Skip to content

Instantly share code, notes, and snippets.

@nikAizuddin

nikAizuddin/FPU instructions example (NASM, for Linux)

Last active March 1, 2024 12:33
Show Gist options
  • Save nikAizuddin/0e307cac142792dcdeba to your computer and use it in GitHub Desktop.
Save nikAizuddin/0e307cac142792dcdeba to your computer and use it in GitHub Desktop.
Download ZIP
Raw
FPU instructions example (NASM, for Linux)
; 1 2 3 4 5 6 7
;01234567890123456789012345678901234567890123456789012345678901234567890
;=======================================================================
;+---------------------------------------------------------------------+
;| |
;| Example using FPU registers for floating point calculations. |
;| |
;| The purpose of this source code is to demonstrate on how to |
;| do floating-point operations using FPU registers. |
;| |
;| To understand this code, make sure you know IEEE-754 Standard |
;| and architecture of FPU registers, especially why it is called |
;| FPU stack. |
;| |
;| When debugging with GNU GDB, make sure you specify the type of |
;| architecture by executing the following command in GDB: |
;| |
;| (gdb) set architecture i386 |
;| |
;| so that the GDB will display the correct value when using |
;| commands: |
;| |
;| (gdb) x/f &single_sum |
;| (gdb) x/fg &double_sum |
;| |
;| to display the floating-point value from memory. |
;| |
;| There are 11 examples in this source code: |
;| example_1: Moving single-precision value to FPU stack |
;| example_2: Moving double-precision value to FPU stack |
;| example_3: Moving extended-precision value to FPU stack |
;| example_4: Add Operation (single + single) |
;| example_5: Add Operation (single + single + single) |
;| example_6: Add Operation (single + double + extended) |
;| example_7: Subtract Operation (single - single) |
;| example_8: Multiply Operation (single * single) |
;| example_9: Divide Operation (single / single) |
;| example_10: Convert single to double |
;| example_11: Compare float numbers (Find the largest value) |
;| |
;| Use these commands to assemble this code: |
;| |
;| $ nasm _start.asm -f elf32 -o _start.o |
;| $ ld _start.o -m elf_i386 -o exe |
;| |
;+---------------------------------------------------------------------+
;| AUTHOR: Nik Mohamad Aizuddin bin Nik Azmi |
;| EMAIL: [email protected] |
;| DATE CREATED: 21/OCT/2014 |
;+---------------------------------------------------------------------+
;| LANGUAGE: x86 Assembly Language |
;| ASSEMBLER: NASM |
;| SYNTAX: Intel |
;| ARCHITECTURE: i386 |
;| KERNEL: Linux 32-bit |
;| FORMAT: ELF32 |
;+---------------------------------------------------------------------+
;| REVISION HISTORY: |
;| |
;| Rev # | Date | Description |
;| -------+-------------+--------------------------------------------- |
;| 1.0.0 | 24/OCT/2014 | First release. |
;| 1.0.1 | 26/OCT/2014 | Fix st0 := st1 <operand> st0 |
;| |
;+---------------------------------------------------------------------+
;| Do whatever you want with this source code :) |
;+---------------------------------------------------------------------+
;=======================================================================
;---- configurations ---------------------------------------------------
[bits 32] ;output format is 32-bit code
cpu 386 ;assemble instructions up to the 386 instruction set only
;---- section unintialized data ----------------------------------------
section .bss noexec write align=4
single_sum: resd 1
single_sub: resd 1
single_mul: resd 1
single_div: resd 1
single2double: resq 1
double_sum: resq 1
;---- section read/write data ------------------------------------------
section .data noexec write align=4
;---- section read-only data -------------------------------------------
section .rodata noexec nowrite align=4
single_value1: dd 12.34
single_value2: dd 102.35
single_value3: dd -52.02
double_value1: dq 12.34
double_value2: dq 102.35
double_value3: dq -52.02
extended_value1: dt 12.34
extended_value2: dt 102.35
extended_value3: dt -52.02
string_1_begin:
db "example_11: " ;where msg belong to?
db "single_value1 is more than single_value2" ;the msg
db 0aH ;newline character
db 00H ;null str terminator
string_1_end:
string_2_begin:
db "example_11: " ;where msg belong to?
db "single_value1 is less than single_value2" ;the msg
db 0aH ;newline character
db 00H ;null str terminator
string_2_end:
string_3_begin:
db "example_11: " ;where msg belong to?
db "single_value1 is equal to single_value2" ;the msg
db 0aH ;newline character
db 00H ;null str terminator
string_3_end:
string_4_begin:
db "example_11 ERROR! " ;where msg belong to?
db "st0 and source are undefined!" ;the msg
db 0aH ;newline character
db 00H ;null str terminator
string_4_end:
;---- section instruction codes ----------------------------------------
section .text exec nowrite align=16
global _start:function
_start:
.example_1:
;+---------------------------------------------------------------------+
;| Moving single-precision value to FPU stack. |
;| |
;| This example 1 will show how to move/copy the single-precision |
;| value from memory to fpu stacks. |
;| |
;| This example will perform these 3 instructions: |
;| STEP 1: push single_value1 to fpu stack |
;| STEP 2: push single_value2 to fpu stack |
;| STEP 3: push single_value3 to fpu stack |
;| |
;| Below are the observations on FPU stacks during these 3 steps: |
;| AT STEP 1: st0 = single_value1 |
;| st1 = 0 |
;| st2 = 0 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| AT STEP 2: st0 = single_value2 |
;| st1 = single_value1 |
;| st2 = 0 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| AT STEP 3: st0 = single_value3 |
;| st1 = single_value2 |
;| st2 = single_value1 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| Note: Before using fpu stacks, make sure to reset them to their |
;| default value. This practice can prevent bugs in the program. |
;+---------------------------------------------------------------------+
finit ;reset fpu registers to default
fld dword [single_value1] ;push value 12.34 to fpu stack
fld dword [single_value2] ;push value 102.35 to fpu stack
fld dword [single_value3] ;push value -52.02 to fpu stack
.example_2:
;+---------------------------------------------------------------------+
;| Moving double-precision value to FPU stack. |
;| |
;| This example 2 will show how to move/copy the double-precision |
;| value from memory to fpu stacks. |
;| |
;| This example will perform these 3 instructions: |
;| STEP 1: push double_value1 to fpu stack |
;| STEP 2: push double_value2 to fpu stack |
;| STEP 3: push double_value3 to fpu stack |
;| |
;| Below are the observations on FPU stacks during these 3 steps: |
;| AT STEP 1: st0 = double_value1 |
;| st1 = 0 |
;| st2 = 0 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| AT STEP 2: st0 = double_value2 |
;| st1 = double_value1 |
;| st2 = 0 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| AT STEP 3: st0 = double_value3 |
;| st1 = double_value2 |
;| st2 = double_value1 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| Note: Before using fpu stacks, make sure to reset them to their |
;| default value. This practice can prevent bugs in the program. |
;+---------------------------------------------------------------------+
finit ;reset fpu registers to default
fld qword [double_value1] ;push value 12.34 to fpu stack
fld qword [double_value2] ;push value 102.35 to fpu stack
fld qword [double_value3] ;push value -52.02 to fpu stack
.example_3:
;+---------------------------------------------------------------------+
;| Moving extended-precision value to FPU stack. |
;| |
;| This example 2 will show how to move/copy the extended-precision |
;| value from memory to fpu stacks. |
;| |
;| This example will perform these 3 instructions: |
;| STEP 1: push extended_value1 to fpu stack |
;| STEP 2: push extended_value2 to fpu stack |
;| STEP 3: push extended_value3 to fpu stack |
;| |
;| Below are the observations on FPU stacks during these 3 steps: |
;| AT STEP 1: st0 = extended_value1 |
;| st1 = 0 |
;| st2 = 0 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| AT STEP 2: st0 = extended_value2 |
;| st1 = extended_value1 |
;| st2 = 0 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| AT STEP 3: st0 = extended_value3 |
;| st1 = extended_value2 |
;| st2 = extended_value1 |
;| st4 = 0 |
;| st5 = 0 |
;| st6 = 0 |
;| st7 = 0 |
;| |
;| Note: Before using fpu stacks, make sure to reset them to their |
;| default value. This practice can prevent bugs in the program. |
;+---------------------------------------------------------------------+
finit ;reset fpu registers to default
fld tword [extended_value1] ;push value 12.34 to fpu stack
fld tword [extended_value2] ;push value 102.35 to fpu stack
fld tword [extended_value3] ;push value -52.02 to fpu stack
.example_4:
;+---------------------------------------------------------------------+
;| Add Operation (single + single) |
;| |
;| The following equation will be executed in this example_4: |
;| |
;| p = x + y |
;| |
;| where, |
;| x = single_value1 |
;| y = single_value2 |
;| p = single_sum |
;| |
;| Behaviour of fpu stacks before and after FAdd: |
;| Before fadd: st0 = 102.34999847412109375 |
;| st1 = 12.340000152587890625 |
;| |
;| After fadd: st0 = 114.689998626708984375 |
;| st1 = 0 |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack
fld dword [single_value2] ;push single_value2 to fpu stack
fadd ;st0 := st1 + st0
fstp dword [single_sum] ;store the summation result into mem
.example_5:
;+---------------------------------------------------------------------+
;| Add Operation (single + single + single) |
;| |
;| The following equation will be executed in this example_5: |
;| |
;| p = x + y + z |
;| |
;| where, |
;| x = single_value1 |
;| y = single_value2 |
;| z = single_value3 |
;| p = single_sum |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack
fld dword [single_value2] ;push single_value2 to fpu stack
fadd ;st0 := st1 + st0
fld dword [single_value3] ;push single_value3 to fpu stack
fadd ;st0 := st1 + st0
fstp dword [single_sum] ;store the summation result into mem
.example_6:
;+---------------------------------------------------------------------+
;| Add Operation (single + double + extended) |
;| |
;| The following equation will be executed in this example_6: |
;| |
;| p = x + y + z |
;| |
;| where, |
;| x = single_value1 |
;| y = double_value2 |
;| z = extended_value3 |
;| p = double_sum |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack
fld qword [double_value2] ;push double_value2 to fpu stack
fadd ;st0 := st1 + st0
fld tword [extended_value3] ;push extended_value3 to fpu stack
fadd ;st0 := st1 + st0
fstp qword [double_sum] ;store the summation result into mem
.example_7:
;+---------------------------------------------------------------------+
;| Subtract Operation (single - single) |
;| |
;| The following equation will be executed in this example_7: |
;| |
;| p = x - y |
;| |
;| where, |
;| x = single_value1 |
;| y = single_value2 |
;| p = single_sum |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack(st1)
fld dword [single_value2] ;push single_value2 to fpu stack(st0)
fsub ;st0 := st1 - st0
fstp dword [single_sub] ;store the subtraction result into mem
.example_8:
;+---------------------------------------------------------------------+
;| Multiply Operation (single * single) |
;| |
;| The following equation will be executed in this example_8: |
;| |
;| p = x * y |
;| |
;| where, |
;| x = single_value1 |
;| y = single_value2 |
;| p = single_mul |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack(st1)
fld dword [single_value2] ;push single_value2 to fpu stack(st0)
fmul ;st0 := st1 * st0
fstp dword [single_mul] ;store the multiplctn. result into mem
.example_9:
;+---------------------------------------------------------------------+
;| Divide Operation (single / single) |
;| |
;| The following equation will be executed in this example_9: |
;| |
;| p = x / y |
;| |
;| where, |
;| x = single_value1 |
;| y = single_value2 |
;| p = single_div |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack(st1)
fld dword [single_value2] ;push single_value2 to fpu stack(st0)
fdiv ;st0 := st1 / st0
fstp dword [single_div] ;store the division result into mem
.example_10:
;+---------------------------------------------------------------------+
;| Convert single to double |
;| |
;| This example shows how to convert/promote single to double. |
;| These steps will be executed for conversion: |
;| STEP 1: Push single_value1 to FPU stack |
;| STEP 2: Store the double value of single_value1 to memory |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value1] ;push single_value1 to fpu stack(st0)
fstp qword [single2double] ;store the double value into memory
.example_11:
;+---------------------------------------------------------------------+
;| Compare float numbers (Find the largest value) |
;| |
;| This example shows how to compare 2 floating-point numbers. |
;| |
;| Given single_value1 and single_value2, |
;| If single_value1 > single_value2 then |
;| Print "single_value1 is bigger than single_value2" |
;| Else If single_value1 < single_value2 Then |
;| Print "single_value1 is less than single_value2" |
;| Else |
;| Print "single_value1 is equal to single_value2" |
;| |
;| |
;| The picture below shows |
;| FPU Status Word 16-bit register (stores general condition of FPU): |
;| |
;| +-----+ |
;| Bit 0 | IE | --> Invalid operation exception |
;| +-----+ |
;| Bit 1 | DE | --> Denormalized exception |
;| +-----+ |
;| Bit 2 | ZE | --> Zero divide exception |
;| +-----+ |
;| Bit 3 | OE | --> Overflow exception |
;| +-----+ |
;| Bit 4 | UE | --> Underflow exception |
;| +-----+ |
;| Bit 5 | PE | --> Precision exception |
;| +-----+ |
;| Bit 6 | SF | --> Stack Fault exception |
;| +-----+ |
;| Bit 7 | IR | --> Interupt Request |
;| +-----+ |
;| Bit 8 | C0 | --> CF (carry flag) [CONDITION CODE] |
;| +-----+ |
;| Bit 9 | C1 | --> =1 (stack overflow), =0 (stack underflow) |
;| +-----+ |
;| Bit 10 | C2 | --> PF (parity flag) [CONDITION CODE] |
;| +-----+ |
;| Bit 11 | | |
;| Bit 12 | TOP | --> Top of stack pointer |
;| Bit 13 | | |
;| +-----+ |
;| Bit 14 | C3 | --> ZF (zero flag) [CONDITION CODE] |
;| +-----+ |
;| Bit 15 | B | --> B=1 (busy), B=0 (idle) |
;| +-----+ |
;| |
;| This table shows FPU Condition Code Bits: |
;| +-------------+-------------------+-------------------------------+ |
;| | |Condition Code bits| | |
;| | Instruction +----+----+----+----+ Condition | |
;| | | C3 | C2 | C1 | C0 | | |
;| +-------------+----+----+----+----+-------------------------------+ |
;| | fcom, | 0 | 0 | | 0 | st0 > source | |
;| | fcomp, +----+----+----+----+-------------------------------+ |
;| | fcompp, | 0 | 0 | | 1 | st0 < source | |
;| | ficom, +----+----+----+----+-------------------------------+ |
;| | ficomp | 1 | 0 | | 0 | st0 = source | |
;| | +----+----+----+----+-------------------------------+ |
;| | | 1 | 1 | | 1 | st0 or source are undefined | |
;| +-------------+----+----+----+----+-------------------------------+ |
;| reference: http://www.plantation-productions.com/Webster/ |
;| www.artofasm.com/Linux/HTML/RealArithmetic.html |
;| |
;+---------------------------------------------------------------------+
finit ;reset fpu stacks to default
fld dword [single_value2] ;push single_value2 to fpu stack(st1)
fld dword [single_value1] ;push single_value1 to fpu stack(st0)
fcom st0, st1 ;compare st0 with st1
fstsw ax ;ax := fpu status register
and eax, 0100011100000000B ;take only condition code flags
cmp eax, 0000000000000000B ;is st0 > source ?
je .example_11_greater
cmp eax, 0000000100000000B ;is st0 < source ?
je .example_11_less
cmp eax, 0100000000000000B ;is st0 = source ?
je .example_11_equal
jmp .example_11_error ;else, st0 or source are undefined
.example_11_greater:
mov ecx, string_1_begin ;ecx := addr string_1
mov edx, (string_1_end - string_1_begin) ;edx := length string_1
jmp .example_11_write
.example_11_less:
mov ecx, string_2_begin ;ecx := addr string_2
mov edx, (string_2_end - string_2_begin) ;edx := length string_2
jmp .example_11_write
.example_11_equal:
mov ecx, string_3_begin ;ecx := addr string_3
mov edx, (string_3_end - string_3_begin) ;edx := length string_3
jmp .example_11_write
.example_11_error:
mov ecx, string_4_begin ;ecx := addr string_4
mov edx, (string_4_end - string_4_begin) ;edx := length string_4
.example_11_write:
mov eax, 04H ;systemcall write
mov ebx, 01H ;write to stdout
int 80H
;+---------------------------------------------------------------------+
;| Thats all folks :) Now exit the program! |
;+---------------------------------------------------------------------+
.exit:
mov eax, 01H ;systemcall exit
mov ebx, 00H ;return 0
int 80H
@pointbazaar
Copy link

Thanks :)

@frogmasta
Copy link

Very helpful!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment