love this and it's timely as I think with AI more of my code will just be native assembly for example: https://github.com/taf2/tic-tac-toe my m series native assembly based tic tac toe windowed game... very portable too you just simply execute codex --yolo port my app to x y z host...
I haven't played with yolo yet. I'll have to use my airgapped, old Lenovo ThinkPad for this! Whenever I see yolo, I can't stop thinking of Joseph Redmon's YOLO and his astounding resume format!
I should try FASM to see if it produces a smaller size exe. COM files are small, but 16-bit DOS was the thing back in the day - 100H yeah and basically a memory dump.
Interesting to bypass CRT and work directly against the Windows API. That said, these rely on USER.32.DLL and KERNEL32.DLL to do the majority of the heavy lifting.
It's a bit like 'smallest hello world in assembly' but the code is just setting up the calling convention and then passing the zero-terminated character string into `write`. Cool, but that could be done a bit more straightforwardly in C, too. A lot of assembly (including in the linked repository) is book-keeping for the platform's calling convention.
> no, you can also call ntdll (or even syscalls although those aren't stable), there's plenty of documentation on the internet
I mean officially. Yes, you can find some on the internet, but nothing on the official MSDN.
> but also why would you, not much point except for very niche functionality
Exactly. Using system DLL API is backward and forward compatible, and just as standardized and well-documented as the POSIX API. I see no problem relying on it.
(Sidenote: as a bonus, MSDN is surprisingly good, understandable, well organized, lots of examples. I rarely say this, but well done MS, that's how a dev doc should be.)
KERNEL32.DLL (in spite of the name) provides user-mode system services for the Windows API. The CRT provides the C99 API. There are many entry points to the same thing on Windows; you can do any of `malloc`/`HeapAlloc`/`new`/`VirtualAlloc` to get a void* memory buffer. `malloc` is from the CRT; `HeapAlloc` is from KERNEL32.DLL, and `new` is from the C++ runtime.
You know? FOR ONE expression our current neuralese-spewing friends could standarize on - I'd not mind this one (and the ilk) becoming "load-bearing", again.-
“Runs a Windows message loop
Has a title bar, minimize, maximize, and close buttons, which all work as expected
Has a system menu with the same
Paints the background and some text centered in the middle, equal to or larger than "Dave's Tiny App"”
That Linux program:
“Let’s take an incredibly simple program, one that does nothing but return a number back to the operating system”
Putting the search terms "Windows" and "assembly language" turns up all sort of books. The one that got me to initially explore the topic (from 1993).... [0]
For a more up to date treatment of assembly and windows.... comes with source code for the IDE also. Ray Seyfarth's book [1].
[0] Windows Assembly Language & Systems Programming: Object Oriented & Low-Level Systems Programming in Assembly Language for Windows 3.X
[1] Introduction to 64 Bit Windows Assembly Language Programming: Fourth Edition
ISBN-13: 978-1543138849, ISBN-10: 1543138845
"smallest possible" is exactly one instruction; possibly 0xc3 [ret] or 0x90 [nop]; the problem lies in how you define "complete Windows application", because the answer really differs depending on if you want to make it ABI compliant. But there's nothing preventing you from creating a custom loader/executor that reads a pure binary file, maps it as exec and runs it.
A .COM file is exactly that, a pure binary that just starts executing at the first byte. Mind you the loader for those has gotten a smidge bigger since its first version.
You can get a sub 20kb executable with C in MSVC and no weird tricks, i.e. just setting a particular combination of compiler/linker flags, like /NODEFAULTIB, /MERGE, /FILEALIGN:512 etc. Small enough that it's basically nothing, downloads in a couple of seconds over 56k dialup, and you get to use normal tooling. This is the compromise that makes the most sense to me, for products you're actually shipping. If you want to try out the nocrt approach, honestly the LLMs do a fine job of writing you some replacements for the parts of the crt that you likely want i.e. memcpy/memcmp/strlen helpers, with x64/arm64 SIMD and all.
Set /HEAP:4096,4096 and /STACK:65536,4096, write yourself a simple 100 line growable arena implementation, use that, avoid the heap entirely. Very quick and easy way to get a real windows program up and running that uses about 500kb of commit at baseline.
You end up with only ntdll.dll, kernelbase.dll, and kernel32.dll loaded. Trying to eliminate any of these becomes pretty painful and you venture into the territory of weird tricks that are expensive to maintain.
Then you load User32.dll, and you get 2MB of commit and 1MB of private bytes. Throw in a Message Box, and it's 272KB more private bytes and 388K more commit.
Windows 10 makes it literally impossible to write something that uses low memory and uses an actual window.
I made my own 64-byte MS-DOS stub that uses a shorter message "Win32 Only!" instead of the usual "This program can only be run in DOS mode" message. By using the shorter stub, using the secret flag to omit the Rich header, and possibly merging sections together, I can usually get the PE header to fit within 512 bytes, and the linker won't pad out the PE header to 1024 bytes.
taf2 | 16 hours ago
eggy | 16 hours ago
eggy | 16 hours ago
[OP] Bluestein | 16 hours ago
... the days! :)
WillAdams | 16 hours ago
https://www.grc.com/freepopular.htm
[OP] Bluestein | 16 hours ago
delta_p_delta_x | 16 hours ago
It's a bit like 'smallest hello world in assembly' but the code is just setting up the calling convention and then passing the zero-terminated character string into `write`. Cool, but that could be done a bit more straightforwardly in C, too. A lot of assembly (including in the linked repository) is book-keeping for the platform's calling convention.
skrellm | 16 hours ago
Since the low-level Windows kernel API is undocumented and non-public, using these system DLLs is your only option on Windows.
It's just like linking against libc.so on any POSIX systems.
Pannoniae | 13 hours ago
but also why would you, not much point except for very niche functionality
skrellm | 10 hours ago
I mean officially. Yes, you can find some on the internet, but nothing on the official MSDN.
> but also why would you, not much point except for very niche functionality
Exactly. Using system DLL API is backward and forward compatible, and just as standardized and well-documented as the POSIX API. I see no problem relying on it.
(Sidenote: as a bonus, MSDN is surprisingly good, understandable, well organized, lots of examples. I rarely say this, but well done MS, that's how a dev doc should be.)
Someone | 9 hours ago
> I mean officially. Yes, you can find some on the internet, but nothing on the official MSDN.
It seems that changed. https://learn.microsoft.com/en-us/windows/win32/devnotes/ntq... says
but it on the official MSDN site, and it does document a call in Ntdll.dll.It’s easy to find many more examples such as https://learn.microsoft.com/en-us/windows/win32/api/winternl..., so I don’t think that’s an accident.
not_a9 | 8 hours ago
> If the call to this function occurs in user mode, you should use the name "NtMapViewOfSection" instead of "ZwMapViewOfSection".
munchler | 16 hours ago
actionfromafar | 15 hours ago
neonz80 | 15 hours ago
actionfromafar | 15 hours ago
delta_p_delta_x | 15 hours ago
haunter | 16 hours ago
[OP] Bluestein | 15 hours ago
p00dles | 15 hours ago
[OP] Bluestein | 15 hours ago
PS. Honest take :)
wolfi1 | 13 hours ago
Someone | 15 hours ago
“Runs a Windows message loop Has a title bar, minimize, maximize, and close buttons, which all work as expected Has a system menu with the same Paints the background and some text centered in the middle, equal to or larger than "Dave's Tiny App"”
That Linux program:
“Let’s take an incredibly simple program, one that does nothing but return a number back to the operating system”
See https://archive.is/w01DO#selection-265.0-265.44 for a fairer comparison (97/133 bytes)
WalterGR | 14 hours ago
ggerules | 14 hours ago
Putting the search terms "Windows" and "assembly language" turns up all sort of books. The one that got me to initially explore the topic (from 1993).... [0]
For a more up to date treatment of assembly and windows.... comes with source code for the IDE also. Ray Seyfarth's book [1].
[0] Windows Assembly Language & Systems Programming: Object Oriented & Low-Level Systems Programming in Assembly Language for Windows 3.X
[1] Introduction to 64 Bit Windows Assembly Language Programming: Fourth Edition ISBN-13: 978-1543138849, ISBN-10: 1543138845
hnea3ekp5i | 14 hours ago
abcd_f | 12 hours ago
smokel | 8 hours ago
And wouldn't a single instruction such as NOP or HLT be simply 1 byte?
AdieuToLogic | 5 hours ago
> Somewhat more meaningful was "INT 19h" (CD 19) ...
I seem to remember the smallest MS-DOS program to be CD 20h saved as a `.com`.
rfgplk | 12 hours ago
QuantumNomad_ | 11 hours ago
chuckadams | 10 hours ago
p0w3n3d | 10 hours ago
ntstatusquo | 9 hours ago
Set /HEAP:4096,4096 and /STACK:65536,4096, write yourself a simple 100 line growable arena implementation, use that, avoid the heap entirely. Very quick and easy way to get a real windows program up and running that uses about 500kb of commit at baseline.
You end up with only ntdll.dll, kernelbase.dll, and kernel32.dll loaded. Trying to eliminate any of these becomes pretty painful and you venture into the territory of weird tricks that are expensive to maintain.
Dave's approach is a fun exercise though.
Dwedit | 6 hours ago
Windows 10 makes it literally impossible to write something that uses low memory and uses an actual window.
Dwedit | 6 hours ago
A copy can be found at https://github.com/Dwedit/NoCopilotKey/blob/main/stub.bin