Globals and start.asm #19

2026-09-06T12:15:42+02:00

Gelthor commented

2026-09-06 12:15:42 +02:00

Added support for global variables.

This requires including start.asm as the first code. This includes a jmp to 0x100 for user code, and a magic number that modules can check to verify that the user has correctly reserved the space.

The magic number is 0xb301534c:

This has the high bit set so is not a valid instruction.
On builds that ignore the high bit this decodes to nor zr, zr, 0x534c a no-op.
The immediate is "SL" for Standard Library.

start.asm also initialises sp so it starts just below the 64KiB LUT memory allocation.

Added some error codes for the upcoming graphics module.

Added support for global variables. This requires including `start.asm` as the first code. This includes a `jmp` to 0x100 for user code, and a magic number that modules can check to verify that the user has correctly reserved the space. The magic number is 0xb301534c: * This has the high bit set so is not a valid instruction. * On builds that ignore the high bit this decodes to `nor zr, zr, 0x534c` a no-op. * The immediate is "SL" for Standard Library. `start.asm` also initialises `sp` so it starts just below the 64KiB LUT memory allocation. Added some error codes for the upcoming graphics module.

Gelthor added 5 commits 2026-09-06 12:15:43 +02:00

Added initial draft of globals and errno 4d752ef67c

Merge remote-tracking branch 'gelthor' into graphics 39ff477f6e

Moved new *.asm files into src/ 04f3f03ad5

Add globals and error codes 46abf8d893

Added support for global variables.

This requires including `start.asm` as the first code. This includes
a `jmp` to 0x100 for user code, and a magic number that modules
can check to verify th the user has correctly reserved the space.

The magic number is 0xb301534c:

* This has the high bit set so is not a valid instruction

* On builds that ignore the high bit this decodes to `nor zr, zr,
  0x534c` a no-op.

* The immediate is "SL" for standardl ibrary.

Due to a bug start pads with U... 0, instead of @addr.

`start.asm` also initialises `sp` so it start just below the 64KiB LUT
memory alocation.

Added some error codes for the upcoming graphics module.

merge upstream b63af2f15a

Mutex requested changes

2026-09-06 12:45:37 +02:00

CONTRIBUTING.md

						
				@@ -16,10 +16,21 @@ All functions in the standard library should follow the following outline:

				; Arguments: <which register contains what argument>

				; Result: <what is the result, and where is it stored>

				; Clobbers: <list of registers that are clobbered>

				; Globals: <list of globals are accessed. OPTIONAL>

Mutex commented

2026-09-06 12:33:47 +02:00

Which errors can occur (if any) should probably have its own section here, probably called "Errors"

Gelthor marked this conversation as resolved

src/graphics_32.asm Outdated

						
				@@ -0,0 +28,4 @@

				; - r5: If non-zero, clears the current framebuffer.

				; - r6: If non-zero, enable double buffering

				; Returns:

				; - r1: Status

Mutex commented

2026-09-06 12:44:08 +02:00

The status is returned in flags, not r1. Errors should probably have their own dedicated section here

The status is returned in `flags`, not `r1`. Errors should probably have their own dedicated section here

Gelthor marked this conversation as resolved

PleegWat reviewed

2026-09-06 13:29:26 +02:00

src/globals.asm

						
				@@ -0,0 +1,28 @@

				; Addresses of global variables

PleegWat commented

2026-09-06 13:28:49 +02:00

I'd personally rather see reserved addresses put in hex, but I don't have a good reasoning for that.

Gelthor marked this conversation as resolved

src/globals.asm Outdated

						
				@@ -0,0 +22,4 @@

				; * 2 => 32 bits per pixel

				pub const FB_LOG_BPP         = 43 ; U8 Specialisations should hardcode this

				pub const MAGIC_ADDRESS_LONG = 12 ; U32 Location of the magic value

PleegWat commented

2026-09-06 13:29:21 +02:00

If we do not declare these addresses in order, I fear we will accidentally introduce double use of an address at some point.

Gelthor marked this conversation as resolved

PleegWat reviewed

2026-09-06 14:02:45 +02:00

src/graphics_32.asm Outdated

						
				@@ -0,0 +23,4 @@

				; Arguments:

				; - r1: Point to allocated framebuffer

				; - r2: Size of FB allocation, in bytes

				; - r3: Bits per pixel: One of `8` or `32`

PleegWat commented

2026-09-06 13:33:09 +02:00

Since this function lives in graphics_32, I'd expect that to imply 32 bits per pixel. If the same function is intended to be used for both modes, it should live in a generic graphics file.

Since this function lives in `graphics_32`, I'd expect that to imply 32 bits per pixel. If the same function is intended to be used for both modes, it should live in a generic graphics file.

Gelthor marked this conversation as resolved

src/graphics_32.asm Outdated

						
				@@ -0,0 +25,4 @@

				; - r2: Size of FB allocation, in bytes

				; - r3: Bits per pixel: One of `8` or `32`

				; - r4: Width - height is set to `width / 4 * 3`

				; - r5: If non-zero, clears the current framebuffer.

PleegWat commented

2026-09-06 13:34:46 +02:00

If you are worried about register pressure, rather than having 2 bool arguments like this you could have a single u32 argument holding a bitwise OR of flags.

Gelthor commented

2026-09-06 18:56:09 +02:00

I'm not sure register pressure is important for a rarely called initialisation function.

src/graphics_32.asm Outdated

						
				@@ -0,0 +67,4 @@

				    mov r8, 0

				    jmp mode_ok

				    invalid_mode:

PleegWat commented

2026-09-06 13:40:28 +02:00

I'd recommend centralizing the function epilogue. Otherwise, if a future refactor changes which registers need to be saved and one of the return sites is missed that could lead to nasty errors.

Gelthor marked this conversation as resolved

src/graphics_32.asm Outdated

						
				@@ -0,0 +81,4 @@

				    ja width_invalid

				    cmp r4, 4

				    jb width_invalid

				    ; Is width a power of two?

PleegWat commented

2026-09-06 13:42:19 +02:00

There is a bit trick for this: !(v & (v - 1)). Though that does not yield the log2.

There is a bit trick for this: `!(v & (v - 1))`. Though that does not yield the log2.

Gelthor commented

2026-09-06 18:54:04 +02:00

Fixed the comment to describe what the code actually does.

Gelthor marked this conversation as resolved

src/graphics_32.asm Outdated

						
				@@ -0,0 +178,4 @@

				    mov flags, 0 ; OK

					jmp r13 ; return

				pub set_text_mode:

PleegWat commented

2026-09-06 13:59:36 +02:00

I think I'd rather see graphics mode and text mode code in separate files.

Gelthor marked this conversation as resolved

src/graphics_32.asm Outdated

						
				@@ -0,0 +415,4 @@

				; Returns:

				; - None

				; Clobbers:

				; - r1: Old display FB => new draw FB

PleegWat commented

2026-09-06 13:56:16 +02:00

Should we document these as returns instead? If any caller is writing to the buffer directly, they could then rely on this return and skip the extra load.

Gelthor marked this conversation as resolved

src/graphics_32.asm Outdated

						
				@@ -0,0 +438,4 @@

				; Framebuffer pointer is stored in FB_PTR.

				;

				; Arguments:

				; - r1 Color 24 bit RGB0   PRESERVED

PleegWat commented

2026-09-06 13:58:34 +02:00

The function code does not match this argument list

Gelthor marked this conversation as resolved

Gelthor added 5 commits 2026-09-06 19:03:34 +02:00

Remove graphics code from this PR 21deca64fd

Global addresses now in hex and listed in order 7368044335

Function premble should list error codes, if any ca5cff0a79

Merge remote-tracking branch 'origin/main' into graphics 1ba70f10cd

Split framebuffer stride into stride and depth 4801520424

Stride is now U16, depth is U16, and stores the bits per pixel of the
buffer.

PleegWat approved these changes

2026-09-06 19:33:46 +02:00

ShatteredMINT merged commit e654d451a7 into main

2026-09-06 20:13:07 +02:00

Sign in to join this conversation.

3 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: TCShenanigans/symphony_stdlib#19

Allow edits from maintainers