simvx.graphics.gpu.pipeline

Graphics pipeline and shader module management.

Module Contents

Functions

create_shader_module	Load a SPIR-V file and create a VkShaderModule.
create_graphics_pipeline	Create a VkPipeline for triangle rendering. Returns (pipeline, layout).
create_forward_pipeline	Create a pipeline for forward rendering with vertex inputs and SSBOs.
create_transparent_pipeline	Create a pipeline for transparent object rendering.
create_skinned_pipeline	Create a pipeline for skinned mesh rendering.
create_pick_pipeline	Create a pipeline for the pick pass (R32_UINT output, position-only vertex input).
create_line_pipeline	Create a pipeline for line rendering with per-vertex color.
create_ui_pipeline	Create a pipeline for solid-color 2D UI rendering.
create_textured_quad_pipeline	Create a pipeline for a textured fullscreen quad (no depth test).
create_gizmo_pipeline	Create a pipeline for gizmo overlay rendering (depth test DISABLED).

Data

__all__
log

API

simvx.graphics.gpu.pipeline.__all__

[‘create_shader_module’, ‘create_graphics_pipeline’, ‘create_forward_pipeline’, ‘create_transparent_…

simvx.graphics.gpu.pipeline.log

‘getLogger(…)’

simvx.graphics.gpu.pipeline.create_shader_module(device: Any, spirv_path: pathlib.Path) → Any

Load a SPIR-V file and create a VkShaderModule.

simvx.graphics.gpu.pipeline.create_graphics_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int]) → tuple[Any, Any]

Create a VkPipeline for triangle rendering. Returns (pipeline, layout).

Uses raw cffi allocation to avoid the vulkan wrapper’s GC lifetime issues with nested pointer structs.

simvx.graphics.gpu.pipeline.create_forward_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int], descriptor_layout: Any, texture_layout: Any | None = None, double_sided: bool = False) → tuple[Any, Any]

Create a pipeline for forward rendering with vertex inputs and SSBOs.

Vertex format: position (vec3), normal (vec3), uv (vec2) = 32 bytes stride. Push constants: view (mat4) + proj (mat4) = 128 bytes. If texture_layout is provided, set 1 is bound to the texture descriptor layout. If double_sided is True, backface culling is disabled. Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_transparent_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int], descriptor_layout: Any, texture_layout: Any | None = None) → tuple[Any, Any]

Create a pipeline for transparent object rendering.

Identical to the forward pipeline except:

• Alpha blending enabled (srcAlpha, oneMinusSrcAlpha)

• Depth write disabled (depth test still enabled)

• Backface culling disabled (transparent objects often need both sides)

Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_skinned_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int], descriptor_layout: Any, texture_layout: Any | None = None, joint_layout: Any | None = None) → tuple[Any, Any]

Create a pipeline for skinned mesh rendering.

Vertex format: position(vec3) + normal(vec3) + uv(vec2) + joints(uvec4) + weights(vec4) = 48 bytes. Descriptor sets: 0=SSBOs, 1=textures, 2=joint matrices SSBO. Push constants: view (mat4) + proj (mat4) = 128 bytes. Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_pick_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int], descriptor_layout: Any) → tuple[Any, Any]

Create a pipeline for the pick pass (R32_UINT output, position-only vertex input).

Push constants: view (mat4) + proj (mat4) = 128 bytes. Vertex input: position (vec3) only, stride 32 (same buffer as forward, skips normal/uv). Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_line_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int]) → tuple[Any, Any]

Create a pipeline for line rendering with per-vertex color.

Vertex format: position (vec3) + color (vec4) = 28 bytes stride. Push constants: view (mat4) + proj (mat4) = 128 bytes. No descriptor sets. Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_ui_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int]) → tuple[Any, Any]

Create a pipeline for solid-color 2D UI rendering.

Vertex format: position (vec2) + uv (vec2) + color (vec4) = 32 bytes stride. Push constants: screen_size (vec2) = 8 bytes. No descriptor sets, no depth test. Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_textured_quad_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int], texture_layout: Any) → tuple[Any, Any]

Create a pipeline for a textured fullscreen quad (no depth test).

Vertex format: position (vec2) + uv (vec2) = 16 bytes stride. Descriptor set 0: combined image sampler (texture). Returns (pipeline, pipeline_layout).

simvx.graphics.gpu.pipeline.create_gizmo_pipeline(device: Any, vert_module: Any, frag_module: Any, render_pass: Any, extent: tuple[int, int], *, topology: int = vk.VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST) → tuple[Any, Any]

Create a pipeline for gizmo overlay rendering (depth test DISABLED).

Same vertex format as line pipeline: position (vec3) + color (vec4) = 28 bytes. Push constants: view (mat4) + proj (mat4) = 128 bytes. Supports both TRIANGLE_LIST and LINE_LIST topology via topology parameter. Returns (pipeline, pipeline_layout).