|
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>Mu's instructions and their table-driven translation</title>
<meta name="Generator" content="Vim/8.1">
<meta name="plugin-version" content="vim8.1_v1">
<meta name="syntax" content="none">
<meta name="settings" content="use_css,no_foldcolumn,expand_tabs,prevent_copy=">
<meta name="colorscheme" content="minimal-light">
<style type="text/css">
<!--
pre { font-family: monospace; color: #000000; background-color: #c6c6c6; }
body { font-family: monospace; color: #000000; background-color: #c6c6c6; }
* { font-size: 1em; }
.muComment { color: #005faf; }
.SpecialChar { color: #d70000; }
.Constant { color: #008787; }
.PreProc { color: #c000c0; }
-->
</style>
</head>
<body>
<pre id='vimCodeElement'>
<span class="muComment">## Mu's instructions and their table-driven translation</span>
See <a href="http://akkartik.name/akkartik-convivial-20200607.pdf">http://akkartik.name/akkartik-convivial-20200607.pdf</a> for the complete
story. In brief: Mu is a statement-oriented language. Blocks consist of flat
lists of instructions. Instructions can have inputs after the operation, and
outputs to the left of a '<span class="SpecialChar"><-</span>'. Inputs and outputs must be variables. They can't
include nested expressions. Variables can be literals ('n'), or live in a
register ('var/reg') or in memory ('var') at some 'stack-offset' from the 'ebp'
register. Outputs must be registers. To modify a variable in memory, pass it in
by reference as an input. (Inputs are more precisely called 'inouts'.)
Conversely, registers that are just read from must not be passed as inputs.
The following chart shows all the instruction forms supported by Mu, along with
the SubX instruction they're translated to.
<span class="muComment">## Integer instructions</span>
These instructions use the general-purpose registers.
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> increment => <span class="Constant">"40/increment-eax"</span>
var/<span class="Constant">ecx</span> <span class="SpecialChar"><-</span> increment => <span class="Constant">"41/increment-ecx"</span>
var/<span class="Constant">edx</span> <span class="SpecialChar"><-</span> increment => <span class="Constant">"42/increment-edx"</span>
var/<span class="Constant">ebx</span> <span class="SpecialChar"><-</span> increment => <span class="Constant">"43/increment-ebx"</span>
var/<span class="Constant">esi</span> <span class="SpecialChar"><-</span> increment => <span class="Constant">"46/increment-esi"</span>
var/<span class="Constant">edi</span> <span class="SpecialChar"><-</span> increment => <span class="Constant">"47/increment-edi"</span>
increment var => <span class="Constant">"ff 0/subop/increment *(ebp+"</span> var.stack-offset <span class="Constant">")"</span>
increment *var/reg => <span class="Constant">"ff 0/subop/increment *"</span> reg
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> decrement => <span class="Constant">"48/decrement-eax"</span>
var/<span class="Constant">ecx</span> <span class="SpecialChar"><-</span> decrement => <span class="Constant">"49/decrement-ecx"</span>
var/<span class="Constant">edx</span> <span class="SpecialChar"><-</span> decrement => <span class="Constant">"4a/decrement-edx"</span>
var/<span class="Constant">ebx</span> <span class="SpecialChar"><-</span> decrement => <span class="Constant">"4b/decrement-ebx"</span>
var/<span class="Constant">esi</span> <span class="SpecialChar"><-</span> decrement => <span class="Constant">"4e/decrement-esi"</span>
var/<span class="Constant">edi</span> <span class="SpecialChar"><-</span> decrement => <span class="Constant">"4f/decrement-edi"</span>
decrement var => <span class="Constant">"ff 1/subop/decrement *(ebp+"</span> var.stack-offset <span class="Constant">")"</span>
decrement *var/reg => <span class="Constant">"ff 1/subop/decrement *"</span> reg
var/reg <span class="SpecialChar"><-</span> add var2/reg2 => <span class="Constant">"01/add-to %"</span> reg <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> add var2 => <span class="Constant">"03/add *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> add *var2/reg2 => <span class="Constant">"03/add *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
add-to var1, var2/reg => <span class="Constant">"01/add-to *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
add-to *var1/reg1, var2/reg2 => <span class="Constant">"01/add-to *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> add n => <span class="Constant">"05/add-to-eax "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> add n => <span class="Constant">"81 0/subop/add %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
add-to var, n => <span class="Constant">"81 0/subop/add *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
add-to *var/reg, n => <span class="Constant">"81 0/subop/add *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> subtract var2/reg2 => <span class="Constant">"29/subtract-from %"</span> reg <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> subtract var2 => <span class="Constant">"2b/subtract *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> subtract *var2/reg2 => <span class="Constant">"2b/subtract *"</span> reg2 <span class="Constant">" "</span> reg1 <span class="Constant">"/r32"</span>
subtract-from var1, var2/reg2 => <span class="Constant">"29/subtract-from *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> reg2 <span class="Constant">"/r32"</span>
subtract-from *var1/reg1, var2/reg2 => <span class="Constant">"29/subtract-from *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> subtract n => <span class="Constant">"2d/subtract-from-eax "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> subtract n => <span class="Constant">"81 5/subop/subtract %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
subtract-from var, n => <span class="Constant">"81 5/subop/subtract *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
subtract-from *var/reg, n => <span class="Constant">"81 5/subop/subtract *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> and var2/reg2 => <span class="Constant">"21/and-with %"</span> reg <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> and var2 => <span class="Constant">"23/and *(ebp+"</span> var2.stack-offset <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> and *var2/reg2 => <span class="Constant">"23/and *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
and-with var1, var2/reg => <span class="Constant">"21/and-with *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
and-with *var1/reg1, var2/reg2 => <span class="Constant">"21/and-with *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> and n => <span class="Constant">"25/and-with-eax "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> and n => <span class="Constant">"81 4/subop/and %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
and-with var, n => <span class="Constant">"81 4/subop/and *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
and-with *var/reg, n => <span class="Constant">"81 4/subop/and *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> or var2/reg2 => <span class="Constant">"09/or-with %"</span> reg <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> or var2 => <span class="Constant">"0b/or *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> or *var2/reg2 => <span class="Constant">"0b/or *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
or-with var1, var2/reg2 => <span class="Constant">"09/or-with *(ebp+"</span> var1.stack-offset <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
or-with *var1/reg1, var2/reg2 => <span class="Constant">"09/or-with *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> or n => <span class="Constant">"0d/or-with-eax "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> or n => <span class="Constant">"81 1/subop/or %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
or-with var, n => <span class="Constant">"81 1/subop/or *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
or-with *var/reg, n => <span class="Constant">"81 1/subop/or *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> xor var2/reg2 => <span class="Constant">"31/xor-with %"</span> reg <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> xor var2 => <span class="Constant">"33/xor *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> xor *var2/reg2 => <span class="Constant">"33/xor *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
xor-with var1, var2/reg => <span class="Constant">"31/xor-with *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
xor-with *var1/reg1, var2/reg2 => <span class="Constant">"31/xor-with *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> xor n => <span class="Constant">"35/xor-with-eax "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> xor n => <span class="Constant">"81 6/subop/xor %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
xor-with var, n => <span class="Constant">"81 6/subop/xor *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
xor-with *var/reg, n => <span class="Constant">"81 6/subop/xor *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> negate => <span class="Constant">"f7 3/subop/negate %"</span> reg
negate var => <span class="Constant">"f7 3/subop/negate *(ebp+"</span> var.stack-offset <span class="Constant">")"</span>
negate *var/reg => <span class="Constant">"f7 3/subop/negate *"</span> reg
var/reg <span class="SpecialChar"><-</span> shift-left n => <span class="Constant">"c1/shift 4/subop/left %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> shift-right n => <span class="Constant">"c1/shift 5/subop/right %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> shift-right-signed n => <span class="Constant">"c1/shift 7/subop/right-signed %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
shift-left var, n => <span class="Constant">"c1/shift 4/subop/left *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
shift-left *var/reg, n => <span class="Constant">"c1/shift 4/subop/left *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
shift-right var, n => <span class="Constant">"c1/shift 5/subop/right *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
shift-right *var/reg, n => <span class="Constant">"c1/shift 5/subop/right *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
shift-right-signed var, n => <span class="Constant">"c1/shift 7/subop/right-signed *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
shift-right-signed *var/reg, n => <span class="Constant">"c1/shift 7/subop/right-signed *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/<span class="Constant">eax</span> <span class="SpecialChar"><-</span> copy n => <span class="Constant">"b8/copy-to-eax "</span> n <span class="Constant">"/imm32"</span>
var/<span class="Constant">ecx</span> <span class="SpecialChar"><-</span> copy n => <span class="Constant">"b9/copy-to-ecx "</span> n <span class="Constant">"/imm32"</span>
var/<span class="Constant">edx</span> <span class="SpecialChar"><-</span> copy n => <span class="Constant">"ba/copy-to-edx "</span> n <span class="Constant">"/imm32"</span>
var/<span class="Constant">ebx</span> <span class="SpecialChar"><-</span> copy n => <span class="Constant">"bb/copy-to-ebx "</span> n <span class="Constant">"/imm32"</span>
var/<span class="Constant">esi</span> <span class="SpecialChar"><-</span> copy n => <span class="Constant">"be/copy-to-esi "</span> n <span class="Constant">"/imm32"</span>
var/<span class="Constant">edi</span> <span class="SpecialChar"><-</span> copy n => <span class="Constant">"bf/copy-to-edi "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> copy var2/reg2 => <span class="Constant">"89/<- %"</span> reg <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
copy-to var1, var2/reg => <span class="Constant">"89/<- *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
copy-to *var1/reg1, var2/reg2 => <span class="Constant">"89/<- *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> copy var2 => <span class="Constant">"8b/-> *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> copy *var2/reg2 => <span class="Constant">"8b/-> *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> copy n => <span class="Constant">"c7 0/subop/copy %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
copy-to var, n => <span class="Constant">"c7 0/subop/copy *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
copy-to *var/reg, n => <span class="Constant">"c7 0/subop/copy *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> copy-byte var2/reg2 => <span class="Constant">"8a/byte-> %"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> copy-byte *var2/reg2 => <span class="Constant">"8a/byte-> *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
copy-byte-to *var1/reg1, var2/reg2 => <span class="Constant">"88/byte<- *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
compare var1, var2/reg2 => <span class="Constant">"39/compare *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> reg2 <span class="Constant">"/r32"</span>
compare *var1/reg1, var2/reg2 => <span class="Constant">"39/compare *"</span> reg1 <span class="Constant">" "</span> reg2 <span class="Constant">"/r32"</span>
compare var1/reg1, var2 => <span class="Constant">"3b/compare<- *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg1 <span class="Constant">"/r32"</span>
compare var/reg, *var2/reg2 => <span class="Constant">"3b/compare<- *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
compare var/<span class="Constant">eax</span>, n => <span class="Constant">"3d/compare-eax-with "</span> n <span class="Constant">"/imm32"</span>
compare var/reg, n => <span class="Constant">"81 7/subop/compare %"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
compare var, n => <span class="Constant">"81 7/subop/compare *(ebp+"</span> var.stack-offset <span class="Constant">") "</span> n <span class="Constant">"/imm32"</span>
compare *var/reg, n => <span class="Constant">"81 7/subop/compare *"</span> reg <span class="Constant">" "</span> n <span class="Constant">"/imm32"</span>
var/reg <span class="SpecialChar"><-</span> multiply var2 => <span class="Constant">"0f af/multiply *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> multiply *var2/reg2 => <span class="Constant">"0f af/multiply *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
<span class="muComment">## Floating-point operations</span>
These instructions operate on either floating-point registers (xreg) or
general-purpose registers (reg) in indirect mode.
var/xreg <span class="SpecialChar"><-</span> add var2/xreg2 => <span class="Constant">"f3 0f 58/add %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> add var2 => <span class="Constant">"f3 0f 58/add *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> add *var2/reg2 => <span class="Constant">"f3 0f 58/add *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> subtract var2/xreg2 => <span class="Constant">"f3 0f 5c/subtract %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> subtract var2 => <span class="Constant">"f3 0f 5c/subtract *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> subtract *var2/reg2 => <span class="Constant">"f3 0f 5c/subtract *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> multiply var2/xreg2 => <span class="Constant">"f3 0f 59/multiply %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> multiply var2 => <span class="Constant">"f3 0f 59/multiply *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> multiply *var2/reg2 => <span class="Constant">"f3 0f 59/multiply *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> divide var2/xreg2 => <span class="Constant">"f3 0f 5e/divide %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> divide var2 => <span class="Constant">"f3 0f 5e/divide *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> divide *var2/reg2 => <span class="Constant">"f3 0f 5e/divide *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
There are also some exclusively floating-point instructions:
var/xreg <span class="SpecialChar"><-</span> reciprocal var2/xreg2 => <span class="Constant">"f3 0f 53/reciprocal %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> reciprocal var2 => <span class="Constant">"f3 0f 53/reciprocal *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> reciprocal *var2/reg2 => <span class="Constant">"f3 0f 53/reciprocal *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> square-root var2/xreg2 => <span class="Constant">"f3 0f 51/square-root %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> square-root var2 => <span class="Constant">"f3 0f 51/square-root *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> square-root *var2/reg2 => <span class="Constant">"f3 0f 51/square-root *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> inverse-square-root var2/xreg2 => <span class="Constant">"f3 0f 52/inverse-square-root %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> inverse-square-root var2 => <span class="Constant">"f3 0f 52/inverse-square-root *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> inverse-square-root *var2/reg2 => <span class="Constant">"f3 0f 52/inverse-square-root *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> min var2/xreg2 => <span class="Constant">"f3 0f 5d/min %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> min var2 => <span class="Constant">"f3 0f 5d/min *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> min *var2/reg2 => <span class="Constant">"f3 0f 5d/min *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> max var2/xreg2 => <span class="Constant">"f3 0f 5f/max %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> max var2 => <span class="Constant">"f3 0f 5f/max *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> max *var2/reg2 => <span class="Constant">"f3 0f 5f/max *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
Remember, when these instructions use indirect mode, they still use an integer
register. Floating-point registers can't hold addresses.
Most instructions operate exclusively on integer or floating-point operands.
The only exceptions are the instructions for converting between integers and
floating-point numbers.
var/xreg <span class="SpecialChar"><-</span> convert var2/reg2 => <span class="Constant">"f3 0f 2a/convert-to-float %"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> convert var2 => <span class="Constant">"f3 0f 2a/convert-to-float *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> convert *var2/reg2 => <span class="Constant">"f3 0f 2a/convert-to-float *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
Converting floats to ints performs rounding by default. (We don't mess with the
MXCSR control register.)
var/reg <span class="SpecialChar"><-</span> convert var2/xreg2 => <span class="Constant">"f3 0f 2d/convert-to-int %"</span> xreg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> convert var2 => <span class="Constant">"f3 0f 2d/convert-to-int *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> convert *var2/reg2 => <span class="Constant">"f3 0f 2d/convert-to-int *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
There's a separate instruction for truncating the fractional part.
var/reg <span class="SpecialChar"><-</span> truncate var2/xreg2 => <span class="Constant">"f3 0f 2c/truncate-to-int %"</span> xreg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> truncate var2 => <span class="Constant">"f3 0f 2c/truncate-to-int *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg <span class="SpecialChar"><-</span> truncate *var2/reg2 => <span class="Constant">"f3 0f 2c/truncate-to-int *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
There are no instructions accepting floating-point literals. To obtain integer
literals in floating-point registers, copy them to general-purpose registers
and then convert them to floating-point.
One pattern you may have noticed above is that the floating-point instructions
above always write to registers. The only exceptions are `copy` instructions,
which can write to memory locations.
var/xreg <span class="SpecialChar"><-</span> copy var2/xreg2 => <span class="Constant">"f3 0f 11/<- %"</span> xreg <span class="Constant">" "</span> xreg2 <span class="Constant">"/x32"</span>
copy-to var1, var2/xreg => <span class="Constant">"f3 0f 11/<- *(ebp+"</span> var1.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> copy var2 => <span class="Constant">"f3 0f 10/-> *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg <span class="Constant">"/x32"</span>
var/xreg <span class="SpecialChar"><-</span> copy *var2/reg2 => <span class="Constant">"f3 0f 10/-> *"</span> reg2 <span class="Constant">" "</span> xreg <span class="Constant">"/x32"</span>
Comparisons must always start with a register:
compare var1/xreg1, var2/xreg2 => <span class="Constant">"0f 2f/compare %"</span> xreg2 <span class="Constant">" "</span> xreg1 <span class="Constant">"/x32"</span>
compare var1/xreg1, var2 => <span class="Constant">"0f 2f/compare *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> xreg1 <span class="Constant">"/x32"</span>
<span class="muComment">## Blocks</span>
In themselves, blocks generate no instructions. However, if a block contains
variable declarations, they must be cleaned up when the block ends.
Clean up var on the stack => <span class="Constant">"81 0/subop/add %esp "</span> size-of(var) <span class="Constant">"/imm32"</span>
Clean up var/reg => <span class="Constant">"8f 0/subop/pop %"</span> reg
Clean up var/xreg => <span class="Constant">"f3 0f 10/-> *esp "</span> xreg <span class="Constant">"/x32"</span>
<span class="Constant">"81 0/subop/add %esp 4/imm32"</span>
<span class="muComment">## Jumps</span>
Besides having to clean up any variable declarations (see above) between
themselves and their target, jumps translate like this:
<span class="PreProc">break</span> => <span class="Constant">"e9/jump break/disp32"</span>
<span class="PreProc">break</span> label => <span class="Constant">"e9/jump "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop</span> => <span class="Constant">"e9/jump loop/disp32"</span>
<span class="PreProc">loop</span> label => <span class="Constant">"e9/jump "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-=</span> => <span class="Constant">"0f 84/jump-if-= break/disp32"</span>
<span class="PreProc">break-if-=</span> label => <span class="Constant">"0f 84/jump-if-= "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-=</span> => <span class="Constant">"0f 84/jump-if-= loop/disp32"</span>
<span class="PreProc">loop-if-=</span> label => <span class="Constant">"0f 84/jump-if-= "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-!=</span> => <span class="Constant">"0f 85/jump-if-!= break/disp32"</span>
<span class="PreProc">break-if-!=</span> label => <span class="Constant">"0f 85/jump-if-!= "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-!=</span> => <span class="Constant">"0f 85/jump-if-!= loop/disp32"</span>
<span class="PreProc">loop-if-!=</span> label => <span class="Constant">"0f 85/jump-if-!= "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-<</span> => <span class="Constant">"0f 8c/jump-if-< break/disp32"</span>
<span class="PreProc">break-if-<</span> label => <span class="Constant">"0f 8c/jump-if-< "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-<</span> => <span class="Constant">"0f 8c/jump-if-< loop/disp32"</span>
<span class="PreProc">loop-if-<</span> label => <span class="Constant">"0f 8c/jump-if-< "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-></span> => <span class="Constant">"0f 8f/jump-if-> break/disp32"</span>
<span class="PreProc">break-if-></span> label => <span class="Constant">"0f 8f/jump-if-> "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-></span> => <span class="Constant">"0f 8f/jump-if-> loop/disp32"</span>
<span class="PreProc">loop-if-></span> label => <span class="Constant">"0f 8f/jump-if-> "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-<=</span> => <span class="Constant">"0f 8e/jump-if-<= break/disp32"</span>
<span class="PreProc">break-if-<=</span> label => <span class="Constant">"0f 8e/jump-if-<= "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-<=</span> => <span class="Constant">"0f 8e/jump-if-<= loop/disp32"</span>
<span class="PreProc">loop-if-<=</span> label => <span class="Constant">"0f 8e/jump-if-<= "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if->=</span> => <span class="Constant">"0f 8d/jump-if->= break/disp32"</span>
<span class="PreProc">break-if->=</span> label => <span class="Constant">"0f 8d/jump-if->= "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if->=</span> => <span class="Constant">"0f 8d/jump-if->= loop/disp32"</span>
<span class="PreProc">loop-if->=</span> label => <span class="Constant">"0f 8d/jump-if->= "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-addr<</span> => <span class="Constant">"0f 82/jump-if-addr< break/disp32"</span>
<span class="PreProc">break-if-addr<</span> label => <span class="Constant">"0f 82/jump-if-addr< "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-addr<</span> => <span class="Constant">"0f 82/jump-if-addr< loop/disp32"</span>
<span class="PreProc">loop-if-addr<</span> label => <span class="Constant">"0f 82/jump-if-addr< "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-addr></span> => <span class="Constant">"0f 87/jump-if-addr> break/disp32"</span>
<span class="PreProc">break-if-addr></span> label => <span class="Constant">"0f 87/jump-if-addr> "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-addr></span> => <span class="Constant">"0f 87/jump-if-addr> loop/disp32"</span>
<span class="PreProc">loop-if-addr></span> label => <span class="Constant">"0f 87/jump-if-addr> "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-addr<=</span> => <span class="Constant">"0f 86/jump-if-addr<= break/disp32"</span>
<span class="PreProc">break-if-addr<=</span> label => <span class="Constant">"0f 86/jump-if-addr<= "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-addr<=</span> => <span class="Constant">"0f 86/jump-if-addr<= loop/disp32"</span>
<span class="PreProc">loop-if-addr<=</span> label => <span class="Constant">"0f 86/jump-if-addr<= "</span> label <span class="Constant">":loop/disp32"</span>
<span class="PreProc">break-if-addr>=</span> => <span class="Constant">"0f 83/jump-if-addr>= break/disp32"</span>
<span class="PreProc">break-if-addr>=</span> label => <span class="Constant">"0f 83/jump-if-addr>= "</span> label <span class="Constant">":break/disp32"</span>
<span class="PreProc">loop-if-addr>=</span> => <span class="Constant">"0f 83/jump-if-addr>= loop/disp32"</span>
<span class="PreProc">loop-if-addr>=</span> label => <span class="Constant">"0f 83/jump-if-addr>= "</span> label <span class="Constant">":loop/disp32"</span>
Similar float variants like `<span class="PreProc">break-if-float<`</span> are aliases for the corresponding
`addr` equivalents. The x86 instruction set stupidly has floating-point
operations only update a subset of flags.
<span class="muComment">## Returns</span>
The `<span class="PreProc">return</span>` instruction cleans up variable declarations just like an unconditional
`<span class="PreProc">jump</span>` to end of function, but also emits a series of copies before the final
`<span class="PreProc">jump</span>`, copying each argument of `<span class="PreProc">return</span>` to the register appropriate to the
respective function output. This doesn't work if a function output register
contains a later `<span class="PreProc">return</span>` argument (e.g. if the registers for two outputs are
swapped in `<span class="PreProc">return</span>`), so you can't do that.
<span class="PreProc">return</span> => <span class="Constant">"c3/return"</span>
---
In the following instructions types are provided for clarity even if they must
be provided in an earlier 'var' declaration.
<span class="muComment"># Address operations</span>
var/reg: (addr T) <span class="SpecialChar"><-</span> address var2: T
=> <span class="Constant">"8d/copy-address *(ebp+"</span> var2.stack-offset <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
<span class="muComment"># Array operations</span>
var/reg: (addr T) <span class="SpecialChar"><-</span> index arr/rega: (addr array T), idx/regi: int
| if size-of(T) is <span class="Constant">1</span>, <span class="Constant">2</span>, <span class="Constant">4</span> or <span class="Constant">8</span>
=> <span class="Constant">"(__check-mu-array-bounds *"</span> rega <span class="Constant">" %"</span> regi <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
<span class="Constant">"8d/copy-address *("</span> rega <span class="Constant">"+"</span> regi <span class="Constant">"<<"</span> log2(size-of(T)) <span class="Constant">"+4) "</span> reg <span class="Constant">"/r32"</span>
var/reg: (addr T) <span class="SpecialChar"><-</span> index arr: (array T len), idx/regi: int
=> <span class="Constant">"(__check-mu-array-bounds *(ebp+"</span> arr.stack-offset <span class="Constant">") %"</span> regi <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
<span class="Constant">"8d/copy-address *(ebp+"</span> regi <span class="Constant">"<<"</span> log2(size-of(T)) <span class="Constant">"+"</span> (arr.stack-offset + <span class="Constant">4</span>) <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg: (addr T) <span class="SpecialChar"><-</span> index arr/rega: (addr array T), n
=> <span class="Constant">"(__check-mu-array-bounds *"</span> rega <span class="Constant">" "</span> n <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
<span class="Constant">"8d/copy-address *("</span> rega <span class="Constant">"+"</span> (n*size-of(T)+<span class="Constant">4</span>) <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg: (addr T) <span class="SpecialChar"><-</span> index arr: (array T len), n
=> <span class="Constant">"(__check-mu-array-bounds *(ebp+"</span> arr.stack-offset <span class="Constant">") "</span> n <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
<span class="Constant">"8d/copy-address *(ebp+"</span> (arr.stack-offset+<span class="Constant">4</span>+n*size-of(T)) <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg: (offset T) <span class="SpecialChar"><-</span> compute-offset arr: (addr array T), idx/regi: int <span class="muComment"># arr can be in reg or mem</span>
=> <span class="Constant">"69/multiply %"</span> regi <span class="Constant">" "</span> size-of(T) <span class="Constant">"/imm32 "</span> reg <span class="Constant">"/r32"</span>
var/reg: (offset T) <span class="SpecialChar"><-</span> compute-offset arr: (addr array T), idx: int <span class="muComment"># arr can be in reg or mem</span>
=> <span class="Constant">"69/multiply *(ebp+"</span> idx.stack-offset <span class="Constant">") "</span> size-of(T) <span class="Constant">"/imm32 "</span> reg <span class="Constant">"/r32"</span>
var/reg: (addr T) <span class="SpecialChar"><-</span> index arr/rega: (addr array T), o/rego: (offset T)
=> <span class="Constant">"(__check-mu-array-bounds %"</span> rega <span class="Constant">" %"</span> rego <span class="Constant">" 1 \"" function-name "</span>\<span class="Constant">")"</span>
<span class="Constant">"8d/copy-address *("</span> rega <span class="Constant">"+"</span> rego <span class="Constant">"+4) "</span> reg <span class="Constant">"/r32"</span>
Computing the length of an array is complex.
var/reg: int <span class="SpecialChar"><-</span> length arr/reg2: (addr array T)
| if T is byte (TODO)
=> <span class="Constant">"8b/-> *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
| if size-of(T) is <span class="Constant">4</span> or <span class="Constant">8</span> or <span class="Constant">16</span> or <span class="Constant">32</span> or <span class="Constant">64</span> or <span class="Constant">128</span>
=> <span class="Constant">"8b/-> *"</span> reg2 <span class="Constant">" "</span> reg <span class="Constant">"/r32"</span>
<span class="Constant">"c1/shift 5/subop/logic-right %"</span> reg <span class="Constant">" "</span> log2(size-of(T)) <span class="Constant">"/imm8"</span>
| otherwise
x86 has no instruction to divide by a literal, so
we need up to <span class="Constant">3</span> extra registers! <span class="Constant">eax</span>/<span class="Constant">edx</span> for division and say <span class="Constant">ecx</span>
=> if reg is not <span class="Constant">eax</span>
<span class="Constant">"50/push-eax"</span>
if reg is not <span class="Constant">ecx</span>
<span class="Constant">"51/push-ecx"</span>
if reg is not <span class="Constant">edx</span>
<span class="Constant">"52/push-edx"</span>
<span class="Constant">"8b/-> *"</span> reg2 <span class="Constant">" eax/r32"</span>
<span class="Constant">"31/xor %edx 2/r32/edx"</span> <span class="muComment"># sign-extend, but array size can't be negative</span>
<span class="Constant">"b9/copy-to-ecx "</span> size-of(T) <span class="Constant">"/imm32"</span>
<span class="Constant">"f7 7/subop/idiv-eax-edx-by %ecx"</span>
if reg is not <span class="Constant">eax</span>
<span class="Constant">"89/<- %"</span> reg <span class="Constant">" 0/r32/eax"</span>
if reg is not <span class="Constant">edx</span>
<span class="Constant">"5a/pop-to-edx"</span>
if reg is not <span class="Constant">ecx</span>
<span class="Constant">"59/pop-to-ecx"</span>
if reg is not <span class="Constant">eax</span>
<span class="Constant">"58/pop-to-eax"</span>
<span class="muComment"># User-defined types</span>
If a record (product) type T was defined to have elements a, b, c, ... of
types T_a, T_b, T_c, ..., then accessing one of those elements f of type T_f:
var/reg: (addr T_f) <span class="SpecialChar"><-</span> get var2/reg2: (addr T), f
=> <span class="Constant">"8d/copy-address *("</span> reg2 <span class="Constant">"+"</span> offset(f) <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
var/reg: (addr T_f) <span class="SpecialChar"><-</span> get var2: T, f
=> <span class="Constant">"8d/copy-address *(ebp+"</span> var2.stack-offset <span class="Constant">"+"</span> offset(f) <span class="Constant">") "</span> reg <span class="Constant">"/r32"</span>
<span class="muComment"># Allocating memory</span>
allocate in: (addr handle T)
=> <span class="Constant">"(allocate Heap "</span> size-of(T) <span class="Constant">" "</span> in <span class="Constant">")"</span>
populate in: (addr handle array T), num <span class="muComment"># can be literal or variable on stack or register</span>
=> <span class="Constant">"(allocate-array2 Heap "</span> size-of(T) <span class="Constant">" "</span> num <span class="Constant">" "</span> in <span class="Constant">")"</span>
populate-stream in: (addr handle stream T), num <span class="muComment"># can be literal or variable on stack or register</span>
=> <span class="Constant">"(new-stream Heap "</span> size-of(T) <span class="Constant">" "</span> num <span class="Constant">" "</span> in <span class="Constant">")"</span>
<span class="muComment"># Some miscellaneous helpers to avoid error-prone size computations</span>
clear x: (addr T)
=> <span class="Constant">"(zero-out "</span> s <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
read-from-stream s: (addr stream T), out: (addr T)
=> <span class="Constant">"(read-from-stream "</span> s <span class="Constant">" "</span> out <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
write-to-stream s: (addr stream T), in: (addr T)
=> <span class="Constant">"(write-to-stream "</span> s <span class="Constant">" "</span> in <span class="Constant">" "</span> size-of(T) <span class="Constant">")"</span>
vim:ft=mu:nowrap:textwidth=<span class="Constant">0</span>
</pre>
</body>
</html>
<!-- vim: set foldmethod=manual : -->
|