0">IBM Linux T<span class="_ _3"></span>echnology Center<span class="_ _4"></span>, IBM Software Labs, India</div><div class="t m0 x2 h5 y5 ff1 fs3 fc1 sc0 ls0 ws0">01 March 2001</div><div class="t m0 x3 h5 y7 ff1 fs3 fc1 sc0 ls0 ws0">Bharata B. Rao offers a guide to the overall use and structure of inline assembly for x86 on the</div><div class="t m0 x3 h5 y8 ff1 fs3 fc1 sc0 ls0 ws0">Linux platform. He covers the basics of inline assembly and its various usages, gives some</div><div class="t m0 x3 h5 y9 ff1 fs3 fc1 sc0 ls0 ws0">basic inline assembly coding guidelines, and explains the instances of inline assembly code in</div><div class="t m0 x3 h5 ya ff1 fs3 fc1 sc0 ls0 ws0">the Linux kernel.</div><div class="t m0 x1 h5 yb ff1 fs3 fc1 sc0 ls0 ws0">If you're a Linux kernel developer<span class="_ _4"></span>, you probably find yourself coding highly architecture-dependent</div><div class="t m0 x1 h5 yc ff1 fs3 fc1 sc0 ls0 ws0">functions or optimizing a code path pretty often. And you probably do this by inserting assembly</div><div class="t m0 x1 h5 yd ff1 fs3 fc1 sc0 ls0 ws0">language instructions into the middle of C statements (a method otherwise known as inline</div><div class="t m0 x1 h5 ye ff1 fs3 fc1 sc0 ls0 ws0">assembly). Let's take a look at the specific usage of inline assembly in Linux. (We'll limit our</div><div class="t m0 x1 h5 yf ff1 fs3 fc1 sc0 ls0 ws0">discussion to the IA32 assembly<span class="_ _4"></span>.)</div><div class="t m0 x1 h4 y10 ff2 fs2 fc2 sc0 ls0 ws0">GNU assembler syntax in brief</div><div class="t m0 x1 h5 y11 ff1 fs3 fc1 sc0 ls0 ws0">Let's first look at the basic assembler syntax used in Linux. GCC, the GNU C Compiler for Linux,</div><div class="t m0 x1 h5 y12 ff1 fs3 fc1 sc0 ls0 ws0">uses A<span class="_ _4"></span>T&T assembly syntax. Some of the basic rules of this syntax are listed below<span class="_ _4"></span>. (The list is by</div><div class="t m0 x1 h5 y13 ff1 fs3 fc1 sc0 ls0 ws0">no means complete; I've included only those rules pertinent to inline assembly<span class="_ _4"></span>.)</div><div class="t m0 x1 h6 y14 ff2 fs3 fc2 sc0 ls0 ws0">Register naming</div><div class="t m0 x1 h5 y15 ff1 fs3 fc1 sc0 ls0 ws0">Register names are prefixed by %. That is, if eax has to be used, it should be used as %eax.</div><div class="t m0 x1 h6 y16 ff2 fs3 fc2 sc0 ls0 ws0">Source and destination ordering</div><div class="t m0 x1 h5 y17 ff1 fs3 fc1 sc0 ls0 ws0">In any instruction, source comes first and destination follows. This differs from Intel syntax, where</div><div class="t m0 x1 h5 y18 ff1 fs3 fc1 sc0 ls0 ws0">source comes after destination.</div><div class="t m0 x1 h7 y19 ff3 fs4 fc3 sc0 ls0 ws0"> mov %eax, %ebx, transfers the contents of eax to ebx.</div><div class="t m0 x1 h6 y1a ff2 fs3 fc2 sc0 ls0 ws0">Size of operand</div><div class="t m0 x1 h5 y1b ff1 fs3 fc1 sc0 ls0 ws0">The instructions are suffixed by b, w<span class="_ _4"></span>, or l, depending on whether the operand is a byte, word, or</div><div class="t m0 x1 h5 y1c ff1 fs3 fc1 sc0 ls0 ws0">long. This is not mandatory; GCC tries provide the appropriate suffix by reading the operands. But</div><div class="t m0 x1 h5 y1d ff1 fs3 fc1 sc0 ls0 ws0">specifying the suffixes manually improves the code readability and eliminates the possibility of the</div><div class="t m0 x1 h5 y1e ff1 fs3 fc1 sc0 ls0 ws0">compilers guessing incorrectly<span class="_ _4"></span>.</div><a class="l" rel='nofollow' onclick='return false;'><div class="d m1"></div></a><a class="l" rel='nofollow' onclick='return false;'><div class="d m1"></div></a><a class="l" rel='nofollow' onclick='return false;'><div class="d m1"></div></a><a class="l" rel='nofollow' onclick='return false;'><div class="d m1"></div></a></div><div class="pi" data-data='{"ctm":[1.612700,0.000000,0.000000,1.612700,0.000000,0.000000]}'></div></div> </body> </html>

<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/626451364c65f41259cae270/bg2.jpg"><div class="t m0 x1 h5 y1f ff1 fs3 fc1 sc0 ls0 ws0">developerWorks®<span class="_ _5"> </span>ibm.com/developerWorks/</div><div class="t m0 x1 h2 y20 ff1 fs0 fc1 sc0 ls0 ws0">Inline assembly for x86 in Linux<span class="_ _2"> </span>Page 2 of 9</div><div class="t m0 x1 h7 y21 ff3 fs4 fc3 sc0 ls0 ws0"> movb %al, %bl -- Byte move</div><div class="t m0 x1 h7 y22 ff3 fs4 fc3 sc0 ls0 ws0"> movw %ax, %bx -- Word move</div><div class="t m0 x1 h7 y23 ff3 fs4 fc3 sc0 ls0 ws0"> movl %eax, %ebx -- Longword move</div><div class="t m0 x1 h6 y24 ff2 fs3 fc2 sc0 ls0 ws0">Immediate operand</div><div class="t m0 x1 h5 y25 ff1 fs3 fc1 sc0 ls0 ws0">An immediate operand is specified by using $.</div><div class="t m0 x1 h7 y26 ff3 fs4 fc3 sc0 ls0 ws0"> movl $0xffff, %eax -- will move the value of 0xffff into eax register.</div><div class="t m0 x1 h6 y27 ff2 fs3 fc2 sc0 ls0 ws0">Indirect memory reference</div><div class="t m0 x1 h5 y28 ff1 fs3 fc1 sc0 ls0 ws0">Any indirect references to memory are done by using ( ).</div><div class="t m0 x1 h7 y29 ff3 fs4 fc3 sc0 ls0 ws0"> movb (%esi), %al -- will transfer the byte in the memory</div><div class="t m0 x1 h7 y2a ff3 fs4 fc3 sc0 ls0 ws0"> pointed by esi into al register</div><div class="t m0 x1 h4 y2b ff2 fs2 fc2 sc0 ls0 ws0">Inline assembly</div><div class="t m0 x1 h5 y2c ff1 fs3 fc1 sc0 ls0 ws0">GCC provides the special construct "asm" for inline assembly<span class="_ _4"></span>, which has the following format:</div><div class="t m0 x1 h7 y2d ff3 fs4 fc3 sc0 ls0 ws0"> asm ( assembler template</div><div class="t m0 x1 h7 y2e ff3 fs4 fc3 sc0 ls0 ws0"> : output operands (optional)</div><div class="t m0 x1 h7 y2f ff3 fs4 fc3 sc0 ls0 ws0"> : input operands (optional)</div><div class="t m0 x1 h7 y30 ff3 fs4 fc3 sc0 ls0 ws0"> : list of clobbered registers (optional)</div><div class="t m0 x1 h7 y31 ff3 fs4 fc3 sc0 ls0 ws0"> );</div><div class="t m0 x1 h5 y32 ff1 fs3 fc1 sc0 ls0 ws0">In this example, the assembler template consists of assembly instructions. The input operands are</div><div class="t m0 x1 h5 y33 ff1 fs3 fc1 sc0 ls0 ws0">the C expressions that serve as input operands to the instructions. The output operands are the C</div><div class="t m0 x1 h5 y34 ff1 fs3 fc1 sc0 ls0 ws0">expressions on which the output of the assembly instructions will be performed.</div><div class="t m0 x1 h7 y35 ff3 fs4 fc3 sc0 ls0 ws0"> asm ("movl %%cr3, %0\n" :"=r"(cr3val));</div><div class="t m0 x1 h7 y36 ff3 fs4 fc3 sc0 ls0 ws0"> a %eax</div><div class="t m0 x1 h7 y37 ff3 fs4 fc3 sc0 ls0 ws0"> b %ebx</div><div class="t m0 x1 h7 y38 ff3 fs4 fc3 sc0 ls0 ws0"> c %ecx</div><div class="t m0 x1 h7 y39 ff3 fs4 fc3 sc0 ls0 ws0"> d %edx</div><div class="t m0 x1 h7 y3a ff3 fs4 fc3 sc0 ls0 ws0"> S %esi</div><div class="t m0 x1 h7 y3b ff3 fs4 fc3 sc0 ls0 ws0"> D %edi</div><div class="t m0 x1 h6 y3c ff2 fs3 fc2 sc0 ls0 ws0">Memory operand constraint(m)</div><div class="t m0 x1 h5 y3d ff1 fs3 fc1 sc0 ls0 ws0">When the operands are in the memory<span class="_ _4"></span>, any operations performed on them will occur directly in</div><div class="t m0 x1 h5 y3e ff1 fs3 fc1 sc0 ls0 ws0">the memory location, as opposed to register constraints, which first store the value in a register</div><div class="t m0 x1 h5 y3f ff1 fs3 fc1 sc0 ls0 ws0">to be modified and then write it back to the memory location. But register constraints are usually</div><div class="t m0 x1 h5 y40 ff1 fs3 fc1 sc0 ls0 ws0">used only when they are absolutely necessary for an instruction or they significantly speed up the</div><div class="t m0 x1 h5 y41 ff1 fs3 fc1 sc0 ls0 ws0">process. Memory constraints can be used most efficiently in cases where a C variable needs to be</div><div class="t m0 x1 h5 y42 ff1 fs3 fc1 sc0 ls0 ws0">updated inside "asm" and you really don't want to use a register to hold its value. For example, the</div><div class="t m0 x1 h5 y43 ff1 fs3 fc1 sc0 ls0 ws0">value of idtr is stored in the memory location loc:</div><div class="t m0 x1 h7 y44 ff3 fs4 fc3 sc0 ls0 ws0"> ("sidt %0\n" : :"m"(loc));</div><div class="t m0 x1 h6 y45 ff2 fs3 fc2 sc0 ls0 ws0">Matching(Digit) constraints</div></div><div class="pi" data-data='{"ctm":[1.612700,0.000000,0.000000,1.612700,0.000000,0.000000]}'></div></div>