cc71d9591d
UIRenderer.flush wrote to mtl_vbuf at byte offset 0 on every flush. Metal records draw commands but reads the buffer at GPU execution time, so a frame with multiple flushes ended up rendering whatever the LAST writer left in the buffer for every draw. Chess UI hit this hard: each of the 32 pieces in the initial position triggers two bind_texture flushes (atlas -> pieces -> atlas), so ~64 mid-frame flushes silently rendered the final info-panel batch over the board and the sprites. New GPU protocol method update_buffer_at(buf, data, size, byte_offset); Metal impl writes at offset via [*]u8 arithmetic on [buf contents]. UIRenderer tracks mtl_buf_offset (reset in begin, advanced per flush, aligned to 16B, wraps on overflow) and draws each batch with vertex_off = byte_off / UI_VERTEX_BYTES. Metal buffer over-allocated 4x the per-flush max (~3 MB) for headroom. GL path untouched — glBufferData already orphans the storage. 71/71 regression tests pass. Metal-clear example, macOS GL chess, and WASM chess all still build.
50 lines
2.4 KiB
Plaintext
50 lines
2.4 KiB
Plaintext
#import "modules/std.sx";
|
|
#import "modules/gpu/types.sx";
|
|
|
|
// GPU is the rendering-API abstraction. Concrete backends live as siblings
|
|
// of this file: `metal.sx` (iOS, eventually macOS), `vulkan.sx` (Linux/
|
|
// Android, plus macOS via MoltenVK), `webgpu.sx` (wasm). The SDL-backed
|
|
// GL renderer used by the desktop+wasm path stays as-is until those
|
|
// backends land.
|
|
|
|
GPU :: protocol {
|
|
// Bind the GPU to a backend-specific render target (e.g. a
|
|
// CAMetalLayer on iOS). pixel_w/pixel_h are the drawable's pixel
|
|
// dimensions; call resize when they change.
|
|
init :: (target: *void, pixel_w: s32, pixel_h: s32) -> bool;
|
|
shutdown :: ();
|
|
|
|
resize :: (pixel_w: s32, pixel_h: s32);
|
|
|
|
begin_frame :: (clear: ClearColor) -> bool;
|
|
|
|
// target_time is the host clock time at which the drawable should be
|
|
// presented (units match the platform's CADisplayLink.targetTimestamp
|
|
// on Apple). Metal forwards it to presentDrawable:atTime: to cap the
|
|
// pipeline at one frame so the inset slide lands on the same vsync as
|
|
// UIKit's keyboard view. GL backends ignore it.
|
|
end_frame :: (target_time: f64);
|
|
|
|
create_shader :: (vsrc: string, fsrc: string) -> ShaderHandle;
|
|
create_buffer :: (size_bytes: s64) -> BufferHandle;
|
|
update_buffer :: (buf: BufferHandle, data: *void, size_bytes: s64);
|
|
// Sub-buffer write at a byte offset. Required for Metal where re-using
|
|
// the same buffer slice across multiple draws in a single command
|
|
// encoder is a race: the GPU executes draws asynchronously and reads
|
|
// shared-storage buffer contents at execution time, so the LAST writer
|
|
// wins if every flush targets offset 0. Renderers that issue more than
|
|
// one draw per frame must advance their write offset between flushes.
|
|
update_buffer_at :: (buf: BufferHandle, data: *void, size_bytes: s64, byte_offset: s64);
|
|
create_texture :: (w: s32, h: s32, format: TextureFormat, pixels: *void) -> TextureHandle;
|
|
update_texture_region :: (tex: TextureHandle, x: s32, y: s32, w: s32, h: s32, pixels: *void);
|
|
|
|
set_shader :: (sh: ShaderHandle);
|
|
set_vertex_buffer :: (buf: BufferHandle);
|
|
set_texture :: (slot: u32, tex: TextureHandle);
|
|
set_vertex_constants :: (slot: u32, data: *void, size_bytes: s64);
|
|
set_scissor :: (x: s32, y: s32, w: s32, h: s32);
|
|
disable_scissor :: ();
|
|
|
|
draw_triangles :: (vertex_offset: s32, vertex_count: s32);
|
|
}
|