akkartik
/
mu
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
mu/mu_instructions

191 lines
11 KiB
Plaintext
Raw Normal View History

5966 - document all supported Mu instructions
2020-02-01 02:39:27 +00:00
## Mu's instructions and their table-driven translation
Mu is a statement-oriented language. Blocks consist of flat lists of instructions.
6159
2020-03-22 00:25:52 +00:00
The following chart shows all the instruction forms supported by Mu, along
with the instruction they're translated to. Variables of the form "var/reg"
live in a register, and other variables are assumed to live on the stack at
some 'stack-offset' from ebp.
var/eax <- increment => "40/increment-eax"
var/ecx <- increment => "41/increment-ecx"
var/edx <- increment => "42/increment-edx"
var/ebx <- increment => "43/increment-ebx"
var/esi <- increment => "46/increment-esi"
var/edi <- increment => "47/increment-edi"
increment var => "ff 0/subop/increment *(ebp+" var.stack-offset ")"
increment *var/reg => "ff 0/subop/increment *" reg
var/eax <- decrement => "48/decrement-eax"
var/ecx <- decrement => "49/decrement-ecx"
var/edx <- decrement => "4a/decrement-edx"
var/ebx <- decrement => "4b/decrement-ebx"
var/esi <- decrement => "4e/decrement-esi"
var/edi <- decrement => "4f/decrement-edi"
decrement var => "ff 1/subop/decrement *(ebp+" var.stack-offset ")"
decrement *var/reg => "ff 1/subop/decrement *" reg
var/reg <- add var2/reg2 => "01/add-to %" reg " " reg2 "/r32"
var/reg <- add var2 => "03/add *(ebp+" var2.stack-offset ") " reg "/r32"
var/reg <- add *var2/reg2 => "03/add *" reg2 " " reg "/r32"
add-to var1, var2/reg => "01/add-to *(ebp+" var1.stack-offset ") " reg "/r32"
var/eax <- add n => "05/add-to-eax " n "/imm32"
var/reg <- add n => "81 0/subop/add %" reg " " n "/imm32"
add-to var, n => "81 0/subop/add *(ebp+" var.stack-offset ") " n "/imm32"
add-to *var/reg, n => "81 0/subop/add *" reg " " n "/imm32"
var/reg <- subtract var2/reg2 => "29/subtract-from %" reg " " reg2 "/r32"
var/reg <- subtract var2 => "2b/subtract *(ebp+" var2.stack-offset ") " reg "/r32"
var/reg <- subtract *var2/reg2 => "2b/subtract *" reg2 " " reg1 "/r32"
subtract-from var1, var2/reg2 => "29/subtract-from *(ebp+" var1.stack-offset ") " reg2 "/r32"
var/eax <- subtract n => "2d/subtract-from-eax " n "/imm32"
var/reg <- subtract n => "81 5/subop/subtract %" reg " " n "/imm32"
subtract-from var, n => "81 5/subop/subtract *(ebp+" var.stack-offset ") " n "/imm32"
subtract-from *var/reg, n => "81 5/subop/subtract *" reg " " n "/imm32"
var/reg <- and var2/reg2 => "21/and-with %" reg " " reg2 "/r32"
var/reg <- and var2 => "23/and *(ebp+" var2.stack-offset " " reg "/r32"
var/reg <- and *var2/reg2 => "23/and *" reg2 " " reg "/r32"
and-with var1, var2/reg => "21/and-with *(ebp+" var1.stack-offset ") " reg "/r32"
var/eax <- and n => "25/and-with-eax " n "/imm32"
var/reg <- and n => "81 4/subop/and %" reg " " n "/imm32"
and-with var, n => "81 4/subop/and *(ebp+" var.stack-offset ") " n "/imm32"
and-with *var/reg, n => "81 4/subop/and *" reg " " n "/imm32"
var/reg <- or var2/reg2 => "09/or-with %" reg " " reg2 "/r32"
var/reg <- or var2 => "0b/or *(ebp+" var2.stack-offset ") " reg "/r32"
var/reg <- or *var2/reg2 => "0b/or *" reg2 " " reg "/r32"
or-with var1, var2/reg2 => "09/or-with *(ebp+" var1.stack-offset " " reg2 "/r32"
var/eax <- or n => "0d/or-with-eax " n "/imm32"
var/reg <- or n => "81 1/subop/or %" reg " " n "/imm32"
or-with var, n => "81 1/subop/or *(ebp+" var.stack-offset ") " n "/imm32"
or-with *var/reg, n => "81 1/subop/or *" reg " " n "/imm32"
var/reg <- xor var2/reg2 => "31/xor-with %" reg " " reg2 "/r32"
var/reg <- xor var2 => "33/xor *(ebp+" var2.stack-offset ") " reg "/r32"
var/reg <- xor *var2/reg2 => "33/xor *" reg2 " " reg "/r32"
xor-with var1, var2/reg => "31/xor-with *(ebp+" var1.stack-offset ") " reg "/r32"
var/eax <- xor n => "35/xor-with-eax " n "/imm32"
var/reg <- xor n => "81 6/subop/xor %" reg " " n "/imm32"
xor-with var, n => "81 6/subop/xor *(ebp+" var.stack-offset ") " n "/imm32"
xor-with *var/reg, n => "81 6/subop/xor *" reg " " n "/imm32"
var/eax <- copy n => "b8/copy-to-eax " n "/imm32"
var/ecx <- copy n => "b9/copy-to-ecx " n "/imm32"
var/edx <- copy n => "ba/copy-to-edx " n "/imm32"
var/ebx <- copy n => "bb/copy-to-ebx " n "/imm32"
var/esi <- copy n => "be/copy-to-esi " n "/imm32"
var/edi <- copy n => "bf/copy-to-edi " n "/imm32"
var/reg <- copy var2/reg2 => "89/<- %" reg " " reg2 "/r32"
copy-to var1, var2/reg => "89/<- *(ebp+" var1.stack-offset ") " reg "/r32"
var/reg <- copy var2 => "8b/-> *(ebp+" var2.stack-offset ") " reg "/r32"
var/reg <- copy *var2/reg2 => "8b/-> *" reg2 " " reg "/r32"
var/reg <- copy n => "c7 0/subop/copy %" reg " " n "/imm32"
copy-to var, n => "c7 0/subop/copy *(ebp+" var.stack-offset ") " n "/imm32"
copy-to *var/reg, n => "c7 0/subop/copy *" reg " " n "/imm32"
compare var1, var2/reg2 => "39/compare *(ebp+" var1.stack-offset ") " reg2 "/r32"
compare *var1/reg1, var2/reg2 => "39/compare *" reg1 " " reg2 "/r32"
compare var1/reg1, var2 => "3b/compare<- *(ebp+" var2.stack-offset ") " reg1 "/r32"
compare var/reg, *var2/reg2 => "3b/compare<- *" reg " " n "/imm32"
compare var/eax, n => "3d/compare-eax-with " n "/imm32"
compare var, n => "81 7/subop/compare *(ebp+" var.stack-offset ") " n "/imm32"
compare *var/reg, n => "81 7/subop/compare *" reg " " n "/imm32"
var/reg <- multiply var2 => "0f af/multiply *(ebp+" var2.stack-offset ") " reg "/r32"
var/reg <- multiply *var2/reg2 => "0f af/multiply *" reg2 " " reg "/r32"
break => "e9/jump break/disp32"
break label => "e9/jump " label ":break/disp32"
loop => "e9/jump loop/disp32"
loop label => "e9/jump " label ":loop/disp32"
break-if-= => "0f 84/jump-if-= break/disp32"
break-if-= label => "0f 84/jump-if-= " label ":break/disp32"
loop-if-= => "0f 84/jump-if-= loop/disp32"
loop-if-= label => "0f 84/jump-if-= " label ":loop/disp32"
break-if-!= => "0f 85/jump-if-!= break/disp32"
break-if-!= label => "0f 85/jump-if-!= " label ":break/disp32"
loop-if-!= => "0f 85/jump-if-!= loop/disp32"
loop-if-!= label => "0f 85/jump-if-!= " label ":loop/disp32"
break-if-< => "0f 8c/jump-if-< break/disp32"
break-if-< label => "0f 8c/jump-if-< " label ":break/disp32"
loop-if-< => "0f 8c/jump-if-< loop/disp32"
loop-if-< label => "0f 8c/jump-if-< " label ":loop/disp32"
break-if-> => "0f 8f/jump-if-> break/disp32"
break-if-> label => "0f 8f/jump-if-> " label ":break/disp32"
loop-if-> => "0f 8f/jump-if-> loop/disp32"
loop-if-> label => "0f 8f/jump-if-> " label ":loop/disp32"
break-if-<= => "0f 8e/jump-if-<= break/disp32"
break-if-<= label => "0f 8e/jump-if-<= " label ":break/disp32"
loop-if-<= => "0f 8e/jump-if-<= loop/disp32"
loop-if-<= label => "0f 8e/jump-if-<= " label ":loop/disp32"
break-if->= => "0f 8d/jump-if->= break/disp32"
break-if->= label => "0f 8d/jump-if->= " label ":break/disp32"
loop-if->= => "0f 8d/jump-if->= loop/disp32"
loop-if->= label => "0f 8d/jump-if->= " label ":loop/disp32"
break-if-addr< => "0f 82/jump-if-addr< break/disp32"
break-if-addr< label => "0f 82/jump-if-addr< " label ":break/disp32"
loop-if-addr< => "0f 82/jump-if-addr< loop/disp32"
loop-if-addr< label => "0f 82/jump-if-addr< " label ":loop/disp32"
break-if-addr> => "0f 87/jump-if-addr> break/disp32"
break-if-addr> label => "0f 87/jump-if-addr> " label ":break/disp32"
loop-if-addr> => "0f 87/jump-if-addr> loop/disp32"
loop-if-addr> label => "0f 87/jump-if-addr> " label ":loop/disp32"
break-if-addr<= => "0f 86/jump-if-addr<= break/disp32"
break-if-addr<= label => "0f 86/jump-if-addr<= " label ":break/disp32"
loop-if-addr<= => "0f 86/jump-if-addr<= loop/disp32"
loop-if-addr<= label => "0f 86/jump-if-addr<= " label ":loop/disp32"
break-if-addr>= => "0f 83/jump-if-addr>= break/disp32"
break-if-addr>= label => "0f 83/jump-if-addr>= " label ":break/disp32"
loop-if-addr>= => "0f 83/jump-if-addr>= loop/disp32"
loop-if-addr>= label => "0f 83/jump-if-addr>= " label ":loop/disp32"
In the following instructions types are provided for clarity even if they must
be provided in an earlier 'var' declaration.
5968
2020-02-01 20:14:12 +00:00
6145 - 'address' operator This could be a can of worms, but I think I have a set of checks that will keep use of addresses type-safe.
2020-03-14 21:46:45 +00:00
Address operations
var/reg: (addr T) <- address var: T
6159
2020-03-22 00:25:52 +00:00
=> "8d/copy-address *(ebp+" var.stack-offset ") " reg "/r32"
6145 - 'address' operator This could be a can of worms, but I think I have a set of checks that will keep use of addresses type-safe.
2020-03-14 21:46:45 +00:00
6041 - array indexing starting to work And we're using it now in factorial.mu! In the process I had to fix a couple of bugs in pointer dereferencing. There are still some limitations: a) Indexing by a literal doesn't work yet. b) Only arrays of ints supported so far. Looking ahead, I'm not sure how I can support indexing arrays by non-literals (variables in registers) unless the element size is a power of 2.
2020-02-21 18:05:04 +00:00
Array operations
6094 - new 'compute-offset' instruction If indexing into a type with power-of-2-sized elements we can access them in one instruction: x/reg1: (addr int) <- index A/reg2: (addr array int), idx/reg3: int This translates to a single instruction because x86 instructions support an addressing mode with left-shifts. For non-powers-of-2, however, we need a multiply. To keep things type-safe, it is performed like this: x/reg1: (offset T) <- compute-offset A: (addr array T), idx: int y/reg2: (addr T) <- index A, x An offset is just an int that is guaranteed to be a multiple of size-of(T). Offsets can only be used in index instructions, and the types will eventually be required to line up. In the process, I have to expand Input-size because mu.subx is growing big.
2020-03-08 01:32:39 +00:00
var/reg <- length arr/reg2: (addr array T)
6159
2020-03-22 00:25:52 +00:00
=> "8b/-> *" reg2 " " reg "/r32"
var/reg <- index arr/rega: (addr array T), idx/regi: int # if size(T) is 4 or 8
=> "8d/copy-address *(" rega "+" regi "<<" log2(size(T)) "+4) " reg "/r32"
6131 - operating on arrays on the stack
2020-03-12 07:17:35 +00:00
var/reg <- index arr: (array T sz), idx/regi: int
6159
2020-03-22 00:25:52 +00:00
=> "8d/copy-address *(ebp+" regi "<<" log2(size(T)) "+" (arr.stack-offset + 4) ") " reg "/r32"
6041 - array indexing starting to work And we're using it now in factorial.mu! In the process I had to fix a couple of bugs in pointer dereferencing. There are still some limitations: a) Indexing by a literal doesn't work yet. b) Only arrays of ints supported so far. Looking ahead, I'm not sure how I can support indexing arrays by non-literals (variables in registers) unless the element size is a power of 2.
2020-02-21 18:05:04 +00:00
var/reg <- index arr/rega: (addr array T), n
6159
2020-03-22 00:25:52 +00:00
=> "8d/copy-address *(" rega "+" (n*size(T)+4) ") " reg "/r32"
6131 - operating on arrays on the stack
2020-03-12 07:17:35 +00:00
var/reg <- index arr: (array T sz), n
6159
2020-03-22 00:25:52 +00:00
=> "8d/copy-address *(ebp+" (arr.stack-offset+4+n*size(T)) ") " reg "/r32"
6094 - new 'compute-offset' instruction If indexing into a type with power-of-2-sized elements we can access them in one instruction: x/reg1: (addr int) <- index A/reg2: (addr array int), idx/reg3: int This translates to a single instruction because x86 instructions support an addressing mode with left-shifts. For non-powers-of-2, however, we need a multiply. To keep things type-safe, it is performed like this: x/reg1: (offset T) <- compute-offset A: (addr array T), idx: int y/reg2: (addr T) <- index A, x An offset is just an int that is guaranteed to be a multiple of size-of(T). Offsets can only be used in index instructions, and the types will eventually be required to line up. In the process, I have to expand Input-size because mu.subx is growing big.
2020-03-08 01:32:39 +00:00
var/reg: (offset T) <- compute-offset arr: (addr array T), idx/regi: int # arr can be in reg or mem
6159
2020-03-22 00:25:52 +00:00
=> "69/multiply %" regi " " size(T) "/imm32 " reg "/r32"
6094 - new 'compute-offset' instruction If indexing into a type with power-of-2-sized elements we can access them in one instruction: x/reg1: (addr int) <- index A/reg2: (addr array int), idx/reg3: int This translates to a single instruction because x86 instructions support an addressing mode with left-shifts. For non-powers-of-2, however, we need a multiply. To keep things type-safe, it is performed like this: x/reg1: (offset T) <- compute-offset A: (addr array T), idx: int y/reg2: (addr T) <- index A, x An offset is just an int that is guaranteed to be a multiple of size-of(T). Offsets can only be used in index instructions, and the types will eventually be required to line up. In the process, I have to expand Input-size because mu.subx is growing big.
2020-03-08 01:32:39 +00:00
var/reg: (offset T) <- compute-offset arr: (addr array T), idx: int # arr can be in reg or mem
6159
2020-03-22 00:25:52 +00:00
=> "69/multiply *(ebp+" idx.stack-offset ") " size(T) "/imm32 " reg "/r32"
6094 - new 'compute-offset' instruction If indexing into a type with power-of-2-sized elements we can access them in one instruction: x/reg1: (addr int) <- index A/reg2: (addr array int), idx/reg3: int This translates to a single instruction because x86 instructions support an addressing mode with left-shifts. For non-powers-of-2, however, we need a multiply. To keep things type-safe, it is performed like this: x/reg1: (offset T) <- compute-offset A: (addr array T), idx: int y/reg2: (addr T) <- index A, x An offset is just an int that is guaranteed to be a multiple of size-of(T). Offsets can only be used in index instructions, and the types will eventually be required to line up. In the process, I have to expand Input-size because mu.subx is growing big.
2020-03-08 01:32:39 +00:00
var/reg <- index arr/rega: (addr array T), o/rego: offset
6159
2020-03-22 00:25:52 +00:00
=> "8d/copy-address *(" rega "+" rego "+4) " reg "/r32"
6041 - array indexing starting to work And we're using it now in factorial.mu! In the process I had to fix a couple of bugs in pointer dereferencing. There are still some limitations: a) Indexing by a literal doesn't work yet. b) Only arrays of ints supported so far. Looking ahead, I'm not sure how I can support indexing arrays by non-literals (variables in registers) unless the element size is a power of 2.
2020-02-21 18:05:04 +00:00
6059
2020-02-28 00:58:16 +00:00
User-defined types
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:
6159
2020-03-22 00:25:52 +00:00
var/reg: (addr T_f) <- get var2/reg2: (addr F), f
=> "8d/copy-address *(" reg2 "+" offset(f) ") " reg "/r32"
var/reg: (addr T_f) <- get var2: (addr F), f
=> "8d/copy-address *(ebp+" var2.stack-offset "+" offset(f) ") " reg "/r32"
6059
2020-02-28 00:58:16 +00:00
6159
2020-03-22 00:25:52 +00:00
vim:ft=mu:nowrap:tw&