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:
-    runs-on: macos-12
-    steps:
+  build-nmage-windows:
+    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

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

@@ -1,9 +1,8 @@
 package buffers
 
 import (
-	"fmt"
-
 	"github.com/bloeys/nmage/assert"
+	"github.com/bloeys/nmage/logging"
 	"github.com/go-gl/gl/v4.1-core/gl"
 )
 
@@ -14,10 +13,11 @@ type Element struct {
 }
 
 // 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:
@@ -65,10 +84,20 @@ 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
 	}
 }
@@ -91,8 +120,19 @@ 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
 	}
 }
@@ -101,6 +141,7 @@ func (dt ElementType) CompCount() int32 {
 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/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...)

engine/engine.go

@@ -9,7 +9,6 @@ import (
 	"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"
@@ -31,7 +30,6 @@ type Window struct {
 	SDLWin         *sdl.Window
 	GlCtx          sdl.GLContext
 	EventCallbacks []func(sdl.Event)
-	Rend           renderer.Render
 }
 
 func (w *Window) handleInputs() {
@@ -109,9 +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(imgui.MouseButtonRight, isSdlButtonRightDown)
-	imIo.SetMouseButtonDown(imgui.MouseButtonMiddle, isSdlButtonMiddleDown)
+	imIo.SetMouseButtonDown(int(imgui.MouseButtonLeft), isSdlButtonLeftDown)
+	imIo.SetMouseButtonDown(int(imgui.MouseButtonRight), isSdlButtonRightDown)
+	imIo.SetMouseButtonDown(int(imgui.MouseButtonMiddle), isSdlButtonMiddleDown)
 }
 
 func (w *Window) handleWindowResize() {
@@ -170,22 +168,21 @@ func initSDL() error {
 	return nil
 }
 
-func CreateOpenGLWindow(title string, x, y, width, height int32, flags WindowFlags, rend renderer.Render) (Window, error) {
-	return createWindow(title, x, y, width, height, WindowFlags_OPENGL|flags, rend)
+func CreateOpenGLWindow(title string, x, y, width, height int32, flags WindowFlags) (Window, error) {
+	return createWindow(title, x, y, width, height, WindowFlags_OPENGL|flags)
 }
 
-func CreateOpenGLWindowCentered(title string, width, height int32, flags WindowFlags, rend renderer.Render) (Window, error) {
-	return createWindow(title, sdl.WINDOWPOS_CENTERED, sdl.WINDOWPOS_CENTERED, width, height, WindowFlags_OPENGL|flags, rend)
+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)
 }
 
-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!")
 
 	win := Window{
 		SDLWin:         nil,
 		EventCallbacks: make([]func(sdl.Event), 0),
-		Rend:           rend,
 	}
 
 	var err error

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 v0.49.0
+	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 v0.49.0 h1:YtbyHpszYhjnxw7KVV0LaCdBktRMqfGx/i37EMomxsE=
-github.com/bloeys/gglm v0.49.0/go.mod h1:qwJQ0WzV191wAMwlGicbfbChbKoSedMk7gFFX6GnyOk=
+github.com/bloeys/gglm v0.50.0 h1:DlGLp9z8KMNx+hNR6PjnPmC0HjDRC19QwAKL1iwhOxs=
+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=

materials/material.go

@@ -1,6 +1,8 @@
 package materials
 
 import (
+	_ "unsafe"
+
 	"github.com/bloeys/gglm/gglm"
 	"github.com/bloeys/nmage/assert"
 	"github.com/bloeys/nmage/assets"
@@ -9,6 +11,16 @@ import (
 	"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 (
@@ -43,6 +55,7 @@ func (ms *MaterialSettings) Has(flags MaterialSettings) bool {
 }
 
 type Material struct {
+	Id         uint32
 	Name       string
 	ShaderProg shaders.ShaderProgram
 	Settings   MaterialSettings
@@ -111,6 +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]
@@ -118,7 +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
@@ -131,7 +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
@@ -154,33 +184,86 @@ 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 getNewMatId() uint32 {
+	lastMatId++
+	return lastMatId
+}
+
 func NewMaterial(matName, shaderPath string) Material {
 
 	shdrProg, err := shaders.LoadAndCompileCombinedShader(shaderPath)
@@ -189,6 +272,7 @@ func NewMaterial(matName, shaderPath string) Material {
 	}
 
 	return Material{
+		Id:         getNewMatId(),
 		Name:       matName,
 		ShaderProg: shdrProg,
 		UnifLocs:   make(map[string]int32),
@@ -209,6 +293,7 @@ func NewMaterialSrc(matName string, shaderSrc []byte) Material {
 	}
 
 	return Material{
+		Id:         getNewMatId(),
 		Name:       matName,
 		ShaderProg: shdrProg,
 		UnifLocs:   make(map[string]int32),

meshes/mesh.go

@@ -151,7 +151,7 @@ func NewMesh(name, modelPath string, postProcessFlags asig.PostProcess) (Mesh, e
 		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

@@ -12,21 +12,21 @@ import (
 var _ renderer.Render = &Rend3DGL{}
 
 type Rend3DGL struct {
-	BoundVao  *buffers.VertexArray
-	BoundMesh *meshes.Mesh
-	BoundMat  *materials.Material
+	BoundVaoId     uint32
+	BoundMatId     uint32
+	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
 	}
 
 	if mat.Settings.Has(materials.MaterialSettings_HasModelMtx) {
@@ -45,14 +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)
@@ -60,14 +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++ {
@@ -76,8 +76,9 @@ func (r *Rend3DGL) DrawCubemap(mesh *meshes.Mesh, mat *materials.Material) {
 }
 
 func (r3d *Rend3DGL) FrameEnd() {
-	r3d.BoundMesh = nil
-	r3d.BoundMat = nil
+	r3d.BoundVaoId = 0
+	r3d.BoundMatId = 0
+	r3d.BoundMeshVaoId = 0
 }
 
 func NewRend3DGL() *Rend3DGL {

res/shaders/simple.glsl

@@ -14,33 +14,25 @@ layout(location=3) in vec2 vertUV0In;
 layout(location=4) in vec3 vertColorIn;
 
 //
-// Uniforms
+// UBOs
 //
-uniform vec3 camPos;
-uniform mat4 modelMat;
-uniform mat3 normalMat;
-uniform mat4 projViewMat;
-uniform mat4 dirLightProjViewMat;
-uniform mat4 spotLightProjViewMats[NUM_SPOT_LIGHTS];
-
 struct DirLight {
     vec3 dir;
     vec3 diffuseColor;
     vec3 specularColor;
-    sampler2D shadowMap;
 };
-uniform DirLight dirLight;
+uniform sampler2D dirLightShadowMap;
 
 struct PointLight {
     vec3 pos;
     vec3 diffuseColor;
     vec3 specularColor;
-    float constant;
-    float linear;
-    float quadratic;
+    float radius;
+    float falloff;
+    float maxBias;
+    float nearPlane;
     float farPlane;
 };
-uniform PointLight pointLights[NUM_POINT_LIGHTS];
 
 struct SpotLight {
     vec3 pos;
@@ -50,7 +42,25 @@ struct SpotLight {
     float innerCutoff;
     float outerCutoff;
 };
-uniform SpotLight spotLights[NUM_SPOT_LIGHTS];
+
+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
@@ -154,20 +164,19 @@ struct DirLight {
     vec3 dir;
     vec3 diffuseColor;
     vec3 specularColor;
-    sampler2D shadowMap;
 };
-uniform DirLight dirLight;
+uniform sampler2D dirLightShadowMap;
 
 struct PointLight {
     vec3 pos;
     vec3 diffuseColor;
     vec3 specularColor;
-    float constant;
-    float linear;
-    float quadratic;
+    float radius;
+    float falloff;
+    float maxBias;
+    float nearPlane;
     float farPlane;
 };
-uniform PointLight pointLights[NUM_POINT_LIGHTS];
 uniform samplerCubeArray pointLightCubeShadowMaps;
 
 struct SpotLight {
@@ -178,10 +187,19 @@ struct SpotLight {
     float innerCutoff;
     float outerCutoff;
 };
-uniform SpotLight spotLights[NUM_SPOT_LIGHTS];
 uniform sampler2DArray spotLightShadowMaps;
 
-uniform vec3 ambientColor = vec3(0.2, 0.2, 0.2);
+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
@@ -248,34 +266,88 @@ vec3 CalcDirLight()
     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 worldLightPos, vec3 tangentLightDir, float farPlane) {
+float CalcPointShadow(int lightIndex, vec3 worldLightPos, vec3 tangentLightDir, float maxBias, float nearPlane, float farPlane) {
 
     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 currentDepth = length(lightToFrag);
+    float bias = max(maxBias * (1 - dot(normalizedVertNorm, tangentLightDir)), 0.005);
 
-    float bias = max(0.05 * (1 - dot(normalizedVertNorm, tangentLightDir)), 0.005);
-
-    float shadow = currentDepth -  bias > closestDepth ? 1.0 : 0.0;
+    float shadow = currentDepth - bias > closestDepth ? 1 : 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
-    if (pointLight.constant == 0){
+    // Ignore inactive lights
+    if (pointLight.radius == 0){
         return vec3(0);
     }
 
@@ -293,10 +365,10 @@ vec3 CalcPointLight(PointLight pointLight, int lightIndex)
 
     // Attenuation
     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, tangentLightDir, pointLight.farPlane);
+    float shadow = CalcPointShadow(lightIndex, pointLight.pos, tangentLightDir, pointLight.maxBias, pointLight.nearPlane, pointLight.farPlane);
 
     return (finalDiffuse + finalSpecular) * attenuation * (1 - shadow);
 }
@@ -343,7 +415,10 @@ float CalcSpotShadow(vec3 tangentLightDir, 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 tangentLightDir = tangentSpotLightDirections[lightIndex];

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"
@@ -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 = `
@@ -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()