Harbor

sharkk/corvette public

Components 3 Issues Requests

Changelog 138e04c61af8

pin gpu instance input parity

@sky · 1 month ago · parent 850ceea8c535

0 added 21 modified 0 deleted

gpu/backend/backend.odin +17 -2 modified

…		280 unchanged lines hidden
281	281
282	282		Vertex_Attribute :: struct {
283	283		location: u32,
	284	+	binding: u32,
284	285		offset: u32,
285	286		format: Vertex_Format,
286	287		}
…		5 unchanged lines hidden
292	293		}
293	294
294	295		Vertex_Binding :: struct {
295		-	binding: u32,
296		-	stride: u32,
	296	+	binding: u32,
	297	+	stride: u32,
	298	+	input_rate: Vertex_Input_Rate,
297	299		}
298	300
	301	+	Vertex_Input_Rate :: enum {
	302	+	Vertex,
	303	+	Instance,
	304	+	}
	305	+
299	306		// --- Vertex layout descriptor ---
300	307
301	308		Vertex_Attrib :: enum u8 {
…		214 unchanged lines hidden
516	523		unmap_buffer: proc(handle: Buffer_Handle),
517	524		get_buffer_mapped: proc(handle: Buffer_Handle) -> rawptr,
518	525		bind_vertex_buffer: proc(ctx: Frame_Context, handle: Buffer_Handle),
	526	+	bind_vertex_buffer_slot: proc(
	527	+	ctx: Frame_Context,
	528	+	slot: u32,
	529	+	handle: Buffer_Handle,
	530	+	offset: u64,
	531	+	stride: u32,
	532	+	),
519	533		bind_index_buffer: proc(ctx: Frame_Context, handle: Buffer_Handle),
520	534
521	535		// --- Textures ---
…		148 unchanged lines hidden
670	684		b.unmap_buffer != nil &&
671	685		b.get_buffer_mapped != nil &&
672	686		b.bind_vertex_buffer != nil &&
	687	+	b.bind_vertex_buffer_slot != nil &&
673	688		b.bind_index_buffer != nil &&
674	689		b.create_texture != nil &&
675	690		b.destroy_texture != nil &&
…		35 unchanged lines hidden

gpu/backend/d3d11/d3d11_backend.odin +1 -0 modified

…		378 unchanged lines hidden
379	379		unmap_buffer = unmap_buffer_d3d11,
380	380		get_buffer_mapped = get_buffer_mapped_d3d11,
381	381		bind_vertex_buffer = bind_vertex_buffer_d3d11,
	382	+	bind_vertex_buffer_slot = bind_vertex_buffer_slot_d3d11,
382	383		bind_index_buffer = bind_index_buffer_d3d11,
383	384
384	385		// Textures
…		212 unchanged lines hidden

gpu/backend/d3d11/d3d11_ops.odin modified

Diff hidden because this file has more than 800 lines.

gpu/backend/d3d12/d3d12_backend.odin modified

Diff hidden because this file has more than 800 lines.

gpu/backend/d3d12/d3d12_ops.odin modified

Diff hidden because this file has more than 800 lines.

gpu/backend/opengl/gl_backend.odin +2 -0 modified

…		74 unchanged lines hidden
75	75		cull_mode: bk.Cull_Mode,
76	76		front_face: bk.Front_Face,
77	77		vertex_stride: u32,
	78	+	vertex_strides: [16]u32,
78	79		push_constant_size: u32,
79	80		push_constant_stages: bk.Shader_Stage_Flags,
80	81		is_compute: bool,
…		116 unchanged lines hidden
197	198		unmap_buffer = unmap_buffer_opengl,
198	199		get_buffer_mapped = get_buffer_mapped_opengl,
199	200		bind_vertex_buffer = bind_vertex_buffer_opengl,
	201	+	bind_vertex_buffer_slot = bind_vertex_buffer_slot_opengl,
200	202		bind_index_buffer = bind_index_buffer_opengl,
201	203		create_texture = create_texture_opengl,
202	204		destroy_texture = destroy_texture_opengl,
…		107 unchanged lines hidden

gpu/backend/opengl/gl_ops.odin modified

Diff hidden because this file has more than 800 lines.

gpu/backend/vertex_layout.odin +17 -2 modified

…		75 unchanged lines hidden
76	76		attributes: [MAX_VERTEX_ATTRIBS]Vertex_Attribute,
77	77		attr_count: int,
78	78		) {
	79	+	return vertex_layout_to_pipeline_attrs_for_binding(layout, 0, .Vertex, 0)
	80	+	}
	81	+
	82	+	vertex_layout_to_pipeline_attrs_for_binding :: proc(
	83	+	layout: ^Vertex_Layout,
	84	+	binding_index: u32,
	85	+	input_rate: Vertex_Input_Rate,
	86	+	first_location: u32,
	87	+	) -> (
	88	+	binding: Vertex_Binding,
	89	+	attributes: [MAX_VERTEX_ATTRIBS]Vertex_Attribute,
	90	+	attr_count: int,
	91	+	) {
79	92		binding = Vertex_Binding {
80		-	binding = 0,
	93	+	binding = binding_index,
81	94		stride = layout.stride,
	95	+	input_rate = input_rate,
82	96		}
83	97		for i in 0 ..< int(layout.count) {
84	98		e := &layout.entries[i]
85	99		attributes[i] = Vertex_Attribute {
86		-	location = u32(i),
	100	+	location = first_location + u32(i),
	101	+	binding = binding_index,
87	102		offset = e.offset,
88	103		format = e.format,
89	104		}
…		109 unchanged lines hidden

gpu/backend/vulkan/vk_backend.odin +1 -0 modified

…		326 unchanged lines hidden
327	327		unmap_buffer = unmap_buffer_vk,
328	328		get_buffer_mapped = get_buffer_mapped_vk,
329	329		bind_vertex_buffer = bind_vertex_buffer_vk,
	330	+	bind_vertex_buffer_slot = bind_vertex_buffer_slot_vk,
330	331		bind_index_buffer = bind_index_buffer_vk,
331	332
332	333		// Textures
…		211 unchanged lines hidden

gpu/backend/vulkan/vk_ops.odin modified

Diff hidden because this file has more than 800 lines.

gpu/compiler/executor.odin modified

Diff hidden because this file has more than 800 lines.

gpu/docs/gpu_intent_api.md +4 -2 modified

…		18 unchanged lines hidden
19	19
20	20		## Execution model
21	21
22		-	The authoritative runtime direction is a planned-IR executor: validation produces diagnostics, planning orders commands, and backend lowering walks the planned IR to issue backend operations. A gated backend-native planned draw path now handles materialized IR buffers, render passes, shaders, graphics pipelines, descriptor sets, direct indexed draws, direct non-indexed draws, single-command indirect indexed/non-indexed draws, supported borrowed imported mesh buffers, lit imported meshes, imported rect clips, the generated 2D rect/line/circle slice, generated textured quads with concrete texture IDs, and generated glyph quads with texture-backed materials. The current runtime translator that maps remaining generated/imported producer forms into legacy 2D/3D renderer calls is migration-only. New backend parity work must not add feature-specific translator branches unless the branch is a temporary, tracked bridge with a removal path.
	22	+	The authoritative runtime direction is a planned-IR executor: validation produces diagnostics, planning orders commands, and backend lowering walks the planned IR to issue backend operations. A gated backend-native planned draw path now handles materialized IR buffers, render passes, shaders, graphics pipelines, descriptor sets, direct indexed draws, direct non-indexed draws, true per-instance vertex buffers for pipelines that declare an instance layout, single-command indirect indexed/non-indexed draws, supported borrowed imported mesh buffers, lit imported meshes, imported rect clips, the generated 2D rect/line/circle slice, generated textured quads with concrete texture IDs, and generated glyph quads with texture-backed materials. The current runtime translator that maps remaining generated/imported producer forms into legacy 2D/3D renderer calls is migration-only. New backend parity work must not add feature-specific translator branches unless the branch is a temporary, tracked bridge with a removal path.
23	23
24	24		Runtime lowering must not fail silently. If a command cannot be lowered, `gpu` records a structured diagnostic with the command handle and reason, then fails submission or mode classification explicitly.
25	25
…		26 unchanged lines hidden
52	52
53	53		`compute_barrier(ctx)` is a conservative/global in-frame barrier for compute-written storage/UAV data consumed by later backend work in the same frame. Its consumer scope is later compute shader reads/writes, vertex/fragment shader reads, vertex/index fetch, and indirect command reads. It is intentionally over-synchronized for correctness; it is not a render-target transition substitute, CPU readback fence, cross-frame lifetime rule, or feedback-loop validator. Backends that lack a complete storage descriptor path must report that separately through capabilities or diagnostics instead of pretending the barrier enables unsupported binding.
54	54
55		-	Indirect draw argument buffers use the backend-neutral `Indirect_Draw_Args` and `Indirect_Draw_Indexed_Args` ABI. The layouts are intentionally byte-identical to Vulkan, D3D11, D3D12, and OpenGL single-command indirect arguments: 16 bytes for non-indexed draws and 20 bytes for indexed draws. In IR, `packet.instances.indirect` is the argument-buffer resource, `packet.instances.offset` is the byte offset, `packet.instances.stride` is zero-or-ABI-size for Phase 7a, and `packet.instances.count` is limited to one command. True per-instance vertex streams remain separate unsupported vertex-input work.
	55	+	Indirect draw argument buffers use the backend-neutral `Indirect_Draw_Args` and `Indirect_Draw_Indexed_Args` ABI. The layouts are intentionally byte-identical to Vulkan, D3D11, D3D12, and OpenGL single-command indirect arguments: 16 bytes for non-indexed draws and 20 bytes for indexed draws. In IR, `packet.instances.indirect` is the argument-buffer resource, `packet.instances.offset` is the byte offset, `packet.instances.stride` is zero-or-ABI-size for indirect arguments, and `packet.instances.count` is limited to one command in the current executor.
56	56
	57	+	Per-instance vertex streams use the normal `packet.instances.resource` field, not the indirect argument field. A planned instance-buffer draw requires a graphics pipeline with `has_instance_layout = true`, a nonzero `packet.instances.stride`, and a prepared instance buffer. Slot 0 is the per-vertex stream; slot 1 is the per-instance stream bound with `packet.instances.offset` and `packet.instances.stride`. Vulkan, D3D11, D3D12, and OpenGL all lower the same backend-neutral `Vertex_Binding.input_rate` and `Vertex_Attribute.binding` metadata; OpenGL uses the DSA vertex-attrib-binding/divisor path internally. Indirect draws may also bind an instance stream: the indirect buffer controls counts/firsts, while the instance buffer controls per-instance attributes.
	58	+
57	59		D3D12 runtime fence behavior cannot be executed by Linux CI. D3D12 lifetime and transition work is accepted by static backend checks, pure fence-order tests where available, and explicit Windows runtime verification debt.
58	60
59	61		Current backend sync matrix:
…		195 unchanged lines hidden

gpu/frame_api.odin modified

Diff hidden because this file has more than 800 lines.

gpu/render_ir/dump.odin modified

Diff hidden because this file has more than 800 lines.

gpu/render_ir/render_ir.odin modified

Diff hidden because this file has more than 800 lines.

gpu/tests/capabilities_test.odin +21 -0 modified

…		52 unchanged lines hidden
53	53		}
54	54
55	55		@(test)
	56	+	test_vertex_layout_can_emit_instance_binding_metadata :: proc(t: ^testing.T) {
	57	+	layout := bk.vertex_layout_build(.Position, .Color)
	58	+	binding, attrs, attr_count := bk.vertex_layout_to_pipeline_attrs_for_binding(
	59	+	&layout,
	60	+	1,
	61	+	.Instance,
	62	+	3,
	63	+	)
	64	+	testing.expect_value(t, binding.binding, u32(1))
	65	+	testing.expect_value(t, binding.input_rate, bk.Vertex_Input_Rate.Instance)
	66	+	testing.expect_value(t, binding.stride, u32(28))
	67	+	testing.expect_value(t, attr_count, 2)
	68	+	testing.expect_value(t, attrs[0].location, u32(3))
	69	+	testing.expect_value(t, attrs[0].binding, u32(1))
	70	+	testing.expect_value(t, attrs[0].offset, u32(0))
	71	+	testing.expect_value(t, attrs[1].location, u32(4))
	72	+	testing.expect_value(t, attrs[1].binding, u32(1))
	73	+	testing.expect_value(t, attrs[1].offset, u32(12))
	74	+	}
	75	+
	76	+	@(test)
56	77		test_backend_handle_index_rejects_null_and_out_of_range :: proc(t: ^testing.T) {
57	78		_, null_ok := bk.handle_index(bk.NULL_BUFFER, 8)
58	79		testing.expect(t, !null_ok)
…		39 unchanged lines hidden

gpu/tests/compiler_executor_test.odin modified

Diff hidden because this file has more than 800 lines.

gpu/tests/recording_backend.odin +40 -2 modified

…		52 unchanged lines hidden
53	53		descriptor: bk.Descriptor_Handle,
54	54		descriptor_index: u32,
55	55		buffer: bk.Buffer_Handle,
	56	+	buffer_slot: u32,
	57	+	buffer_offset: u64,
	58	+	buffer_stride: u32,
56	59		argument_buffer: bk.Buffer_Handle,
57	60		argument_offset: u64,
58	61		draw_count: u32,
…		2 unchanged lines hidden
61	64		color_formats: [bk.MAX_COLOR_TARGETS]bk.Format,
62	65		color_views: [bk.MAX_COLOR_TARGETS]bk.Texture_Handle,
63	66		clear_colors: [bk.MAX_COLOR_TARGETS][4]f32,
	67	+	vertex_binding_count: u32,
	68	+	vertex_bindings: [16]bk.Vertex_Binding,
	69	+	vertex_attribute_count: u32,
	70	+	vertex_attributes: [16]bk.Vertex_Attribute,
64	71
65	72		push_constant_size: u32,
66	73		push_constant_bytes: [compiler.IMPORTED_PUSH_CONSTANT_CAP]u8,
…		115 unchanged lines hidden
182	189		push_constants = recording_push_constants,
183	190		bind_descriptor_set = recording_bind_descriptor_set,
184	191		bind_vertex_buffer = recording_bind_vertex_buffer,
	192	+	bind_vertex_buffer_slot = recording_bind_vertex_buffer_slot,
185	193		bind_index_buffer = recording_bind_index_buffer,
186	194		draw = recording_draw,
187	195		draw_indexed = recording_draw_indexed,
…		204 unchanged lines hidden
392	400		pipeline = handle,
393	401		color_count = desc.color_attachment_count,
394	402		color_formats = desc.color_formats,
	403	+	vertex_binding_count = u32(len(desc.vertex_bindings)),
	404	+	vertex_attribute_count = u32(len(desc.vertex_attributes)),
395	405		})
	406	+	call := &rec.calls[len(rec.calls) - 1]
	407	+	for binding, i in desc.vertex_bindings {
	408	+	if i >= len(call.vertex_bindings) {break}
	409	+	call.vertex_bindings[i] = binding
	410	+	}
	411	+	for attr, i in desc.vertex_attributes {
	412	+	if i >= len(call.vertex_attributes) {break}
	413	+	call.vertex_attributes[i] = attr
	414	+	}
396	415		return handle, true
397	416		}
398	417
…		51 unchanged lines hidden
450	469		}
451	470
452	471		recording_bind_vertex_buffer :: proc(ctx: bk.Frame_Context, handle: bk.Buffer_Handle) {
	472	+	recording_bind_vertex_buffer_slot(ctx, 0, handle, 0, 0)
	473	+	}
	474	+
	475	+	recording_bind_vertex_buffer_slot :: proc(
	476	+	ctx: bk.Frame_Context,
	477	+	slot: u32,
	478	+	handle: bk.Buffer_Handle,
	479	+	offset: u64,
	480	+	stride: u32,
	481	+	) {
453	482		if recording_active != nil {
454		-	recording_active.bound_vertex_buffer = handle
	483	+	if slot == 0 {
	484	+	recording_active.bound_vertex_buffer = handle
	485	+	}
455	486		}
456		-	recording_record({kind = .Bind_Vertex_Buffer, frame_index = ctx.frame_index, buffer = handle})
	487	+	recording_record({
	488	+	kind = .Bind_Vertex_Buffer,
	489	+	frame_index = ctx.frame_index,
	490	+	buffer = handle,
	491	+	buffer_slot = slot,
	492	+	buffer_offset = offset,
	493	+	buffer_stride = stride,
	494	+	})
457	495		}
458	496
459	497		recording_bind_index_buffer :: proc(ctx: bk.Frame_Context, handle: bk.Buffer_Handle) {
…		160 unchanged lines hidden

gpu/tests/render_ir/dump_test.odin +1 -1 modified

…		95 unchanged lines hidden
96	96		" #1 name=\"material\" layout=1 bindings=1\n" +
97	97		" binding=0 type=Combined_Image_Sampler resource=2 sampler=3 size=0\n" +
98	98		"pipelines 2\n" +
99		-	" #1 Graphics name=\"rect\" key=\"gfx:vs=1;fs=2;layouts=1;layout=PNU;topology=Triangle_List;cull=Back;front=Counter_Clockwise;blend=true/Alpha;depth=false/false;color=B8G8R8A8_SRGB;depth_format=Undefined;pc=0;pc_stages=None\"\n" +
	99	+	" #1 Graphics name=\"rect\" key=\"gfx:vs=1;fs=2;layouts=1;layout=PNU;instance_layout=false/PNU;topology=Triangle_List;cull=Back;front=Counter_Clockwise;blend=true/Alpha;depth=false/false;color=B8G8R8A8_SRGB;depth_format=Undefined;pc=0;pc_stages=None\"\n" +
100	100		" #2 Compute name=\"particles\" key=\"compute:shader=0;layouts=None;pc=0;pc_stages=None\"\n" +
101	101		"commands 2\n" +
102	102		" #1 Draw pass=1 pipeline=1 vertex_count=6 instance_count=1\n" +
…		3 unchanged lines hidden

gpu/tests/render_ir/pipeline_desc_test.odin +2 -2 modified

…		83 unchanged lines hidden
84	84		testing.expect_value(
85	85		t,
86	86		gfx_key,
87		-	"gfx:vs=1;fs=2;layouts=None;layout=PNU;topology=Triangle_List;cull=Back;front=Counter_Clockwise;blend=true/Alpha;depth=true/true;color=B8G8R8A8_SRGB;depth_format=D32_SFLOAT;pc=64;pc_stages=Vertex\|Fragment",
	87	+	"gfx:vs=1;fs=2;layouts=None;layout=PNU;instance_layout=false/PNU;topology=Triangle_List;cull=Back;front=Counter_Clockwise;blend=true/Alpha;depth=true/true;color=B8G8R8A8_SRGB;depth_format=D32_SFLOAT;pc=64;pc_stages=Vertex\|Fragment",
88	88		)
89	89
90	90		compute_key := ir.format_pipeline_cache_key(
…		23 unchanged lines hidden
114	114		testing.expect_value(
115	115		t,
116	116		key,
117		-	"gfx:vs=0;fs=0;layouts=1\|2;layout=PNU;topology=Triangle_List;cull=Back;front=Counter_Clockwise;blend=true/Alpha;depth=false/false;color=B8G8R8A8_SRGB;depth_format=Undefined;pc=0;pc_stages=None",
	117	+	"gfx:vs=0;fs=0;layouts=1\|2;layout=PNU;instance_layout=false/PNU;topology=Triangle_List;cull=Back;front=Counter_Clockwise;blend=true/Alpha;depth=false/false;color=B8G8R8A8_SRGB;depth_format=Undefined;pc=0;pc_stages=None",
118	118		)
119	119		}

gpu/types.odin +1 -0 modified

…		206 unchanged lines hidden
207	207		Vertex_Binding :: bk.Vertex_Binding
208	208		Vertex_Attribute :: bk.Vertex_Attribute
209	209		Vertex_Format :: bk.Vertex_Format
	210	+	Vertex_Input_Rate :: bk.Vertex_Input_Rate
210	211		Vertex_Attrib :: bk.Vertex_Attrib
211	212		Vertex_Layout_Entry :: bk.Vertex_Layout_Entry
212	213		Vertex_Layout :: bk.Vertex_Layout
…		26 unchanged lines hidden