Working mac arm+mac x86+windows x86 builds

.github/workflows/build-nmage.yml

@@ -1,24 +1,69 @@
 name: build-nmage
 
 on:
+  push:
+    branches:
+      - dev
   create:
   workflow_dispatch:
 
 jobs:
-  build-nmage-macos:
+  build-nmage-windows:
-    runs-on: macos-12
+    runs-on: windows-latest
+
     steps:
-    
       - name: Install golang
-        uses: actions/setup-go@v3
+        uses: actions/setup-go@v5
         with:
-          go-version: '>=1.22'
+          go-version: ">=1.23"
+
+      - name: Install assimp-go dll
+        run: |
+          Invoke-WebRequest -Uri "https://github.com/bloeys/assimp-go/releases/download/v0.4.2/libassimp-5.dll" -OutFile "C:\Windows\System32\libassimp-5.dll"
+
+      - name: Download and setup SDL2
+        run: |
+          Invoke-WebRequest -Uri "https://github.com/libsdl-org/SDL/releases/download/release-2.30.7/SDL2-devel-2.30.7-mingw.zip" -OutFile "SDL2.zip"
+          Expand-Archive -Path "SDL2.zip" -DestinationPath "SDL2"
+          Copy-Item -Path "SDL2\SDL2-2.30.7\x86_64-w64-mingw32" -Destination "C:\mingw64" -Recurse -Force
+
+      - name: Clone nmage
+        run: git clone https://github.com/bloeys/nmage
+
+      - name: Build nmage
+        working-directory: nmage
+        run: go build .
+
+  build-nmage-macos:
+    runs-on: ${{ matrix.os }}
+
+    strategy:
+      matrix:
+        # Based on: https://github.com/actions/runner-images?tab=readme-ov-file#available-images
+        os:
+          - macos-13 # x86
+          - macos-14 # Arm
+
+    steps:
+      - name: Install golang
+        uses: actions/setup-go@v5
+        with:
+          go-version: ">=1.23"
+
+      - name: Determine architecture
+        id: arch
+        run: |
+          if [ "$(uname -m)" = "arm64" ]; then
+            echo "arch=arm64" >> "$GITHUB_OUTPUT"
+          else
+            echo "arch=amd64" >> "$GITHUB_OUTPUT"
+          fi
 
       - name: Install assimp-go dylib
-        run: sudo mkdir -p /usr/local/lib && sudo wget https://github.com/bloeys/assimp-go/releases/download/v0.4.2/libassimp_darwin_amd64.dylib -O /usr/local/lib/libassimp.5.dylib
+        run: sudo mkdir -p /usr/local/lib && sudo wget https://github.com/bloeys/assimp-go/releases/download/v0.4.2/libassimp_darwin_${{ steps.arch.outputs.arch }}.dylib -O /usr/local/lib/libassimp.5.dylib
 
       - name: Install SDL2
-        run: brew install sdl2{,_image,_mixer,_ttf,_gfx} pkg-config
+        run: brew install sdl2{,_image,_ttf,_gfx} pkg-config
 
       - name: Clone nmage
         run: git clone https://github.com/bloeys/nmage

.gitignore

@@ -15,4 +15,7 @@
 vendor/
 .vscode/
 imgui.ini
-*~
+*~
+
+# Custom
+*.pprof

assets/textures.go

@@ -23,6 +23,15 @@ const (
 	ColorFormat_RGBA8
 )
 
+var (
+	DefaultBlackTexId    Texture
+	DefaultWhiteTexId    Texture
+	DefaultDiffuseTexId  Texture
+	DefaultSpecularTexId Texture
+	DefaultNormalTexId   Texture
+	DefaultEmissionTexId Texture
+)
+
 type Texture struct {
 	// Path only exists for textures loaded from disk
 	Path string

buffers/buf_usage.go

@@ -9,25 +9,48 @@ import (
 
 type BufUsage int
 
+// Full docs for buffer usage can be found here: https://registry.khronos.org/OpenGL-Refpages/gl4/html/glBufferData.xhtml
 const (
 	BufUsage_Unknown BufUsage = iota
 
 	//Buffer is set only once and used many times
-	BufUsage_Static
+	BufUsage_Static_Draw
 	//Buffer is changed a lot and used many times
-	BufUsage_Dynamic
+	BufUsage_Dynamic_Draw
 	//Buffer is set only once and used by the GPU at most a few times
-	BufUsage_Stream
+	BufUsage_Stream_Draw
+
+	BufUsage_Static_Read
+	BufUsage_Dynamic_Read
+	BufUsage_Stream_Read
+
+	BufUsage_Static_Copy
+	BufUsage_Dynamic_Copy
+	BufUsage_Stream_Copy
 )
 
 func (b BufUsage) ToGL() uint32 {
 	switch b {
-	case BufUsage_Static:
+	case BufUsage_Static_Draw:
 		return gl.STATIC_DRAW
-	case BufUsage_Dynamic:
+	case BufUsage_Dynamic_Draw:
 		return gl.DYNAMIC_DRAW
-	case BufUsage_Stream:
+	case BufUsage_Stream_Draw:
 		return gl.STREAM_DRAW
+
+	case BufUsage_Static_Read:
+		return gl.STATIC_READ
+	case BufUsage_Dynamic_Read:
+		return gl.DYNAMIC_READ
+	case BufUsage_Stream_Read:
+		return gl.STREAM_READ
+
+	case BufUsage_Static_Copy:
+		return gl.STATIC_COPY
+	case BufUsage_Dynamic_Copy:
+		return gl.DYNAMIC_COPY
+	case BufUsage_Stream_Copy:
+		return gl.STREAM_COPY
 	}
 
 	assert.T(false, fmt.Sprintf("Unexpected BufUsage value '%v'", b))

buffers/element.go

@@ -1,23 +1,23 @@
 package buffers
 
 import (
-	"fmt"
-
 	"github.com/bloeys/nmage/assert"
+	"github.com/bloeys/nmage/logging"
 	"github.com/go-gl/gl/v4.1-core/gl"
 )
 
-//Element represents an element that makes up a buffer (e.g. Vec3 at an offset of 12 bytes)
+// Element represents an element that makes up a buffer (e.g. Vec3 at an offset of 12 bytes)
 type Element struct {
 	Offset int
 	ElementType
 }
 
-//ElementType is the type of an element thats makes up a buffer (e.g. Vec3)
+// ElementType is the type of an element thats makes up a buffer (e.g. Vec3)
-type ElementType int
+type ElementType uint8
 
 const (
 	DataTypeUnknown ElementType = iota
+
 	DataTypeUint32
 	DataTypeInt32
 	DataTypeFloat32
@@ -25,35 +25,54 @@ const (
 	DataTypeVec2
 	DataTypeVec3
 	DataTypeVec4
+
+	DataTypeMat2
+	DataTypeMat3
+	DataTypeMat4
+
+	DataTypeStruct
 )
 
 func (dt ElementType) GLType() uint32 {
 
 	switch dt {
+
 	case DataTypeUint32:
 		return gl.UNSIGNED_INT
 	case DataTypeInt32:
 		return gl.INT
-
 	case DataTypeFloat32:
 		fallthrough
+
 	case DataTypeVec2:
 		fallthrough
 	case DataTypeVec3:
 		fallthrough
 	case DataTypeVec4:
+		fallthrough
+	case DataTypeMat2:
+		fallthrough
+	case DataTypeMat3:
+		fallthrough
+	case DataTypeMat4:
 		return gl.FLOAT
 
+	case DataTypeStruct:
+		logging.ErrLog.Fatalf("ElementType.GLType of DataTypeStruct is not supported")
+		return 0
+
 	default:
-		assert.T(false, fmt.Sprintf("Unknown data type passed. DataType '%v'", dt))
+		assert.T(false, "Unknown data type passed. DataType '%d'", dt)
 		return 0
 	}
 }
 
-//CompSize returns the size in bytes for one component of the type (e.g. for Vec2 its 4)
+// CompSize returns the size in bytes for one component of the type (e.g. for Vec2 its 4).
+// Bools return 1, although in layout=std140 its 4
 func (dt ElementType) CompSize() int32 {
 
 	switch dt {
+
 	case DataTypeUint32:
 		fallthrough
 	case DataTypeFloat32:
@@ -65,15 +84,25 @@ func (dt ElementType) CompSize() int32 {
 	case DataTypeVec3:
 		fallthrough
 	case DataTypeVec4:
+		fallthrough
+	case DataTypeMat2:
+		fallthrough
+	case DataTypeMat3:
+		fallthrough
+	case DataTypeMat4:
 		return 4
 
+	case DataTypeStruct:
+		logging.ErrLog.Fatalf("ElementType.CompSize of DataTypeStruct is not supported")
+		return 0
+
 	default:
-		assert.T(false, fmt.Sprintf("Unknown data type passed. DataType '%v'", dt))
+		assert.T(false, "Unknown data type passed. DataType '%d'", dt)
 		return 0
 	}
 }
 
-//CompCount returns the number of components in the element (e.g. for Vec2 its 2)
+// CompCount returns the number of components in the element (e.g. for Vec2 its 2)
 func (dt ElementType) CompCount() int32 {
 
 	switch dt {
@@ -91,16 +120,28 @@ func (dt ElementType) CompCount() int32 {
 	case DataTypeVec4:
 		return 4
 
+	case DataTypeMat2:
+		return 2 * 2
+	case DataTypeMat3:
+		return 3 * 3
+	case DataTypeMat4:
+		return 4 * 4
+
+	case DataTypeStruct:
+		logging.ErrLog.Fatalf("ElementType.CompCount of DataTypeStruct is not supported")
+		return 0
+
 	default:
-		assert.T(false, fmt.Sprintf("Unknown data type passed. DataType '%v'", dt))
+		assert.T(false, "Unknown data type passed. DataType '%d'", dt)
 		return 0
 	}
 }
 
-//Size returns the total size in bytes (e.g. for vec3 its 3*4=12 bytes)
+// Size returns the total size in bytes (e.g. for vec3 its 3*4=12 bytes)
 func (dt ElementType) Size() int32 {
 
 	switch dt {
+
 	case DataTypeUint32:
 		fallthrough
 	case DataTypeFloat32:
@@ -115,8 +156,123 @@ func (dt ElementType) Size() int32 {
 	case DataTypeVec4:
 		return 4 * 4
 
+	case DataTypeMat2:
+		return 2 * 2 * 4
+	case DataTypeMat3:
+		return 3 * 3 * 4
+	case DataTypeMat4:
+		return 4 * 4 * 4
+
+	case DataTypeStruct:
+		logging.ErrLog.Fatalf("ElementType.Size of DataTypeStruct is not supported")
+		return 0
+
 	default:
-		assert.T(false, fmt.Sprintf("Unknown data type passed. DataType '%v'", dt))
+		assert.T(false, "Unknown data type passed. DataType '%d'", dt)
 		return 0
 	}
 }
+
+func (dt ElementType) GlStd140SizeBytes() uint8 {
+
+	switch dt {
+
+	case DataTypeUint32:
+		fallthrough
+	case DataTypeFloat32:
+		fallthrough
+	case DataTypeInt32:
+		return 4
+
+	case DataTypeVec2:
+		return 4 * 2
+
+	case DataTypeVec3:
+		return 4 * 3
+
+	case DataTypeVec4:
+		return 4 * 4
+
+		// Matrices follow: (vec4Alignment) * numColumns
+	case DataTypeMat2:
+		return 2 * 2 * 4
+	case DataTypeMat3:
+		return 3 * 3 * 4
+	case DataTypeMat4:
+		return 4 * 4 * 4
+
+	case DataTypeStruct:
+		logging.ErrLog.Fatalf("ElementType.GlStd140SizeBytes of DataTypeStruct is not supported")
+		return 0
+
+	default:
+		assert.T(false, "Unknown data type passed. DataType '%d'", dt)
+		return 0
+	}
+}
+
+func (dt ElementType) GlStd140AlignmentBoundary() uint16 {
+
+	switch dt {
+
+	case DataTypeUint32:
+		fallthrough
+	case DataTypeFloat32:
+		fallthrough
+	case DataTypeInt32:
+		return 4
+
+	case DataTypeVec2:
+		return 8
+
+	case DataTypeVec3:
+		fallthrough
+	case DataTypeVec4:
+		fallthrough
+	case DataTypeMat2:
+		fallthrough
+	case DataTypeMat3:
+		fallthrough
+	case DataTypeMat4:
+		fallthrough
+	case DataTypeStruct:
+		return 16
+
+	default:
+		assert.T(false, "Unknown data type passed. DataType '%d'", dt)
+		return 0
+	}
+}
+
+func (dt ElementType) String() string {
+
+	switch dt {
+
+	case DataTypeUint32:
+		return "uint32"
+	case DataTypeFloat32:
+		return "float32"
+	case DataTypeInt32:
+		return "int32"
+
+	case DataTypeVec2:
+		return "Vec2"
+	case DataTypeVec3:
+		return "Vec3"
+	case DataTypeVec4:
+		return "Vec4"
+
+	case DataTypeMat2:
+		return "Mat2"
+	case DataTypeMat3:
+		return "Mat3"
+	case DataTypeMat4:
+		return "Mat4"
+
+	case DataTypeStruct:
+		return "Struct"
+
+	default:
+		return "Unknown"
+	}
+}

buffers/framebuffer.go

@@ -42,6 +42,7 @@ const (
 	FramebufferAttachmentDataFormat_Unknown FramebufferAttachmentDataFormat = iota
 	FramebufferAttachmentDataFormat_R32Int
 	FramebufferAttachmentDataFormat_RGBA8
+	FramebufferAttachmentDataFormat_RGBAF16
 	FramebufferAttachmentDataFormat_SRGBA
 	FramebufferAttachmentDataFormat_DepthF32
 	FramebufferAttachmentDataFormat_Depth24Stencil8
@@ -50,7 +51,8 @@ const (
 func (f FramebufferAttachmentDataFormat) IsColorFormat() bool {
 	return f == FramebufferAttachmentDataFormat_R32Int ||
 		f == FramebufferAttachmentDataFormat_RGBA8 ||
-		f == FramebufferAttachmentDataFormat_SRGBA
+		f == FramebufferAttachmentDataFormat_SRGBA ||
+		f == FramebufferAttachmentDataFormat_RGBAF16
 }
 
 func (f FramebufferAttachmentDataFormat) IsDepthFormat() bool {
@@ -65,6 +67,8 @@ func (f FramebufferAttachmentDataFormat) GlInternalFormat() int32 {
 		return gl.R32I
 	case FramebufferAttachmentDataFormat_RGBA8:
 		return gl.RGB8
+	case FramebufferAttachmentDataFormat_RGBAF16:
+		return gl.RGBA16F
 	case FramebufferAttachmentDataFormat_SRGBA:
 		return gl.SRGB_ALPHA
 	case FramebufferAttachmentDataFormat_DepthF32:
@@ -85,6 +89,8 @@ func (f FramebufferAttachmentDataFormat) GlFormat() uint32 {
 
 	case FramebufferAttachmentDataFormat_RGBA8:
 		fallthrough
+	case FramebufferAttachmentDataFormat_RGBAF16:
+		fallthrough
 	case FramebufferAttachmentDataFormat_SRGBA:
 		return gl.RGBA
 
@@ -100,6 +106,33 @@ func (f FramebufferAttachmentDataFormat) GlFormat() uint32 {
 	}
 }
 
+func (f FramebufferAttachmentDataFormat) GlComponentType() uint32 {
+
+	switch f {
+
+	case FramebufferAttachmentDataFormat_R32Int:
+		return gl.INT
+
+	case FramebufferAttachmentDataFormat_RGBA8:
+		fallthrough
+	case FramebufferAttachmentDataFormat_SRGBA:
+		return gl.UNSIGNED_BYTE
+
+	case FramebufferAttachmentDataFormat_RGBAF16:
+		// Seems this is fine to be float instead of half float
+		fallthrough
+	case FramebufferAttachmentDataFormat_DepthF32:
+		return gl.FLOAT
+
+	case FramebufferAttachmentDataFormat_Depth24Stencil8:
+		return gl.UNSIGNED_INT_24_8
+
+	default:
+		logging.ErrLog.Fatalf("unknown framebuffer attachment data format. Format=%d\n", f)
+		return 0
+	}
+}
+
 type FramebufferAttachment struct {
 	Id     uint32
 	Type   FramebufferAttachmentType
@@ -124,7 +157,7 @@ func (fbo *Framebuffer) BindWithViewport() {
 	gl.Viewport(0, 0, int32(fbo.Width), int32(fbo.Height))
 }
 
-// Clear calls gl.Clear with the fob's clear flags.
+// Clear calls gl.Clear with the fbo's clear flags.
 // Note that the fbo must be complete and bound.
 // Calling this without a bound fbo will clear something else, like your screen.
 func (fbo *Framebuffer) Clear() {
@@ -207,7 +240,17 @@ func (fbo *Framebuffer) NewColorAttachment(
 		}
 
 		gl.BindTexture(gl.TEXTURE_2D, a.Id)
-		gl.TexImage2D(gl.TEXTURE_2D, 0, attachFormat.GlInternalFormat(), int32(fbo.Width), int32(fbo.Height), 0, attachFormat.GlFormat(), gl.UNSIGNED_BYTE, nil)
+		gl.TexImage2D(
+			gl.TEXTURE_2D,
+			0,
+			attachFormat.GlInternalFormat(),
+			int32(fbo.Width),
+			int32(fbo.Height),
+			0,
+			attachFormat.GlFormat(),
+			attachFormat.GlComponentType(),
+			nil,
+		)
 
 		gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.LINEAR)
 		gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.LINEAR)
@@ -298,7 +341,17 @@ func (fbo *Framebuffer) NewDepthAttachment(
 		}
 
 		gl.BindTexture(gl.TEXTURE_2D, a.Id)
-		gl.TexImage2D(gl.TEXTURE_2D, 0, attachFormat.GlInternalFormat(), int32(fbo.Width), int32(fbo.Height), 0, attachFormat.GlFormat(), gl.FLOAT, nil)
+		gl.TexImage2D(
+			gl.TEXTURE_2D,
+			0,
+			attachFormat.GlInternalFormat(),
+			int32(fbo.Width),
+			int32(fbo.Height),
+			0,
+			attachFormat.GlFormat(),
+			attachFormat.GlComponentType(),
+			nil,
+		)
 
 		gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.NEAREST)
 		gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.NEAREST)
@@ -341,7 +394,17 @@ func (fbo *Framebuffer) NewDepthAttachment(
 
 		gl.BindTexture(gl.TEXTURE_CUBE_MAP, a.Id)
 		for i := 0; i < 6; i++ {
-			gl.TexImage2D(uint32(gl.TEXTURE_CUBE_MAP_POSITIVE_X+i), 0, attachFormat.GlInternalFormat(), int32(fbo.Width), int32(fbo.Height), 0, attachFormat.GlFormat(), gl.FLOAT, nil)
+			gl.TexImage2D(
+				uint32(gl.TEXTURE_CUBE_MAP_POSITIVE_X+i),
+				0,
+				attachFormat.GlInternalFormat(),
+				int32(fbo.Width),
+				int32(fbo.Height),
+				0,
+				attachFormat.GlFormat(),
+				attachFormat.GlComponentType(),
+				nil,
+			)
 		}
 
 		gl.TexParameteri(gl.TEXTURE_CUBE_MAP, gl.TEXTURE_MIN_FILTER, gl.NEAREST)
@@ -398,7 +461,7 @@ func (fbo *Framebuffer) NewDepthCubemapArrayAttachment(
 		6*numCubemaps,
 		0,
 		attachFormat.GlFormat(),
-		gl.FLOAT,
+		attachFormat.GlComponentType(),
 		nil,
 	)
 
@@ -455,7 +518,7 @@ func (fbo *Framebuffer) NewDepthTextureArrayAttachment(
 		numTextures,
 		0,
 		attachFormat.GlFormat(),
-		gl.FLOAT,
+		attachFormat.GlComponentType(),
 		nil,
 	)
 
@@ -513,7 +576,17 @@ func (fbo *Framebuffer) NewDepthStencilAttachment(
 		}
 
 		gl.BindTexture(gl.TEXTURE_2D, a.Id)
-		gl.TexImage2D(gl.TEXTURE_2D, 0, attachFormat.GlInternalFormat(), int32(fbo.Width), int32(fbo.Height), 0, attachFormat.GlFormat(), gl.UNSIGNED_INT_24_8, nil)
+		gl.TexImage2D(
+			gl.TEXTURE_2D,
+			0,
+			attachFormat.GlInternalFormat(),
+			int32(fbo.Width),
+			int32(fbo.Height),
+			0,
+			attachFormat.GlFormat(),
+			attachFormat.GlComponentType(),
+			nil,
+		)
 
 		gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.NEAREST)
 		gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.NEAREST)

buffers/index_buffer.go

@@ -27,9 +27,9 @@ func (ib *IndexBuffer) SetData(values []uint32) {
 	ib.IndexBufCount = int32(len(values))
 
 	if sizeInBytes == 0 {
-		gl.BufferData(gl.ELEMENT_ARRAY_BUFFER, 0, gl.Ptr(nil), BufUsage_Static.ToGL())
+		gl.BufferData(gl.ELEMENT_ARRAY_BUFFER, 0, gl.Ptr(nil), BufUsage_Static_Draw.ToGL())
 	} else {
-		gl.BufferData(gl.ELEMENT_ARRAY_BUFFER, sizeInBytes, gl.Ptr(&values[0]), BufUsage_Static.ToGL())
+		gl.BufferData(gl.ELEMENT_ARRAY_BUFFER, sizeInBytes, gl.Ptr(&values[0]), BufUsage_Static_Draw.ToGL())
 	}
 }
 

buffers/uniform_buffer.go

@@ -0,0 +1,747 @@
+package buffers
+
+import (
+	"math"
+	"reflect"
+
+	"github.com/bloeys/gglm/gglm"
+	"github.com/bloeys/nmage/assert"
+	"github.com/bloeys/nmage/logging"
+	"github.com/go-gl/gl/v4.1-core/gl"
+)
+
+type UniformBufferFieldInput struct {
+	Id   uint16
+	Type ElementType
+	// Count should be set in case this field is an array of type `[Count]Type`.
+	// Count=0 is valid and is equivalent to Count=1, which means the type is NOT an array, but a single field.
+	Count uint16
+
+	// Subfields is used when type is a struct, in which case it holds the fields of the struct.
+	// Ids do not have to be unique across structs.
+	Subfields []UniformBufferFieldInput
+}
+
+type UniformBufferField struct {
+	Id            uint16
+	AlignedOffset uint16
+	// Count should be set in case this field is an array of type `[Count]Type`.
+	// Count=0 is valid and is equivalent to Count=1, which means the type is NOT an array, but a single field.
+	Count uint16
+	Type  ElementType
+
+	// Subfields is used when type is a struct, in which case it holds the fields of the struct.
+	// Ids do not have to be unique across structs.
+	Subfields []UniformBufferField
+}
+
+type UniformBuffer struct {
+	Id uint32
+	// Size is the allocated memory in bytes on the GPU for this uniform buffer
+	Size   uint32
+	Fields []UniformBufferField
+}
+
+func (ub *UniformBuffer) Bind() {
+	gl.BindBuffer(gl.UNIFORM_BUFFER, ub.Id)
+}
+
+func (ub *UniformBuffer) UnBind() {
+	gl.BindBuffer(gl.UNIFORM_BUFFER, 0)
+}
+
+func (ub *UniformBuffer) SetBindPoint(bindPointIndex uint32) {
+	gl.BindBufferBase(gl.UNIFORM_BUFFER, bindPointIndex, ub.Id)
+}
+
+func addUniformBufferFieldsToArray(startAlignedOffset uint16, arrayToAddTo *[]UniformBufferField, fieldsToAdd []UniformBufferFieldInput) (totalSize uint32) {
+
+	if len(fieldsToAdd) == 0 {
+		return 0
+	}
+
+	// This function is recursive so only size the array once
+	if cap(*arrayToAddTo) == 0 {
+		*arrayToAddTo = make([]UniformBufferField, 0, len(fieldsToAdd))
+	}
+
+	var alignedOffset uint16 = 0
+	fieldIdToTypeMap := make(map[uint16]ElementType, len(fieldsToAdd))
+
+	for i := 0; i < len(fieldsToAdd); i++ {
+
+		f := fieldsToAdd[i]
+		if f.Count == 0 {
+			f.Count = 1
+		}
+
+		existingFieldType, ok := fieldIdToTypeMap[f.Id]
+		assert.T(!ok, "Uniform buffer field id is reused within the same uniform buffer. FieldId=%d was first used on a field with type=%s and then used on a different field with type=%s\n", f.Id, existingFieldType.String(), f.Type.String())
+
+		// To understand this take an example. Say we have a total offset of 100 and we are adding a vec4.
+		// Vec4s must be aligned to a 16 byte boundary but 100 is not (100 % 16 != 0).
+		//
+		// To fix this, we take the alignment error which is alignErr=100 % 16=4, but this is error to the nearest
+		// boundary, which is below the offset.
+		//
+		// To get the nearest boundary larger than the offset we can:
+		// offset + (boundary - alignErr) == 100 + (16 - 4) == 112; 112 % 16 == 0, meaning its a boundary
+		//
+		// Note that arrays of scalars/vectors are always aligned to 16 bytes, like a vec4
+		//
+		// Official spec and full details in subsection 'Standard Uniform Block Layout' at http://www.opengl.org/registry/specs/ARB/uniform_buffer_object.txt
+		var alignmentBoundary uint16 = 16
+		if f.Count == 1 {
+			alignmentBoundary = f.Type.GlStd140AlignmentBoundary()
+		}
+
+		alignmentError := alignedOffset % alignmentBoundary
+		if alignmentError != 0 {
+			alignedOffset += alignmentBoundary - alignmentError
+		}
+
+		newField := UniformBufferField{Id: f.Id, Type: f.Type, AlignedOffset: startAlignedOffset + alignedOffset, Count: f.Count}
+		*arrayToAddTo = append(*arrayToAddTo, newField)
+
+		// Prepare aligned offset for the next field.
+		//
+		// Matrices are treated as an array of column vectors, where each column is a vec4,
+		// that's why we have a multiplier depending on how many columns we have when calculating
+		// the offset
+		multiplier := uint16(1)
+		if f.Type == DataTypeMat2 {
+			multiplier = 2
+		} else if f.Type == DataTypeMat3 {
+			multiplier = 3
+		} else if f.Type == DataTypeMat4 {
+			multiplier = 4
+		}
+
+		if f.Type == DataTypeStruct {
+
+			subfieldsAlignedOffset := uint16(addUniformBufferFieldsToArray(startAlignedOffset+alignedOffset, arrayToAddTo, f.Subfields))
+
+			// Pad structs to 16 byte boundary
+			padTo16Boundary(&subfieldsAlignedOffset)
+			alignedOffset += subfieldsAlignedOffset * f.Count
+
+		} else {
+
+			// Elements advance the alignedOffset by their actual byte size.
+			// Aligned offset is padded if the place its at is not aligned to the boundary required by the next element.
+			//
+			// The exception is structs, because fields after a struct field are always aligned at a 16 byte boundary.
+			//
+			// For example, a vec3 starting at offset 80, taking 12 bytes, would put the aligned offset at 92.
+			// If the next element is a float32 (alignment boundary = 4) then no padding is required and
+			// the float will start at 92 and end at 96.
+			// However, if the element after the vec3 is a vec3 (alignment boundary = 16), then it would require
+			// a padding of 4 bytes so that it can start at 96, which is aligned to 16. In this case the second vec3
+			// would start at 96 and end at 96+12=108.
+			alignedOffset = newField.AlignedOffset + uint16(f.Type.GlStd140SizeBytes())*f.Count*multiplier - startAlignedOffset
+		}
+	}
+
+	return uint32(alignedOffset)
+}
+
+func padTo16Boundary[T uint16 | int | int32](val *T) {
+	alignmentError := *val % 16
+	if alignmentError != 0 {
+		*val += 16 - alignmentError
+	}
+}
+
+func (ub *UniformBuffer) getField(fieldId uint16, fieldType ElementType) UniformBufferField {
+
+	for i := 0; i < len(ub.Fields); i++ {
+
+		f := ub.Fields[i]
+
+		if f.Id != fieldId {
+			continue
+		}
+
+		assert.T(f.Type == fieldType, "Uniform buffer field id is reused within the same uniform buffer. FieldId=%d was first used on a field with type=%v, but is now being used on a field with type=%v\n", fieldId, f.Type.String(), fieldType.String())
+
+		return f
+	}
+
+	logging.ErrLog.Panicf("couldn't find uniform buffer field of id=%d and type=%s\n", fieldId, fieldType.String())
+	return UniformBufferField{}
+}
+
+func (ub *UniformBuffer) SetInt32(fieldId uint16, val int32) {
+
+	f := ub.getField(fieldId, DataTypeInt32)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4, gl.Ptr(&val))
+}
+
+func (ub *UniformBuffer) SetUint32(fieldId uint16, val uint32) {
+
+	f := ub.getField(fieldId, DataTypeUint32)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4, gl.Ptr(&val))
+}
+
+func (ub *UniformBuffer) SetFloat32(fieldId uint16, val float32) {
+
+	f := ub.getField(fieldId, DataTypeFloat32)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4, gl.Ptr(&val))
+}
+
+func (ub *UniformBuffer) SetVec2(fieldId uint16, val *gglm.Vec2) {
+	f := ub.getField(fieldId, DataTypeVec2)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4*2, gl.Ptr(&val.Data[0]))
+}
+
+func (ub *UniformBuffer) SetVec3(fieldId uint16, val *gglm.Vec3) {
+	f := ub.getField(fieldId, DataTypeVec3)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4*3, gl.Ptr(&val.Data[0]))
+}
+
+func (ub *UniformBuffer) SetVec4(fieldId uint16, val *gglm.Vec4) {
+	f := ub.getField(fieldId, DataTypeVec4)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4*4, gl.Ptr(&val.Data[0]))
+}
+
+func (ub *UniformBuffer) SetMat2(fieldId uint16, val *gglm.Mat2) {
+	f := ub.getField(fieldId, DataTypeMat2)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4*4, gl.Ptr(&val.Data[0][0]))
+}
+
+func (ub *UniformBuffer) SetMat3(fieldId uint16, val *gglm.Mat3) {
+	f := ub.getField(fieldId, DataTypeMat3)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4*9, gl.Ptr(&val.Data[0][0]))
+}
+
+func (ub *UniformBuffer) SetMat4(fieldId uint16, val *gglm.Mat4) {
+	f := ub.getField(fieldId, DataTypeMat4)
+	gl.BufferSubData(gl.UNIFORM_BUFFER, int(f.AlignedOffset), 4*16, gl.Ptr(&val.Data[0][0]))
+}
+
+func (ub *UniformBuffer) SetStruct(inputStruct any) {
+	setStruct(ub.Fields, make([]byte, ub.Size), inputStruct, 1000_000, false, 0)
+}
+
+func setStruct(fields []UniformBufferField, buf []byte, inputStruct any, maxFieldsToConsume int, onlyBufWrite bool, writeOffset int) (bytesWritten, fieldsConsumed int) {
+
+	if len(fields) == 0 {
+		return
+	}
+
+	if inputStruct == nil {
+		logging.ErrLog.Panicf("UniformBuffer.SetStruct called with a value that is nil")
+	}
+
+	structVal := reflect.ValueOf(inputStruct)
+	if structVal.Kind() != reflect.Struct {
+		logging.ErrLog.Panicf("UniformBuffer.SetStruct called with a value that is not a struct. Val=%v\n", inputStruct)
+	}
+
+	// Needed because fieldIndex can move faster than struct fields in case of struct fields
+	structFieldIndex := 0
+	for fieldIndex := 0; fieldIndex < len(fields) && fieldIndex < maxFieldsToConsume; fieldIndex++ {
+
+		ubField := &fields[fieldIndex]
+		valField := structVal.Field(structFieldIndex)
+
+		fieldsConsumed++
+		structFieldIndex++
+
+		kind := valField.Kind()
+		if kind == reflect.Pointer {
+			valField = valField.Elem()
+		}
+
+		var elementType reflect.Type
+		isArray := kind == reflect.Slice || kind == reflect.Array
+		if isArray {
+			elementType = valField.Type().Elem()
+			kind = elementType.Kind()
+		} else {
+			elementType = valField.Type()
+		}
+
+		if isArray {
+			assert.T(valField.Len() == int(ubField.Count), "ubo field of id=%d is an array/slice field of length=%d but got input of length=%d\n", ubField.Id, ubField.Count, valField.Len())
+		}
+
+		typeMatches := false
+		bytesWritten = int(ubField.AlignedOffset) + writeOffset
+
+		switch ubField.Type {
+
+		case DataTypeUint32:
+
+			typeMatches = elementType.Name() == "uint32"
+			if typeMatches {
+
+				if isArray {
+					Write32BitIntegerSliceToByteBufWithAlignment(buf, &bytesWritten, 16, valField.Slice(0, valField.Len()).Interface().([]uint32))
+				} else {
+					Write32BitIntegerToByteBuf(buf, &bytesWritten, uint32(valField.Uint()))
+				}
+			}
+
+		case DataTypeFloat32:
+
+			typeMatches = elementType.Name() == "float32"
+			if typeMatches {
+
+				if isArray {
+					WriteF32SliceToByteBufWithAlignment(buf, &bytesWritten, 16, valField.Slice(0, valField.Len()).Interface().([]float32))
+				} else {
+					WriteF32ToByteBuf(buf, &bytesWritten, float32(valField.Float()))
+				}
+			}
+
+		case DataTypeInt32:
+
+			typeMatches = elementType.Name() == "int32"
+			if typeMatches {
+
+				if isArray {
+					Write32BitIntegerSliceToByteBufWithAlignment(buf, &bytesWritten, 16, valField.Slice(0, valField.Len()).Interface().([]int32))
+				} else {
+					Write32BitIntegerToByteBuf(buf, &bytesWritten, uint32(valField.Int()))
+				}
+			}
+
+		case DataTypeVec2:
+
+			typeMatches = elementType.Name() == "Vec2"
+
+			if typeMatches {
+
+				if isArray {
+					WriteVec2SliceToByteBufWithAlignment(buf, &bytesWritten, 16, valField.Slice(0, valField.Len()).Interface().([]gglm.Vec2))
+				} else {
+					v2 := valField.Interface().(gglm.Vec2)
+					WriteF32SliceToByteBuf(buf, &bytesWritten, v2.Data[:])
+				}
+			}
+
+		case DataTypeVec3:
+
+			typeMatches = elementType.Name() == "Vec3"
+
+			if typeMatches {
+
+				if isArray {
+					WriteVec3SliceToByteBufWithAlignment(buf, &bytesWritten, 16, valField.Slice(0, valField.Len()).Interface().([]gglm.Vec3))
+				} else {
+					v3 := valField.Interface().(gglm.Vec3)
+					WriteF32SliceToByteBuf(buf, &bytesWritten, v3.Data[:])
+				}
+			}
+
+		case DataTypeVec4:
+
+			typeMatches = elementType.Name() == "Vec4"
+
+			if typeMatches {
+
+				if isArray {
+					WriteVec4SliceToByteBufWithAlignment(buf, &bytesWritten, 16, valField.Slice(0, valField.Len()).Interface().([]gglm.Vec4))
+				} else {
+					v3 := valField.Interface().(gglm.Vec4)
+					WriteF32SliceToByteBuf(buf, &bytesWritten, v3.Data[:])
+				}
+			}
+
+		case DataTypeMat2:
+
+			typeMatches = elementType.Name() == "Mat2"
+
+			if typeMatches {
+
+				if isArray {
+					m2Arr := valField.Interface().([]gglm.Mat2)
+					WriteMat2SliceToByteBufWithAlignment(buf, &bytesWritten, 16*2, m2Arr)
+				} else {
+					m := valField.Interface().(gglm.Mat2)
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[0][:])
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[1][:])
+				}
+			}
+
+		case DataTypeMat3:
+
+			typeMatches = elementType.Name() == "Mat3"
+
+			if typeMatches {
+
+				if isArray {
+					m3Arr := valField.Interface().([]gglm.Mat3)
+					WriteMat3SliceToByteBufWithAlignment(buf, &bytesWritten, 16*3, m3Arr)
+				} else {
+					m := valField.Interface().(gglm.Mat3)
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[0][:])
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[1][:])
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[2][:])
+				}
+			}
+
+		case DataTypeMat4:
+
+			typeMatches = elementType.Name() == "Mat4"
+
+			if typeMatches {
+
+				if isArray {
+					m4Arr := valField.Interface().([]gglm.Mat4)
+					WriteMat4SliceToByteBufWithAlignment(buf, &bytesWritten, 16*4, m4Arr)
+				} else {
+					m := valField.Interface().(gglm.Mat4)
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[0][:])
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[1][:])
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[2][:])
+					WriteF32SliceToByteBuf(buf, &bytesWritten, m.Data[3][:])
+				}
+			}
+
+		case DataTypeStruct:
+
+			typeMatches = kind == reflect.Struct
+
+			if typeMatches {
+
+				if isArray {
+
+					offset := 0
+					arrSize := valField.Len()
+					fieldsToUse := fields[fieldIndex+1:]
+					for i := 0; i < arrSize; i++ {
+
+						setStructBytesWritten, setStructFieldsConsumed := setStruct(fieldsToUse, buf, valField.Index(i).Interface(), elementType.NumField(), true, offset*i)
+
+						if offset == 0 {
+							offset = setStructBytesWritten
+							padTo16Boundary(&offset)
+
+							bytesWritten += offset * arrSize
+
+							// Tracking consumed fields is needed because if we have a struct inside another struct
+							// elementType.NumField() will only give us the fields consumed by the first struct,
+							// but we need to count all fields of all nested structs inside this one
+							fieldIndex += setStructFieldsConsumed
+							fieldsConsumed += setStructFieldsConsumed
+						}
+					}
+
+				} else {
+
+					setStructBytesWritten, setStructFieldsConsumed := setStruct(fields[fieldIndex+1:], buf, valField.Interface(), valField.NumField(), true, writeOffset)
+
+					bytesWritten += setStructBytesWritten
+					fieldIndex += setStructFieldsConsumed
+					fieldsConsumed += setStructFieldsConsumed
+				}
+			}
+
+		default:
+			assert.T(false, "Unknown uniform buffer data type passed. DataType '%d'", ubField.Type)
+		}
+
+		if !typeMatches {
+			logging.ErrLog.Panicf("Struct field ordering and types must match uniform buffer fields, but at field index %d got UniformBufferField=%v but a struct field of type %s\n", fieldIndex, ubField, valField.String())
+		}
+	}
+
+	if bytesWritten == 0 {
+		return 0, fieldsConsumed
+	}
+
+	if !onlyBufWrite {
+		gl.BufferSubData(gl.UNIFORM_BUFFER, 0, bytesWritten, gl.Ptr(&buf[0]))
+	}
+
+	return bytesWritten - int(fields[0].AlignedOffset) - writeOffset, fieldsConsumed
+}
+
+func Write32BitIntegerToByteBuf[T uint32 | int32](buf []byte, startIndex *int, val T) {
+
+	assert.T(*startIndex+4 <= len(buf), "failed to write uint32/int32 to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d", *startIndex, len(buf))
+
+	buf[*startIndex] = byte(val)
+	buf[*startIndex+1] = byte(val >> 8)
+	buf[*startIndex+2] = byte(val >> 16)
+	buf[*startIndex+3] = byte(val >> 24)
+
+	*startIndex += 4
+}
+
+func Write32BitIntegerSliceToByteBufWithAlignment[T uint32 | int32](buf []byte, startIndex *int, alignmentPerField int, vals []T) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerField <= len(buf), "failed to write uint32/int32 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerField, *startIndex, len(buf), len(vals)*alignmentPerField)
+
+	for i := 0; i < len(vals); i++ {
+
+		val := vals[i]
+
+		buf[*startIndex] = byte(val)
+		buf[*startIndex+1] = byte(val >> 8)
+		buf[*startIndex+2] = byte(val >> 16)
+		buf[*startIndex+3] = byte(val >> 24)
+
+		*startIndex += alignmentPerField
+	}
+}
+
+func WriteF32ToByteBuf(buf []byte, startIndex *int, val float32) {
+
+	assert.T(*startIndex+4 <= len(buf), "failed to write float32 to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d", *startIndex, len(buf))
+
+	bits := math.Float32bits(val)
+
+	buf[*startIndex] = byte(bits)
+	buf[*startIndex+1] = byte(bits >> 8)
+	buf[*startIndex+2] = byte(bits >> 16)
+	buf[*startIndex+3] = byte(bits >> 24)
+
+	*startIndex += 4
+}
+
+func WriteF32SliceToByteBuf(buf []byte, startIndex *int, vals []float32) {
+
+	assert.T(*startIndex+len(vals)*4 <= len(buf), "failed to write slice of float32 to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", *startIndex, len(buf), len(vals)*4)
+
+	for i := 0; i < len(vals); i++ {
+
+		bits := math.Float32bits(vals[i])
+
+		buf[*startIndex] = byte(bits)
+		buf[*startIndex+1] = byte(bits >> 8)
+		buf[*startIndex+2] = byte(bits >> 16)
+		buf[*startIndex+3] = byte(bits >> 24)
+
+		*startIndex += 4
+	}
+}
+
+func WriteF32SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerField int, vals []float32) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerField <= len(buf), "failed to write slice of float32 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerField, *startIndex, len(buf), len(vals)*alignmentPerField)
+
+	for i := 0; i < len(vals); i++ {
+
+		bits := math.Float32bits(vals[i])
+
+		buf[*startIndex] = byte(bits)
+		buf[*startIndex+1] = byte(bits >> 8)
+		buf[*startIndex+2] = byte(bits >> 16)
+		buf[*startIndex+3] = byte(bits >> 24)
+
+		*startIndex += alignmentPerField
+	}
+}
+
+func WriteVec2SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerVector int, vals []gglm.Vec2) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerVector <= len(buf), "failed to write slice of gglm.Vec2 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerVector, *startIndex, len(buf), len(vals)*alignmentPerVector)
+
+	for i := 0; i < len(vals); i++ {
+
+		bitsX := math.Float32bits(vals[i].X())
+		bitsY := math.Float32bits(vals[i].Y())
+
+		buf[*startIndex] = byte(bitsX)
+		buf[*startIndex+1] = byte(bitsX >> 8)
+		buf[*startIndex+2] = byte(bitsX >> 16)
+		buf[*startIndex+3] = byte(bitsX >> 24)
+
+		buf[*startIndex+4] = byte(bitsY)
+		buf[*startIndex+5] = byte(bitsY >> 8)
+		buf[*startIndex+6] = byte(bitsY >> 16)
+		buf[*startIndex+7] = byte(bitsY >> 24)
+
+		*startIndex += alignmentPerVector
+	}
+}
+
+func WriteVec3SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerVector int, vals []gglm.Vec3) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerVector <= len(buf), "failed to write slice of gglm.Vec3 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerVector, *startIndex, len(buf), len(vals)*alignmentPerVector)
+
+	for i := 0; i < len(vals); i++ {
+
+		bitsX := math.Float32bits(vals[i].X())
+		bitsY := math.Float32bits(vals[i].Y())
+		bitsZ := math.Float32bits(vals[i].Z())
+
+		buf[*startIndex] = byte(bitsX)
+		buf[*startIndex+1] = byte(bitsX >> 8)
+		buf[*startIndex+2] = byte(bitsX >> 16)
+		buf[*startIndex+3] = byte(bitsX >> 24)
+
+		buf[*startIndex+4] = byte(bitsY)
+		buf[*startIndex+5] = byte(bitsY >> 8)
+		buf[*startIndex+6] = byte(bitsY >> 16)
+		buf[*startIndex+7] = byte(bitsY >> 24)
+
+		buf[*startIndex+8] = byte(bitsZ)
+		buf[*startIndex+9] = byte(bitsZ >> 8)
+		buf[*startIndex+10] = byte(bitsZ >> 16)
+		buf[*startIndex+11] = byte(bitsZ >> 24)
+
+		*startIndex += alignmentPerVector
+	}
+}
+
+func WriteVec4SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerVector int, vals []gglm.Vec4) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerVector <= len(buf), "failed to write slice of gglm.Vec4 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerVector, *startIndex, len(buf), len(vals)*alignmentPerVector)
+
+	for i := 0; i < len(vals); i++ {
+
+		bitsX := math.Float32bits(vals[i].X())
+		bitsY := math.Float32bits(vals[i].Y())
+		bitsZ := math.Float32bits(vals[i].Z())
+		bitsW := math.Float32bits(vals[i].W())
+
+		buf[*startIndex] = byte(bitsX)
+		buf[*startIndex+1] = byte(bitsX >> 8)
+		buf[*startIndex+2] = byte(bitsX >> 16)
+		buf[*startIndex+3] = byte(bitsX >> 24)
+
+		buf[*startIndex+4] = byte(bitsY)
+		buf[*startIndex+5] = byte(bitsY >> 8)
+		buf[*startIndex+6] = byte(bitsY >> 16)
+		buf[*startIndex+7] = byte(bitsY >> 24)
+
+		buf[*startIndex+8] = byte(bitsZ)
+		buf[*startIndex+9] = byte(bitsZ >> 8)
+		buf[*startIndex+10] = byte(bitsZ >> 16)
+		buf[*startIndex+11] = byte(bitsZ >> 24)
+
+		buf[*startIndex+12] = byte(bitsW)
+		buf[*startIndex+13] = byte(bitsW >> 8)
+		buf[*startIndex+14] = byte(bitsW >> 16)
+		buf[*startIndex+15] = byte(bitsW >> 24)
+
+		*startIndex += alignmentPerVector
+	}
+}
+
+func WriteMat2SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerMatrix int, vals []gglm.Mat2) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerMatrix <= len(buf), "failed to write slice of gglm.Mat2 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerMatrix, *startIndex, len(buf), len(vals)*alignmentPerMatrix)
+
+	for i := 0; i < len(vals); i++ {
+
+		m := &vals[i]
+
+		WriteVec2SliceToByteBufWithAlignment(
+			buf,
+			startIndex,
+			16,
+			[]gglm.Vec2{
+				{Data: m.Data[0]},
+				{Data: m.Data[1]},
+			},
+		)
+	}
+}
+
+func WriteMat3SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerMatrix int, vals []gglm.Mat3) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerMatrix <= len(buf), "failed to write slice of gglm.Mat3 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerMatrix, *startIndex, len(buf), len(vals)*alignmentPerMatrix)
+
+	for i := 0; i < len(vals); i++ {
+
+		m := &vals[i]
+
+		WriteVec3SliceToByteBufWithAlignment(
+			buf,
+			startIndex,
+			16,
+			[]gglm.Vec3{
+				{Data: m.Data[0]},
+				{Data: m.Data[1]},
+				{Data: m.Data[2]},
+			},
+		)
+	}
+}
+
+func WriteMat4SliceToByteBufWithAlignment(buf []byte, startIndex *int, alignmentPerMatrix int, vals []gglm.Mat4) {
+
+	assert.T(*startIndex+len(vals)*alignmentPerMatrix <= len(buf), "failed to write slice of gglm.Mat2 with custom alignment=%d to buffer because the buffer doesn't have enough space. Start index=%d, Buffer length=%d, but needs %d bytes free", alignmentPerMatrix, *startIndex, len(buf), len(vals)*alignmentPerMatrix)
+
+	for i := 0; i < len(vals); i++ {
+
+		m := &vals[i]
+
+		WriteVec4SliceToByteBufWithAlignment(
+			buf,
+			startIndex,
+			16,
+			[]gglm.Vec4{
+				{Data: m.Data[0]},
+				{Data: m.Data[1]},
+				{Data: m.Data[2]},
+				{Data: m.Data[3]},
+			},
+		)
+	}
+}
+
+func ReflectValueMatchesUniformBufferField(v reflect.Value, ubField *UniformBufferField) bool {
+
+	if v.Kind() == reflect.Pointer {
+		v = v.Elem()
+	}
+
+	switch ubField.Type {
+
+	case DataTypeUint32:
+		t := v.Type()
+		return t.Name() == "uint32"
+	case DataTypeFloat32:
+		t := v.Type()
+		return t.Name() == "float32"
+	case DataTypeInt32:
+		t := v.Type()
+		return t.Name() == "int32"
+	case DataTypeVec2:
+		_, ok := v.Interface().(gglm.Vec2)
+		return ok
+	case DataTypeVec3:
+		_, ok := v.Interface().(gglm.Vec3)
+		return ok
+	case DataTypeVec4:
+		_, ok := v.Interface().(gglm.Vec4)
+		return ok
+	case DataTypeMat2:
+		_, ok := v.Interface().(gglm.Mat2)
+		return ok
+	case DataTypeMat3:
+		_, ok := v.Interface().(gglm.Mat3)
+		return ok
+	case DataTypeMat4:
+		_, ok := v.Interface().(gglm.Mat4)
+		return ok
+
+	default:
+		assert.T(false, "Unknown uniform buffer data type passed. DataType '%d'", ubField.Type)
+		return false
+	}
+}
+
+func NewUniformBuffer(fields []UniformBufferFieldInput, usage BufUsage) UniformBuffer {
+
+	ub := UniformBuffer{}
+
+	ub.Size = addUniformBufferFieldsToArray(0, &ub.Fields, fields)
+
+	gl.GenBuffers(1, &ub.Id)
+	if ub.Id == 0 {
+		logging.ErrLog.Panicln("Failed to create OpenGL buffer for a uniform buffer")
+	}
+
+	ub.Bind()
+	gl.BufferData(gl.UNIFORM_BUFFER, int(ub.Size), gl.Ptr(nil), usage.ToGL())
+	ub.UnBind()
+
+	return ub
+}

buffers/vertex_buffer.go

@@ -55,7 +55,7 @@ func NewVertexBuffer(layout ...Element) VertexBuffer {
 
 	gl.GenBuffers(1, &vb.Id)
 	if vb.Id == 0 {
-		logging.ErrLog.Println("Failed to create OpenGL buffer")
+		logging.ErrLog.Panicln("Failed to create OpenGL buffer")
 	}
 
 	vb.SetLayout(layout...)

camera/camera.go

@@ -41,7 +41,7 @@ func (c *Camera) Update() {
 	c.ViewMat = gglm.LookAtRH(&c.Pos, c.Pos.Clone().Add(&c.Forward), &c.WorldUp).Mat4
 
 	if c.Type == Type_Perspective {
-		c.ProjMat = *gglm.Perspective(c.Fov, c.AspectRatio, c.NearClip, c.FarClip)
+		c.ProjMat = gglm.Perspective(c.Fov, c.AspectRatio, c.NearClip, c.FarClip)
 	} else {
 		c.ProjMat = gglm.Ortho(c.Left, c.Right, c.Top, c.Bottom, c.NearClip, c.FarClip).Mat4
 	}
@@ -59,9 +59,9 @@ func (c *Camera) UpdateRotation(pitch, yaw float32) {
 	c.Update()
 }
 
-func NewPerspective(pos, forward, worldUp *gglm.Vec3, nearClip, farClip, fovRadians, aspectRatio float32) *Camera {
+func NewPerspective(pos, forward, worldUp *gglm.Vec3, nearClip, farClip, fovRadians, aspectRatio float32) Camera {
 
-	cam := &Camera{
+	cam := Camera{
 		Type:    Type_Perspective,
 		Pos:     *pos,
 		Forward: *forward,
@@ -78,9 +78,9 @@ func NewPerspective(pos, forward, worldUp *gglm.Vec3, nearClip, farClip, fovRadi
 	return cam
 }
 
-func NewOrthographic(pos, forward, worldUp *gglm.Vec3, nearClip, farClip, left, right, top, bottom float32) *Camera {
+func NewOrthographic(pos, forward, worldUp *gglm.Vec3, nearClip, farClip, left, right, top, bottom float32) Camera {
 
-	cam := &Camera{
+	cam := Camera{
 		Type:    Type_Orthographic,
 		Pos:     *pos,
 		Forward: *forward,

engine/engine.go

@@ -1,12 +1,14 @@
 package engine
 
 import (
+	"image"
+	"image/color"
 	"runtime"
 
 	imgui "github.com/AllenDang/cimgui-go"
 	"github.com/bloeys/nmage/assert"
+	"github.com/bloeys/nmage/assets"
 	"github.com/bloeys/nmage/input"
-	"github.com/bloeys/nmage/renderer"
 	"github.com/bloeys/nmage/timing"
 	nmageimgui "github.com/bloeys/nmage/ui/imgui"
 	"github.com/go-gl/gl/v4.1-core/gl"
@@ -28,30 +30,16 @@ type Window struct {
 	SDLWin         *sdl.Window
 	GlCtx          sdl.GLContext
 	EventCallbacks []func(sdl.Event)
-	Rend           renderer.Render
 }
 
 func (w *Window) handleInputs() {
 
-	input.EventLoopStart()
 	imIo := imgui.CurrentIO()
 
 	imguiCaptureMouse := imIo.WantCaptureMouse()
 	imguiCaptureKeyboard := imIo.WantCaptureKeyboard()
 
-	// These two are to fix a bug where state isn't cleared
+	input.EventLoopStart(imguiCaptureMouse, imguiCaptureKeyboard)
-	// even after imgui captures the keyboard/mouse.
-	//
-	// For example, if player is moving due to key held and then imgui captures the keyboard,
-	// the player keeps moving even when the key is no longer pressed because the input system never
-	// receives the key up event.
-	if imguiCaptureMouse {
-		input.ClearMouseState()
-	}
-
-	if imguiCaptureKeyboard {
-		input.ClearKeyboardState()
-	}
 
 	for event := sdl.PollEvent(); event != nil; event = sdl.PollEvent() {
 
@@ -65,46 +53,27 @@ func (w *Window) handleInputs() {
 
 		case *sdl.MouseWheelEvent:
 
-			if !imguiCaptureMouse {
+			input.HandleMouseWheelEvent(e)
-				input.HandleMouseWheelEvent(e)
-			}
-
 			imIo.AddMouseWheelDelta(float32(e.X), float32(e.Y))
 
 		case *sdl.KeyboardEvent:
 
-			if !imguiCaptureKeyboard {
+			input.HandleKeyboardEvent(e)
-				input.HandleKeyboardEvent(e)
+
-			}
+			// Send modifier key updates to imgui (based on the imgui SDL backend)
+			imIo.AddKeyEvent(imgui.ModCtrl, e.Keysym.Mod&sdl.KMOD_CTRL != 0)
+			imIo.AddKeyEvent(imgui.ModShift, e.Keysym.Mod&sdl.KMOD_SHIFT != 0)
+			imIo.AddKeyEvent(imgui.ModAlt, e.Keysym.Mod&sdl.KMOD_ALT != 0)
+			imIo.AddKeyEvent(imgui.ModSuper, e.Keysym.Mod&sdl.KMOD_GUI != 0)
 
 			imIo.AddKeyEvent(nmageimgui.SdlScancodeToImGuiKey(e.Keysym.Scancode), e.Type == sdl.KEYDOWN)
 
-			// Send modifier key updates to imgui
-			if e.Keysym.Sym == sdl.K_LCTRL || e.Keysym.Sym == sdl.K_RCTRL {
-				imIo.SetKeyCtrl(e.Type == sdl.KEYDOWN)
-			}
-
-			if e.Keysym.Sym == sdl.K_LSHIFT || e.Keysym.Sym == sdl.K_RSHIFT {
-				imIo.SetKeyShift(e.Type == sdl.KEYDOWN)
-			}
-
-			if e.Keysym.Sym == sdl.K_LALT || e.Keysym.Sym == sdl.K_RALT {
-				imIo.SetKeyAlt(e.Type == sdl.KEYDOWN)
-			}
-
-			if e.Keysym.Sym == sdl.K_LGUI || e.Keysym.Sym == sdl.K_RGUI {
-				imIo.SetKeySuper(e.Type == sdl.KEYDOWN)
-			}
-
 		case *sdl.TextInputEvent:
 			imIo.AddInputCharactersUTF8(e.GetText())
 
 		case *sdl.MouseButtonEvent:
 
-			if !imguiCaptureMouse {
+			input.HandleMouseBtnEvent(e)
-				input.HandleMouseBtnEvent(e)
-			}
-
 			isPressed := e.State == sdl.PRESSED
 
 			if e.Button == sdl.BUTTON_LEFT {
@@ -117,9 +86,7 @@ func (w *Window) handleInputs() {
 
 		case *sdl.MouseMotionEvent:
 
-			if !imguiCaptureMouse {
+			input.HandleMouseMotionEvent(e)
-				input.HandleMouseMotionEvent(e)
-			}
 
 		case *sdl.WindowEvent:
 
@@ -140,9 +107,9 @@ func (w *Window) handleInputs() {
 	}
 
 	// If a mouse press event came, always pass it as "mouse held this frame", so we don't miss click-release events that are shorter than 1 frame.
-	imIo.SetMouseButtonDown(imgui.MouseButtonLeft, isSdlButtonLeftDown)
+	imIo.SetMouseButtonDown(int(imgui.MouseButtonLeft), isSdlButtonLeftDown)
-	imIo.SetMouseButtonDown(imgui.MouseButtonRight, isSdlButtonRightDown)
+	imIo.SetMouseButtonDown(int(imgui.MouseButtonRight), isSdlButtonRightDown)
-	imIo.SetMouseButtonDown(imgui.MouseButtonMiddle, isSdlButtonMiddleDown)
+	imIo.SetMouseButtonDown(int(imgui.MouseButtonMiddle), isSdlButtonMiddleDown)
 }
 
 func (w *Window) handleWindowResize() {
@@ -201,42 +168,45 @@ func initSDL() error {
 	return nil
 }
 
-func CreateOpenGLWindow(title string, x, y, width, height int32, flags WindowFlags, rend renderer.Render) (*Window, error) {
+func CreateOpenGLWindow(title string, x, y, width, height int32, flags WindowFlags) (Window, error) {
-	return createWindow(title, x, y, width, height, WindowFlags_OPENGL|flags, rend)
+	return createWindow(title, x, y, width, height, WindowFlags_OPENGL|flags)
 }
 
-func CreateOpenGLWindowCentered(title string, width, height int32, flags WindowFlags, rend renderer.Render) (*Window, error) {
+func CreateOpenGLWindowCentered(title string, width, height int32, flags WindowFlags) (Window, error) {
-	return createWindow(title, sdl.WINDOWPOS_CENTERED, sdl.WINDOWPOS_CENTERED, width, height, WindowFlags_OPENGL|flags, rend)
+	return createWindow(title, sdl.WINDOWPOS_CENTERED, sdl.WINDOWPOS_CENTERED, width, height, WindowFlags_OPENGL|flags)
 }
 
-func createWindow(title string, x, y, width, height int32, flags WindowFlags, rend renderer.Render) (*Window, error) {
+func createWindow(title string, x, y, width, height int32, flags WindowFlags) (Window, error) {
 
 	assert.T(isInited, "engine.Init() was not called!")
 
-	sdlWin, err := sdl.CreateWindow(title, x, y, width, height, uint32(flags))
+	win := Window{
-	if err != nil {
+		SDLWin:         nil,
-		return nil, err
-	}
-
-	win := &Window{
-		SDLWin:         sdlWin,
 		EventCallbacks: make([]func(sdl.Event), 0),
-		Rend:           rend,
 	}
 
-	win.GlCtx, err = sdlWin.GLCreateContext()
+	var err error
+
+	win.SDLWin, err = sdl.CreateWindow(title, x, y, width, height, uint32(flags))
 	if err != nil {
-		return nil, err
+		return win, err
+	}
+
+	win.GlCtx, err = win.SDLWin.GLCreateContext()
+	if err != nil {
+		return win, err
 	}
 
 	err = initOpenGL()
 	if err != nil {
-		return nil, err
+		return win, err
 	}
 
+	setupDefaultTextures()
+
 	// Get rid of the blinding white startup screen (unfortunately there is still one frame of white)
 	gl.Clear(gl.COLOR_BUFFER_BIT | gl.DEPTH_BUFFER_BIT | gl.STENCIL_BUFFER_BIT)
-	sdlWin.GLSwap()
+	win.SDLWin.GLSwap()
 
 	return win, err
 }
@@ -263,6 +233,57 @@ func initOpenGL() error {
 	return nil
 }
 
+func setupDefaultTextures() error {
+
+	// 1x1 black texture
+	defaultBlackImg := image.NewNRGBA(image.Rect(0, 0, 1, 1))
+	defaultBlackImg.Set(0, 0, color.NRGBA{R: 0, G: 0, B: 0, A: 1})
+	defaultBlackImgTex, err := assets.LoadTextureInMemPngImg(defaultBlackImg, &assets.TextureLoadOptions{NoSrgba: true})
+	if err != nil {
+		return err
+	}
+	assets.DefaultBlackTexId = defaultBlackImgTex
+
+	// 1x1 white texture
+	defaultWhiteImg := image.NewNRGBA(image.Rect(0, 0, 1, 1))
+	defaultWhiteImg.Set(0, 0, color.NRGBA{R: 255, G: 255, B: 255, A: 1})
+	defaultWhiteImgTex, err := assets.LoadTextureInMemPngImg(defaultWhiteImg, &assets.TextureLoadOptions{NoSrgba: true})
+	if err != nil {
+		return err
+	}
+	assets.DefaultWhiteTexId = defaultWhiteImgTex
+
+	// Default diffuse
+	assets.DefaultDiffuseTexId = defaultWhiteImgTex
+
+	// Default specular
+	assets.DefaultSpecularTexId = defaultBlackImgTex
+
+	// Default Normal map which is created to be RGB(0.5,0.5,1), which when multiplied by TBN matrix gives the vertex normal.
+	// 128 is better than 127 for normal maps. See 'Flat Color' section here: http://wiki.polycount.com/wiki/Normal_map
+	// Basically, 127 can create seams while 128 looks correct
+	defaultNormalMapImg := image.NewNRGBA(image.Rect(0, 0, 1, 1))
+	defaultNormalMapImg.Set(0, 0, color.NRGBA{R: 128, G: 128, B: 255, A: 1})
+	defaultNormalMapTex, err := assets.LoadTextureInMemPngImg(defaultNormalMapImg, &assets.TextureLoadOptions{NoSrgba: true})
+	if err != nil {
+		return err
+	}
+
+	assets.DefaultNormalTexId = defaultNormalMapTex
+
+	// Default emission
+	assets.DefaultEmissionTexId = defaultBlackImgTex
+
+	assert.T(assets.DefaultBlackTexId.TexID != 0, "The default black texture handle is zero. Either texture wasn't created or handle wasn't updated")
+	assert.T(assets.DefaultWhiteTexId.TexID != 0, "The default white texture handle is zero. Either texture wasn't created or handle wasn't updated")
+	assert.T(assets.DefaultDiffuseTexId.TexID != 0, "The default diffuse texture handle is zero. Either texture wasn't created or handle wasn't updated")
+	assert.T(assets.DefaultSpecularTexId.TexID != 0, "The default specular texture handle is zero. Either texture wasn't created or handle wasn't updated")
+	assert.T(assets.DefaultNormalTexId.TexID != 0, "The default normal texture handle is zero. Either texture wasn't created or handle wasn't updated")
+	assert.T(assets.DefaultEmissionTexId.TexID != 0, "The default emission texture handle is zero. Either texture wasn't created or handle wasn't updated")
+
+	return nil
+}
+
 func SetSrgbFramebuffer(isEnabled bool) {
 
 	if isEnabled {

engine/game.go

@@ -1,6 +1,7 @@
 package engine
 
 import (
+	"github.com/bloeys/nmage/renderer"
 	"github.com/bloeys/nmage/timing"
 	nmageimgui "github.com/bloeys/nmage/ui/imgui"
 	"github.com/go-gl/gl/v4.1-core/gl"
@@ -20,7 +21,7 @@ type Game interface {
 	DeInit()
 }
 
-func Run(g Game, w *Window, ui nmageimgui.ImguiInfo) {
+func Run(g Game, w *Window, rend renderer.Render, ui nmageimgui.ImguiInfo) {
 
 	isRunning = true
 
@@ -40,7 +41,6 @@ func Run(g Game, w *Window, ui nmageimgui.ImguiInfo) {
 
 	for isRunning {
 
-		//PERF: Cache these
 		width, height = w.SDLWin.GetSize()
 		fbWidth, fbHeight = w.SDLWin.GLGetDrawableSize()
 
@@ -56,7 +56,7 @@ func Run(g Game, w *Window, ui nmageimgui.ImguiInfo) {
 		w.SDLWin.GLSwap()
 
 		g.FrameEnd()
-		w.Rend.FrameEnd()
+		rend.FrameEnd()
 		timing.FrameEnded()
 	}
 

go.mod

@@ -1,6 +1,6 @@
 module github.com/bloeys/nmage
 
-go 1.22
+go 1.23
 
 require github.com/veandco/go-sdl2 v0.4.35
 
@@ -8,12 +8,15 @@ require github.com/go-gl/gl v0.0.0-20211210172815-726fda9656d6
 
 require (
 	github.com/bloeys/assimp-go v0.4.4
-	github.com/bloeys/gglm v0.43.0
+	github.com/bloeys/gglm v0.50.0
 )
 
 require (
-	github.com/AllenDang/cimgui-go v0.0.0-20230720025235-f2ff398a66b2
+	github.com/AllenDang/cimgui-go v0.0.0-20240912193335-545751598105
 	github.com/mandykoh/prism v0.35.1
 )
 
-require github.com/mandykoh/go-parallel v0.1.0 // indirect
+require (
+	github.com/mandykoh/go-parallel v0.1.0 // indirect
+	golang.org/x/exp v0.0.0-20240506185415-9bf2ced13842 // indirect
+)

go.sum

@@ -1,9 +1,11 @@
 github.com/AllenDang/cimgui-go v0.0.0-20230720025235-f2ff398a66b2 h1:3HA/5qD8Rimxz/y1sLyVaM7ws1dzjXzMt4hOBiwHggo=
 github.com/AllenDang/cimgui-go v0.0.0-20230720025235-f2ff398a66b2/go.mod h1:iNfbIyOBN8k3XScMxULbrwYbPsXEAUD0Jb6UwrspQb8=
+github.com/AllenDang/cimgui-go v0.0.0-20240912193335-545751598105 h1:bhXqv2EG5YCLdgkLSFCpqeVz4cCoNbi4RDFrHrwxQ1o=
+github.com/AllenDang/cimgui-go v0.0.0-20240912193335-545751598105/go.mod h1:CYfBRenCaNtSvKVzChYh6gswUSo6c5IUcYeV6eCCRw0=
 github.com/bloeys/assimp-go v0.4.4 h1:Yn5e/RpE0Oes0YMBy8O7KkwAO4R/RpgrZPJCt08dVIU=
 github.com/bloeys/assimp-go v0.4.4/go.mod h1:my3yRxT7CfOztmvi+0svmwbaqw0KFrxaHxncoyaEIP0=
-github.com/bloeys/gglm v0.43.0 h1:ZpOghR3PHfpkigTDh+FqxLsF0gN8CD6s/bWoei6LyxI=
+github.com/bloeys/gglm v0.50.0 h1:DlGLp9z8KMNx+hNR6PjnPmC0HjDRC19QwAKL1iwhOxs=
-github.com/bloeys/gglm v0.43.0/go.mod h1:qwJQ0WzV191wAMwlGicbfbChbKoSedMk7gFFX6GnyOk=
+github.com/bloeys/gglm v0.50.0/go.mod h1:5s2U/NiOrtJyrSup1j8wK+QOBmGIO03ub0LHMvuNSK8=
 github.com/go-gl/gl v0.0.0-20211210172815-726fda9656d6 h1:zDw5v7qm4yH7N8C8uWd+8Ii9rROdgWxQuGoJ9WDXxfk=
 github.com/go-gl/gl v0.0.0-20211210172815-726fda9656d6/go.mod h1:9YTyiznxEY1fVinfM7RvRcjRHbw2xLBJ3AAGIT0I4Nw=
 github.com/mandykoh/go-parallel v0.1.0 h1:7vJMNMC4dsbgZdkAb2A8tV5ENY1v7VxIO1wzQWZoT8k=
@@ -15,6 +17,8 @@ github.com/veandco/go-sdl2 v0.4.35/go.mod h1:OROqMhHD43nT4/i9crJukyVecjPNYYuCofe
 github.com/yuin/goldmark v1.4.13/go.mod h1:6yULJ656Px+3vBD8DxQVa3kxgyrAnzto9xy5taEt/CY=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20210921155107-089bfa567519/go.mod h1:GvvjBRRGRdwPK5ydBHafDWAxML/pGHZbMvKqRZ5+Abc=
+golang.org/x/exp v0.0.0-20240506185415-9bf2ced13842 h1:vr/HnozRka3pE4EsMEg1lgkXJkTFJCVUX+S/ZT6wYzM=
+golang.org/x/exp v0.0.0-20240506185415-9bf2ced13842/go.mod h1:XtvwrStGgqGPLc4cjQfWqZHG1YFdYs6swckp8vpsjnc=
 golang.org/x/image v0.5.0 h1:5JMiNunQeQw++mMOz48/ISeNu3Iweh/JaZU8ZLqHRrI=
 golang.org/x/image v0.5.0/go.mod h1:FVC7BI/5Ym8R25iw5OLsgshdUBbT1h5jZTpA+mvAdZ4=
 golang.org/x/mod v0.6.0-dev.0.20220419223038-86c51ed26bb4/go.mod h1:jJ57K6gSWd91VN4djpZkiMVwK6gcyfeH4XE8wZrZaV4=

input/input.go

@@ -1,6 +1,23 @@
+// The input package provides an interface to mouse and keyboard inputs
+// like key clicks and releases, along with some higher level constructs like
+// pressed/released this frames, double clicks, and normalized inputs.
+//
+// The input package has two sets of functions for most cases, where one
+// is in the form 'xy' and the other 'xyCaptured'. The captured form
+// always returns normal events even if the mouse or keyboard are captured
+// by the UI system. The 'xy' form however will return zero/false if the
+// respective input device is currently captured (with the exception of mouse position, that is always correctly returned).
+//
+// For most cases, you want to use the 'xy' form. For example, you only want to receive
+// key down events for game character movement when the UI isn't capturing the keyboard,
+// because otherwise the character will move while typing in a UI textbox.
+//
+// The functions IsMouseCaptured and IsKeyboardCaptured are also available.
 package input
 
-import "github.com/veandco/go-sdl2/sdl"
+import (
+	"github.com/veandco/go-sdl2/sdl"
+)
 
 type keyState struct {
 	Key                 sdl.Keycode
@@ -31,15 +48,22 @@ type mouseWheelState struct {
 }
 
 var (
-	keyMap        = make(map[sdl.Keycode]keyState)
+	mouseWheel  = mouseWheelState{}
-	mouseBtnMap   = make(map[int]mouseBtnState)
+	mouseMotion = mouseMotionState{}
-	mouseMotion   = mouseMotionState{}
+	mouseBtnMap = make(map[int]mouseBtnState)
-	mouseWheel    = mouseWheelState{}
+	keyMap      = make(map[sdl.Keycode]keyState)
-	quitRequested bool
+
+	isQuitRequested    bool
+	isMouseCaptured    bool
+	isKeyboardCaptured bool
 )
 
-func EventLoopStart() {
+func EventLoopStart(mouseGotCaptured, keyboardGotCaptured bool) {
 
+	isMouseCaptured = mouseGotCaptured
+	isKeyboardCaptured = keyboardGotCaptured
+
+	// Update per-frame state
 	for k, v := range keyMap {
 		v.IsPressedThisFrame = false
 		v.IsReleasedThisFrame = false
@@ -59,7 +83,7 @@ func EventLoopStart() {
 	mouseWheel.XDelta = 0
 	mouseWheel.YDelta = 0
 
-	quitRequested = false
+	isQuitRequested = false
 }
 
 func ClearKeyboardState() {
@@ -73,11 +97,19 @@ func ClearMouseState() {
 }
 
 func HandleQuitEvent(e *sdl.QuitEvent) {
-	quitRequested = true
+	isQuitRequested = true
+}
+
+func IsMouseCaptured() bool {
+	return isMouseCaptured
+}
+
+func IsKeyboardCaptured() bool {
+	return isKeyboardCaptured
 }
 
 func IsQuitClicked() bool {
-	return quitRequested
+	return isQuitRequested
 }
 
 func HandleKeyboardEvent(e *sdl.KeyboardEvent) {
@@ -123,18 +155,36 @@ func HandleMouseWheelEvent(e *sdl.MouseWheelEvent) {
 	mouseWheel.YDelta = e.Y
 }
 
-// GetMousePos returns the window coordinates of the mouse
+// GetMousePos returns the window coordinates of the mouse regardless of whether the mouse is captured or not
 func GetMousePos() (x, y int32) {
 	return mouseMotion.XPos, mouseMotion.YPos
 }
 
 // GetMouseMotion returns how many pixels were moved last frame
 func GetMouseMotion() (xDelta, yDelta int32) {
+
+	if isMouseCaptured {
+		return 0, 0
+	}
+
+	return GetMouseMotionCaptured()
+}
+
+func GetMouseMotionCaptured() (xDelta, yDelta int32) {
 	return mouseMotion.XDelta, mouseMotion.YDelta
 }
 
 func GetMouseMotionNorm() (xDelta, yDelta int32) {
 
+	if isMouseCaptured {
+		return 0, 0
+	}
+
+	return GetMouseMotionNormCaptured()
+}
+
+func GetMouseMotionNormCaptured() (xDelta, yDelta int32) {
+
 	x, y := mouseMotion.XDelta, mouseMotion.YDelta
 	if x > 0 {
 		x = 1
@@ -152,12 +202,31 @@ func GetMouseMotionNorm() (xDelta, yDelta int32) {
 }
 
 func GetMouseWheelMotion() (xDelta, yDelta int32) {
+
+	if isMouseCaptured {
+		return 0, 0
+	}
+
+	return GetMouseWheelMotionCaptured()
+}
+
+func GetMouseWheelMotionCaptured() (xDelta, yDelta int32) {
 	return mouseWheel.XDelta, mouseWheel.YDelta
 }
 
 // GetMouseWheelXNorm returns 1 if mouse wheel xDelta > 0, -1 if xDelta < 0, and 0 otherwise
 func GetMouseWheelXNorm() int32 {
 
+	if isMouseCaptured {
+		return 0
+	}
+
+	return GetMouseWheelXNormCaptured()
+}
+
+// GetMouseWheelXNormCaptured returns 1 if mouse wheel xDelta > 0, -1 if xDelta < 0, and 0 otherwise
+func GetMouseWheelXNormCaptured() int32 {
+
 	if mouseWheel.XDelta > 0 {
 		return 1
 	} else if mouseWheel.XDelta < 0 {
@@ -167,9 +236,19 @@ func GetMouseWheelXNorm() int32 {
 	return 0
 }
 
-// returns 1 if mouse wheel yDelta > 0, -1 if yDelta < 0, and 0 otherwise
+// GetMouseWheelYNorm returns 1 if mouse wheel yDelta > 0, -1 if yDelta < 0, and 0 otherwise
 func GetMouseWheelYNorm() int32 {
 
+	if isMouseCaptured {
+		return 0
+	}
+
+	return GetMouseWheelYNormCaptured()
+}
+
+// GetMouseWheelYNormCaptured returns 1 if mouse wheel yDelta > 0, -1 if yDelta < 0, and 0 otherwise
+func GetMouseWheelYNormCaptured() int32 {
+
 	if mouseWheel.YDelta > 0 {
 		return 1
 	} else if mouseWheel.YDelta < 0 {
@@ -181,6 +260,15 @@ func GetMouseWheelYNorm() int32 {
 
 func KeyClicked(kc sdl.Keycode) bool {
 
+	if isKeyboardCaptured {
+		return false
+	}
+
+	return KeyClickedCaptured(kc)
+}
+
+func KeyClickedCaptured(kc sdl.Keycode) bool {
+
 	ks, ok := keyMap[kc]
 	if !ok {
 		return false
@@ -191,6 +279,15 @@ func KeyClicked(kc sdl.Keycode) bool {
 
 func KeyReleased(kc sdl.Keycode) bool {
 
+	if isKeyboardCaptured {
+		return false
+	}
+
+	return KeyReleasedCaptured(kc)
+}
+
+func KeyReleasedCaptured(kc sdl.Keycode) bool {
+
 	ks, ok := keyMap[kc]
 	if !ok {
 		return false
@@ -201,6 +298,15 @@ func KeyReleased(kc sdl.Keycode) bool {
 
 func KeyDown(kc sdl.Keycode) bool {
 
+	if isKeyboardCaptured {
+		return false
+	}
+
+	return KeyDownCaptured(kc)
+}
+
+func KeyDownCaptured(kc sdl.Keycode) bool {
+
 	ks, ok := keyMap[kc]
 	if !ok {
 		return false
@@ -211,6 +317,15 @@ func KeyDown(kc sdl.Keycode) bool {
 
 func KeyUp(kc sdl.Keycode) bool {
 
+	if isKeyboardCaptured {
+		return false
+	}
+
+	return KeyUpCaptured(kc)
+}
+
+func KeyUpCaptured(kc sdl.Keycode) bool {
+
 	ks, ok := keyMap[kc]
 	if !ok {
 		return true
@@ -221,6 +336,15 @@ func KeyUp(kc sdl.Keycode) bool {
 
 func MouseClicked(mb int) bool {
 
+	if isMouseCaptured {
+		return false
+	}
+
+	return MouseClickedCaptued(mb)
+}
+
+func MouseClickedCaptued(mb int) bool {
+
 	btn, ok := mouseBtnMap[mb]
 	if !ok {
 		return false
@@ -231,6 +355,15 @@ func MouseClicked(mb int) bool {
 
 func MouseDoubleClicked(mb int) bool {
 
+	if isMouseCaptured {
+		return false
+	}
+
+	return MouseDoubleClickedCaptured(mb)
+}
+
+func MouseDoubleClickedCaptured(mb int) bool {
+
 	btn, ok := mouseBtnMap[mb]
 	if !ok {
 		return false
@@ -240,6 +373,16 @@ func MouseDoubleClicked(mb int) bool {
 }
 
 func MouseReleased(mb int) bool {
+
+	if isMouseCaptured {
+		return false
+	}
+
+	return MouseReleasedCaptured(mb)
+}
+
+func MouseReleasedCaptured(mb int) bool {
+
 	btn, ok := mouseBtnMap[mb]
 	if !ok {
 		return false
@@ -250,6 +393,15 @@ func MouseReleased(mb int) bool {
 
 func MouseDown(mb int) bool {
 
+	if isMouseCaptured {
+		return false
+	}
+
+	return MouseDownCaptued(mb)
+}
+
+func MouseDownCaptued(mb int) bool {
+
 	btn, ok := mouseBtnMap[mb]
 	if !ok {
 		return false
@@ -260,6 +412,15 @@ func MouseDown(mb int) bool {
 
 func MouseUp(mb int) bool {
 
+	if isMouseCaptured {
+		return false
+	}
+
+	return MouseUpCaptured(mb)
+}
+
+func MouseUpCaptured(mb int) bool {
+
 	btn, ok := mouseBtnMap[mb]
 	if !ok {
 		return true

materials/material.go

@@ -1,13 +1,26 @@
 package materials
 
 import (
+	_ "unsafe"
+
 	"github.com/bloeys/gglm/gglm"
 	"github.com/bloeys/nmage/assert"
+	"github.com/bloeys/nmage/assets"
 	"github.com/bloeys/nmage/logging"
 	"github.com/bloeys/nmage/shaders"
 	"github.com/go-gl/gl/v4.1-core/gl"
 )
 
+// @TODO: This noescape magic is to avoid heap allocations done when
+// passing vectors or matrices into cgo via set uniform calls.
+//
+// But I would rather this kind of stuff is done on the gl wrapper level.
+// Should we wrap the OpenGL APIs we use ourself?
+
+var (
+	lastMatId uint32
+)
+
 type TextureSlot uint32
 
 const (
@@ -21,14 +34,36 @@ const (
 	TextureSlot_ShadowMap_Array1 TextureSlot = 13
 )
 
+type MaterialSettings uint64
+
+const (
+	MaterialSettings_None        MaterialSettings = iota
+	MaterialSettings_HasModelMtx MaterialSettings = 1 << (iota - 1)
+	MaterialSettings_HasNormalMtx
+)
+
+func (ms *MaterialSettings) Set(flags MaterialSettings) {
+	*ms |= flags
+}
+
+func (ms *MaterialSettings) Remove(flags MaterialSettings) {
+	*ms &= ^flags
+}
+
+func (ms *MaterialSettings) Has(flags MaterialSettings) bool {
+	return *ms&flags == flags
+}
+
 type Material struct {
+	Id         uint32
 	Name       string
 	ShaderProg shaders.ShaderProgram
+	Settings   MaterialSettings
 
 	UnifLocs   map[string]int32
 	AttribLocs map[string]int32
 
-	// @TODO do this in a better way. Perhaps something like how we do fbo attachments
+	// @TODO: Do this in a better way?. Perhaps something like how we do fbo attachments? Or keep it?
 	// Phong shading
 	DiffuseTex  uint32
 	SpecularTex uint32
@@ -51,26 +86,19 @@ func (m *Material) Bind() {
 
 	m.ShaderProg.Bind()
 
-	if m.DiffuseTex != 0 {
+	gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Diffuse))
-		gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Diffuse))
+	gl.BindTexture(gl.TEXTURE_2D, m.DiffuseTex)
-		gl.BindTexture(gl.TEXTURE_2D, m.DiffuseTex)
-	}
 
-	if m.SpecularTex != 0 {
+	gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Specular))
-		gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Specular))
+	gl.BindTexture(gl.TEXTURE_2D, m.SpecularTex)
-		gl.BindTexture(gl.TEXTURE_2D, m.SpecularTex)
-	}
 
-	if m.NormalTex != 0 {
+	gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Normal))
-		gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Normal))
+	gl.BindTexture(gl.TEXTURE_2D, m.NormalTex)
-		gl.BindTexture(gl.TEXTURE_2D, m.NormalTex)
-	}
 
-	if m.EmissionTex != 0 {
+	gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Emission))
-		gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Emission))
+	gl.BindTexture(gl.TEXTURE_2D, m.EmissionTex)
-		gl.BindTexture(gl.TEXTURE_2D, m.EmissionTex)
-	}
 
+	// @TODO: Have defaults for these
 	if m.CubemapTex != 0 {
 		gl.ActiveTexture(uint32(gl.TEXTURE0 + TextureSlot_Cubemap))
 		gl.BindTexture(gl.TEXTURE_CUBE_MAP, m.CubemapTex)
@@ -96,6 +124,21 @@ func (m *Material) UnBind() {
 	gl.UseProgram(0)
 }
 
+func (m *Material) SetUniformBlockBindingPoint(uniformBlockName string, bindPointIndex uint32) {
+
+	nullStr := gl.Str(uniformBlockName + "\x00")
+	index := gl.GetUniformBlockIndex(m.ShaderProg.Id, nullStr)
+	assert.T(
+		index != gl.INVALID_INDEX,
+		"SetUniformBlockBindingPoint for material=%s (matId=%d; shaderId=%d) failed because the uniform block=%s wasn't found",
+		m.Name,
+		m.Id,
+		m.ShaderProg.Id,
+		uniformBlockName,
+	)
+	gl.UniformBlockBinding(m.ShaderProg.Id, index, bindPointIndex)
+}
+
 func (m *Material) GetAttribLoc(attribName string) int32 {
 
 	loc, ok := m.AttribLocs[attribName]
@@ -103,7 +146,8 @@ func (m *Material) GetAttribLoc(attribName string) int32 {
 		return loc
 	}
 
-	loc = gl.GetAttribLocation(m.ShaderProg.Id, gl.Str(attribName+"\x00"))
+	name := gl.Str(attribName + "\x00")
+	loc = gl.GetAttribLocation(m.ShaderProg.Id, name)
 	assert.T(loc != -1, "Attribute '"+attribName+"' doesn't exist on material "+m.Name)
 	m.AttribLocs[attribName] = loc
 	return loc
@@ -116,7 +160,8 @@ func (m *Material) GetUnifLoc(uniformName string) int32 {
 		return loc
 	}
 
-	loc = gl.GetUniformLocation(m.ShaderProg.Id, gl.Str(uniformName+"\x00"))
+	name := gl.Str(uniformName + "\x00")
+	loc = gl.GetUniformLocation(m.ShaderProg.Id, name)
 	assert.T(loc != -1, "Uniform '"+uniformName+"' doesn't exist on material "+m.Name)
 	m.UnifLocs[uniformName] = loc
 	return loc
@@ -139,49 +184,124 @@ func (m *Material) SetUnifFloat32(uniformName string, val float32) {
 }
 
 func (m *Material) SetUnifVec2(uniformName string, vec2 *gglm.Vec2) {
-	gl.ProgramUniform2fv(m.ShaderProg.Id, m.GetUnifLoc(uniformName), 1, &vec2.Data[0])
+	internalSetUnifVec2(m.ShaderProg.Id, m.GetUnifLoc(uniformName), vec2)
+}
+
+//go:noescape
+//go:linkname internalSetUnifVec2 github.com/bloeys/nmage/materials.SetUnifVec2
+func internalSetUnifVec2(shaderProgId uint32, unifLoc int32, vec2 *gglm.Vec2)
+
+func SetUnifVec2(shaderProgId uint32, unifLoc int32, vec2 *gglm.Vec2) {
+	gl.ProgramUniform2fv(shaderProgId, unifLoc, 1, &vec2.Data[0])
 }
 
 func (m *Material) SetUnifVec3(uniformName string, vec3 *gglm.Vec3) {
-	gl.ProgramUniform3fv(m.ShaderProg.Id, m.GetUnifLoc(uniformName), 1, &vec3.Data[0])
+	internalSetUnifVec3(m.ShaderProg.Id, m.GetUnifLoc(uniformName), vec3)
+}
+
+//go:noescape
+//go:linkname internalSetUnifVec3 github.com/bloeys/nmage/materials.SetUnifVec3
+func internalSetUnifVec3(shaderProgId uint32, unifLoc int32, vec3 *gglm.Vec3)
+
+func SetUnifVec3(shaderProgId uint32, unifLoc int32, vec3 *gglm.Vec3) {
+	gl.ProgramUniform3fv(shaderProgId, unifLoc, 1, &vec3.Data[0])
 }
 
 func (m *Material) SetUnifVec4(uniformName string, vec4 *gglm.Vec4) {
-	gl.ProgramUniform4fv(m.ShaderProg.Id, m.GetUnifLoc(uniformName), 1, &vec4.Data[0])
+	internalSetUnifVec4(m.ShaderProg.Id, m.GetUnifLoc(uniformName), vec4)
+}
+
+//go:noescape
+//go:linkname internalSetUnifVec4 github.com/bloeys/nmage/materials.SetUnifVec4
+func internalSetUnifVec4(shaderProgId uint32, unifLoc int32, vec4 *gglm.Vec4)
+
+func SetUnifVec4(shaderProgId uint32, unifLoc int32, vec4 *gglm.Vec4) {
+	gl.ProgramUniform4fv(shaderProgId, unifLoc, 1, &vec4.Data[0])
 }
 
 func (m *Material) SetUnifMat2(uniformName string, mat2 *gglm.Mat2) {
-	gl.ProgramUniformMatrix2fv(m.ShaderProg.Id, m.GetUnifLoc(uniformName), 1, false, &mat2.Data[0][0])
+	internalSetUnifMat2(m.ShaderProg.Id, m.GetUnifLoc(uniformName), mat2)
+}
+
+//go:noescape
+//go:linkname internalSetUnifMat2 github.com/bloeys/nmage/materials.SetUnifMat2
+func internalSetUnifMat2(shaderProgId uint32, unifLoc int32, mat2 *gglm.Mat2)
+
+func SetUnifMat2(shaderProgId uint32, unifLoc int32, mat2 *gglm.Mat2) {
+	gl.ProgramUniformMatrix2fv(shaderProgId, unifLoc, 1, false, &mat2.Data[0][0])
 }
 
 func (m *Material) SetUnifMat3(uniformName string, mat3 *gglm.Mat3) {
-	gl.ProgramUniformMatrix3fv(m.ShaderProg.Id, m.GetUnifLoc(uniformName), 1, false, &mat3.Data[0][0])
+	internalSetUnifMat3(m.ShaderProg.Id, m.GetUnifLoc(uniformName), mat3)
+}
+
+//go:noescape
+//go:linkname internalSetUnifMat3 github.com/bloeys/nmage/materials.SetUnifMat3
+func internalSetUnifMat3(shaderProgId uint32, unifLoc int32, mat3 *gglm.Mat3)
+
+func SetUnifMat3(shaderProgId uint32, unifLoc int32, mat3 *gglm.Mat3) {
+	gl.ProgramUniformMatrix3fv(shaderProgId, unifLoc, 1, false, &mat3.Data[0][0])
 }
 
 func (m *Material) SetUnifMat4(uniformName string, mat4 *gglm.Mat4) {
-	gl.ProgramUniformMatrix4fv(m.ShaderProg.Id, m.GetUnifLoc(uniformName), 1, false, &mat4.Data[0][0])
+	internalSetUnifMat4(m.ShaderProg.Id, m.GetUnifLoc(uniformName), mat4)
+}
+
+//go:noescape
+//go:linkname internalSetUnifMat4 github.com/bloeys/nmage/materials.SetUnifMat4
+func internalSetUnifMat4(shaderProgId uint32, unifLoc int32, mat4 *gglm.Mat4)
+
+func SetUnifMat4(shaderProgId uint32, unifLoc int32, mat4 *gglm.Mat4) {
+	gl.ProgramUniformMatrix4fv(shaderProgId, unifLoc, 1, false, &mat4.Data[0][0])
 }
 
 func (m *Material) Delete() {
 	gl.DeleteProgram(m.ShaderProg.Id)
 }
 
-func NewMaterial(matName, shaderPath string) *Material {
+func getNewMatId() uint32 {
+	lastMatId++
+	return lastMatId
+}
+
+func NewMaterial(matName, shaderPath string) Material {
 
 	shdrProg, err := shaders.LoadAndCompileCombinedShader(shaderPath)
 	if err != nil {
 		logging.ErrLog.Fatalf("Failed to create new material '%s'. Err: %s\n", matName, err.Error())
 	}
 
-	return &Material{Name: matName, ShaderProg: shdrProg, UnifLocs: make(map[string]int32), AttribLocs: make(map[string]int32)}
+	return Material{
+		Id:         getNewMatId(),
+		Name:       matName,
+		ShaderProg: shdrProg,
+		UnifLocs:   make(map[string]int32),
+		AttribLocs: make(map[string]int32),
+
+		DiffuseTex:  assets.DefaultDiffuseTexId.TexID,
+		SpecularTex: assets.DefaultSpecularTexId.TexID,
+		NormalTex:   assets.DefaultNormalTexId.TexID,
+		EmissionTex: assets.DefaultEmissionTexId.TexID,
+	}
 }
 
-func NewMaterialSrc(matName string, shaderSrc []byte) *Material {
+func NewMaterialSrc(matName string, shaderSrc []byte) Material {
 
 	shdrProg, err := shaders.LoadAndCompileCombinedShaderSrc(shaderSrc)
 	if err != nil {
 		logging.ErrLog.Fatalf("Failed to create new material '%s'. Err: %s\n", matName, err.Error())
 	}
 
-	return &Material{Name: matName, ShaderProg: shdrProg, UnifLocs: make(map[string]int32), AttribLocs: make(map[string]int32)}
+	return Material{
+		Id:         getNewMatId(),
+		Name:       matName,
+		ShaderProg: shdrProg,
+		UnifLocs:   make(map[string]int32),
+		AttribLocs: make(map[string]int32),
+
+		DiffuseTex:  assets.DefaultDiffuseTexId.TexID,
+		SpecularTex: assets.DefaultSpecularTexId.TexID,
+		NormalTex:   assets.DefaultNormalTexId.TexID,
+		EmissionTex: assets.DefaultEmissionTexId.TexID,
+	}
 }

meshes/mesh.go

@@ -16,24 +16,48 @@ type SubMesh struct {
 }
 
 type Mesh struct {
-	Name      string
+	Name string
+	/*
+		Vao has the following shader attribute layout:
+			- Loc0: Pos
+			- Loc1: Normal
+			- Loc2: UV0
+			- Loc3: Tangent
+			- (Optional) Color
+
+		Optional stuff appear in the order in this list, depending on what other optional stuff exists.
+
+		For example:
+			- If color exists it will be in Loc3, otherwise it is unset
+	*/
 	Vao       buffers.VertexArray
 	SubMeshes []SubMesh
 }
 
-func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (*Mesh, error) {
+var (
+	// DefaultMeshLoadFlags are the flags always applied when loading a new mesh regardless
+	// of what post process flags are used when loading a mesh.
+	//
+	// Defaults to: asig.PostProcessTriangulate | asig.PostProcessCalcTangentSpace;
+	// Note: changing this will break the normal lit shaders, which expect tangents to be there
+	DefaultMeshLoadFlags asig.PostProcess = asig.PostProcessTriangulate | asig.PostProcessCalcTangentSpace
+)
 
-	scene, release, err := asig.ImportFile(modelPath, asig.PostProcessTriangulate|postProcessFlags)
+func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (Mesh, error) {
+
+	finalPostProcessFlags := DefaultMeshLoadFlags | postProcessFlags
+
+	scene, release, err := asig.ImportFile(modelPath, finalPostProcessFlags)
 	if err != nil {
-		return nil, errors.New("Failed to load model. Err: " + err.Error())
+		return Mesh{}, errors.New("Failed to load model. Err: " + err.Error())
 	}
 	defer release()
 
 	if len(scene.Meshes) == 0 {
-		return nil, errors.New("No meshes found in file: " + modelPath)
+		return Mesh{}, errors.New("No meshes found in file: " + modelPath)
 	}
 
-	mesh := &Mesh{
+	mesh := Mesh{
 		Name:      name,
 		Vao:       buffers.NewVertexArray(),
 		SubMeshes: make([]SubMesh, 0, 1),
@@ -42,8 +66,17 @@ func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (*Mesh,
 	vbo := buffers.NewVertexBuffer()
 	ibo := buffers.NewIndexBuffer()
 
-	// Initial sizes assuming one submesh that has vertex pos+normals+texCoords, and 3 indices per face
+	// Estimate a useful prealloc capacity based on the first submesh that has vertex pos+normals+tangents+texCoords
-	var vertexBufData []float32 = make([]float32, 0, len(scene.Meshes[0].Vertices)*3*3*2)
+	vertexBufDataCapacity := len(scene.Meshes[0].Vertices) * 3 * 3 * 3 * 2
+
+	// Increase capacity depending on what the mesh has
+	if len(scene.Meshes[0].ColorSets) > 0 && len(scene.Meshes[0].ColorSets[0]) > 0 {
+		vertexBufDataCapacity *= 4
+	}
+
+	var vertexBufData []float32 = make([]float32, 0, vertexBufDataCapacity)
+
+	// Initial size assumes 3 indices per face
 	var indexBufData []uint32 = make([]uint32, 0, len(scene.Meshes[0].Faces)*3)
 
 	// fmt.Printf("\nMesh %s has %d meshe(s) with first mesh having %d vertices\n", name, len(scene.Meshes), len(scene.Meshes[0].Vertices))
@@ -52,12 +85,25 @@ func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (*Mesh,
 
 		sceneMesh := scene.Meshes[i]
 
+		// We always want tangents and UV0
+		if len(sceneMesh.Tangents) == 0 {
+			sceneMesh.Tangents = make([]gglm.Vec3, len(sceneMesh.Vertices))
+		}
+
 		if len(sceneMesh.TexCoords[0]) == 0 {
 			sceneMesh.TexCoords[0] = make([]gglm.Vec3, len(sceneMesh.Vertices))
 		}
 
-		layoutToUse := []buffers.Element{{ElementType: buffers.DataTypeVec3}, {ElementType: buffers.DataTypeVec3}, {ElementType: buffers.DataTypeVec2}}
+		hasColorSet0 := len(sceneMesh.ColorSets) > 0 && len(sceneMesh.ColorSets[0]) > 0
-		if len(sceneMesh.ColorSets) > 0 && len(sceneMesh.ColorSets[0]) > 0 {
+
+		layoutToUse := []buffers.Element{
+			{ElementType: buffers.DataTypeVec3}, // Position
+			{ElementType: buffers.DataTypeVec3}, // Normals
+			{ElementType: buffers.DataTypeVec3}, // Tangents
+			{ElementType: buffers.DataTypeVec2}, // UV0
+		}
+
+		if hasColorSet0 {
 			layoutToUse = append(layoutToUse, buffers.Element{ElementType: buffers.DataTypeVec4})
 		}
 
@@ -79,8 +125,14 @@ func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (*Mesh,
 			}
 		}
 
-		arrs := []arrToInterleave{{V3s: sceneMesh.Vertices}, {V3s: sceneMesh.Normals}, {V2s: v3sToV2s(sceneMesh.TexCoords[0])}}
+		arrs := []arrToInterleave{
-		if len(sceneMesh.ColorSets) > 0 && len(sceneMesh.ColorSets[0]) > 0 {
+			{V3s: sceneMesh.Vertices},
+			{V3s: sceneMesh.Normals},
+			{V3s: sceneMesh.Tangents},
+			{V2s: v3sToV2s(sceneMesh.TexCoords[0])},
+		}
+
+		if hasColorSet0 {
 			arrs = append(arrs, arrToInterleave{V4s: sceneMesh.ColorSets[0]})
 		}
 
@@ -99,7 +151,7 @@ func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (*Mesh,
 		indexBufData = append(indexBufData, indices...)
 	}
 
-	vbo.SetData(vertexBufData, buffers.BufUsage_Static)
+	vbo.SetData(vertexBufData, buffers.BufUsage_Static_Draw)
 	ibo.SetData(indexBufData)
 
 	mesh.Vao.AddVertexBuffer(vbo)

renderer/rend3dgl/rend3dgl.go

@@ -12,24 +12,31 @@ import (
 var _ renderer.Render = &Rend3DGL{}
 
 type Rend3DGL struct {
-	BoundVao  *buffers.VertexArray
+	BoundVaoId     uint32
-	BoundMesh *meshes.Mesh
+	BoundMatId     uint32
-	BoundMat  *materials.Material
+	BoundMeshVaoId uint32
 }
 
 func (r *Rend3DGL) DrawMesh(mesh *meshes.Mesh, modelMat *gglm.TrMat, mat *materials.Material) {
 
-	if mesh != r.BoundMesh {
+	if mesh.Vao.Id != r.BoundMeshVaoId {
 		mesh.Vao.Bind()
-		r.BoundMesh = mesh
+		r.BoundMeshVaoId = mesh.Vao.Id
 	}
 
-	if mat != r.BoundMat {
+	if mat.Id != r.BoundMatId {
 		mat.Bind()
-		r.BoundMat = mat
+		r.BoundMatId = mat.Id
 	}
 
-	mat.SetUnifMat4("modelMat", &modelMat.Mat4)
+	if mat.Settings.Has(materials.MaterialSettings_HasModelMtx) {
+		mat.SetUnifMat4("modelMat", &modelMat.Mat4)
+	}
+
+	if mat.Settings.Has(materials.MaterialSettings_HasNormalMtx) {
+		normalMat := modelMat.Clone().InvertAndTranspose().ToMat3()
+		mat.SetUnifMat3("normalMat", &normalMat)
+	}
 
 	for i := 0; i < len(mesh.SubMeshes); i++ {
 		gl.DrawElementsBaseVertexWithOffset(gl.TRIANGLES, mesh.SubMeshes[i].IndexCount, gl.UNSIGNED_INT, uintptr(mesh.SubMeshes[i].BaseIndex), mesh.SubMeshes[i].BaseVertex)
@@ -38,14 +45,14 @@ func (r *Rend3DGL) DrawMesh(mesh *meshes.Mesh, modelMat *gglm.TrMat, mat *materi
 
 func (r *Rend3DGL) DrawVertexArray(mat *materials.Material, vao *buffers.VertexArray, firstElement int32, elementCount int32) {
 
-	if vao != r.BoundVao {
+	if vao.Id != r.BoundVaoId {
 		vao.Bind()
-		r.BoundVao = vao
+		r.BoundVaoId = vao.Id
 	}
 
-	if mat != r.BoundMat {
+	if mat.Id != r.BoundMatId {
 		mat.Bind()
-		r.BoundMat = mat
+		r.BoundMatId = mat.Id
 	}
 
 	gl.DrawArrays(gl.TRIANGLES, firstElement, elementCount)
@@ -53,14 +60,14 @@ func (r *Rend3DGL) DrawVertexArray(mat *materials.Material, vao *buffers.VertexA
 
 func (r *Rend3DGL) DrawCubemap(mesh *meshes.Mesh, mat *materials.Material) {
 
-	if mesh != r.BoundMesh {
+	if mesh.Vao.Id != r.BoundMeshVaoId {
 		mesh.Vao.Bind()
-		r.BoundMesh = mesh
+		r.BoundMeshVaoId = mesh.Vao.Id
 	}
 
-	if mat != r.BoundMat {
+	if mat.Id != r.BoundMatId {
 		mat.Bind()
-		r.BoundMat = mat
+		r.BoundMatId = mat.Id
 	}
 
 	for i := 0; i < len(mesh.SubMeshes); i++ {
@@ -69,8 +76,9 @@ func (r *Rend3DGL) DrawCubemap(mesh *meshes.Mesh, mat *materials.Material) {
 }
 
 func (r3d *Rend3DGL) FrameEnd() {
-	r3d.BoundMesh = nil
+	r3d.BoundVaoId = 0
-	r3d.BoundMat = nil
+	r3d.BoundMatId = 0
+	r3d.BoundMeshVaoId = 0
 }
 
 func NewRend3DGL() *Rend3DGL {

res/shaders/debug-depth.glsl

@@ -2,22 +2,19 @@
 #version 410
 
 layout(location=0) in vec3 vertPosIn;
-layout(location=1) in vec3 vertNormalIn;
+layout(location=2) in vec3 vertTangentIn;
-layout(location=2) in vec2 vertUV0In;
+layout(location=3) in vec2 vertUV0In;
-layout(location=3) in vec3 vertColorIn;
+layout(location=4) in vec3 vertColorIn;
 
-out vec3 vertNormal;
 out vec2 vertUV0;
 out vec3 vertColor;
 out vec3 fragPos;
 
-//MVP = Model View Projection
 uniform mat4 modelMat;
 uniform mat4 projViewMat;
 
 void main()
 {
-    vertNormal = mat3(transpose(inverse(modelMat))) * vertNormalIn;
     vertUV0 = vertUV0In;
     vertColor = vertColorIn;
 
@@ -31,7 +28,6 @@ void main()
 #version 410
 
 in vec3 vertColor;
-in vec3 vertNormal;
 in vec2 vertUV0;
 in vec3 fragPos;
 

res/shaders/simple-unlit.glsl

@@ -3,26 +3,19 @@
 
 layout(location=0) in vec3 vertPosIn;
 layout(location=1) in vec3 vertNormalIn;
-layout(location=2) in vec2 vertUV0In;
+layout(location=2) in vec3 vertTangentIn;
-layout(location=3) in vec3 vertColorIn;
+layout(location=3) in vec2 vertUV0In;
+layout(location=4) in vec3 vertColorIn;
 
-out vec3 vertNormal;
 out vec2 vertUV0;
 out vec3 vertColor;
 out vec3 fragPos;
 
-//MVP = Model View Projection
 uniform mat4 modelMat;
 uniform mat4 projViewMat;
 
 void main()
 {
-    // @TODO: Calculate this on the CPU and send it as a uniform
-    //
-    // This produces the normal matrix that multiplies with the model normal to produce the
-    // world space normal. Based on 'One last thing' section from: https://learnopengl.com/Lighting/Basic-Lighting
-    vertNormal = mat3(transpose(inverse(modelMat))) * vertNormalIn;
-    
     vertUV0 = vertUV0In;
     vertColor = vertColorIn;
 
@@ -41,7 +34,6 @@ struct Material {
 uniform Material material;
 
 in vec3 vertColor;
-in vec3 vertNormal;
 in vec2 vertUV0;
 in vec3 fragPos;
 

res/shaders/simple.glsl

@@ -1,82 +1,39 @@
 //shader:vertex
 #version 410
 
+#define NUM_SPOT_LIGHTS 4
+#define NUM_POINT_LIGHTS 8
+
+//
+// Inputs
+//
 layout(location=0) in vec3 vertPosIn;
 layout(location=1) in vec3 vertNormalIn;
-layout(location=2) in vec2 vertUV0In;
+layout(location=2) in vec3 vertTangentIn;
-layout(location=3) in vec3 vertColorIn;
+layout(location=3) in vec2 vertUV0In;
-
+layout(location=4) in vec3 vertColorIn;
-uniform mat4 modelMat;
-uniform mat4 projViewMat;
-uniform mat4 dirLightProjViewMat;
-
-#define NUM_SPOT_LIGHTS 4
-uniform mat4 spotLightProjViewMats[NUM_SPOT_LIGHTS];
-
-out vec3 vertNormal;
-out vec2 vertUV0;
-out vec3 vertColor;
-out vec3 fragPos;
-out vec4 fragPosDirLight;
-out vec4 fragPosSpotLight[NUM_SPOT_LIGHTS];
-
-void main()
-{
-    // @TODO: Calculate this on the CPU and send it as a uniform
-    //
-    // This produces the normal matrix that multiplies with the model normal to produce the
-    // world space normal. Based on 'One last thing' section from: https://learnopengl.com/Lighting/Basic-Lighting
-    vertNormal = mat3(transpose(inverse(modelMat))) * vertNormalIn;
-    
-    vertUV0 = vertUV0In;
-    vertColor = vertColorIn;
-
-    vec4 modelVert = modelMat * vec4(vertPosIn, 1);
-    fragPos = modelVert.xyz;
-    fragPosDirLight = dirLightProjViewMat * vec4(fragPos, 1);
-
-    for (int i = 0; i < NUM_SPOT_LIGHTS; i++)
-        fragPosSpotLight[i] = spotLightProjViewMats[i] * vec4(fragPos, 1);
-
-    gl_Position = projViewMat * modelVert;
-}
-
-//shader:fragment
-#version 410
-
-struct Material {
-    sampler2D diffuse;
-    sampler2D specular;
-    // sampler2D normal;
-    sampler2D emission;
-    float shininess;
-};
-
-uniform Material material;
 
+//
+// UBOs
+//
 struct DirLight {
     vec3 dir;
     vec3 diffuseColor;
     vec3 specularColor;
-    sampler2D shadowMap;
 };
-
+uniform sampler2D dirLightShadowMap;
-uniform DirLight dirLight;
 
 struct PointLight {
     vec3 pos;
     vec3 diffuseColor;
     vec3 specularColor;
-    float constant;
+    float radius;
-    float linear;
+    float falloff;
-    float quadratic;
+    float maxBias;
+    float nearPlane;
     float farPlane;
 };
 
-#define NUM_POINT_LIGHTS 8
-uniform PointLight pointLights[NUM_POINT_LIGHTS];
-uniform samplerCubeArray pointLightCubeShadowMaps;
-
 struct SpotLight {
     vec3 pos;
     vec3 dir;
@@ -86,36 +43,182 @@ struct SpotLight {
     float outerCutoff;
 };
 
+layout (std140) uniform GlobalMatrices {
+    vec3 camPos;
+    mat4 projViewMat;
+};
+
+layout (std140) uniform Lights {
+    DirLight dirLight;
+    PointLight pointLights[NUM_POINT_LIGHTS];
+    SpotLight spotLights[NUM_SPOT_LIGHTS];
+    vec3 ambientColor;
+};
+
+//
+// Uniforms
+//
+uniform mat4 modelMat;
+uniform mat4 dirLightProjViewMat;
+uniform mat4 spotLightProjViewMats[NUM_SPOT_LIGHTS];
+
+//
+// Outputs
+//
+out vec2 vertUV0;
+out vec3 vertColor;
+
+out vec3 fragPos;
+out vec3 fragPosDirLight;
+out vec4 fragPosSpotLight[NUM_SPOT_LIGHTS];
+
+out vec3 tangentCamPos;
+out vec3 tangentFragPos;
+out vec3 tangentDirLightDir;
+out vec3 tangentSpotLightPositions[NUM_SPOT_LIGHTS];
+out vec3 tangentSpotLightDirections[NUM_SPOT_LIGHTS];
+out vec3 tangentPointLightPositions[NUM_POINT_LIGHTS];
+
+void main()
+{
+    vertUV0 = vertUV0In;
+    vertColor = vertColorIn;
+    vec4 modelVert = modelMat * vec4(vertPosIn, 1);
+
+    // Tangent-BiTangent-Normal matrix for normal mapping
+    vec3 T = normalize(vec3(modelMat * vec4(vertTangentIn,   0.0)));
+    vec3 N = normalize(vec3(modelMat * vec4(vertNormalIn,    0.0)));
+
+    // Ensure T is orthogonal with respect to N
+    T = normalize(T - dot(T, N) * N);
+
+    vec3 B = cross(N, T);
+    mat3 tbnMtx = transpose(mat3(T, B, N));
+
+    // Lighting related
+    fragPos = modelVert.xyz;
+    fragPosDirLight = vec3(dirLightProjViewMat * vec4(fragPos, 1));
+
+    tangentCamPos = tbnMtx * camPos;
+    tangentFragPos = tbnMtx * fragPos;
+    tangentDirLightDir = tbnMtx * dirLight.dir;
+
+    for (int i = 0; i < NUM_POINT_LIGHTS; i++)
+        tangentPointLightPositions[i] = tbnMtx * pointLights[i].pos;
+
+    for (int i = 0; i < NUM_SPOT_LIGHTS; i++)
+    {
+        fragPosSpotLight[i] = spotLightProjViewMats[i] * vec4(fragPos, 1);
+
+        tangentSpotLightPositions[i] = tbnMtx * spotLights[i].pos;
+        tangentSpotLightDirections[i] = tbnMtx * spotLights[i].dir;
+    }
+
+    gl_Position = projViewMat * modelVert;
+}
+
+//shader:fragment
+#version 410
+
+/*
+    Note that while all lighting calculations are done in tangent space,
+    shadow mapping is done in world space.
+
+    The exception is the bias calculation. Since the bias relies on the normal
+    and the normal is in tangent space, we use a tangent space fragment position
+    with it, but the rest of shadow processing is in world space.
+*/
+
 #define NUM_SPOT_LIGHTS 4
-uniform SpotLight spotLights[NUM_SPOT_LIGHTS];
+#define NUM_POINT_LIGHTS 8
-uniform sampler2DArray spotLightShadowMaps;
 
-uniform vec3 camPos;
+//
-uniform vec3 ambientColor = vec3(0.2, 0.2, 0.2);
+// Inputs
-
+//
-in vec3 vertColor;
-in vec3 vertNormal;
-in vec2 vertUV0;
 in vec3 fragPos;
-in vec4 fragPosDirLight;
+in vec2 vertUV0;
+in vec3 vertColor;
+in vec3 fragPosDirLight;
 in vec4 fragPosSpotLight[NUM_SPOT_LIGHTS];
 
+in vec3 tangentCamPos;
+in vec3 tangentFragPos;
+in vec3 tangentDirLightDir;
+in vec3 tangentSpotLightPositions[NUM_SPOT_LIGHTS];
+in vec3 tangentSpotLightDirections[NUM_SPOT_LIGHTS];
+in vec3 tangentPointLightPositions[NUM_POINT_LIGHTS];
+
+//
+// Uniforms
+//
+struct Material {
+    sampler2D diffuse;
+    sampler2D specular;
+    sampler2D normal;
+    sampler2D emission;
+    float shininess;
+};
+uniform Material material;
+
+struct DirLight {
+    vec3 dir;
+    vec3 diffuseColor;
+    vec3 specularColor;
+};
+uniform sampler2D dirLightShadowMap;
+
+struct PointLight {
+    vec3 pos;
+    vec3 diffuseColor;
+    vec3 specularColor;
+    float radius;
+    float falloff;
+    float maxBias;
+    float nearPlane;
+    float farPlane;
+};
+uniform samplerCubeArray pointLightCubeShadowMaps;
+
+struct SpotLight {
+    vec3 pos;
+    vec3 dir;
+    vec3 diffuseColor;
+    vec3 specularColor;
+    float innerCutoff;
+    float outerCutoff;
+};
+uniform sampler2DArray spotLightShadowMaps;
+
+layout (std140) uniform GlobalMatrices {
+    vec3 camPos;
+    mat4 projViewMat;
+};
+
+layout (std140) uniform Lights {
+    DirLight dirLight;
+    PointLight pointLights[NUM_POINT_LIGHTS];
+    SpotLight spotLights[NUM_SPOT_LIGHTS];
+    vec3 ambientColor;
+};
+
+//
+// Outputs
+//
 out vec4 fragColor;
 
+//
 // Global variables used as cache for lighting calculations
+//
+vec3 tangentViewDir;
 vec4 diffuseTexColor;
 vec4 specularTexColor;
 vec4 emissionTexColor;
 vec3 normalizedVertNorm;
-vec3 viewDir;
 
-float CalcDirShadow(sampler2D shadowMap, vec3 lightDir)
+float CalcDirShadow(sampler2D shadowMap, vec3 tangentLightDir)
 {
-    // Move from clip space to NDC
-    vec3 projCoords = fragPosDirLight.xyz / fragPosDirLight.w;
-
     // Move from [-1,1] to [0, 1]
-    projCoords = projCoords * 0.5 + 0.5;
+    vec3 projCoords = fragPosDirLight * 0.5 + 0.5;
 
     // If sampling outside the depth texture then force 'no shadow'
     if(projCoords.z > 1)
@@ -126,7 +229,7 @@ float CalcDirShadow(sampler2D shadowMap, vec3 lightDir)
 
     // Bias in the range [0.005, 0.05] depending on the angle, where a higher
     // angle gives a higher bias, as shadow acne gets worse with angle
-    float bias = max(0.05 * (1 - dot(normalizedVertNorm, lightDir)), 0.005);
+    float bias = max(0.05 * (1 - dot(normalizedVertNorm, tangentLightDir)), 0.005);
 
     // 'Percentage Close Filtering'.
     // Basically get soft shadows by averaging this texel and surrounding ones
@@ -151,70 +254,126 @@ float CalcDirShadow(sampler2D shadowMap, vec3 lightDir)
 
 vec3 CalcDirLight()
 {
-    vec3 lightDir = normalize(-dirLight.dir);
+    vec3 lightDir = normalize(-tangentDirLightDir);
 
     // Diffuse
     float diffuseAmount = max(0.0, dot(normalizedVertNorm, lightDir));
     vec3 finalDiffuse = diffuseAmount * dirLight.diffuseColor * diffuseTexColor.rgb;
 
     // Specular
-    vec3 halfwayDir = normalize(lightDir + viewDir);
+    vec3 halfwayDir = normalize(lightDir + tangentViewDir);
     float specularAmount = pow(max(dot(normalizedVertNorm, halfwayDir), 0.0), material.shininess);
     vec3 finalSpecular = specularAmount * dirLight.specularColor * specularTexColor.rgb;
 
     // Shadow
-    float shadow = CalcDirShadow(dirLight.shadowMap, lightDir);
+    float shadow = CalcDirShadow(dirLightShadowMap, lightDir);
 
     return (finalDiffuse + finalSpecular) * (1 - shadow);
 }
 
-float CalcPointShadow(int lightIndex, vec3 lightPos, vec3 lightDir, float farPlane) {
+float CalcPointShadow(int lightIndex, vec3 worldLightPos, vec3 tangentLightDir, float maxBias, float nearPlane, float farPlane) {
 
-    vec3 lightToFrag = fragPos - lightPos;
+    vec3 lightToFrag = fragPos - worldLightPos;
+
+    // Get depth of current fragment
+    float currentDepth = length(lightToFrag);
+
+    if (currentDepth < nearPlane) {
+        return 0;
+    }
 
     float closestDepth = texture(pointLightCubeShadowMaps, vec4(lightToFrag, lightIndex)).r;
 
     // We stored depth in the cubemap in the range [0, 1], so now we move back to [0, farPlane]
     closestDepth *= farPlane;
 
-    // Get depth of current fragment
+    float bias = max(maxBias * (1 - dot(normalizedVertNorm, tangentLightDir)), 0.005);
-    float currentDepth = length(lightToFrag);
 
-    float bias = max(0.05 * (1 - dot(normalizedVertNorm, lightDir)), 0.005);
+    float shadow = currentDepth - bias > closestDepth ? 1 : 0;
-    float shadow = currentDepth -  bias > closestDepth ? 1.0 : 0.0;
 
     return shadow;
 }
 
+//
+// The following point light attenuation formulas
+// are from https://lisyarus.github.io/blog/posts/point-light-attenuation.html
+//
+// I found them more intuitive than the standard implementation and it also ensures
+// we have zero light at the selected distance.
+//
+float sqr(float x)
+{
+    return x * x;
+}
+
+// This version doesn't have a harsh cutoff at radius
+float AttenuateNoCusp(float dist, float radius, float falloff)
+{
+    // Since we only use this as attenuation and max intensity defines
+    // the max output value, anything more than 1 would increase
+    // the output of the light, which I don't think makes sense for
+    // our attenuation purposes.
+    //
+    // Seems to me this can be done simply by increasing color values above 255. 
+    //
+    // Forcing to 1 for now.
+    #define MAX_INTENSITY 1
+
+    float s = dist / radius;
+
+    if (s >= 1.0)
+        return 0.0;
+
+    float s2 = sqr(s);
+
+    return MAX_INTENSITY * sqr(1 - s2) / (1 + falloff * s2);
+}
+
+// This version has a harsh/immediate cutoff at radius
+float AttenuateCusp(float dist, float radius, float falloff)
+{
+    #define MAX_INTENSITY 1
+
+    float s = dist / radius;
+
+    if (s >= 1.0)
+        return 0.0;
+
+    float s2 = sqr(s);
+
+    return MAX_INTENSITY * sqr(1 - s2) / (1 + falloff * s);
+}
+
 vec3 CalcPointLight(PointLight pointLight, int lightIndex)
 {
-    // Ignore unset lights
+    // Ignore inactive lights
-    if (pointLight.constant == 0){
+    if (pointLight.radius == 0){
         return vec3(0);
     }
 
-    vec3 lightDir = normalize(pointLight.pos - fragPos);
+    vec3 tangentLightPos = tangentPointLightPositions[lightIndex];
+    vec3 tangentLightDir = normalize(tangentLightPos - tangentFragPos);
 
     // Diffuse
-    float diffuseAmount = max(0.0, dot(normalizedVertNorm, lightDir));
+    float diffuseAmount = max(0.0, dot(normalizedVertNorm, tangentLightDir));
     vec3 finalDiffuse = diffuseAmount * pointLight.diffuseColor * diffuseTexColor.rgb;
 
     // Specular
-    vec3 halfwayDir = normalize(lightDir + viewDir);
+    vec3 halfwayDir = normalize(tangentLightDir + tangentViewDir);
     float specularAmount = pow(max(dot(normalizedVertNorm, halfwayDir), 0.0), material.shininess);
     vec3 finalSpecular = specularAmount * pointLight.specularColor * specularTexColor.rgb;
 
     // Attenuation
-    float distToLight = length(pointLight.pos - fragPos);
+    float distToLight = length(tangentLightPos - tangentFragPos);
-    float attenuation = 1 / (pointLight.constant + pointLight.linear * distToLight + pointLight.quadratic * (distToLight * distToLight));  
+    float attenuation = AttenuateNoCusp(distToLight, pointLight.radius, pointLight.falloff);
 
     // Shadow
-    float shadow = CalcPointShadow(lightIndex, pointLight.pos, lightDir, pointLight.farPlane);
+    float shadow = CalcPointShadow(lightIndex, pointLight.pos, tangentLightDir, pointLight.maxBias, pointLight.nearPlane, pointLight.farPlane);
 
     return (finalDiffuse + finalSpecular) * attenuation * (1 - shadow);
 }
 
-float CalcSpotShadow(vec3 lightDir, int lightIndex)
+float CalcSpotShadow(vec3 tangentLightDir, int lightIndex)
 {
     // Move from clip space to NDC
     vec3 projCoords = fragPosSpotLight[lightIndex].xyz / fragPosSpotLight[lightIndex].w;
@@ -231,7 +390,7 @@ float CalcSpotShadow(vec3 lightDir, int lightIndex)
 
     // Bias in the range [0.005, 0.05] depending on the angle, where a higher
     // angle gives a higher bias, as shadow acne gets worse with angle
-    float bias = max(0.05 * (1 - dot(normalizedVertNorm, lightDir)), 0.005);
+    float bias = max(0.05 * (1 - dot(normalizedVertNorm, tangentLightDir)), 0.005);
 
     // 'Percentage Close Filtering'.
     // Basically get soft shadows by averaging this texel and surrounding ones
@@ -256,15 +415,19 @@ float CalcSpotShadow(vec3 lightDir, int lightIndex)
 
 vec3 CalcSpotLight(SpotLight light, int lightIndex)
 {
-    if (light.innerCutoff == 0)
+    // The inner/outer cutoffs are cosine values,
+    // which means a value of 1 is mainly produced when the input
+    // is 0 degrees or radians. cos(180) will also be 1, but that's too much :)
+    if (light.innerCutoff == 1)
         return vec3(0);
 
-    vec3 fragToLightDir = normalize(light.pos - fragPos);
+    vec3 tangentLightDir = tangentSpotLightDirections[lightIndex];
+    vec3 fragToLightDir = normalize(tangentSpotLightPositions[lightIndex] - tangentFragPos);
 
     // Spot light cone with full intensity within inner cutoff,
     // and falloff between inner-outer cutoffs, and zero
     // light after outer cutoff
-    float theta = dot(fragToLightDir, normalize(-light.dir));
+    float theta = dot(fragToLightDir, normalize(-tangentLightDir));
     float epsilon = (light.innerCutoff - light.outerCutoff);
     float intensity = clamp((theta - light.outerCutoff) / epsilon, float(0), float(1));
 
@@ -276,7 +439,7 @@ vec3 CalcSpotLight(SpotLight light, int lightIndex)
     vec3 finalDiffuse = diffuseAmount * light.diffuseColor * diffuseTexColor.rgb;
 
     // Specular
-    vec3 halfwayDir = normalize(fragToLightDir + viewDir);
+    vec3 halfwayDir = normalize(fragToLightDir + tangentViewDir);
     float specularAmount = pow(max(dot(normalizedVertNorm, halfwayDir), 0.0), material.shininess);
     vec3 finalSpecular = specularAmount * light.specularColor * specularTexColor.rgb;
 
@@ -286,15 +449,21 @@ vec3 CalcSpotLight(SpotLight light, int lightIndex)
     return (finalDiffuse + finalSpecular) * intensity * (1 - shadow);
 }
 
+#define DRAW_NORMALS false
+
 void main()
 {
     // Shared values
+    tangentViewDir = normalize(tangentCamPos - tangentFragPos);
     diffuseTexColor = texture(material.diffuse, vertUV0);
     specularTexColor = texture(material.specular, vertUV0);
     emissionTexColor = texture(material.emission, vertUV0);
 
-    normalizedVertNorm = normalize(vertNormal);
+    // Read normal data encoded [0,1]
-    viewDir = normalize(camPos - fragPos);
+    normalizedVertNorm = texture(material.normal, vertUV0).rgb;
+
+    // Remap normal to [-1,1]
+    normalizedVertNorm = normalize(normalizedVertNorm * 2.0 - 1.0);
 
     // Light contributions
     vec3 finalColor = CalcDirLight();
@@ -313,4 +482,9 @@ void main()
     vec3 finalAmbient = ambientColor * diffuseTexColor.rgb;
 
     fragColor = vec4(finalColor + finalAmbient + finalEmission, 1);
+
+    if (DRAW_NORMALS)
+    {
+        fragColor = vec4(texture(material.normal, vertUV0).rgb, 1);
+    }
 }

res/shaders/skybox.glsl

@@ -3,8 +3,9 @@
 
 layout(location=0) in vec3 vertPosIn;
 layout(location=1) in vec3 vertNormalIn;
-layout(location=2) in vec2 vertUV0In;
+layout(location=2) in vec3 vertTangentIn;
-layout(location=3) in vec3 vertColorIn;
+layout(location=3) in vec2 vertUV0In;
+layout(location=4) in vec3 vertColorIn;
 
 out vec3 vertUV0;
 

res/shaders/tonemapped-screen-quad.glsl

@@ -0,0 +1,50 @@
+//shader:vertex
+#version 410
+
+out vec2 vertUV0;
+
+// Hardcoded vertex positions for a fullscreen quad.
+// Format: vec4(pos.x, pos.y, uv0.x, uv0.y)
+vec4 quadData[6] = vec4[](
+    vec4(-1.0,  1.0, 0.0, 1.0),
+    vec4(-1.0, -1.0, 0.0, 0.0),
+    vec4(1.0, -1.0, 1.0, 0.0),
+    vec4(-1.0,  1.0, 0.0, 1.0),
+    vec4(1.0, -1.0, 1.0, 0.0),
+    vec4(1.0,  1.0, 1.0, 1.0)
+);
+
+void main()
+{
+    vec4 vertData = quadData[gl_VertexID];
+
+    vertUV0 = vertData.zw;
+    gl_Position = vec4(vertData.xy, 0.0, 1.0);
+}
+
+//shader:fragment
+#version 410
+
+struct Material {
+    sampler2D diffuse;
+};
+
+uniform float exposure = 1;
+uniform Material material;
+
+in vec2 vertUV0;
+
+out vec4 fragColor;
+
+void main()
+{
+    vec4 diffuseTexColor = texture(material.diffuse, vertUV0);
+
+    // Reinhard tone mapping
+    // vec3 mappedColor = diffuseTexColor.rgb / (diffuseTexColor.rgb + vec3(1.0));
+
+    // Exposure tone mapping
+    vec3 mappedColor = vec3(1.0) - exp(-diffuseTexColor.rgb * exposure);
+
+    fragColor = vec4(mappedColor, 1);
+}

ui/imgui/imgui.go

@@ -1,6 +1,17 @@
 package nmageimgui
 
 import (
+	"unsafe"
+
+	// The following is included just so we can get
+	// the c imports and cgo configs defined here: https://github.com/AllenDang/cimgui-go/blob/main/sdlbackend/sdl_backend.go
+	//
+	// This is needed because AllenDang/cimgui-go links sdl2 statically, which requires us
+	// to explicitly define all the libs it needs, which isn't normally needed if we were dynamically linking.
+	//
+	// Without this, we get compilation errors as sdl2 can't find libs it relies on.
+	_ "github.com/AllenDang/cimgui-go/sdlbackend"
+
 	imgui "github.com/AllenDang/cimgui-go"
 	"github.com/bloeys/gglm/gglm"
 	"github.com/bloeys/nmage/materials"
@@ -12,7 +23,7 @@ import (
 type ImguiInfo struct {
 	ImCtx imgui.Context
 
-	Mat        *materials.Material
+	Mat        materials.Material
 	VaoID      uint32
 	VboID      uint32
 	IndexBufID uint32
@@ -129,7 +140,7 @@ func (i *ImguiInfo) AddFontTTF(fontPath string, fontSize float32, fontConfig *im
 
 	fontConfigToUse := imgui.NewFontConfig()
 	if fontConfig != nil {
-		fontConfigToUse = *fontConfig
+		fontConfigToUse = fontConfig
 	}
 
 	glyphRangesToUse := imgui.NewGlyphRange()
@@ -141,13 +152,13 @@ func (i *ImguiInfo) AddFontTTF(fontPath string, fontSize float32, fontConfig *im
 
 	a := imIO.Fonts()
 	f := a.AddFontFromFileTTFV(fontPath, fontSize, fontConfigToUse, glyphRangesToUse.Data())
-	pixels, width, height, _ := a.GetTextureDataAsAlpha8()
+	pixels, width, height, _ := a.TextureDataAsAlpha8()
 
 	gl.ActiveTexture(gl.TEXTURE0)
 	gl.BindTexture(gl.TEXTURE_2D, *i.TexID)
 	gl.TexImage2D(gl.TEXTURE_2D, 0, gl.RED, int32(width), int32(height), 0, gl.RED, gl.UNSIGNED_BYTE, pixels)
 
-	return f
+	return *f
 }
 
 const DefaultImguiShader = `
@@ -204,7 +215,7 @@ void main()
 // If the path is empty a default nMage shader is used
 func NewImGui(shaderPath string) ImguiInfo {
 
-	var imguiMat *materials.Material
+	var imguiMat materials.Material
 	if shaderPath == "" {
 		imguiMat = materials.NewMaterialSrc("ImGUI Mat", []byte(DefaultImguiShader))
 	} else {
@@ -212,7 +223,7 @@ func NewImGui(shaderPath string) ImguiInfo {
 	}
 
 	imguiInfo := ImguiInfo{
-		ImCtx: imgui.CreateContext(),
+		ImCtx: *imgui.CreateContext(),
 		Mat:   imguiMat,
 		TexID: new(uint32),
 	}
@@ -233,11 +244,12 @@ func NewImGui(shaderPath string) ImguiInfo {
 	gl.TexParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.LINEAR)
 	gl.PixelStorei(gl.UNPACK_ROW_LENGTH, 0)
 
-	pixels, width, height, _ := io.Fonts().GetTextureDataAsAlpha8()
+	pixels, width, height, _ := io.Fonts().TextureDataAsAlpha8()
 	gl.TexImage2D(gl.TEXTURE_2D, 0, gl.RED, int32(width), int32(height), 0, gl.RED, gl.UNSIGNED_BYTE, pixels)
 
 	// Store our identifier
-	io.Fonts().SetTexID(imgui.TextureID(imguiInfo.TexID))
+
+	io.Fonts().SetTexID(imgui.TextureID{Data: uintptr(unsafe.Pointer(imguiInfo.TexID))})
 
 	//Shader attributes
 	imguiInfo.Mat.Bind()