diff --git a/.gitignore b/.gitignore
index c1b78b3..f445fc5 100644
--- a/.gitignore
+++ b/.gitignore
@@ -6,4 +6,5 @@
 *.img
 dc-build.sh
 .buildconfig
-GL/version.h
+build/*
+builddir/*
diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index 2c21ac7..060ccd6 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -6,5 +6,18 @@ build:sh4-gcc:
   image: kazade/dreamcast-sdk
   script:
     - source /etc/bash.bashrc
-    - make clean
-    - make samples
+    - mkdir builddir
+    - cd builddir
+    - cmake -DCMAKE_TOOLCHAIN_FILE=../toolchains/Dreamcast.cmake ..
+    - make
+    
+build:x86-gcc:
+  stage: build
+  image: fedora:33
+  before_script:
+    - sudo dnf install -y cmake gcc gcc-c++ SDL2-devel.i686 glibc-devel.i686
+  script:
+    - mkdir builddir
+    - cd builddir
+    - cmake ..
+    - make
diff --git a/CMakeLists.txt b/CMakeLists.txt
new file mode 100644
index 0000000..ffc6089
--- /dev/null
+++ b/CMakeLists.txt
@@ -0,0 +1,146 @@
+cmake_minimum_required(VERSION 3.0)
+project(GLdc)
+
+# set the default backend
+if(PLATFORM_DREAMCAST)
+    set(BACKEND "kospvr" CACHE STRING "Backend to use")
+else()
+    set(BACKEND "software" CACHE STRING "Backend to use")
+endif()
+
+# List of possible backends
+set_property(CACHE BACKEND PROPERTY STRINGS kospvr software)
+
+message("\nCompiling using backend: ${BACKEND}\n")
+
+string(TOUPPER ${BACKEND} BACKEND_UPPER)
+add_definitions(-DBACKEND_${BACKEND_UPPER})
+
+set(CMAKE_C_STANDARD 99)
+
+include_directories(include)
+
+if(NOT PLATFORM_DREAMCAST)
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -m32")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -m32")
+endif()
+
+set(
+    SOURCES
+    containers/aligned_vector.c
+    containers/named_array.c
+    containers/stack.c
+    GL/clip.c
+    GL/draw.c
+    GL/error.c
+    GL/flush.c
+    GL/fog.c
+    GL/framebuffer.c
+    GL/glu.c
+    GL/immediate.c
+    GL/lighting.c
+    GL/matrix.c
+    GL/profiler.c
+    GL/state.c
+    GL/texture.c
+    GL/util.c
+    GL/yalloc/yalloc.c
+    ${CMAKE_CURRENT_BINARY_DIR}/version.c
+)
+
+execute_process(
+    COMMAND git describe --abbrev=4 --dirty --always --tags
+    OUTPUT_VARIABLE GLDC_VERSION
+    OUTPUT_STRIP_TRAILING_WHITESPACE
+)
+
+configure_file(GL/version.c.in ${CMAKE_CURRENT_BINARY_DIR}/version.c)
+
+if(PLATFORM_DREAMCAST)
+    set(SOURCES ${SOURCES} GL/platforms/sh4.c)
+else()
+    find_package(SDL2 REQUIRED)
+    include_directories(${SDL2_INCLUDE_DIRS})
+    link_libraries(${SDL2_LIBRARIES})
+    set(
+        SOURCES
+        ${SOURCES}
+        GL/platforms/software.c
+        GL/platforms/software/edge_equation.c
+        GL/platforms/software/parameter_equation.c
+    )
+endif()
+
+add_library(GLdc STATIC ${SOURCES})
+link_libraries(m)
+
+include_directories(include)
+
+link_libraries(GLdc)
+
+function(gen_sample sample)
+    set(SAMPLE_SRCS ${ARGN})
+    set(GENROMFS "$ENV{KOS_BASE}/utils/genromfs/genromfs")
+    set(BIN2O $ENV{KOS_BASE}/utils/bin2o/bin2o)
+    set(ROMDISK_IMG "${CMAKE_SOURCE_DIR}/samples/${sample}/romdisk.img")
+    set(ROMDISK_O "${CMAKE_SOURCE_DIR}/samples/${sample}/romdisk.o")
+    set(ROMDISK_DIR "${CMAKE_SOURCE_DIR}/samples/${sample}/romdisk")
+
+    add_executable(${sample} ${SAMPLE_SRCS})
+
+    if(PLATFORM_DREAMCAST)
+        if(EXISTS "${CMAKE_SOURCE_DIR}/samples/${sample}/romdisk")
+            message("Generating romdisk for sample: ${sample}")
+            add_custom_command(
+                OUTPUT ${ROMDISK_IMG}
+                COMMAND ${GENROMFS} -f ${ROMDISK_IMG} -d ${ROMDISK_DIR} -v
+            )
+
+            add_custom_command(
+                OUTPUT ${ROMDISK_O}
+                COMMAND ${BIN2O} romdisk.img romdisk romdisk.o
+                DEPENDS ${ROMDISK_IMG}
+                WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}/samples/${sample}"
+            )
+
+            add_custom_target(${sample}_romdisk DEPENDS ${ROMDISK_O})
+            add_dependencies(${sample} ${sample}_romdisk)
+            target_link_libraries(${sample} ${ROMDISK_O})
+
+        else()
+            message("No such romdisk for sample: ${sample} at 'samples/${sample}/romdisk'")
+        endif()
+    endif()
+endfunction()
+
+gen_sample(blend_test samples/blend_test/main.c)
+gen_sample(depth_funcs samples/depth_funcs/main.c)
+gen_sample(depth_funcs_alpha_testing samples/depth_funcs_alpha_testing/main.c samples/depth_funcs_alpha_testing/gl_png.c)
+gen_sample(depth_funcs_ortho samples/depth_funcs_ortho/main.c)
+gen_sample(lerabot01 samples/lerabot01/main.c)
+gen_sample(lerabot_blend_test samples/lerabot_blend_test/main.c)
+gen_sample(lights samples/lights/main.c)
+gen_sample(mipmap samples/mipmap/main.c)
+gen_sample(multitexture_arrays samples/multitexture_arrays/main.c samples/multitexture_arrays/pvr-texture.c)
+gen_sample(nehe02 samples/nehe02/main.c)
+gen_sample(nehe02de samples/nehe02de/main.c)
+gen_sample(nehe02va samples/nehe02va/main.c)
+gen_sample(nehe03 samples/nehe03/main.c)
+gen_sample(nehe04 samples/nehe04/main.c)
+gen_sample(nehe05 samples/nehe05/main.c)
+gen_sample(nehe06 samples/nehe06/main.c)
+gen_sample(nehe06_vq samples/nehe06_vq/main.c)
+gen_sample(ortho2d samples/ortho2d/main.c)
+gen_sample(paletted samples/paletted/main.c)
+gen_sample(paletted_pcx samples/paletted_pcx/main.c)
+gen_sample(polygon_offset samples/polygon_offset/main.c)
+gen_sample(terrain samples/terrain/main.c)
+gen_sample(zclip samples/zclip/main.c)
+gen_sample(zclip_triangle samples/zclip_triangle/main.c)
+gen_sample(zclip_trianglestrip samples/zclip_trianglestrip/main.c)
+
+if(PLATFORM_DREAMCAST)
+    gen_sample(polymark samples/polymark/main.c)
+    gen_sample(quadmark samples/quadmark/main.c)
+    gen_sample(trimark samples/trimark/main.c)
+endif()
diff --git a/GL/draw.c b/GL/draw.c
index fc99dee..8a6f73c 100644
--- a/GL/draw.c
+++ b/GL/draw.c
@@ -4,14 +4,10 @@
 #include <stdlib.h>
 #include <math.h>
 #include <assert.h>
-#include <dc/vec3f.h>
 
-#include "../include/gl.h"
-#include "../include/glext.h"
 #include "private.h"
 #include "profiler.h"
-#include "sh4_math.h"
-
+#include "platform.h"
 
 static AttribPointer VERTEX_POINTER;
 static AttribPointer UV_POINTER;
@@ -439,7 +435,7 @@ GL_FORCE_INLINE void transformNormalToEyeSpace(GLfloat* normal) {
     mat_trans_normal3(normal[0], normal[1], normal[2]);
 }
 
-PVRHeader* _glSubmissionTargetHeader(SubmissionTarget* target) {
+PolyHeader *_glSubmissionTargetHeader(SubmissionTarget* target) {
     assert(target->header_offset < target->output->vector.size);
     return aligned_vector_at(&target->output->vector, target->header_offset);
 }
@@ -458,7 +454,7 @@ static inline void genTriangles(Vertex* output, GLuint count) {
 
     GLuint i;
     for(i = 0; i < count; i += 3) {
-        it->flags = PVR_CMD_VERTEX_EOL;
+        it->flags = GPU_CMD_VERTEX_EOL;
         it += 3;
     }
 }
@@ -468,13 +464,13 @@ static inline void genQuads(Vertex* output, GLuint count) {
     GLuint i;
     for(i = 0; i < count; i += 4) {
         swapVertex((final - 1), final);
-        final->flags = PVR_CMD_VERTEX_EOL;
+        final->flags = GPU_CMD_VERTEX_EOL;
         final += 4;
     }
 }
 
 static void genTriangleStrip(Vertex* output, GLuint count) {
-    output[count - 1].flags = PVR_CMD_VERTEX_EOL;
+    output[count - 1].flags = GPU_CMD_VERTEX_EOL;
 }
 
 static void genTriangleFan(Vertex* output, GLuint count) {
@@ -486,7 +482,7 @@ static void genTriangleFan(Vertex* output, GLuint count) {
     GLubyte i = count - 2;
     while(i--) {
         *dst = *src--;
-        (*dst--).flags = PVR_CMD_VERTEX_EOL;
+        (*dst--).flags = GPU_CMD_VERTEX_EOL;
         *dst-- = *src;
         *dst-- = *output;
     }
@@ -655,7 +651,7 @@ GL_FORCE_INLINE void _readSTData(const GLuint first, const GLuint count, VertexE
     const GLubyte ststride = (ST_POINTER.stride) ? ST_POINTER.stride : ST_POINTER.size * byte_size(ST_POINTER.type);
     const void* stptr = ((GLubyte*) ST_POINTER.ptr + (first * ststride));
 
-    ReadUVFunc func = calcReadUVFunc();
+    ReadUVFunc func = calcReadSTFunc();
     GLubyte* out = (GLubyte*) extra[0].st;
 
     ITERATE(count) {
@@ -697,7 +693,8 @@ GL_FORCE_INLINE void _readNormalData(const GLuint first, const GLuint count, Ver
 }
 
 GL_FORCE_INLINE void _readDiffuseData(const GLuint first, const GLuint count, Vertex* output) {
-    const GLuint cstride = (DIFFUSE_POINTER.stride) ? DIFFUSE_POINTER.stride : DIFFUSE_POINTER.size * byte_size(DIFFUSE_POINTER.type);
+    const GLuint size = (DIFFUSE_POINTER.size == GL_BGRA) ? 4 : DIFFUSE_POINTER.size;
+    const GLuint cstride = (DIFFUSE_POINTER.stride) ? DIFFUSE_POINTER.stride : size * byte_size(DIFFUSE_POINTER.type);
     const GLubyte* cptr = ((GLubyte*) DIFFUSE_POINTER.ptr) + (first * cstride);
 
     ReadDiffuseFunc func = calcReadDiffuseFunc();
@@ -765,7 +762,7 @@ static void generateElements(
         st_func(st, (GLubyte*) ve->st);
         normal_func(nxyz, (GLubyte*) ve->nxyz);
 
-        output->flags = PVR_CMD_VERTEX;
+        output->flags = GPU_CMD_VERTEX;
         ++output;
         ++ve;
     }
@@ -786,7 +783,7 @@ static void generate(SubmissionTarget* target, const GLenum mode, const GLsizei
             const GLubyte* pos = VERTEX_POINTER.ptr;
             Vertex* it = start;
             ITERATE(count) {
-                it->flags = PVR_CMD_VERTEX;
+                it->flags = GPU_CMD_VERTEX;
                 memcpy(it->xyz, pos, FAST_PATH_BYTE_SIZE);
                 it++;
                 pos += VERTEX_POINTER.stride;
@@ -799,7 +796,7 @@ static void generate(SubmissionTarget* target, const GLenum mode, const GLsizei
             Vertex* it = _glSubmissionTargetStart(target);
 
             ITERATE(count) {
-                it->flags = PVR_CMD_VERTEX;
+                it->flags = GPU_CMD_VERTEX;
                 ++it;
             }
         }
@@ -843,25 +840,7 @@ static void transform(SubmissionTarget* target) {
 
     _glApplyRenderMatrix(); /* Apply the Render Matrix Stack */
 
-    ITERATE(target->count) {
-        register float __x __asm__("fr12") = (vertex->xyz[0]);
-        register float __y __asm__("fr13") = (vertex->xyz[1]);
-        register float __z __asm__("fr14") = (vertex->xyz[2]);
-        register float __w __asm__("fr15") = (vertex->w);
-
-        __asm__ __volatile__(
-            "fldi1 fr15\n"
-            "ftrv   xmtrx,fv12\n"
-            : "=f" (__x), "=f" (__y), "=f" (__z), "=f" (__w)
-            : "0" (__x), "1" (__y), "2" (__z), "3" (__w)
-        );
-
-        vertex->xyz[0] = __x;
-        vertex->xyz[1] = __y;
-        vertex->xyz[2] = __z;
-        vertex->w = __w;
-        ++vertex;
-    }
+    TransformVertices(vertex, target->count);
 }
 
 static void clip(SubmissionTarget* target) {
@@ -875,14 +854,17 @@ static void clip(SubmissionTarget* target) {
 }
 
 static void mat_transform3(const float* xyz, const float* xyzOut, const uint32_t count, const uint32_t inStride, const uint32_t outStride) {
-    uint8_t* dataIn = (uint8_t*) xyz;
+    const uint8_t* dataIn = (const uint8_t*) xyz;
     uint8_t* dataOut = (uint8_t*) xyzOut;
 
     ITERATE(count) {
-        float* in = (float*) dataIn;
+        const float* in = (const float*) dataIn;
         float* out = (float*) dataOut;
 
-        mat_trans_single3_nodiv_nomod(in[0], in[1], in[2], out[0], out[1], out[2]);
+        TransformVec3NoMod(
+            in,
+            out
+        );
 
         dataIn += inStride;
         dataOut += outStride;
@@ -890,14 +872,14 @@ static void mat_transform3(const float* xyz, const float* xyzOut, const uint32_t
 }
 
 static void mat_transform_normal3(const float* xyz, const float* xyzOut, const uint32_t count, const uint32_t inStride, const uint32_t outStride) {
-    uint8_t* dataIn = (uint8_t*) xyz;
+    const uint8_t* dataIn = (const uint8_t*) xyz;
     uint8_t* dataOut = (uint8_t*) xyzOut;
 
     ITERATE(count) {
-        float* in = (float*) dataIn;
+        const float* in = (const float*) dataIn;
         float* out = (float*) dataOut;
 
-        mat_trans_normal3_nomod(in[0], in[1], in[2], out[0], out[1], out[2]);
+        TransformNormalNoMod(in, out);
 
         dataIn += inStride;
         dataOut += outStride;
@@ -930,8 +912,8 @@ static void light(SubmissionTarget* target) {
     _glPerformLighting(vertex, ES, target->count);
 }
 
-#define PVR_MIN_Z 0.2f
-#define PVR_MAX_Z 1.0 + PVR_MIN_Z
+#define GPU_MIN_Z 0.2f
+#define GPU_MAX_Z 1.0 + GPU_MIN_Z
 
 GL_FORCE_INLINE void divide(SubmissionTarget* target) {
     TRACE();
@@ -949,26 +931,26 @@ GL_FORCE_INLINE void divide(SubmissionTarget* target) {
     }
 }
 
-GL_FORCE_INLINE void push(PVRHeader* header, GLboolean multiTextureHeader, PolyList* activePolyList, GLshort textureUnit) {
+GL_FORCE_INLINE void push(PolyHeader* header, GLboolean multiTextureHeader, PolyList* activePolyList, GLshort textureUnit) {
     TRACE();
 
     // Compile the header
-    pvr_poly_cxt_t cxt = *_glGetPVRContext();
+    PolyContext cxt = *_glGetPVRContext();
     cxt.list_type = activePolyList->list_type;
 
     _glUpdatePVRTextureContext(&cxt, textureUnit);
 
     if(multiTextureHeader) {
-        assert(cxt.list_type == PVR_LIST_TR_POLY);
+        assert(cxt.list_type == GPU_LIST_TR_POLY);
 
-        cxt.gen.alpha = PVR_ALPHA_ENABLE;
-        cxt.txr.alpha = PVR_TXRALPHA_ENABLE;
-        cxt.blend.src = PVR_BLEND_ZERO;
-        cxt.blend.dst = PVR_BLEND_DESTCOLOR;
-        cxt.depth.comparison = PVR_DEPTHCMP_EQUAL;
+        cxt.gen.alpha = GPU_ALPHA_ENABLE;
+        cxt.txr.alpha = GPU_TXRALPHA_ENABLE;
+        cxt.blend.src = GPU_BLEND_ZERO;
+        cxt.blend.dst = GPU_BLEND_DESTCOLOR;
+        cxt.depth.comparison = GPU_DEPTHCMP_EQUAL;
     }
 
-    pvr_poly_compile(&header->hdr, &cxt);
+    CompilePolyHeader(header, &cxt);
 
     /* Post-process the vertex list */
     /*
@@ -1136,7 +1118,7 @@ GL_FORCE_INLINE void submitVertices(GLenum mode, GLsizei first, GLuint count, GL
 
     assert(vertex);
 
-    PVRHeader* mtHeader = (PVRHeader*) vertex++;
+    PolyHeader* mtHeader = (PolyHeader*) vertex++;
 
     /* Replace the UV coordinates with the ST ones */
     VertexExtra* ve = aligned_vector_at(target->extras, 0);
diff --git a/GL/error.c b/GL/error.c
index 49732e5..c964fec 100644
--- a/GL/error.c
+++ b/GL/error.c
@@ -8,10 +8,10 @@
     KOS Open GL State Machine Error Code Implementation.
 */
 
-#include "gl.h"
-
 #include <stdio.h>
 
+#include "private.h"
+
 static GLenum last_error = GL_NO_ERROR;
 static char error_function[64] = { '\0' };
 
diff --git a/GL/flush.c b/GL/flush.c
index 95a6dde..8829533 100644
--- a/GL/flush.c
+++ b/GL/flush.c
@@ -1,57 +1,12 @@
 
-
-#include <kos.h>
-
-#include "../include/glkos.h"
 #include "../containers/aligned_vector.h"
 #include "private.h"
 #include "profiler.h"
-#include "version.h"
-
-#define TA_SQ_ADDR (unsigned int *)(void *) \
-    (0xe0000000 | (((unsigned long)0x10000000) & 0x03ffffe0))
 
 static PolyList OP_LIST;
 static PolyList PT_LIST;
 static PolyList TR_LIST;
 
-static void pvr_list_submit(void *src, int n) {
-    GLuint *d = TA_SQ_ADDR;
-    GLuint *s = src;
-
-    /* fill/write queues as many times necessary */
-    while(n--) {
-        __asm__("pref @%0" : : "r"(s + 8));  /* prefetch 32 bytes for next loop */
-        d[0] = *(s++);
-        d[1] = *(s++);
-        d[2] = *(s++);
-        d[3] = *(s++);
-        d[4] = *(s++);
-        d[5] = *(s++);
-        d[6] = *(s++);
-        d[7] = *(s++);
-        __asm__("pref @%0" : : "r"(d));
-        d += 8;
-    }
-
-    /* Wait for both store queues to complete */
-    d = (GLuint *)0xe0000000;
-    d[0] = d[8] = 0;
-}
-
-static void _glInitPVR(GLboolean autosort, GLboolean fsaa) {
-    pvr_init_params_t params = {
-        /* Enable opaque and translucent polygons with size 32 and 32 */
-        {PVR_BINSIZE_32, PVR_BINSIZE_0, PVR_BINSIZE_32, PVR_BINSIZE_0, PVR_BINSIZE_32},
-        PVR_VERTEX_BUF_SIZE, /* Vertex buffer size */
-        0, /* No DMA */
-        fsaa, /* No FSAA */
-        (autosort) ? 0 : 1 /* Disable translucent auto-sorting to match traditional GL */
-    };
-
-    pvr_init(&params);
-}
-
 
 PolyList* _glActivePolyList() {
     if(_glIsBlendingEnabled()) {
@@ -92,7 +47,7 @@ void APIENTRY glKosInitEx(GLdcConfig* config) {
 
     printf("\nWelcome to GLdc! Git revision: %s\n\n", GLDC_VERSION);
 
-    _glInitPVR(config->autosort_enabled, config->fsaa_enabled);
+    InitGPU(config->autosort_enabled, config->fsaa_enabled);
 
     _glInitMatrices();
     _glInitAttributePointers();
@@ -105,9 +60,9 @@ void APIENTRY glKosInitEx(GLdcConfig* config) {
 
     _glInitTextures();
 
-    OP_LIST.list_type = PVR_LIST_OP_POLY;
-    PT_LIST.list_type = PVR_LIST_PT_POLY;
-    TR_LIST.list_type = PVR_LIST_TR_POLY;
+    OP_LIST.list_type = GPU_LIST_OP_POLY;
+    PT_LIST.list_type = GPU_LIST_PT_POLY;
+    TR_LIST.list_type = GPU_LIST_TR_POLY;
 
     aligned_vector_init(&OP_LIST.vector, sizeof(Vertex));
     aligned_vector_init(&PT_LIST.vector, sizeof(Vertex));
@@ -124,7 +79,6 @@ void APIENTRY glKosInit() {
     glKosInitEx(&config);
 }
 
-#define QACRTA ((((unsigned int)0x10000000)>>26)<<2)&0x1c
 
 void APIENTRY glKosSwapBuffers() {
     static int frame_count = 0;
@@ -133,24 +87,19 @@ void APIENTRY glKosSwapBuffers() {
 
     profiler_push(__func__);
 
-    pvr_wait_ready();
+    SceneBegin();
+        SceneListBegin(GPU_LIST_OP_POLY);
+        SceneListSubmit(OP_LIST.vector.data, OP_LIST.vector.size);
+        SceneListFinish();
 
-    pvr_scene_begin();
-        QACR0 = QACRTA;
-        QACR1 = QACRTA;
+        SceneListBegin(GPU_LIST_PT_POLY);
+        SceneListSubmit(PT_LIST.vector.data, PT_LIST.vector.size);
+        SceneListFinish();
 
-        pvr_list_begin(PVR_LIST_OP_POLY);
-        pvr_list_submit(OP_LIST.vector.data, OP_LIST.vector.size);
-        pvr_list_finish();
-
-        pvr_list_begin(PVR_LIST_PT_POLY);
-        pvr_list_submit(PT_LIST.vector.data, PT_LIST.vector.size);
-        pvr_list_finish();
-
-        pvr_list_begin(PVR_LIST_TR_POLY);
-        pvr_list_submit(TR_LIST.vector.data, TR_LIST.vector.size);
-        pvr_list_finish();
-    pvr_scene_finish();
+        SceneListBegin(GPU_LIST_TR_POLY);
+        SceneListSubmit(TR_LIST.vector.data, TR_LIST.vector.size);
+        SceneListFinish();
+    SceneFinish();
 
     aligned_vector_clear(&OP_LIST.vector);
     aligned_vector_clear(&PT_LIST.vector);
diff --git a/GL/fog.c b/GL/fog.c
index ccf0662..08dc21d 100644
--- a/GL/fog.c
+++ b/GL/fog.c
@@ -10,13 +10,13 @@ static GLfloat FOG_COLOR [] = {0.0f, 0.0f, 0.0f, 0.0f};
 
 static void updatePVRFog() {
     if(FOG_MODE == GL_LINEAR) {
-        pvr_fog_table_linear(FOG_START, FOG_END);
+        GPUSetFogLinear(FOG_START, FOG_END);
     } else if(FOG_MODE == GL_EXP) {
-        pvr_fog_table_exp(FOG_DENSITY);
+        GPUSetFogExp(FOG_DENSITY);
     } else if(FOG_MODE == GL_EXP2) {
-        pvr_fog_table_exp2(FOG_DENSITY);
+        GPUSetFogExp2(FOG_DENSITY);
     }
-    pvr_fog_table_color(FOG_COLOR[3], FOG_COLOR[0], FOG_COLOR[1], FOG_COLOR[2]);
+    GPUSetFogColor(FOG_COLOR[3], FOG_COLOR[0], FOG_COLOR[1], FOG_COLOR[2]);
 }
 
 void APIENTRY glFogf(GLenum pname,  GLfloat param) {
diff --git a/GL/framebuffer.c b/GL/framebuffer.c
index 96c1d43..0df1893 100644
--- a/GL/framebuffer.c
+++ b/GL/framebuffer.c
@@ -2,8 +2,6 @@
 #include <assert.h>
 
 #include "private.h"
-#include "../include/glkos.h"
-#include "../include/glext.h"
 
 typedef struct {
     GLuint index;
@@ -160,7 +158,7 @@ static GL_NO_INSTRUMENT GLboolean _glCalculateAverageTexel(GLuint pvrFormat, con
     const GLubyte ARGB4444 = 1;
     const GLubyte RGB565 = 2;
 
-    if((pvrFormat & PVR_TXRFMT_PAL8BPP) == PVR_TXRFMT_PAL8BPP) {
+    if((pvrFormat & GPU_TXRFMT_PAL8BPP) == GPU_TXRFMT_PAL8BPP) {
         /* Paletted... all we can do really is just pick one of the
          * 4 texels.. unless we want to change the palette (bad) or
          * pick the closest available colour (slow, and probably bad)
@@ -232,7 +230,7 @@ GLboolean _glGenerateMipmapTwiddled(const GLuint pvrFormat, const GLubyte* prevD
     uint32_t i, j;
     uint32_t stride = 0;
 
-    if((pvrFormat & PVR_TXRFMT_PAL8BPP) == PVR_TXRFMT_PAL8BPP) {
+    if((pvrFormat & GPU_TXRFMT_PAL8BPP) == GPU_TXRFMT_PAL8BPP) {
         stride = 1;
     } else {
         stride = 2;
@@ -281,7 +279,7 @@ void APIENTRY glGenerateMipmapEXT(GLenum target) {
         return;
     }
 
-    if((tex->color & PVR_TXRFMT_NONTWIDDLED) == PVR_TXRFMT_NONTWIDDLED) {
+    if((tex->color & GPU_TXRFMT_NONTWIDDLED) == GPU_TXRFMT_NONTWIDDLED) {
         /* glTexImage2D should twiddle internally textures in nearly all cases
          * so this error is unlikely */
 
diff --git a/GL/glu.c b/GL/glu.c
index 9b3e8cf..717b273 100644
--- a/GL/glu.c
+++ b/GL/glu.c
@@ -6,7 +6,7 @@ void APIENTRY gluPerspective(GLfloat angle, GLfloat aspect,
                     GLfloat znear, GLfloat zfar) {
     GLfloat fW, fH;
 
-    fH = tanf(angle * (F_PI / 360.0f)) * znear;
+    fH = tanf(angle * (M_PI / 360.0f)) * znear;
     fW = fH * aspect;
 
     glFrustum(-fW, fW, -fH, fH, znear, zfar);
diff --git a/GL/immediate.c b/GL/immediate.c
index e1f530b..f9e1958 100644
--- a/GL/immediate.c
+++ b/GL/immediate.c
@@ -10,11 +10,7 @@
 #include <string.h>
 #include <stdio.h>
 
-#include "../include/gl.h"
-#include "../include/glext.h"
-#include "../include/glkos.h"
 #include "profiler.h"
-
 #include "private.h"
 
 static GLboolean IMMEDIATE_MODE_ACTIVE = GL_FALSE;
diff --git a/GL/lighting.c b/GL/lighting.c
index a7d2094..de0e15f 100644
--- a/GL/lighting.c
+++ b/GL/lighting.c
@@ -3,8 +3,9 @@
 #include <string.h>
 #include <math.h>
 #include <limits.h>
-#include <dc/vec3f.h>
+
 #include "private.h"
+#include "platform.h"
 
 #define _MIN(x, y) (x < y) ? x : y
 
@@ -201,12 +202,7 @@ void APIENTRY glLightfv(GLenum light, GLenum pname, const GLfloat *params) {
             if(LIGHTS[idx].isDirectional) {
                 //FIXME: Do we need to rotate directional lights?
             } else {
-                mat_trans_single4(
-                    LIGHTS[idx].position[0],
-                    LIGHTS[idx].position[1],
-                    LIGHTS[idx].position[2],
-                    LIGHTS[idx].position[3]
-                );
+                TransformVec3(LIGHTS[idx].position);
             }
         }
         break;
@@ -387,10 +383,6 @@ GL_FORCE_INLINE GLboolean isSpecularColorMaterial() {
     return (COLOR_MATERIAL_MODE == GL_SPECULAR);
 }
 
-GL_FORCE_INLINE void initVec3(struct vec3f* v, const GLfloat* src) {
-    memcpy(v, src, sizeof(GLfloat) * 3);
-}
-
 /*
  * Implementation from here (MIT):
  * https://github.com/appleseedhq/appleseed/blob/master/src/appleseed/foundation/math/fastmath.h
@@ -455,7 +447,7 @@ GL_FORCE_INLINE void _glLightVertexPoint(
 #undef _PROCESS_COMPONENT
 }
 
-void _glPerformLighting(Vertex* vertices, EyeSpaceData* es, const int32_t count) {
+void _glPerformLighting(Vertex* vertices, EyeSpaceData* es, const uint32_t count) {
     GLubyte i;
     GLuint j;
 
@@ -503,7 +495,7 @@ void _glPerformLighting(Vertex* vertices, EyeSpaceData* es, const int32_t count)
         float Vx = -data->xyz[0];
         float Vy = -data->xyz[1];
         float Vz = -data->xyz[2];
-        vec3f_normalize(Vx, Vy, Vz);
+        VEC3_NORMALIZE(Vx, Vy, Vz);
 
         const float Nx = data->n[0];
         const float Ny = data->n[1];
@@ -523,15 +515,15 @@ void _glPerformLighting(Vertex* vertices, EyeSpaceData* es, const int32_t count)
                 float Hy = (Ly + 0);
                 float Hz = (Lz + 1);
 
-                vec3f_normalize(Lx, Ly, Lz);
-                vec3f_normalize(Hx, Hy, Hz);
+                VEC3_NORMALIZE(Lx, Ly, Lz);
+                VEC3_NORMALIZE(Hx, Hy, Hz);
 
                 float LdotN, NdotH;
-                vec3f_dot(
+                VEC3_DOT(
                     Nx, Ny, Nz, Lx, Ly, Lz, LdotN
                 );
 
-                vec3f_dot(
+                VEC3_DOT(
                     Nx, Ny, Nz, Hx, Hy, Hz, NdotH
                 );
 
@@ -544,8 +536,7 @@ void _glPerformLighting(Vertex* vertices, EyeSpaceData* es, const int32_t count)
                 );
             } else {
                 float D;
-
-                vec3f_length(Lx, Ly, Lz, D);
+                VEC3_LENGTH(Lx, Ly, Lz, D);
 
                 float att = (
                     LIGHTS[i].constant_attenuation + (
@@ -563,15 +554,15 @@ void _glPerformLighting(Vertex* vertices, EyeSpaceData* es, const int32_t count)
                     float Hy = (Ly + Vy);
                     float Hz = (Lz + Vz);
 
-                    vec3f_normalize(Lx, Ly, Lz);
-                    vec3f_normalize(Hx, Hy, Hz);
+                    VEC3_NORMALIZE(Lx, Ly, Lz);
+                    VEC3_NORMALIZE(Hx, Hy, Hz);
 
                     float LdotN, NdotH;
-                    vec3f_dot(
+                    VEC3_DOT(
                         Nx, Ny, Nz, Lx, Ly, Lz, LdotN
                     );
 
-                    vec3f_dot(
+                    VEC3_DOT(
                         Nx, Ny, Nz, Hx, Hy, Hz, NdotH
                     );
 
diff --git a/GL/matrix.c b/GL/matrix.c
index b0b21d5..275b607 100644
--- a/GL/matrix.c
+++ b/GL/matrix.c
@@ -1,13 +1,9 @@
 #include <string.h>
-
+#include <math.h>
 #include <stdio.h>
-#include <dc/fmath.h>
-#include <dc/matrix.h>
-#include <dc/matrix3d.h>
-#include <dc/vec3f.h>
 
 #include "private.h"
-#include "../include/gl.h"
+
 #include "../containers/stack.h"
 
 #define DEG2RAD (0.01745329251994329576923690768489)
@@ -36,18 +32,6 @@ static const Matrix4x4 IDENTITY = {
 
 GLfloat NEAR_PLANE_DISTANCE = 0.0f;
 
-static inline void upload_matrix(Matrix4x4* m) {
-    mat_load((matrix_t*) m);
-}
-
-static inline void multiply_matrix(Matrix4x4* m) {
-    mat_apply((matrix_t*) m);
-}
-
-static inline void download_matrix(Matrix4x4* m) {
-    mat_store((matrix_t*) m);
-}
-
 Matrix4x4* _glGetProjectionMatrix() {
     return (Matrix4x4*) stack_top(&MATRIX_STACKS[1]);
 }
@@ -65,8 +49,10 @@ void _glInitMatrices() {
     stack_push(&MATRIX_STACKS[1], IDENTITY);
     stack_push(&MATRIX_STACKS[2], IDENTITY);
 
-    memcpy4(NORMAL_MATRIX, IDENTITY, sizeof(Matrix4x4));
-    memcpy4(SCREENVIEW_MATRIX, IDENTITY, sizeof(Matrix4x4));
+    FASTCPY4(NORMAL_MATRIX, IDENTITY, sizeof(Matrix4x4));
+    FASTCPY4(SCREENVIEW_MATRIX, IDENTITY, sizeof(Matrix4x4));
+
+    const VideoMode* vid_mode = GetVideoMode();
 
     glDepthRange(0.0f, 1.0f);
     glViewport(0, 0, vid_mode->width, vid_mode->height);
@@ -110,7 +96,7 @@ static void transpose(GLfloat* m) {
 }
 
 static void recalculateNormalMatrix() {
-    memcpy4(NORMAL_MATRIX, stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)), sizeof(Matrix4x4));
+    FASTCPY4(NORMAL_MATRIX, stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)), sizeof(Matrix4x4));
     inverse((GLfloat*) NORMAL_MATRIX);
     transpose((GLfloat*) NORMAL_MATRIX);
 }
@@ -136,20 +122,16 @@ void APIENTRY glLoadIdentity() {
 }
 
 void APIENTRY glTranslatef(GLfloat x, GLfloat y, GLfloat z) {
-    static Matrix4x4 trn __attribute__((aligned(32))) = {
+    const Matrix4x4 trn __attribute__((aligned(32))) = {
         1.0f, 0.0f, 0.0f, 0.0f,
         0.0f, 1.0f, 0.0f, 0.0f,
         0.0f, 0.0f, 1.0f, 0.0f,
-        0.0f, 0.0f, 0.0f, 1.0f
+        x, y, z, 1.0f
     };
 
-    trn[M12] = x;
-    trn[M13] = y;
-    trn[M14] = z;
-
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix(&trn);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4(&trn);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 
     if(MATRIX_MODE == GL_MODELVIEW) {
         recalculateNormalMatrix();
@@ -158,20 +140,16 @@ void APIENTRY glTranslatef(GLfloat x, GLfloat y, GLfloat z) {
 
 
 void APIENTRY glScalef(GLfloat x, GLfloat y, GLfloat z) {
-    static Matrix4x4 scale __attribute__((aligned(32))) = {
-        1.0f, 0.0f, 0.0f, 0.0f,
-        0.0f, 1.0f, 0.0f, 0.0f,
-        0.0f, 0.0f, 1.0f, 0.0f,
+    const Matrix4x4 scale __attribute__((aligned(32))) = {
+        x, 0.0f, 0.0f, 0.0f,
+        0.0f, y, 0.0f, 0.0f,
+        0.0f, 0.0f, z, 0.0f,
         0.0f, 0.0f, 0.0f, 1.0f
     };
 
-    scale[M0] = x;
-    scale[M5] = y;
-    scale[M10] = z;
-
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix(&scale);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4(&scale);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 
     if(MATRIX_MODE == GL_MODELVIEW) {
         recalculateNormalMatrix();
@@ -179,7 +157,7 @@ void APIENTRY glScalef(GLfloat x, GLfloat y, GLfloat z) {
 }
 
 void APIENTRY glRotatef(GLfloat angle, GLfloat x, GLfloat  y, GLfloat z) {
-    static Matrix4x4 rotate __attribute__((aligned(32))) = {
+    Matrix4x4 rotate __attribute__((aligned(32))) = {
         1.0f, 0.0f, 0.0f, 0.0f,
         0.0f, 1.0f, 0.0f, 0.0f,
         0.0f, 0.0f, 1.0f, 0.0f,
@@ -190,7 +168,7 @@ void APIENTRY glRotatef(GLfloat angle, GLfloat x, GLfloat  y, GLfloat z) {
     float c = cos(r);
     float s = sin(r);
 
-    vec3f_normalize(x, y, z);
+    VEC3_NORMALIZE(x, y, z);
 
     float invc = 1.0f - c;
     float xs = x * s;
@@ -212,9 +190,9 @@ void APIENTRY glRotatef(GLfloat angle, GLfloat x, GLfloat  y, GLfloat z) {
     rotate[M9] = yz * invc - xs;
     rotate[M10] = (z * z) * invc + c;
 
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix(&rotate);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4((const Matrix4x4*) &rotate);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 
     if(MATRIX_MODE == GL_MODELVIEW) {
         recalculateNormalMatrix();
@@ -258,7 +236,7 @@ void APIENTRY glOrtho(GLfloat left, GLfloat right,
              GLfloat znear, GLfloat zfar) {
 
     /* Ortho Matrix */
-    static Matrix4x4 OrthoMatrix __attribute__((aligned(32))) = {
+    Matrix4x4 OrthoMatrix __attribute__((aligned(32))) = {
         1.0f, 0.0f, 0.0f, 0.0f,
         0.0f, 1.0f, 0.0f, 0.0f,
         0.0f, 0.0f, 1.0f, 0.0f,
@@ -272,9 +250,9 @@ void APIENTRY glOrtho(GLfloat left, GLfloat right,
     OrthoMatrix[M13] = -(top + bottom) / (top - bottom);
     OrthoMatrix[M14] = -(zfar + znear) / (zfar - znear);
 
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix(&OrthoMatrix);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4((const Matrix4x4*) &OrthoMatrix);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 }
 
 
@@ -284,9 +262,9 @@ void APIENTRY glFrustum(GLfloat left, GLfloat right,
                GLfloat znear, GLfloat zfar) {
 
     /* Frustum Matrix */
-    static Matrix4x4 FrustumMatrix __attribute__((aligned(32)));
+    Matrix4x4 FrustumMatrix __attribute__((aligned(32)));
 
-    memset(FrustumMatrix, 0, sizeof(float) * 16);
+    MEMSET(FrustumMatrix, 0, sizeof(float) * 16);
 
     const float near2 = 2.0f * znear;
     const float A = (right + left) / (right - left);
@@ -303,39 +281,20 @@ void APIENTRY glFrustum(GLfloat left, GLfloat right,
     FrustumMatrix[M11] = -1.0f;
     FrustumMatrix[M14] = D;
 
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix(&FrustumMatrix);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4((const Matrix4x4*) &FrustumMatrix);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 }
 
 
 /* Multiply the current matrix by an arbitrary matrix */
 void glMultMatrixf(const GLfloat *m) {
-    static Matrix4x4 TEMP;
+    Matrix4x4 TEMP;
+    FASTCPY4(TEMP, m, sizeof(Matrix4x4));
 
-    TEMP[M0] = m[0];
-    TEMP[M1] = m[1];
-    TEMP[M2] = m[2];
-    TEMP[M3] = m[3];
-
-    TEMP[M4] = m[4];
-    TEMP[M5] = m[5];
-    TEMP[M6] = m[6];
-    TEMP[M7] = m[7];
-
-    TEMP[M8] = m[8];
-    TEMP[M9] = m[9];
-    TEMP[M10] = m[10];
-    TEMP[M11] = m[11];
-
-    TEMP[M12] = m[12];
-    TEMP[M13] = m[13];
-    TEMP[M14] = m[14];
-    TEMP[M15] = m[15];
-
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix((Matrix4x4*) &TEMP);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4((const Matrix4x4*) &TEMP);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 
     if(MATRIX_MODE == GL_MODELVIEW) {
         recalculateNormalMatrix();
@@ -400,9 +359,9 @@ void glMultTransposeMatrixf(const GLfloat *m) {
     TEMP[M14] = m[11];
     TEMP[M15] = m[15];
 
-    upload_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
-    multiply_matrix(&TEMP);
-    download_matrix(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    UploadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
+    MultiplyMatrix4x4((const Matrix4x4*) &TEMP);
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + MATRIX_IDX));
 
     if(MATRIX_MODE == GL_MODELVIEW) {
         recalculateNormalMatrix();
@@ -411,6 +370,8 @@ void glMultTransposeMatrixf(const GLfloat *m) {
 
 /* Set the GL viewport */
 void APIENTRY glViewport(GLint x, GLint y, GLsizei width, GLsizei height) {
+    const VideoMode* vid_mode = GetVideoMode();
+
     gl_viewport_x1 = x;
     gl_viewport_y1 = y;
     gl_viewport_width = width;
@@ -453,40 +414,40 @@ static inline void vec3f_cross(const GLfloat* v1, const GLfloat* v2, GLfloat* re
 GL_FORCE_INLINE void vec3f_normalize_sh4(float *v){
     float length, ilength;
 
-	ilength = MATH_fsrra(v[0]*v[0] + v[1]*v[1] + v[2]*v[2]);
+    ilength = MATH_fsrra(v[0]*v[0] + v[1]*v[1] + v[2]*v[2]);
     length = MATH_Fast_Invert(ilength);
-	if (length)
-	{
-		v[0] *= ilength;
-		v[1] *= ilength;
-		v[2] *= ilength;
-	}
+    if (length)
+    {
+        v[0] *= ilength;
+        v[1] *= ilength;
+        v[2] *= ilength;
+    }
 }
 
 void gluLookAt(GLfloat eyex, GLfloat eyey, GLfloat eyez, GLfloat centerx,
                GLfloat centery, GLfloat centerz, GLfloat upx, GLfloat upy,
                GLfloat upz) {
     GLfloat m [16];
-   	GLfloat f [3];
-	GLfloat u [3];
-	GLfloat s [3];
+    GLfloat f [3];
+    GLfloat u [3];
+    GLfloat s [3];
 
-	f[0] = centerx - eyex;
-	f[1] = centery - eyey;
-	f[2] = centerz - eyez;
+    f[0] = centerx - eyex;
+    f[1] = centery - eyey;
+    f[2] = centerz - eyez;
 
-	u[0] = upx;
-	u[1] = upy;
-	u[2] = upz;
+    u[0] = upx;
+    u[1] = upy;
+    u[2] = upz;
 
     vec3f_normalize_sh4(f);
-	vec3f_cross(f, u, s);
+    vec3f_cross(f, u, s);
     vec3f_normalize_sh4(s);
-	vec3f_cross(s, f, u);
+    vec3f_cross(s, f, u);
 
-	m[0] =  s[0]; m[4] =  s[1]; m[8] =   s[2]; m[12] = 0.0f;
-	m[1] =  u[0]; m[5] =  u[1]; m[9] =   u[2]; m[13] = 0.0f;
-	m[2] = -f[0]; m[6] = -f[1]; m[10] = -f[2]; m[14] = 0.0f;
+    m[0] =  s[0]; m[4] =  s[1]; m[8] =   s[2]; m[12] = 0.0f;
+    m[1] =  u[0]; m[5] =  u[1]; m[9] =   u[2]; m[13] = 0.0f;
+    m[2] = -f[0]; m[6] = -f[1]; m[10] = -f[2]; m[14] = 0.0f;
     m[3] =  0.0f; m[7] =  0.0f; m[11] =  0.0f; m[15] = 1.0f;
 
     static Matrix4x4 trn __attribute__((aligned(32))) = {
@@ -501,26 +462,26 @@ void gluLookAt(GLfloat eyex, GLfloat eyey, GLfloat eyez, GLfloat centerx,
     trn[M14] = -eyez;
 
     // Does not modify internal Modelview matrix
-    upload_matrix(&m);
-    multiply_matrix(&trn);
-    multiply_matrix(stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
-    download_matrix(stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
+    UploadMatrix4x4((const Matrix4x4*) &m);
+    MultiplyMatrix4x4((const Matrix4x4*) &trn);
+    MultiplyMatrix4x4(stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
+    DownloadMatrix4x4(stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
 }
 
 void _glApplyRenderMatrix() {
-    upload_matrix(&SCREENVIEW_MATRIX);
-    multiply_matrix(stack_top(MATRIX_STACKS + (GL_PROJECTION & 0xF)));
-    multiply_matrix(stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
+    UploadMatrix4x4((const Matrix4x4*) &SCREENVIEW_MATRIX);
+    MultiplyMatrix4x4((const Matrix4x4*) stack_top(MATRIX_STACKS + (GL_PROJECTION & 0xF)));
+    MultiplyMatrix4x4((const Matrix4x4*) stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
 }
 
 void _glMatrixLoadTexture() {
-    upload_matrix(stack_top(MATRIX_STACKS + (GL_TEXTURE & 0xF)));
+    UploadMatrix4x4((const Matrix4x4*) stack_top(MATRIX_STACKS + (GL_TEXTURE & 0xF)));
 }
 
 void _glMatrixLoadModelView() {
-    upload_matrix(stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
+    UploadMatrix4x4((const Matrix4x4*) stack_top(MATRIX_STACKS + (GL_MODELVIEW & 0xF)));
 }
 
 void _glMatrixLoadNormal() {
-    upload_matrix(&NORMAL_MATRIX);
+    UploadMatrix4x4((const Matrix4x4*) &NORMAL_MATRIX);
 }
diff --git a/GL/platform.h b/GL/platform.h
new file mode 100644
index 0000000..b8d4db5
--- /dev/null
+++ b/GL/platform.h
@@ -0,0 +1,498 @@
+#pragma once
+
+#include <stdint.h>
+#include <stdbool.h>
+#include <assert.h>
+
+#define MEMSET(dst, v, size) memset((dst), (v), (size))
+
+typedef enum GPUAlpha {
+    GPU_ALPHA_DISABLE = 0,
+    GPU_ALPHA_ENABLE = 1
+} GPUAlpha;
+
+typedef enum GPUTexture {
+    GPU_TEXTURE_DISABLE = 0,
+    GPU_TEXTURE_ENABLE = 1
+} GPUTexture;
+
+typedef enum GPUTextureAlpha {
+    GPU_TXRALPHA_DISABLE = 1,
+    GPU_TXRALPHA_ENABLE = 0
+} GPUTextureAlpha;
+
+typedef enum GPUList {
+    GPU_LIST_OP_POLY       = 0,
+    GPU_LIST_OP_MOD        = 1,
+    GPU_LIST_TR_POLY       = 2,
+    GPU_LIST_TR_MOD        = 3,
+    GPU_LIST_PT_POLY       = 4
+} GPUList;
+
+typedef enum GPUBlend {
+    GPU_BLEND_ZERO          = 0,
+    GPU_BLEND_ONE           = 1,
+    GPU_BLEND_DESTCOLOR     = 2,
+    GPU_BLEND_INVDESTCOLOR  = 3,
+    GPU_BLEND_SRCALPHA      = 4,
+    GPU_BLEND_INVSRCALPHA   = 5,
+    GPU_BLEND_DESTALPHA     = 6,
+    GPU_BLEND_INVDESTALPHA  = 7
+} GPUBlend;
+
+typedef enum GPUDepthCompare {
+    GPU_DEPTHCMP_NEVER      = 0,
+    GPU_DEPTHCMP_LESS       = 1,
+    GPU_DEPTHCMP_EQUAL      = 2,
+    GPU_DEPTHCMP_LEQUAL     = 3,
+    GPU_DEPTHCMP_GREATER    = 4,
+    GPU_DEPTHCMP_NOTEQUAL   = 5,
+    GPU_DEPTHCMP_GEQUAL     = 6,
+    GPU_DEPTHCMP_ALWAYS     = 7
+} GPUDepthCompare;
+
+typedef enum GPUTextureFormat {
+    GPU_TXRFMT_NONE,
+    GPU_TXRFMT_VQ_DISABLE = (0 << 30),
+    GPU_TXRFMT_VQ_ENABLE = (1 << 30),
+    GPU_TXRFMT_ARGB1555 = (0 << 27),
+    GPU_TXRFMT_RGB565 = (1 << 27),
+    GPU_TXRFMT_ARGB4444 = (2 << 27),
+    GPU_TXRFMT_YUV422 = (3 << 27),
+    GPU_TXRFMT_BUMP = (4 << 27),
+    GPU_TXRFMT_PAL4BPP = (5 << 27),
+    GPU_TXRFMT_PAL8BPP = (6 << 27),
+    GPU_TXRFMT_TWIDDLED = (0 << 26),
+    GPU_TXRFMT_NONTWIDDLED = (1 << 26),
+    GPU_TXRFMT_NOSTRIDE = (0 << 21),
+    GPU_TXRFMT_STRIDE = (1 << 21)
+} GPUTextureFormat;
+
+static inline uint32_t GPUPaletteSelect8BPP(uint32_t x) {
+    return x << 25;
+}
+
+static inline uint32_t GPUPaletteSelect4BPP(uint32_t x) {
+    return x << 21;
+}
+
+typedef enum GPUCulling {
+    GPU_CULLING_NONE = 0,
+    GPU_CULLING_SMALL = 1,
+    GPU_CULLING_CCW = 2,
+    GPU_CULLING_CW = 3
+} GPUCulling;
+
+typedef enum GPUUVFlip {
+    GPU_UVFLIP_NONE = 0,
+    GPU_UVFLIP_V = 1,
+    GPU_UVFLIP_U = 2,
+    GPU_UVFLIP_UV = 3
+} GPUUVFlip;
+
+typedef enum GPUUVClamp {
+    GPU_UVCLAMP_NONE = 0,
+    GPU_UVCLAMP_V = 1,
+    GPU_UVCLAMP_U = 2,
+    GPU_UVCLAMP_UV = 3
+} GPUUVClamp;
+
+typedef enum GPUColorClamp {
+    GPU_CLRCLAMP_DISABLE = 0,
+    GPU_CLRCLAMP_ENABLE = 1
+} GPUColorClamp;
+
+typedef enum GPUFilter {
+    GPU_FILTER_NEAREST = 0,
+    GPU_FILTER_BILINEAR = 2,
+    GPU_FILTER_TRILINEAR1 = 4,
+    GPU_FILTER_TRILINEAR2 = 6
+} GPUFilter;
+
+typedef enum GPUDepthWrite {
+    GPU_DEPTHWRITE_ENABLE = 0,
+    GPU_DEPTHWRITE_DISABLE = 1
+} GPUDepthWrite;
+
+typedef enum GPUUserClip {
+    GPU_USERCLIP_DISABLE = 0,
+    GPU_USERCLIP_INSIDE = 2,
+    GPU_USERCLIP_OUTSIDE = 3
+} GPUUserClip;
+
+typedef enum GPUColorFormat {
+    GPU_CLRFMT_ARGBPACKED = 0,
+    GPU_CLRFMT_4FLOATS = 1,
+    GPU_CLRFMT_INTENSITY = 2,
+    GPU_CLRFMT_INTENSITY_PREV = 3
+} GPUColorFormat;
+
+typedef enum GPUUVFormat {
+    GPU_UVFMT_32BIT = 0,
+    GPU_UVFMT_16BIT = 1
+} GPUUVFormat;
+
+typedef enum GPUPaletteFormat {
+    GPU_PAL_ARGB1555 = 0,
+    GPU_PAL_RGB565 = 1,
+    GPU_PAL_ARGB4444 = 2,
+    GPU_PAL_ARGB8888 = 3
+} GPUPaletteFormat;
+
+typedef enum GPUFog {
+    GPU_FOG_TABLE = 0,
+    GPU_FOG_VERTEX = 1,
+    GPU_FOG_DISABLE = 2,
+    GPU_FOG_TABLE2 = 3
+} GPUFog;
+
+typedef enum GPUShade {
+    GPU_SHADE_FLAT = 0,
+    GPU_SHADE_GOURAUD = 1
+} GPUShade;
+
+typedef enum GPUTextureEnv {
+    GPU_TXRENV_REPLACE = 0,
+    GPU_TXRENV_MODULATE = 1,
+    GPU_TXRENV_DECAL = 2,
+    GPU_TXRENV_MODULATEALPHA = 3
+} GPUTextureEnv;
+
+typedef struct VideoMode {
+    uint16_t width;
+    uint16_t height;
+} VideoMode;
+
+const VideoMode* GetVideoMode();
+
+/* Duplication of pvr_poly_cxt_t from KOS so that we can
+ * compile on non-KOS platforms for testing */
+
+typedef struct {
+    GPUList     list_type;
+
+    struct {
+        int     alpha;
+        int     shading;
+        int     fog_type;
+        int     culling;
+        int     color_clamp;
+        int     clip_mode;
+        int     modifier_mode;
+        int     specular;
+        int     alpha2;
+        int     fog_type2;
+        int     color_clamp2;
+    } gen;
+    struct {
+        int     src;
+        int     dst;
+        int     src_enable;
+        int     dst_enable;
+        int     src2;
+        int     dst2;
+        int     src_enable2;
+        int     dst_enable2;
+    } blend;
+    struct {
+        int     color;
+        int     uv;
+        int     modifier;
+    } fmt;
+    struct {
+        int     comparison;
+        int     write;
+    } depth;
+    struct {
+        int     enable;
+        int     filter;
+        int     mipmap;
+        int     mipmap_bias;
+        int     uv_flip;
+        int     uv_clamp;
+        int     alpha;
+        int     env;
+        int     width;
+        int     height;
+        int     format;
+        void*   base;
+    } txr;
+    struct {
+        int     enable;
+        int     filter;
+        int     mipmap;
+        int     mipmap_bias;
+        int     uv_flip;
+        int     uv_clamp;
+        int     alpha;
+        int     env;
+        int     width;
+        int     height;
+        int     format;
+        void*   base;
+    } txr2;
+} PolyContext;
+
+typedef struct {
+    uint32_t cmd;
+    uint32_t mode1;
+    uint32_t mode2;
+    uint32_t mode3;
+    uint32_t d1;
+    uint32_t d2;
+    uint32_t d3;
+    uint32_t d4;
+} PolyHeader;
+
+enum GPUCommand {
+    GPU_CMD_POLYHDR = 0x80840000,
+    GPU_CMD_VERTEX = 0xe0000000,
+    GPU_CMD_VERTEX_EOL = 0xf0000000,
+    GPU_CMD_USERCLIP = 0x20000000,
+    GPU_CMD_MODIFIER = 0x80000000,
+    GPU_CMD_SPRITE = 0xA0000000
+};
+
+typedef float Matrix4x4[16];
+
+void SceneBegin();
+
+void SceneListBegin(GPUList list);
+void SceneListSubmit(void* src, int n);
+void SceneListFinish();
+
+void SceneFinish();
+
+#define GPU_TA_CMD_TYPE_SHIFT       24
+#define GPU_TA_CMD_TYPE_MASK        (7 << GPU_TA_CMD_TYPE_SHIFT)
+
+#define GPU_TA_CMD_USERCLIP_SHIFT   16
+#define GPU_TA_CMD_USERCLIP_MASK    (3 << GPU_TA_CMD_USERCLIP_SHIFT)
+
+#define GPU_TA_CMD_CLRFMT_SHIFT     4
+#define GPU_TA_CMD_CLRFMT_MASK      (7 << GPU_TA_CMD_CLRFMT_SHIFT)
+
+#define GPU_TA_CMD_SPECULAR_SHIFT   2
+#define GPU_TA_CMD_SPECULAR_MASK    (1 << GPU_TA_CMD_SPECULAR_SHIFT)
+
+#define GPU_TA_CMD_SHADE_SHIFT      1
+#define GPU_TA_CMD_SHADE_MASK       (1 << GPU_TA_CMD_SHADE_SHIFT)
+
+#define GPU_TA_CMD_UVFMT_SHIFT      0
+#define GPU_TA_CMD_UVFMT_MASK       (1 << GPU_TA_CMD_UVFMT_SHIFT)
+
+#define GPU_TA_CMD_MODIFIER_SHIFT   7
+#define GPU_TA_CMD_MODIFIER_MASK    (1 <<  GPU_TA_CMD_MODIFIER_SHIFT)
+
+#define GPU_TA_CMD_MODIFIERMODE_SHIFT   6
+#define GPU_TA_CMD_MODIFIERMODE_MASK    (1 <<  GPU_TA_CMD_MODIFIERMODE_SHIFT)
+
+#define GPU_TA_PM1_DEPTHCMP_SHIFT   29
+#define GPU_TA_PM1_DEPTHCMP_MASK    (7 << GPU_TA_PM1_DEPTHCMP_SHIFT)
+
+#define GPU_TA_PM1_CULLING_SHIFT    27
+#define GPU_TA_PM1_CULLING_MASK     (3 << GPU_TA_PM1_CULLING_SHIFT)
+
+#define GPU_TA_PM1_DEPTHWRITE_SHIFT 26
+#define GPU_TA_PM1_DEPTHWRITE_MASK  (1 << GPU_TA_PM1_DEPTHWRITE_SHIFT)
+
+#define GPU_TA_PM1_TXRENABLE_SHIFT  25
+#define GPU_TA_PM1_TXRENABLE_MASK   (1 << GPU_TA_PM1_TXRENABLE_SHIFT)
+
+#define GPU_TA_PM1_MODIFIERINST_SHIFT   29
+#define GPU_TA_PM1_MODIFIERINST_MASK    (3 <<  GPU_TA_PM1_MODIFIERINST_SHIFT)
+
+#define GPU_TA_PM2_SRCBLEND_SHIFT   29
+#define GPU_TA_PM2_SRCBLEND_MASK    (7 << GPU_TA_PM2_SRCBLEND_SHIFT)
+
+#define GPU_TA_PM2_DSTBLEND_SHIFT   26
+#define GPU_TA_PM2_DSTBLEND_MASK    (7 << GPU_TA_PM2_DSTBLEND_SHIFT)
+
+#define GPU_TA_PM2_SRCENABLE_SHIFT  25
+#define GPU_TA_PM2_SRCENABLE_MASK   (1 << GPU_TA_PM2_SRCENABLE_SHIFT)
+
+#define GPU_TA_PM2_DSTENABLE_SHIFT  24
+#define GPU_TA_PM2_DSTENABLE_MASK   (1 << GPU_TA_PM2_DSTENABLE_SHIFT)
+
+#define GPU_TA_PM2_FOG_SHIFT        22
+#define GPU_TA_PM2_FOG_MASK     (3 << GPU_TA_PM2_FOG_SHIFT)
+
+#define GPU_TA_PM2_CLAMP_SHIFT      21
+#define GPU_TA_PM2_CLAMP_MASK       (1 << GPU_TA_PM2_CLAMP_SHIFT)
+
+#define GPU_TA_PM2_ALPHA_SHIFT      20
+#define GPU_TA_PM2_ALPHA_MASK       (1 << GPU_TA_PM2_ALPHA_SHIFT)
+
+#define GPU_TA_PM2_TXRALPHA_SHIFT   19
+#define GPU_TA_PM2_TXRALPHA_MASK    (1 << GPU_TA_PM2_TXRALPHA_SHIFT)
+
+#define GPU_TA_PM2_UVFLIP_SHIFT     17
+#define GPU_TA_PM2_UVFLIP_MASK      (3 << GPU_TA_PM2_UVFLIP_SHIFT)
+
+#define GPU_TA_PM2_UVCLAMP_SHIFT    15
+#define GPU_TA_PM2_UVCLAMP_MASK     (3 << GPU_TA_PM2_UVCLAMP_SHIFT)
+
+#define GPU_TA_PM2_FILTER_SHIFT     12
+#define GPU_TA_PM2_FILTER_MASK      (7 << GPU_TA_PM2_FILTER_SHIFT)
+
+#define GPU_TA_PM2_MIPBIAS_SHIFT    8
+#define GPU_TA_PM2_MIPBIAS_MASK     (15 << GPU_TA_PM2_MIPBIAS_SHIFT)
+
+#define GPU_TA_PM2_TXRENV_SHIFT     6
+#define GPU_TA_PM2_TXRENV_MASK      (3 << GPU_TA_PM2_TXRENV_SHIFT)
+
+#define GPU_TA_PM2_USIZE_SHIFT      3
+#define GPU_TA_PM2_USIZE_MASK       (7 << GPU_TA_PM2_USIZE_SHIFT)
+
+#define GPU_TA_PM2_VSIZE_SHIFT      0
+#define GPU_TA_PM2_VSIZE_MASK       (7 << GPU_TA_PM2_VSIZE_SHIFT)
+
+#define GPU_TA_PM3_MIPMAP_SHIFT     31
+#define GPU_TA_PM3_MIPMAP_MASK      (1 << GPU_TA_PM3_MIPMAP_SHIFT)
+
+#define GPU_TA_PM3_TXRFMT_SHIFT     0
+#define GPU_TA_PM3_TXRFMT_MASK      0xffffffff
+
+/* Compile a polygon context into a polygon header */
+static inline void CompilePolyHeader(PolyHeader *dst, const PolyContext *src) {
+    int u, v;
+    uint32_t  txr_base;
+
+    /* Basically we just take each parameter, clip it, shift it
+       into place, and OR it into the final result. */
+
+    /* The base values for CMD */
+    dst->cmd = GPU_CMD_POLYHDR;
+
+    if(src->txr.enable == GPU_TEXTURE_ENABLE)
+        dst->cmd |= 8;
+
+    /* Or in the list type, shading type, color and UV formats */
+    dst->cmd |= (src->list_type << GPU_TA_CMD_TYPE_SHIFT) & GPU_TA_CMD_TYPE_MASK;
+    dst->cmd |= (src->fmt.color << GPU_TA_CMD_CLRFMT_SHIFT) & GPU_TA_CMD_CLRFMT_MASK;
+    dst->cmd |= (src->gen.shading << GPU_TA_CMD_SHADE_SHIFT) & GPU_TA_CMD_SHADE_MASK;
+    dst->cmd |= (src->fmt.uv << GPU_TA_CMD_UVFMT_SHIFT) & GPU_TA_CMD_UVFMT_MASK;
+    dst->cmd |= (src->gen.clip_mode << GPU_TA_CMD_USERCLIP_SHIFT) & GPU_TA_CMD_USERCLIP_MASK;
+    dst->cmd |= (src->fmt.modifier << GPU_TA_CMD_MODIFIER_SHIFT) & GPU_TA_CMD_MODIFIER_MASK;
+    dst->cmd |= (src->gen.modifier_mode << GPU_TA_CMD_MODIFIERMODE_SHIFT) & GPU_TA_CMD_MODIFIERMODE_MASK;
+    dst->cmd |= (src->gen.specular << GPU_TA_CMD_SPECULAR_SHIFT) & GPU_TA_CMD_SPECULAR_MASK;
+
+    /* Polygon mode 1 */
+    dst->mode1  = (src->depth.comparison << GPU_TA_PM1_DEPTHCMP_SHIFT) & GPU_TA_PM1_DEPTHCMP_MASK;
+    dst->mode1 |= (src->gen.culling << GPU_TA_PM1_CULLING_SHIFT) & GPU_TA_PM1_CULLING_MASK;
+    dst->mode1 |= (src->depth.write << GPU_TA_PM1_DEPTHWRITE_SHIFT) & GPU_TA_PM1_DEPTHWRITE_MASK;
+    dst->mode1 |= (src->txr.enable << GPU_TA_PM1_TXRENABLE_SHIFT) & GPU_TA_PM1_TXRENABLE_MASK;
+
+    /* Polygon mode 2 */
+    dst->mode2  = (src->blend.src << GPU_TA_PM2_SRCBLEND_SHIFT) & GPU_TA_PM2_SRCBLEND_MASK;
+    dst->mode2 |= (src->blend.dst << GPU_TA_PM2_DSTBLEND_SHIFT) & GPU_TA_PM2_DSTBLEND_MASK;
+    dst->mode2 |= (src->blend.src_enable << GPU_TA_PM2_SRCENABLE_SHIFT) & GPU_TA_PM2_SRCENABLE_MASK;
+    dst->mode2 |= (src->blend.dst_enable << GPU_TA_PM2_DSTENABLE_SHIFT) & GPU_TA_PM2_DSTENABLE_MASK;
+    dst->mode2 |= (src->gen.fog_type << GPU_TA_PM2_FOG_SHIFT) & GPU_TA_PM2_FOG_MASK;
+    dst->mode2 |= (src->gen.color_clamp << GPU_TA_PM2_CLAMP_SHIFT) & GPU_TA_PM2_CLAMP_MASK;
+    dst->mode2 |= (src->gen.alpha << GPU_TA_PM2_ALPHA_SHIFT) & GPU_TA_PM2_ALPHA_MASK;
+
+    if(src->txr.enable == GPU_TEXTURE_DISABLE) {
+        dst->mode3 = 0;
+    }
+    else {
+        dst->mode2 |= (src->txr.alpha << GPU_TA_PM2_TXRALPHA_SHIFT) & GPU_TA_PM2_TXRALPHA_MASK;
+        dst->mode2 |= (src->txr.uv_flip << GPU_TA_PM2_UVFLIP_SHIFT) & GPU_TA_PM2_UVFLIP_MASK;
+        dst->mode2 |= (src->txr.uv_clamp << GPU_TA_PM2_UVCLAMP_SHIFT) & GPU_TA_PM2_UVCLAMP_MASK;
+        dst->mode2 |= (src->txr.filter << GPU_TA_PM2_FILTER_SHIFT) & GPU_TA_PM2_FILTER_MASK;
+        dst->mode2 |= (src->txr.mipmap_bias << GPU_TA_PM2_MIPBIAS_SHIFT) & GPU_TA_PM2_MIPBIAS_MASK;
+        dst->mode2 |= (src->txr.env << GPU_TA_PM2_TXRENV_SHIFT) & GPU_TA_PM2_TXRENV_MASK;
+
+        switch(src->txr.width) {
+            case 8:
+                u = 0;
+                break;
+            case 16:
+                u = 1;
+                break;
+            case 32:
+                u = 2;
+                break;
+            case 64:
+                u = 3;
+                break;
+            case 128:
+                u = 4;
+                break;
+            case 256:
+                u = 5;
+                break;
+            case 512:
+                u = 6;
+                break;
+            case 1024:
+                u = 7;
+                break;
+            default:
+                assert(0 && "Invalid texture U size");
+                u = 0;
+                break;
+        }
+
+        switch(src->txr.height) {
+            case 8:
+                v = 0;
+                break;
+            case 16:
+                v = 1;
+                break;
+            case 32:
+                v = 2;
+                break;
+            case 64:
+                v = 3;
+                break;
+            case 128:
+                v = 4;
+                break;
+            case 256:
+                v = 5;
+                break;
+            case 512:
+                v = 6;
+                break;
+            case 1024:
+                v = 7;
+                break;
+            default:
+                assert(0 && "Invalid texture V size");
+                v = 0;
+                break;
+        }
+
+        dst->mode2 |= (u << GPU_TA_PM2_USIZE_SHIFT) & GPU_TA_PM2_USIZE_MASK;
+        dst->mode2 |= (v << GPU_TA_PM2_VSIZE_SHIFT) & GPU_TA_PM2_VSIZE_MASK;
+
+        /* Polygon mode 3 */
+        dst->mode3  = (src->txr.mipmap << GPU_TA_PM3_MIPMAP_SHIFT) & GPU_TA_PM3_MIPMAP_MASK;
+        dst->mode3 |= (src->txr.format << GPU_TA_PM3_TXRFMT_SHIFT) & GPU_TA_PM3_TXRFMT_MASK;
+
+        /* Convert the texture address */
+        txr_base = (uint32_t) src->txr.base;
+        txr_base = (txr_base & 0x00fffff8) >> 3;
+        dst->mode3 |= txr_base;
+    }
+
+    if(src->fmt.modifier && src->gen.modifier_mode) {
+        /* If we're affected by a modifier volume, silently promote the header
+           to the one that is affected by a modifier volume. */
+        dst->d1 = dst->mode2;
+        dst->d2 = dst->mode3;
+    }
+    else {
+        dst->d1 = dst->d2 = 0xffffffff;
+    }
+
+    dst->d3 = dst->d4 = 0xffffffff;
+}
+
+#ifdef BACKEND_KOSPVR
+#include "platforms/sh4.h"
+#else
+#include "platforms/software.h"
+#endif
diff --git a/GL/platforms/sh4.c b/GL/platforms/sh4.c
new file mode 100644
index 0000000..23d8c7a
--- /dev/null
+++ b/GL/platforms/sh4.c
@@ -0,0 +1,72 @@
+#include "../platform.h"
+#include "sh4.h"
+
+#define TA_SQ_ADDR (unsigned int *)(void *) \
+    (0xe0000000 | (((unsigned long)0x10000000) & 0x03ffffe0))
+
+#define QACRTA ((((unsigned int)0x10000000)>>26)<<2)&0x1c
+
+#define PVR_VERTEX_BUF_SIZE 2560 * 256
+
+void InitGPU(_Bool autosort, _Bool fsaa) {
+    pvr_init_params_t params = {
+        /* Enable opaque and translucent polygons with size 32 and 32 */
+        {PVR_BINSIZE_32, PVR_BINSIZE_0, PVR_BINSIZE_32, PVR_BINSIZE_0, PVR_BINSIZE_32},
+        PVR_VERTEX_BUF_SIZE, /* Vertex buffer size */
+        0, /* No DMA */
+        fsaa, /* No FSAA */
+        (autosort) ? 0 : 1 /* Disable translucent auto-sorting to match traditional GL */
+    };
+
+    pvr_init(&params);
+}
+
+void SceneBegin() {
+    pvr_wait_ready();
+    pvr_scene_begin();
+    QACR0 = QACRTA;
+    QACR1 = QACRTA;
+}
+
+void SceneListBegin(GPUList list) {
+    pvr_list_begin(list);
+}
+
+void SceneListSubmit(void* src, int n) {
+    uint32_t *d = (uint32_t*) TA_SQ_ADDR;
+    uint32_t *s = src;
+
+    /* fill/write queues as many times necessary */
+    while(n--) {
+        __asm__("pref @%0" : : "r"(s + 8));  /* prefetch 32 bytes for next loop */
+        d[0] = *(s++);
+        d[1] = *(s++);
+        d[2] = *(s++);
+        d[3] = *(s++);
+        d[4] = *(s++);
+        d[5] = *(s++);
+        d[6] = *(s++);
+        d[7] = *(s++);
+        __asm__("pref @%0" : : "r"(d));
+        d += 8;
+    }
+
+    /* Wait for both store queues to complete */
+    d = (uint32_t *)0xe0000000;
+    d[0] = d[8] = 0;
+}
+
+void SceneListFinish() {
+    pvr_list_finish();
+}
+
+void SceneFinish() {
+    pvr_scene_finish();
+}
+
+const VideoMode* GetVideoMode() {
+    static VideoMode mode;
+    mode.width = vid_mode->width;
+    mode.height = vid_mode->height;
+    return &mode;
+}
diff --git a/GL/platforms/sh4.h b/GL/platforms/sh4.h
new file mode 100644
index 0000000..0bc7dd6
--- /dev/null
+++ b/GL/platforms/sh4.h
@@ -0,0 +1,124 @@
+#pragma once
+
+#include <kos.h>
+#include <dc/matrix.h>
+#include <dc/pvr.h>
+#include <dc/vec3f.h>
+#include <dc/fmath.h>
+#include <dc/matrix3d.h>
+
+#include "../types.h"
+#include "sh4_math.h"
+
+#define FASTCPY(dst, src, bytes) \
+    (bytes % 32 == 0) ? sq_cpy(dst, src, bytes) : memcpy(dst, src, bytes)
+
+#define FASTCPY4(dst, src, bytes) \
+    (bytes % 32 == 0) ? sq_cpy(dst, src, bytes) : memcpy4(dst, src, bytes)
+
+#define MEMSET4(dst, v, size) memset4((dst), (v), (size))
+
+#define VEC3_NORMALIZE(x, y, z) vec3f_normalize((x), (y), (z))
+#define VEC3_LENGTH(x, y, z, l) vec3f_length((x), (y), (z), (l))
+#define VEC3_DOT(x1, y1, z1, x2, y2, z2, d) vec3f_dot((x1), (y1), (z1), (x2), (y2), (z2), (d))
+
+static inline void UploadMatrix4x4(const Matrix4x4* mat) {
+    mat_load((matrix_t*) mat);
+}
+
+static inline void DownloadMatrix4x4(Matrix4x4* mat) {
+    mat_store((matrix_t*) mat);
+}
+
+static inline void MultiplyMatrix4x4(const Matrix4x4* mat) {
+    mat_apply((matrix_t*) mat);
+}
+
+static inline void TransformVec3(float* x) {
+    mat_trans_single4(x[0], x[1], x[2], x[3]);
+}
+
+/* Transform a 3-element vector using the stored matrix (w == 1) */
+static inline void TransformVec3NoMod(const float* xIn, float* xOut) {
+    mat_trans_single3_nodiv_nomod(xIn[0], xIn[1], xIn[2], xOut[0], xOut[1], xOut[2]);
+}
+
+/* Transform a 3-element normal using the stored matrix (w == 0)*/
+static inline void TransformNormalNoMod(const float* in, float* out) {
+    mat_trans_normal3_nomod(in[0], in[1], in[2], out[0], out[1], out[2]);
+}
+
+/* Transform a 4-element vector in-place by the stored matrix */
+inline void TransformVec4(float* x) {
+
+}
+
+static inline void TransformVertices(Vertex* vertices, const int count) {
+    Vertex* it = vertices;
+    for(int i = 0; i < count; ++i, ++it) {
+        register float __x __asm__("fr12") = (it->xyz[0]);
+        register float __y __asm__("fr13") = (it->xyz[1]);
+        register float __z __asm__("fr14") = (it->xyz[2]);
+        register float __w __asm__("fr15") = (it->w);
+
+        __asm__ __volatile__(
+            "fldi1 fr15\n"
+            "ftrv   xmtrx,fv12\n"
+            : "=f" (__x), "=f" (__y), "=f" (__z), "=f" (__w)
+            : "0" (__x), "1" (__y), "2" (__z), "3" (__w)
+        );
+
+        it->xyz[0] = __x;
+        it->xyz[1] = __y;
+        it->xyz[2] = __z;
+        it->w = __w;
+    }
+}
+
+void InitGPU(_Bool autosort, _Bool fsaa);
+
+static inline size_t GPUMemoryAvailable() {
+    return pvr_mem_available();
+}
+
+static inline void* GPUMemoryAlloc(size_t size) {
+    return pvr_mem_malloc(size);
+}
+
+static inline void GPUSetPaletteFormat(GPUPaletteFormat format) {
+    pvr_set_pal_format(format);
+}
+
+static inline void GPUSetPaletteEntry(uint32_t idx, uint32_t value) {
+    pvr_set_pal_entry(idx, value);
+}
+
+static inline void GPUSetBackgroundColour(float r, float g, float b) {
+    pvr_set_bg_color(r, g, b);
+}
+
+#define PT_ALPHA_REF 0x011c
+
+static inline void GPUSetAlphaCutOff(uint8_t val) {
+    PVR_SET(PT_ALPHA_REF, val);
+}
+
+static inline void GPUSetClearDepth(float v) {
+    pvr_set_zclip(v);
+}
+
+static inline void GPUSetFogLinear(float start, float end) {
+    pvr_fog_table_linear(start, end);
+}
+
+static inline void GPUSetFogExp(float density) {
+    pvr_fog_table_exp(density);
+}
+
+static inline void GPUSetFogExp2(float density) {
+    pvr_fog_table_exp2(density);
+}
+
+static inline void GPUSetFogColor(float r, float g, float b, float a) {
+    pvr_fog_table_color(r, g, b, a);
+}
diff --git a/GL/sh4_math.h b/GL/platforms/sh4_math.h
similarity index 100%
rename from GL/sh4_math.h
rename to GL/platforms/sh4_math.h
diff --git a/GL/platforms/software.c b/GL/platforms/software.c
new file mode 100644
index 0000000..7e904a0
--- /dev/null
+++ b/GL/platforms/software.c
@@ -0,0 +1,330 @@
+#include <SDL.h>
+
+#include <stdlib.h>
+#include <string.h>
+
+#include "../platform.h"
+#include "software.h"
+#include "software/edge_equation.h"
+#include "software/parameter_equation.h"
+
+static size_t AVAILABLE_VRAM = 16 * 1024 * 1024;
+static Matrix4x4 MATRIX;
+
+static SDL_Window* WINDOW = NULL;
+static SDL_Renderer* RENDERER = NULL;
+
+static uint8_t BACKGROUND_COLOR[3] = {0, 0, 0};
+
+GPUCulling CULL_MODE = GPU_CULLING_CCW;
+
+
+static VideoMode vid_mode = {
+    640, 480
+};
+
+
+typedef struct GPUVertex {
+    uint32_t flags;
+    float x;
+    float y;
+    float z;
+    float u;
+    float v;
+    uint8_t bgra[4];
+    uint8_t obgra[4];
+} GPUVertex;
+
+#define MIN(x, y) ((x) < (y) ? (x) : (y))
+#define MAX(x, y) ((x) > (y) ? (x) : (y))
+
+static void DrawTriangle(GPUVertex* v0, GPUVertex* v1, GPUVertex* v2) {
+    // Compute triangle bounding box.
+
+    int minX = MIN(MIN(v0->x, v1->x), v2->x);
+    int maxX = MAX(MAX(v0->x, v1->x), v2->x);
+    int minY = MIN(MIN(v0->y, v1->y), v2->y);
+    int maxY = MAX(MAX(v0->y, v1->y), v2->y);
+
+    // Clip to scissor rect.
+
+    minX = MAX(minX, 0);
+    maxX = MIN(maxX, vid_mode.width);
+    minY = MAX(minY, 0);
+    maxY = MIN(maxY, vid_mode.height);
+
+    // Compute edge equations.
+
+    EdgeEquation e0, e1, e2;
+    EdgeEquationInit(&e0, &v0->x, &v1->x);
+    EdgeEquationInit(&e1, &v1->x, &v2->x);
+    EdgeEquationInit(&e2, &v2->x, &v0->x);
+
+    float area = 0.5 * (e0.c + e1.c + e2.c);
+
+    /* This is very ugly. I don't understand the math properly
+     * so I just swap the vertex order if something is back-facing
+     * and we want to render it. Patches welcome! */
+#define REVERSE_WINDING() \
+    GPUVertex* tv = v0; \
+    v0 = v1; \
+    v1 = tv; \
+    EdgeEquationInit(&e0, &v0->x, &v1->x); \
+    EdgeEquationInit(&e1, &v1->x, &v2->x); \
+    EdgeEquationInit(&e2, &v2->x, &v0->x); \
+    area = 0.5f * (e0.c + e1.c + e2.c) \
+
+    // Check if triangle is backfacing.
+    if(CULL_MODE == GPU_CULLING_CCW) {
+        if(area < 0) {
+            return;
+        }
+    } else if(CULL_MODE == GPU_CULLING_CW) {
+        if(area < 0) {
+            // We only draw front-facing polygons, so swap
+            // the back to front and draw
+            REVERSE_WINDING();
+        } else {
+            // Front facing, so bail
+            return;
+        }
+    } else if(area < 0) {
+        /* We're not culling, but this is backfacing, so swap vertices and edges */
+        REVERSE_WINDING();
+    }
+
+    ParameterEquation r, g, b;
+
+    ParameterEquationInit(&r, v0->bgra[2], v1->bgra[2], v2->bgra[2], &e0, &e1, &e2, area);
+    ParameterEquationInit(&g, v0->bgra[1], v1->bgra[1], v2->bgra[1], &e0, &e1, &e2, area);
+    ParameterEquationInit(&b, v0->bgra[0], v1->bgra[0], v2->bgra[0], &e0, &e1, &e2, area);
+
+    // Add 0.5 to sample at pixel centers.
+    for (float x = minX + 0.5f, xm = maxX + 0.5f; x <= xm; x += 1.0f)
+    for (float y = minY + 0.5f, ym = maxY + 0.5f; y <= ym; y += 1.0f)
+    {
+      if (EdgeEquationTestPoint(&e0, x, y) && EdgeEquationTestPoint(&e1, x, y) && EdgeEquationTestPoint(&e2, x, y)) {
+        int rint = ParameterEquationEvaluate(&r, x, y);
+        int gint = ParameterEquationEvaluate(&g, x, y);
+        int bint = ParameterEquationEvaluate(&b, x, y);
+        SDL_SetRenderDrawColor(RENDERER, rint, gint, bint, 255);
+        SDL_RenderDrawPoint(RENDERER, x, y);
+      }
+    }
+}
+
+
+void InitGPU(_Bool autosort, _Bool fsaa) {
+    SDL_Init(SDL_INIT_VIDEO | SDL_INIT_EVENTS);
+
+    WINDOW = SDL_CreateWindow(
+        "GLdc",
+        SDL_WINDOWPOS_UNDEFINED,
+        SDL_WINDOWPOS_UNDEFINED,
+        vid_mode.width, vid_mode.height,
+        SDL_WINDOW_SHOWN
+    );
+
+    RENDERER = SDL_CreateRenderer(
+        WINDOW, -1, SDL_RENDERER_ACCELERATED
+    );
+}
+
+void SceneBegin() {
+    SDL_SetRenderDrawColor(RENDERER, BACKGROUND_COLOR[0], BACKGROUND_COLOR[1], BACKGROUND_COLOR[2], 0);
+    SDL_RenderClear(RENDERER);
+}
+
+void SceneListBegin(GPUList list) {
+
+}
+
+void SceneListSubmit(void* src, int n) {
+    uint32_t vertex_counter = 0;
+    const uint32_t* flags = (const uint32_t*) src;
+    uint32_t step = sizeof(GPUVertex) / sizeof(uint32_t);
+
+    for(int i = 0; i < n; ++i, flags += step) {
+        if((*flags & GPU_CMD_POLYHDR) == GPU_CMD_POLYHDR) {
+            vertex_counter = 0;
+
+            uint32_t mode1 = *(flags + 1);
+            // Extract culling mode
+            uint32_t mask = mode1 & GPU_TA_PM1_CULLING_MASK;
+            CULL_MODE = mask >> GPU_TA_PM1_CULLING_SHIFT;
+
+        } else {
+            switch(*flags) {
+            case GPU_CMD_VERTEX_EOL:
+            case GPU_CMD_VERTEX: // Fallthrough
+                vertex_counter++;
+            break;
+            default:
+                break;
+            }
+        }
+
+        if(vertex_counter > 2) {
+            GPUVertex* v0 = (GPUVertex*) (flags - step - step);
+            GPUVertex* v1 = (GPUVertex*) (flags - step);
+            GPUVertex* v2 = (GPUVertex*) (flags);
+            (vertex_counter % 2 == 0) ? DrawTriangle(v0, v1, v2) : DrawTriangle(v1, v0, v2);
+        }
+
+        if((*flags) == GPU_CMD_VERTEX_EOL) {
+            vertex_counter = 0;
+        }
+    }
+}
+
+void SceneListFinish() {
+
+}
+
+void SceneFinish() {
+    SDL_RenderPresent(RENDERER);
+
+    /* Only sensible place to hook the quit signal */
+
+    SDL_Event e = {0};
+
+    while (SDL_PollEvent(&e))
+      switch (e.type) {
+        case SDL_QUIT:
+          exit(0);
+          break;
+        default:
+          break;
+    }
+}
+
+void UploadMatrix4x4(const Matrix4x4* mat) {
+    memcpy(&MATRIX, mat, sizeof(Matrix4x4));
+}
+
+void MultiplyMatrix4x4(const Matrix4x4* mat) {
+    Matrix4x4 product;
+
+    product[0] = MATRIX[0] * (*mat)[0] + MATRIX[4] * (*mat)[1] + MATRIX[8] * (*mat)[2] + MATRIX[12] * (*mat)[3];
+    product[1] = MATRIX[1] * (*mat)[0] + MATRIX[5] * (*mat)[1] + MATRIX[9] * (*mat)[2] + MATRIX[13] * (*mat)[3];
+    product[2] = MATRIX[2] * (*mat)[0] + MATRIX[6] * (*mat)[1] + MATRIX[10] * (*mat)[2] + MATRIX[14] * (*mat)[3];
+    product[3] = MATRIX[3] * (*mat)[0] + MATRIX[7] * (*mat)[1] + MATRIX[11] * (*mat)[2] + MATRIX[15] * (*mat)[3];
+
+    product[4] = MATRIX[0] * (*mat)[4] + MATRIX[4] * (*mat)[5] + MATRIX[8] * (*mat)[6] + MATRIX[12] * (*mat)[7];
+    product[5] = MATRIX[1] * (*mat)[4] + MATRIX[5] * (*mat)[5] + MATRIX[9] * (*mat)[6] + MATRIX[13] * (*mat)[7];
+    product[6] = MATRIX[2] * (*mat)[4] + MATRIX[6] * (*mat)[5] + MATRIX[10] * (*mat)[6] + MATRIX[14] * (*mat)[7];
+    product[7] = MATRIX[3] * (*mat)[4] + MATRIX[7] * (*mat)[5] + MATRIX[11] * (*mat)[6] + MATRIX[15] * (*mat)[7];
+
+    product[8] = MATRIX[0] * (*mat)[8] + MATRIX[4] * (*mat)[9] + MATRIX[8] * (*mat)[10] + MATRIX[12] * (*mat)[11];
+    product[9] = MATRIX[1] * (*mat)[8] + MATRIX[5] * (*mat)[9] + MATRIX[9] * (*mat)[10] + MATRIX[13] * (*mat)[11];
+    product[10] = MATRIX[2] * (*mat)[8] + MATRIX[6] * (*mat)[9] + MATRIX[10] * (*mat)[10] + MATRIX[14] * (*mat)[11];
+    product[11] = MATRIX[3] * (*mat)[8] + MATRIX[7] * (*mat)[9] + MATRIX[11] * (*mat)[10] + MATRIX[15] * (*mat)[11];
+
+    product[12] = MATRIX[0] * (*mat)[12] + MATRIX[4] * (*mat)[13] + MATRIX[8] * (*mat)[14] + MATRIX[12] * (*mat)[15];
+    product[13] = MATRIX[1] * (*mat)[12] + MATRIX[5] * (*mat)[13] + MATRIX[9] * (*mat)[14] + MATRIX[13] * (*mat)[15];
+    product[14] = MATRIX[2] * (*mat)[12] + MATRIX[6] * (*mat)[13] + MATRIX[10] * (*mat)[14] + MATRIX[14] * (*mat)[15];
+    product[15] = MATRIX[3] * (*mat)[12] + MATRIX[7] * (*mat)[13] + MATRIX[11] * (*mat)[14] + MATRIX[15] * (*mat)[15];
+
+    UploadMatrix4x4(&product);
+}
+
+void DownloadMatrix4x4(Matrix4x4* mat) {
+    memcpy(mat, &MATRIX, sizeof(Matrix4x4));
+}
+
+const VideoMode* GetVideoMode() {
+    return &vid_mode;
+}
+
+size_t GPUMemoryAvailable() {
+    return AVAILABLE_VRAM;
+}
+
+void* GPUMemoryAlloc(size_t size) {
+    if(size > AVAILABLE_VRAM) {
+        return NULL;
+    } else {
+        AVAILABLE_VRAM -= size;
+        return malloc(size);
+    }
+}
+
+void GPUSetPaletteFormat(GPUPaletteFormat format) {
+
+}
+
+void GPUSetPaletteEntry(uint32_t idx, uint32_t value) {
+
+}
+
+void GPUSetBackgroundColour(float r, float g, float b) {
+    BACKGROUND_COLOR[0] = r * 255.0f;
+    BACKGROUND_COLOR[1] = g * 255.0f;
+    BACKGROUND_COLOR[2] = b * 255.0f;
+}
+
+void GPUSetAlphaCutOff(uint8_t v) {
+
+}
+
+void GPUSetClearDepth(float v) {
+
+}
+
+void GPUSetFogLinear(float start, float end) {
+
+}
+
+void GPUSetFogExp(float density) {
+
+}
+
+void GPUSetFogExp2(float density) {
+
+}
+
+void GPUSetFogColor(float r, float g, float b, float a) {
+
+}
+
+void TransformVec3NoMod(const float* v, float* ret) {
+    ret[0] = v[0] * MATRIX[0] + v[1] * MATRIX[4] + v[2] * MATRIX[8] + 1.0f * MATRIX[12];
+    ret[1] = v[0] * MATRIX[1] + v[1] * MATRIX[5] + v[2] * MATRIX[9] + 1.0f * MATRIX[13];
+    ret[2] = v[0] * MATRIX[2] + v[1] * MATRIX[6] + v[2] * MATRIX[10] + 1.0f * MATRIX[14];
+}
+
+void TransformVec4NoMod(const float* v, float* ret) {
+    ret[0] = v[0] * MATRIX[0] + v[1] * MATRIX[4] + v[2] * MATRIX[8] + v[3] * MATRIX[12];
+    ret[1] = v[0] * MATRIX[1] + v[1] * MATRIX[5] + v[2] * MATRIX[9] + v[3] * MATRIX[13];
+    ret[2] = v[0] * MATRIX[2] + v[1] * MATRIX[6] + v[2] * MATRIX[10] + v[3] * MATRIX[14];
+    ret[3] = v[0] * MATRIX[3] + v[1] * MATRIX[7] + v[2] * MATRIX[11] + v[3] * MATRIX[15];
+}
+
+void TransformVec3(float* v) {
+    float ret[3];
+    TransformVec3NoMod(v, ret);
+    FASTCPY(v, ret, sizeof(float) * 3);
+}
+
+void TransformVec4(float* v) {
+    float ret[4];
+    TransformVec4NoMod(v, ret);
+    FASTCPY(v, ret, sizeof(float) * 4);
+}
+
+void TransformVertices(Vertex* vertices, const int count) {
+    float ret[4];
+    for(int i = 0; i < count; ++i, ++vertices) {
+        ret[0] = vertices->xyz[0];
+        ret[1] = vertices->xyz[1];
+        ret[2] = vertices->xyz[2];
+        ret[3] = 1.0f;
+
+        TransformVec4(ret);
+
+        vertices->xyz[0] = ret[0];
+        vertices->xyz[1] = ret[1];
+        vertices->xyz[2] = ret[2];
+        vertices->w = ret[3];
+    }
+}
diff --git a/GL/platforms/software.h b/GL/platforms/software.h
new file mode 100644
index 0000000..5bac618
--- /dev/null
+++ b/GL/platforms/software.h
@@ -0,0 +1,69 @@
+#pragma once
+
+#include <math.h>
+#include <memory.h>
+
+#include "../types.h"
+
+#define MATH_Fast_Divide(n, d) (n / d)
+#define MATH_fmac(a, b, c) (a * b + c)
+#define MATH_Fast_Sqrt(x) sqrtf((x))
+#define MATH_fsrra(x) (1.0f / sqrtf((x)))
+#define MATH_Fast_Invert(x) (1.0f / (x))
+
+#define FASTCPY(dst, src, bytes) memcpy(dst, src, bytes)
+#define FASTCPY4(dst, src, bytes) memcpy(dst, src, bytes)
+#define MEMSET4(dst, v, size) memset((dst), (v), (size))
+
+#define VEC3_NORMALIZE(x, y, z) \
+    do { \
+        float l = MATH_fsrra((x) * (x) + (y) * (y) + (z) * (z)); \
+        x *= l; \
+        y *= l; \
+        z *= l; \
+    } while(0)
+
+#define VEC3_LENGTH(x, y, z, d) \
+    d = MATH_Fast_Sqrt((x) * (x) + (y) * (y) + (z) * (z))
+
+#define VEC3_DOT(x1, y1, z1, x2, y2, z2, d) \
+    d = (x1 * x2) + (y1 * y2) + (z1 * z2)
+
+struct PolyHeader;
+struct PolyContext;
+
+void UploadMatrix4x4(const Matrix4x4* mat);
+void MultiplyMatrix4x4(const Matrix4x4* mat);
+void DownloadMatrix4x4(Matrix4x4* mat);
+
+/* Transform a 3-element vector in-place using the stored matrix (w == 1) */
+void TransformVec3(float* v);
+
+/* Transform a 3-element vector using the stored matrix (w == 1) */
+void TransformVec3NoMod(const float* v, float* ret);
+
+/* Transform a 3-element normal using the stored matrix (w == 0)*/
+static inline void TransformNormalNoMod(const float* xIn, float* xOut) {
+
+}
+
+void TransformVertices(Vertex* vertices, const int count);
+
+void InitGPU(_Bool autosort, _Bool fsaa);
+
+enum GPUPaletteFormat;
+
+size_t GPUMemoryAvailable();
+void* GPUMemoryAlloc(size_t size);
+void GPUSetPaletteFormat(GPUPaletteFormat format);
+void GPUSetPaletteEntry(uint32_t idx, uint32_t value);
+
+void GPUSetBackgroundColour(float r, float g, float b);
+void GPUSetAlphaCutOff(uint8_t v);
+void GPUSetClearDepth(float v);
+
+void GPUSetFogLinear(float start, float end);
+void GPUSetFogExp(float density);
+void GPUSetFogExp2(float density);
+void GPUSetFogColor(float r, float g, float b, float a);
+
diff --git a/GL/platforms/software/edge_equation.c b/GL/platforms/software/edge_equation.c
new file mode 100644
index 0000000..6a6c7cb
--- /dev/null
+++ b/GL/platforms/software/edge_equation.c
@@ -0,0 +1,20 @@
+#include "edge_equation.h"
+
+void EdgeEquationInit(EdgeEquation* edge, const float* v0, const float* v1) {
+    edge->a = v0[1] - v1[1];
+    edge->b = v1[0] - v0[0];
+    edge->c = -(edge->a * (v0[0] + v1[0]) + edge->b * (v0[1] + v1[1])) / 2;
+    edge->tie = edge->a != 0 ? edge->a > 0 : edge->b > 0;
+}
+
+float EdgeEquationEvaluate(const EdgeEquation* edge, float x, float y) {
+    return edge->a * x + edge->b * y + edge->c;
+}
+
+bool EdgeEquationTestValue(const EdgeEquation* edge, float value) {
+    return (value >= 0 || (value == 0 && edge->tie));
+}
+
+bool EdgeEquationTestPoint(const EdgeEquation* edge, float x, float y) {
+    return EdgeEquationTestValue(edge, EdgeEquationEvaluate(edge, x, y));
+}
diff --git a/GL/platforms/software/edge_equation.h b/GL/platforms/software/edge_equation.h
new file mode 100644
index 0000000..2bd1bde
--- /dev/null
+++ b/GL/platforms/software/edge_equation.h
@@ -0,0 +1,17 @@
+#pragma once
+
+#include <stdint.h>
+#include <stdbool.h>
+
+typedef struct EdgeEquation {
+    float a;
+    float b;
+    float c;
+    bool tie;
+} EdgeEquation;
+
+void EdgeEquationInit(EdgeEquation* edge, const float* v0, const float* v1);
+float EdgeEquationEvaluate(const EdgeEquation* edge, float x, float y);
+bool EdgeEquationTestValue(const EdgeEquation* edge, float value);
+bool EdgeEquationTestPoint(const EdgeEquation* edge, float x, float y);
+
diff --git a/GL/platforms/software/parameter_equation.c b/GL/platforms/software/parameter_equation.c
new file mode 100644
index 0000000..cf46299
--- /dev/null
+++ b/GL/platforms/software/parameter_equation.c
@@ -0,0 +1,16 @@
+#include "parameter_equation.h"
+#include "edge_equation.h"
+
+void ParameterEquationInit(ParameterEquation* equation, float p0, float p1, float p2, const EdgeEquation* e0, const EdgeEquation* e1, const EdgeEquation* e2, float area) {
+
+    float factor = 1.0f / (2.0f * area);
+
+    equation->a = factor * (p0 * e0->a + p1 * e1->a + p2 * e2->a);
+    equation->b = factor * (p0 * e0->b + p1 * e1->b + p2 * e2->b);
+    equation->c = factor * (p0 * e0->c + p1 * e1->c + p2 * e2->c);
+}
+
+
+float ParameterEquationEvaluate(const ParameterEquation* equation, float x, float y) {
+    return equation->a * x + equation->b * y + equation->c;
+}
diff --git a/GL/platforms/software/parameter_equation.h b/GL/platforms/software/parameter_equation.h
new file mode 100644
index 0000000..6c3beb7
--- /dev/null
+++ b/GL/platforms/software/parameter_equation.h
@@ -0,0 +1,16 @@
+#pragma once
+
+typedef struct ParameterEquation {
+    float a;
+    float b;
+    float c;
+} ParameterEquation;
+
+struct EdgeEquation;
+
+void ParameterEquationInit(
+    ParameterEquation* equation,
+    float p0, float p1, float p2,
+    const struct EdgeEquation* e0, const struct EdgeEquation* e1, const struct EdgeEquation* e2, float area);
+
+float ParameterEquationEvaluate(const ParameterEquation* equation, float x, float y);
diff --git a/GL/private.h b/GL/private.h
index cf6f789..2908508 100644
--- a/GL/private.h
+++ b/GL/private.h
@@ -2,16 +2,17 @@
 #define PRIVATE_H
 
 #include <stdint.h>
-#include <dc/matrix.h>
-#include <dc/pvr.h>
-#include <dc/vec3f.h>
-#include <dc/fmath.h>
-#include <dc/matrix3d.h>
+#include <stdio.h>
+
+#include "platform.h"
+#include "types.h"
+
+#include "../include/GL/gl.h"
+#include "../include/GL/glext.h"
+#include "../include/GL/glkos.h"
 
-#include "../include/gl.h"
 #include "../containers/aligned_vector.h"
 #include "../containers/named_array.h"
-#include "sh4_math.h"
 
 extern void* memcpy4 (void *dest, const void *src, size_t count);
 
@@ -20,12 +21,6 @@ extern void* memcpy4 (void *dest, const void *src, size_t count);
 #define GL_FORCE_INLINE static GL_INLINE_DEBUG
 #define _GL_UNUSED(x) (void)(x)
 
-#define FASTCPY(dst, src, bytes) \
-    (bytes % 32 == 0) ? sq_cpy(dst, src, bytes) : memcpy(dst, src, bytes);
-
-#define FASTCPY4(dst, src, bytes) \
-    (bytes % 32 == 0) ? sq_cpy(dst, src, bytes) : memcpy4(dst, src, bytes);
-
 #define _PACK4(v) ((v * 0xF) / 0xFF)
 #define PACK_ARGB4444(a,r,g,b) (_PACK4(a) << 12) | (_PACK4(r) << 8) | (_PACK4(g) << 4) | (_PACK4(b))
 #define PACK_ARGB8888(a,r,g,b) ( ((a & 0xFF) << 24) | ((r & 0xFF) << 16) | ((g & 0xFF) << 8) | (b & 0xFF) )
@@ -46,7 +41,6 @@ extern void* memcpy4 (void *dest, const void *src, size_t count);
 
 #define MAX_TEXTURE_SIZE 1024
 
-typedef float Matrix4x4[16];
 
 /* This gives us an easy way to switch
  * internal matrix order if necessary */
@@ -89,9 +83,6 @@ typedef float Matrix4x4[16];
 #define M15 15
 #endif
 
-typedef struct {
-    pvr_poly_hdr_t hdr;
-} PVRHeader;
 
 typedef struct {
     unsigned int flags;      /* Constant PVR_CMD_USERCLIP */
@@ -192,19 +183,6 @@ typedef struct {
     GLfloat specularMaterial[4];
 } LightSource;
 
-typedef struct {
-    /* Same 32 byte layout as pvr_vertex_t */
-    uint32_t flags;
-    float xyz[3];
-    float uv[2];
-    uint8_t bgra[4];
-
-    /* In the pvr_vertex_t structure, this next 4 bytes is oargb
-     * but we're not using that for now, so having W here makes the code
-     * simpler */
-    float w;
-} Vertex;
-
 
 #define argbcpy(dst, src) \
     *((GLuint*) dst) = *((GLuint*) src) \
@@ -263,7 +241,7 @@ typedef struct {
     AlignedVector* extras;
 } SubmissionTarget;
 
-PVRHeader* _glSubmissionTargetHeader(SubmissionTarget* target);
+PolyHeader* _glSubmissionTargetHeader(SubmissionTarget* target);
 Vertex* _glSubmissionTargetStart(SubmissionTarget* target);
 Vertex* _glSubmissionTargetEnd(SubmissionTarget* target);
 
@@ -310,10 +288,10 @@ Matrix4x4* _glGetModelViewMatrix();
 void _glWipeTextureOnFramebuffers(GLuint texture);
 GLubyte _glCheckImmediateModeInactive(const char* func);
 
-pvr_poly_cxt_t* _glGetPVRContext();
+PolyContext* _glGetPVRContext();
 GLubyte _glInitTextures();
 
-void _glUpdatePVRTextureContext(pvr_poly_cxt_t* context, GLshort textureUnit);
+void _glUpdatePVRTextureContext(PolyContext* context, GLshort textureUnit);
 void _glAllocateSpaceForMipmaps(TextureObject* active);
 
 typedef struct {
@@ -365,7 +343,7 @@ typedef struct {
     float finalColour[4]; //16 bytes (to 40)
 } EyeSpaceData;
 
-extern void _glPerformLighting(Vertex* vertices, EyeSpaceData *es, const int32_t count);
+extern void _glPerformLighting(Vertex* vertices, EyeSpaceData *es, const uint32_t count);
 
 unsigned char _glIsClippingEnabled();
 void _glEnableClipping(unsigned char v);
@@ -378,7 +356,6 @@ GLuint _glFreeTextureMemory();
 GLuint _glUsedTextureMemory();
 GLuint _glFreeContiguousTextureMemory();
 
-#define PVR_VERTEX_BUF_SIZE 2560 * 256
 #define MAX_TEXTURE_UNITS 2
 #define MAX_LIGHTS 8
 
diff --git a/GL/profiler.c b/GL/profiler.c
index c74d84a..6a27981 100644
--- a/GL/profiler.c
+++ b/GL/profiler.c
@@ -1,4 +1,3 @@
-#include <kos.h>
 #include <stdint.h>
 #include <string.h>
 #include <stdio.h>
diff --git a/GL/state.c b/GL/state.c
index 4f08ea5..9487e6a 100644
--- a/GL/state.c
+++ b/GL/state.c
@@ -2,19 +2,11 @@
 #include <string.h>
 #include <stdio.h>
 
-#include <dc/pvr.h>
-#include <dc/vec3f.h>
-#include <dc/video.h>
-
-#include "../include/gl.h"
-#include "../include/glext.h"
-#include "../include/glkos.h"
-
 #include "private.h"
 
-static pvr_poly_cxt_t GL_CONTEXT;
+static PolyContext GL_CONTEXT;
 
-pvr_poly_cxt_t* _glGetPVRContext() {
+PolyContext *_glGetPVRContext() {
     return &GL_CONTEXT;
 }
 
@@ -43,12 +35,12 @@ GLboolean _glIsSharedTexturePaletteEnabled() {
 
 static int _calc_pvr_face_culling() {
     if(!CULLING_ENABLED) {
-        return PVR_CULLING_NONE;
+        return GPU_CULLING_NONE;
     } else {
         if(CULL_FACE == GL_BACK) {
-            return (FRONT_FACE == GL_CW) ? PVR_CULLING_CCW : PVR_CULLING_CW;
+            return (FRONT_FACE == GL_CW) ? GPU_CULLING_CCW : GPU_CULLING_CW;
         } else {
-            return (FRONT_FACE == GL_CCW) ? PVR_CULLING_CCW : PVR_CULLING_CW;
+            return (FRONT_FACE == GL_CCW) ? GPU_CULLING_CCW : GPU_CULLING_CW;
         }
     }
 }
@@ -58,28 +50,28 @@ static GLboolean DEPTH_TEST_ENABLED = GL_FALSE;
 
 static int _calc_pvr_depth_test() {
     if(!DEPTH_TEST_ENABLED) {
-        return PVR_DEPTHCMP_ALWAYS;
+        return GPU_DEPTHCMP_ALWAYS;
     }
 
     switch(DEPTH_FUNC) {
         case GL_NEVER:
-            return PVR_DEPTHCMP_NEVER;
+            return GPU_DEPTHCMP_NEVER;
         case GL_LESS:
-            return PVR_DEPTHCMP_GREATER;
+            return GPU_DEPTHCMP_GREATER;
         case GL_EQUAL:
-            return PVR_DEPTHCMP_EQUAL;
+            return GPU_DEPTHCMP_EQUAL;
         case GL_LEQUAL:
-            return PVR_DEPTHCMP_GEQUAL;
+            return GPU_DEPTHCMP_GEQUAL;
         case GL_GREATER:
-            return PVR_DEPTHCMP_LESS;
+            return GPU_DEPTHCMP_LESS;
         case GL_NOTEQUAL:
-            return PVR_DEPTHCMP_NOTEQUAL;
+            return GPU_DEPTHCMP_NOTEQUAL;
         case GL_GEQUAL:
-            return PVR_DEPTHCMP_LEQUAL;
+            return GPU_DEPTHCMP_LEQUAL;
         break;
         case GL_ALWAYS:
         default:
-            return PVR_DEPTHCMP_ALWAYS;
+            return GPU_DEPTHCMP_ALWAYS;
     }
 }
 
@@ -105,32 +97,32 @@ GLboolean _glIsAlphaTestEnabled() {
 static int _calcPVRBlendFactor(GLenum factor) {
     switch(factor) {
     case GL_ZERO:
-        return PVR_BLEND_ZERO;
+        return GPU_BLEND_ZERO;
     case GL_SRC_ALPHA:
-        return PVR_BLEND_SRCALPHA;
+        return GPU_BLEND_SRCALPHA;
     case GL_DST_COLOR:
-        return PVR_BLEND_DESTCOLOR;
+        return GPU_BLEND_DESTCOLOR;
     case GL_DST_ALPHA:
-        return PVR_BLEND_DESTALPHA;
+        return GPU_BLEND_DESTALPHA;
     case GL_ONE_MINUS_DST_COLOR:
-        return PVR_BLEND_INVDESTCOLOR;
+        return GPU_BLEND_INVDESTCOLOR;
     case GL_ONE_MINUS_SRC_ALPHA:
-        return PVR_BLEND_INVSRCALPHA;
+        return GPU_BLEND_INVSRCALPHA;
     case GL_ONE_MINUS_DST_ALPHA:
-        return PVR_BLEND_INVDESTALPHA;
+        return GPU_BLEND_INVDESTALPHA;
     case GL_ONE:
-        return PVR_BLEND_ONE;
+        return GPU_BLEND_ONE;
     default:
         fprintf(stderr, "Invalid blend mode: %u\n", (unsigned int) factor);
-        return PVR_BLEND_ONE;
+        return GPU_BLEND_ONE;
     }
 }
 
-static void _updatePVRBlend(pvr_poly_cxt_t* context) {
+static void _updatePVRBlend(PolyContext* context) {
     if(BLEND_ENABLED || ALPHA_TEST_ENABLED) {
-        context->gen.alpha = PVR_ALPHA_ENABLE;
+        context->gen.alpha = GPU_ALPHA_ENABLE;
     } else {
-        context->gen.alpha = PVR_ALPHA_DISABLE;
+        context->gen.alpha = GPU_ALPHA_DISABLE;
     }
 
     context->blend.src = _calcPVRBlendFactor(BLEND_SFACTOR);
@@ -158,21 +150,21 @@ GLboolean _glCheckValidEnum(GLint param, GLint* values, const char* func) {
 
 GLboolean TEXTURES_ENABLED [] = {GL_FALSE, GL_FALSE};
 
-void _glUpdatePVRTextureContext(pvr_poly_cxt_t* context, GLshort textureUnit) {
+void _glUpdatePVRTextureContext(PolyContext *context, GLshort textureUnit) {
     const TextureObject *tx1 = (textureUnit == 0) ? _glGetTexture0() : _glGetTexture1();
 
     /* Disable all texturing to start with */
-    context->txr.enable = PVR_TEXTURE_DISABLE;
-    context->txr2.enable = PVR_TEXTURE_DISABLE;
-    context->txr2.alpha = PVR_TXRALPHA_DISABLE;
+    context->txr.enable = GPU_TEXTURE_DISABLE;
+    context->txr2.enable = GPU_TEXTURE_DISABLE;
+    context->txr2.alpha = GPU_TXRALPHA_DISABLE;
 
     if(!TEXTURES_ENABLED[textureUnit] || !tx1) {
         return;
     }
 
-    context->txr.alpha = (BLEND_ENABLED || ALPHA_TEST_ENABLED) ? PVR_TXRALPHA_ENABLE : PVR_TXRALPHA_DISABLE;
+    context->txr.alpha = (BLEND_ENABLED || ALPHA_TEST_ENABLED) ? GPU_TXRALPHA_ENABLE : GPU_TXRALPHA_DISABLE;
 
-    GLuint filter = PVR_FILTER_NEAREST;
+    GLuint filter = GPU_FILTER_NEAREST;
     GLboolean enableMipmaps = GL_FALSE;
 
     switch(tx1->minFilter) {
@@ -199,17 +191,17 @@ void _glUpdatePVRTextureContext(pvr_poly_cxt_t* context, GLshort textureUnit) {
 
     if(enableMipmaps) {
         if(tx1->minFilter == GL_LINEAR_MIPMAP_NEAREST) {
-            filter = PVR_FILTER_TRILINEAR1;
+            filter = GPU_FILTER_TRILINEAR1;
         } else if(tx1->minFilter == GL_LINEAR_MIPMAP_LINEAR) {
-            filter = PVR_FILTER_TRILINEAR2;
+            filter = GPU_FILTER_TRILINEAR2;
         } else if(tx1->minFilter == GL_NEAREST_MIPMAP_LINEAR) {
-            filter = PVR_FILTER_BILINEAR;
+            filter = GPU_FILTER_BILINEAR;
         } else {
-            filter = PVR_FILTER_NEAREST;
+            filter = GPU_FILTER_NEAREST;
         }
     } else {
         if(tx1->minFilter == GL_LINEAR && tx1->magFilter == GL_LINEAR) {
-            filter = PVR_FILTER_BILINEAR;
+            filter = GPU_FILTER_BILINEAR;
         }
     }
 
@@ -221,7 +213,7 @@ void _glUpdatePVRTextureContext(pvr_poly_cxt_t* context, GLshort textureUnit) {
     }
 
     if(tx1->data) {
-        context->txr.enable = PVR_TEXTURE_ENABLE;
+        context->txr.enable = GPU_TEXTURE_ENABLE;
         context->txr.filter = filter;
         context->txr.width = tx1->width;
         context->txr.height = tx1->height;
@@ -239,14 +231,14 @@ void _glUpdatePVRTextureContext(pvr_poly_cxt_t* context, GLshort textureUnit) {
         if(tx1->isPaletted) {
             if(_glIsSharedTexturePaletteEnabled()) {
                 TexturePalette* palette = _glGetSharedPalette(tx1->shared_bank);
-                context->txr.format |= PVR_TXRFMT_8BPP_PAL(palette->bank);
+                context->txr.format |= GPUPaletteSelect8BPP(palette->bank);
             } else {
-                context->txr.format |= PVR_TXRFMT_8BPP_PAL((tx1->palette) ? tx1->palette->bank : 0);
+                context->txr.format |= GPUPaletteSelect8BPP((tx1->palette) ? tx1->palette->bank : 0);
             }
         }
 
         context->txr.env = tx1->env;
-        context->txr.uv_flip = PVR_UVFLIP_NONE;
+        context->txr.uv_flip = GPU_UVFLIP_NONE;
         context->txr.uv_clamp = tx1->uv_clamp;
     }
 }
@@ -262,12 +254,12 @@ GLboolean _glIsColorMaterialEnabled() {
 static GLfloat CLEAR_COLOUR[3];
 
 void _glInitContext() {
-    memset(&GL_CONTEXT, 0, sizeof(pvr_poly_cxt_t));
+    memset(&GL_CONTEXT, 0, sizeof(PolyContext));
 
-    GL_CONTEXT.list_type = PVR_LIST_OP_POLY;
-    GL_CONTEXT.fmt.color = PVR_CLRFMT_ARGBPACKED;
-    GL_CONTEXT.fmt.uv = PVR_UVFMT_32BIT;
-    GL_CONTEXT.gen.color_clamp = PVR_CLRCLAMP_DISABLE;
+    GL_CONTEXT.list_type = GPU_LIST_OP_POLY;
+    GL_CONTEXT.fmt.color = GPU_CLRFMT_ARGBPACKED;
+    GL_CONTEXT.fmt.uv = GPU_UVFMT_32BIT;
+    GL_CONTEXT.gen.color_clamp = GPU_CLRCLAMP_DISABLE;
 
     glClearDepth(1.0f);
     glDepthFunc(GL_LESS);
@@ -309,13 +301,13 @@ GLAPI void APIENTRY glEnable(GLenum cap) {
             _updatePVRBlend(&GL_CONTEXT);
         } break;
         case GL_SCISSOR_TEST: {
-            GL_CONTEXT.gen.clip_mode = PVR_USERCLIP_INSIDE;
+            GL_CONTEXT.gen.clip_mode = GPU_USERCLIP_INSIDE;
         } break;
         case GL_LIGHTING: {
             LIGHTING_ENABLED = GL_TRUE;
         } break;
         case GL_FOG:
-            GL_CONTEXT.gen.fog_type = PVR_FOG_TABLE;
+            GL_CONTEXT.gen.fog_type = GPU_FOG_TABLE;
         break;
         case GL_COLOR_MATERIAL:
             COLOR_MATERIAL_ENABLED = GL_TRUE;
@@ -372,13 +364,13 @@ GLAPI void APIENTRY glDisable(GLenum cap) {
             _updatePVRBlend(&GL_CONTEXT);
         break;
         case GL_SCISSOR_TEST: {
-            GL_CONTEXT.gen.clip_mode = PVR_USERCLIP_DISABLE;
+            GL_CONTEXT.gen.clip_mode = GPU_USERCLIP_DISABLE;
         } break;
         case GL_LIGHTING: {
             LIGHTING_ENABLED = GL_FALSE;
         } break;
         case GL_FOG:
-            GL_CONTEXT.gen.fog_type = PVR_FOG_DISABLE;
+            GL_CONTEXT.gen.fog_type = GPU_FOG_DISABLE;
         break;
         case GL_COLOR_MATERIAL:
             COLOR_MATERIAL_ENABLED = GL_FALSE;
@@ -419,7 +411,7 @@ GLAPI void APIENTRY glDisable(GLenum cap) {
 /* Clear Caps */
 GLAPI void APIENTRY glClear(GLuint mode) {
     if(mode & GL_COLOR_BUFFER_BIT) {
-        pvr_set_bg_color(CLEAR_COLOUR[0], CLEAR_COLOUR[1], CLEAR_COLOUR[2]);
+        GPUSetBackgroundColour(CLEAR_COLOUR[0], CLEAR_COLOUR[1], CLEAR_COLOUR[2]);
     }
 }
 
@@ -429,9 +421,12 @@ GLAPI void APIENTRY glClearColor(GLfloat r, GLfloat g, GLfloat b, GLfloat a) {
     if(b > 1) b = 1;
     if(a > 1) a = 1;
 
-    CLEAR_COLOUR[0] = r * a;
-    CLEAR_COLOUR[1] = g * a;
-    CLEAR_COLOUR[2] = b * a;
+    /* FIXME: The background-poly doesn't take an alpha value */
+    _GL_UNUSED(a);
+
+    CLEAR_COLOUR[0] = r;
+    CLEAR_COLOUR[1] = g;
+    CLEAR_COLOUR[2] = b;
 }
 
 /* Depth Testing */
@@ -441,7 +436,7 @@ GLAPI void APIENTRY glClearDepthf(GLfloat depth) {
 
 GLAPI void APIENTRY glClearDepth(GLfloat depth) {
     /* We reverse because using invW means that farther Z == lower number */
-    pvr_set_zclip(1.0f - depth);
+    GPUSetClearDepth(1.0f - depth);
 }
 
 GLAPI void APIENTRY glDrawBuffer(GLenum mode) {
@@ -455,7 +450,7 @@ GLAPI void APIENTRY glReadBuffer(GLenum mode) {
 }
 
 GLAPI void APIENTRY glDepthMask(GLboolean flag) {
-    GL_CONTEXT.depth.write = (flag == GL_TRUE) ? PVR_DEPTHWRITE_ENABLE : PVR_DEPTHWRITE_DISABLE;
+    GL_CONTEXT.depth.write = (flag == GL_TRUE) ? GPU_DEPTHWRITE_ENABLE : GPU_DEPTHWRITE_DISABLE;
 }
 
 GLAPI void APIENTRY glDepthFunc(GLenum func) {
@@ -484,12 +479,12 @@ GLAPI void APIENTRY glCullFace(GLenum mode) {
 }
 
 GLenum _glGetShadeModel() {
-    return (GL_CONTEXT.gen.shading == PVR_SHADE_FLAT) ? GL_FLAT : GL_SMOOTH;
+    return (GL_CONTEXT.gen.shading == GPU_SHADE_FLAT) ? GL_FLAT : GL_SMOOTH;
 }
 
 /* Shading - Flat or Goraud */
 GLAPI void APIENTRY glShadeModel(GLenum mode) {
-    GL_CONTEXT.gen.shading = (mode == GL_SMOOTH) ? PVR_SHADE_GOURAUD : PVR_SHADE_FLAT;
+    GL_CONTEXT.gen.shading = (mode == GL_SMOOTH) ? GPU_SHADE_GOURAUD : GPU_SHADE_FLAT;
 }
 
 /* Blending */
@@ -499,7 +494,6 @@ GLAPI void APIENTRY glBlendFunc(GLenum sfactor, GLenum dfactor) {
     _updatePVRBlend(&GL_CONTEXT);
 }
 
-#define PT_ALPHA_REF 0x011c
 
 GLAPI void APIENTRY glAlphaFunc(GLenum func, GLclampf ref) {
     GLint validFuncs[] = {
@@ -512,7 +506,7 @@ GLAPI void APIENTRY glAlphaFunc(GLenum func, GLclampf ref) {
     }
 
     GLubyte val = (GLubyte)(ref * 255.0f);
-    PVR_SET(PT_ALPHA_REF, val);
+    GPUSetAlphaCutOff(val);
 }
 
 void glLineWidth(GLfloat width) {
@@ -561,6 +555,9 @@ void APIENTRY glScissor(GLint x, GLint y, GLsizei width, GLsizei height) {
     PVRTileClipCommand *c = aligned_vector_extend(&_glActivePolyList()->vector, 1);
 
     GLint miny, maxx, maxy;
+
+    const VideoMode* vid_mode = GetVideoMode();
+
     GLsizei gl_scissor_width = MAX( MIN(width, vid_mode->width), 0 );
     GLsizei gl_scissor_height = MAX( MIN(height, vid_mode->height), 0 );
 
@@ -570,7 +567,7 @@ void APIENTRY glScissor(GLint x, GLint y, GLsizei width, GLsizei height) {
     maxy = (gl_scissor_height + miny);
 
     /* load command structure while mapping screen coords to TA tiles */
-    c->flags = PVR_CMD_USERCLIP;
+    c->flags = GPU_CMD_USERCLIP;
     c->d1 = c->d2 = c->d3 = 0;
     c->sx = CLAMP(x / 32, 0, vid_mode->width / 32);
     c->sy = CLAMP(miny / 32, 0, vid_mode->height / 32);
@@ -583,7 +580,7 @@ GLboolean APIENTRY glIsEnabled(GLenum cap) {
     case GL_DEPTH_TEST:
         return DEPTH_TEST_ENABLED;
     case GL_SCISSOR_TEST:
-        return GL_CONTEXT.gen.clip_mode == PVR_USERCLIP_INSIDE;
+        return GL_CONTEXT.gen.clip_mode == GPU_USERCLIP_INSIDE;
     case GL_CULL_FACE:
         return CULLING_ENABLED;
     case GL_LIGHTING:
@@ -643,10 +640,10 @@ void APIENTRY glGetBooleanv(GLenum pname, GLboolean* params) {
 void APIENTRY glGetFloatv(GLenum pname, GLfloat* params) {
     switch(pname) {
         case GL_PROJECTION_MATRIX:
-            memcpy4(params, _glGetProjectionMatrix(), sizeof(float) * 16);
+            FASTCPY4(params, _glGetProjectionMatrix(), sizeof(float) * 16);
         break;
         case GL_MODELVIEW_MATRIX:
-            memcpy4(params, _glGetModelViewMatrix(), sizeof(float) * 16);
+            FASTCPY4(params, _glGetModelViewMatrix(), sizeof(float) * 16);
         break;
         case GL_POLYGON_OFFSET_FACTOR:
             *params = OFFSET_FACTOR;
diff --git a/GL/texture.c b/GL/texture.c
index e052e83..fca7c07 100644
--- a/GL/texture.c
+++ b/GL/texture.c
@@ -7,13 +7,10 @@
 #include <string.h>
 
 #include "config.h"
-#include "../include/glext.h"
-#include "../include/glkos.h"
+#include "platform.h"
 
 #include "yalloc/yalloc.h"
 
-#include <kos/string.h>
-
 /* We always leave this amount of vram unallocated to prevent
  * issues with the allocator */
 #define PVR_MEM_BUFFER_SIZE (64 * 1024)
@@ -55,7 +52,7 @@ static TexturePalette* _initTexturePalette() {
     TexturePalette* palette = (TexturePalette*) malloc(sizeof(TexturePalette));
     assert(palette);
 
-    memset4(palette, 0x0, sizeof(TexturePalette));
+    MEMSET4(palette, 0x0, sizeof(TexturePalette));
     palette->bank = -1;
     return palette;
 }
@@ -115,6 +112,54 @@ static void _glReleasePaletteSlot(GLshort slot, GLushort size) {
     }
 }
 
+/* Linear/iterative twiddling algorithm from Marcus' tatest */
+#define TWIDTAB(x) ( (x&1)|((x&2)<<1)|((x&4)<<2)|((x&8)<<3)|((x&16)<<4)| \
+                     ((x&32)<<5)|((x&64)<<6)|((x&128)<<7)|((x&256)<<8)|((x&512)<<9) )
+#define TWIDOUT(x, y) ( TWIDTAB((y)) | (TWIDTAB((x)) << 1) )
+
+
+static void GPUTextureTwiddle8PPP(void* src, void* dst, uint32_t w, uint32_t h) {
+    uint32_t x, y, yout, min, mask;
+
+    min = MIN(w, h);
+    mask = min - 1;
+
+    uint8_t* pixels;
+    uint16_t* vtex;
+    pixels = (uint8_t*) src;
+    vtex = (uint16_t*) dst;
+
+    for(y = 0; y < h; y += 2) {
+        yout = y;
+        for(x = 0; x < w; x++) {
+            vtex[TWIDOUT((yout & mask) / 2, x & mask) +
+                 (x / min + yout / min)*min * min / 2] =
+                     pixels[y * w + x] | (pixels[(y + 1) * w + x] << 8);
+        }
+    }
+}
+
+static void GPUTextureTwiddle16BPP(void * src, void* dst, uint32_t w, uint32_t h) {
+    uint32_t x, y, yout, min, mask;
+
+    min = MIN(w, h);
+    mask = min - 1;
+
+    uint16_t* pixels;
+    uint16_t* vtex;
+    pixels = (uint16_t*) src;
+    vtex = (uint16_t*) dst;
+
+    for(y = 0; y < h; y++) {
+        yout = y;
+
+        for(x = 0; x < w; x++) {
+            vtex[TWIDOUT(x & mask, yout & mask) +
+                 (x / min + yout / min)*min * min] = pixels[y * w + x];
+        }
+    }
+}
+
 TexturePalette* _glGetSharedPalette(GLshort bank) {
     assert(bank >= 0 && bank < 4);
     return SHARED_PALETTES[bank];
@@ -124,10 +169,10 @@ void _glSetInternalPaletteFormat(GLenum val) {
     INTERNAL_PALETTE_FORMAT = val;
 
     if(INTERNAL_PALETTE_FORMAT == GL_RGBA4) {
-        pvr_set_pal_format(PVR_PAL_ARGB4444);
+        GPUSetPaletteFormat(GPU_PAL_ARGB4444);
     } else {
         assert(INTERNAL_PALETTE_FORMAT == GL_RGBA8);
-        pvr_set_pal_format(PVR_PAL_ARGB8888);
+        GPUSetPaletteFormat(GPU_PAL_ARGB8888);
     }
 }
 
@@ -146,9 +191,9 @@ void _glApplyColorTable(TexturePalette* src) {
     for(i = 0; i < src->width; ++i) {
         GLubyte* entry = &src->data[i * 4];
         if(INTERNAL_PALETTE_FORMAT == GL_RGBA8) {
-            pvr_set_pal_entry(offset + i, PACK_ARGB8888(entry[3], entry[0], entry[1], entry[2]));
+            GPUSetPaletteEntry(offset + i, PACK_ARGB8888(entry[3], entry[0], entry[1], entry[2]));
         } else {
-            pvr_set_pal_entry(offset + i, PACK_ARGB4444(entry[3], entry[0], entry[1], entry[2]));
+            GPUSetPaletteEntry(offset + i, PACK_ARGB4444(entry[3], entry[0], entry[1], entry[2]));
         }
     }
 }
@@ -333,9 +378,9 @@ GLubyte _glInitTextures() {
     memset((void*) BANKS_USED, 0x0, sizeof(BANKS_USED));
     memset((void*) SUBBANKS_USED, 0x0, sizeof(SUBBANKS_USED));
 
-    size_t vram_free = pvr_mem_available();
+    size_t vram_free = GPUMemoryAvailable();
     YALLOC_SIZE = vram_free - PVR_MEM_BUFFER_SIZE; /* Take all but 64kb VRAM */
-    YALLOC_BASE = pvr_mem_malloc(YALLOC_SIZE);
+    YALLOC_BASE = GPUMemoryAlloc(YALLOC_SIZE);
     yalloc_init(YALLOC_BASE, YALLOC_SIZE);
     return 1;
 }
@@ -375,7 +420,7 @@ static void _glInitializeTextureObject(TextureObject* txr, unsigned int id) {
     txr->width = txr->height = 0;
     txr->mipmap = 0;
     txr->uv_clamp = 0;
-    txr->env = PVR_TXRENV_MODULATEALPHA;
+    txr->env = GPU_TXRENV_MODULATEALPHA;
     txr->data = NULL;
     txr->mipmapCount = 0;
     txr->minFilter = GL_NEAREST;
@@ -497,13 +542,13 @@ void APIENTRY glTexEnvi(GLenum target, GLenum pname, GLint param) {
 
             switch(param) {
                 case GL_MODULATE:
-                    active->env = PVR_TXRENV_MODULATEALPHA;
+                    active->env = GPU_TXRENV_MODULATEALPHA;
                 break;
                 case GL_DECAL:
-                    active->env = PVR_TXRENV_DECAL;
+                    active->env = GPU_TXRENV_DECAL;
                 break;
                 case GL_REPLACE:
-                    active->env = PVR_TXRENV_REPLACE;
+                    active->env = GPU_TXRENV_REPLACE;
                 break;
                 default:
                     break;
@@ -631,7 +676,7 @@ void APIENTRY glCompressedTexImage2DARB(GLenum target,
     active->data = yalloc_alloc_and_defrag(imageSize);
 
     if(data) {
-        sq_cpy(active->data, data, imageSize);
+        FASTCPY(active->data, data, imageSize);
     }
 }
 
@@ -722,47 +767,47 @@ static GLint _cleanInternalFormat(GLint internalFormat) {
          * the type was already 1555 (1-bit alpha) in which case we return that
          */
             if(type == GL_UNSIGNED_SHORT_1_5_5_5_REV) {
-                return PVR_TXRFMT_ARGB1555 | PVR_TXRFMT_NONTWIDDLED;
+                return GPU_TXRFMT_ARGB1555 | GPU_TXRFMT_NONTWIDDLED;
             } else if(type == GL_UNSIGNED_SHORT_1_5_5_5_REV_TWID_KOS) {
-                return PVR_TXRFMT_ARGB1555 | PVR_TXRFMT_TWIDDLED;
+                return GPU_TXRFMT_ARGB1555 | GPU_TXRFMT_TWIDDLED;
             } else if(type == GL_UNSIGNED_SHORT_4_4_4_4_REV_TWID_KOS) {
-                return PVR_TXRFMT_ARGB4444 | PVR_TXRFMT_TWIDDLED;
+                return GPU_TXRFMT_ARGB4444 | GPU_TXRFMT_TWIDDLED;
             } else {
-                return PVR_TXRFMT_ARGB4444 | PVR_TXRFMT_NONTWIDDLED;
+                return GPU_TXRFMT_ARGB4444 | GPU_TXRFMT_NONTWIDDLED;
             }
         case GL_RED:
         case GL_RGB:
             /* No alpha? Return RGB565 which is the best we can do without using palettes */
-            return PVR_TXRFMT_RGB565 | PVR_TXRFMT_NONTWIDDLED;
+            return GPU_TXRFMT_RGB565 | GPU_TXRFMT_NONTWIDDLED;
         /* Compressed and twiddled versions */
         case GL_UNSIGNED_SHORT_5_6_5_TWID_KOS:
-            return PVR_TXRFMT_RGB565 | PVR_TXRFMT_TWIDDLED;
+            return GPU_TXRFMT_RGB565 | GPU_TXRFMT_TWIDDLED;
         case GL_UNSIGNED_SHORT_4_4_4_4_REV_TWID_KOS:
-            return PVR_TXRFMT_ARGB4444 | PVR_TXRFMT_TWIDDLED;
+            return GPU_TXRFMT_ARGB4444 | GPU_TXRFMT_TWIDDLED;
         case GL_UNSIGNED_SHORT_1_5_5_5_REV_TWID_KOS:
-            return PVR_TXRFMT_ARGB1555 | PVR_TXRFMT_TWIDDLED;
+            return GPU_TXRFMT_ARGB1555 | GPU_TXRFMT_TWIDDLED;
         case GL_COMPRESSED_RGB_565_VQ_KOS:
         case GL_COMPRESSED_RGB_565_VQ_MIPMAP_KOS:
-            return PVR_TXRFMT_RGB565 | PVR_TXRFMT_NONTWIDDLED | PVR_TXRFMT_VQ_ENABLE;
+            return GPU_TXRFMT_RGB565 | GPU_TXRFMT_NONTWIDDLED | GPU_TXRFMT_VQ_ENABLE;
         case GL_COMPRESSED_RGB_565_VQ_TWID_KOS:
         case GL_COMPRESSED_RGB_565_VQ_MIPMAP_TWID_KOS:
-            return PVR_TXRFMT_RGB565 | PVR_TXRFMT_TWIDDLED | PVR_TXRFMT_VQ_ENABLE;
+            return GPU_TXRFMT_RGB565 | GPU_TXRFMT_TWIDDLED | GPU_TXRFMT_VQ_ENABLE;
         case GL_COMPRESSED_ARGB_4444_VQ_TWID_KOS:
         case GL_COMPRESSED_ARGB_4444_VQ_MIPMAP_TWID_KOS:
-            return PVR_TXRFMT_ARGB4444 | PVR_TXRFMT_TWIDDLED | PVR_TXRFMT_VQ_ENABLE;
+            return GPU_TXRFMT_ARGB4444 | GPU_TXRFMT_TWIDDLED | GPU_TXRFMT_VQ_ENABLE;
         case GL_COMPRESSED_ARGB_4444_VQ_KOS:
         case GL_COMPRESSED_ARGB_4444_VQ_MIPMAP_KOS:
-            return PVR_TXRFMT_ARGB4444 | PVR_TXRFMT_NONTWIDDLED | PVR_TXRFMT_VQ_ENABLE;
+            return GPU_TXRFMT_ARGB4444 | GPU_TXRFMT_NONTWIDDLED | GPU_TXRFMT_VQ_ENABLE;
         case GL_COMPRESSED_ARGB_1555_VQ_KOS:
         case GL_COMPRESSED_ARGB_1555_VQ_MIPMAP_KOS:
-            return PVR_TXRFMT_ARGB1555 | PVR_TXRFMT_NONTWIDDLED | PVR_TXRFMT_VQ_ENABLE;
+            return GPU_TXRFMT_ARGB1555 | GPU_TXRFMT_NONTWIDDLED | GPU_TXRFMT_VQ_ENABLE;
         case GL_COMPRESSED_ARGB_1555_VQ_TWID_KOS:
         case GL_COMPRESSED_ARGB_1555_VQ_MIPMAP_TWID_KOS:
-            return PVR_TXRFMT_ARGB1555 | PVR_TXRFMT_TWIDDLED | PVR_TXRFMT_VQ_ENABLE;
+            return GPU_TXRFMT_ARGB1555 | GPU_TXRFMT_TWIDDLED | GPU_TXRFMT_VQ_ENABLE;
         case GL_COLOR_INDEX8_EXT:
-            return PVR_TXRFMT_PAL8BPP | PVR_TXRFMT_TWIDDLED;
+            return GPU_TXRFMT_PAL8BPP | GPU_TXRFMT_TWIDDLED;
         case GL_COLOR_INDEX4_EXT:
-            return PVR_TXRFMT_PAL4BPP | PVR_TXRFMT_TWIDDLED;
+            return GPU_TXRFMT_PAL4BPP | GPU_TXRFMT_TWIDDLED;
         default:
             return 0;
     }
@@ -1228,9 +1273,9 @@ void APIENTRY glTexImage2D(GLenum target, GLint level, GLint internalFormat,
         const GLubyte *pixels = (GLubyte*) (conversionBuffer) ? conversionBuffer : data;
 
         if(internalFormat == GL_COLOR_INDEX8_EXT) {
-            pvr_txr_load_ex((void*) pixels, targetData, width, height, PVR_TXRLOAD_8BPP);
+            GPUTextureTwiddle8PPP((void*) pixels, targetData, width, height);
         } else {
-            pvr_txr_load_ex((void*) pixels, targetData, width, height, PVR_TXRLOAD_16BPP);
+            GPUTextureTwiddle16BPP((void*) pixels, targetData, width, height);
         }
 
         /* We make sure we remove nontwiddled and add twiddled. We could always
diff --git a/GL/types.h b/GL/types.h
new file mode 100644
index 0000000..99758e2
--- /dev/null
+++ b/GL/types.h
@@ -0,0 +1,16 @@
+#pragma once
+
+#include <stdint.h>
+
+typedef struct {
+    /* Same 32 byte layout as pvr_vertex_t */
+    uint32_t flags;
+    float xyz[3];
+    float uv[2];
+    uint8_t bgra[4];
+
+    /* In the pvr_vertex_t structure, this next 4 bytes is oargb
+     * but we're not using that for now, so having W here makes the code
+     * simpler */
+    float w;
+} Vertex;
diff --git a/GL/util.c b/GL/util.c
index eaf2e17..835dd98 100644
--- a/GL/util.c
+++ b/GL/util.c
@@ -1,4 +1,4 @@
-#include "../include/glkos.h"
+#include "private.h"
 
 void APIENTRY glVertexPackColor3fKOS(GLVertexKOS* vertex, float r, float g, float b) {
     vertex->bgra[3] = 255;
diff --git a/GL/version.c.in b/GL/version.c.in
new file mode 100644
index 0000000..7e37b06
--- /dev/null
+++ b/GL/version.c.in
@@ -0,0 +1,2 @@
+
+const char* GLDC_VERSION = "@GLDC_VERSION@";
diff --git a/Makefile b/Makefile
deleted file mode 100644
index 0fa72f6..0000000
--- a/Makefile
+++ /dev/null
@@ -1,37 +0,0 @@
-# KallistiOS ##version##
-#
-# kos-ports/libgl Makefile
-# Copyright (C) 2013, 2014 Josh Pearson
-# Copyright (C) 2014 Lawrence Sebald
-# Copyright (C) 2020 Luke Benstead
-
-TARGET = libGLdc.a
-OBJS = GL/draw.o GL/flush.o GL/framebuffer.o GL/immediate.o GL/lighting.o GL/state.o GL/texture.o GL/glu.o GL/version.h
-OBJS += GL/matrix.o GL/fog.o GL/error.o GL/clip.o containers/stack.o containers/named_array.o containers/aligned_vector.o GL/profiler.o
-OBJS += GL/yalloc/yalloc.o
-
-SUBDIRS =
-
-KOS_CFLAGS += -ffast-math -Ofast -Iinclude
-
-GL/version.h:
-	rm -f $@
-	@printf '#pragma once\n#define GLDC_VERSION "$(shell git describe --abbrev=4 --dirty --always --tags)"\n' > $@
-
-link:
-	$(KOS_AR) rcs $(TARGET) $(OBJS)
-
-build: GL/version.h $(OBJS) link
-
-
-samples: build
-	$(KOS_MAKE) -C samples all
-
-defaultall: create_kos_link $(OBJS) subdirs linklib samples
-
-include $(KOS_BASE)/addons/Makefile.prefab
-
-# creates the kos link to the headers
-create_kos_link:
-	rm -f ../include/GL
-	ln -s ../GLdc/include ../include/GL
diff --git a/containers/aligned_vector.h b/containers/aligned_vector.h
index dbbcda4..b64d089 100644
--- a/containers/aligned_vector.h
+++ b/containers/aligned_vector.h
@@ -1,14 +1,12 @@
 #pragma once
 
-#ifndef ALIGNED_VECTOR_H
-#define ALIGNED_VECTOR_H
+#include <assert.h>
+#include <stdlib.h>
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-#include <assert.h>
-
 typedef struct {
     unsigned int size;
     unsigned int capacity;
@@ -39,5 +37,3 @@ static inline void* aligned_vector_back(AlignedVector* vector){
 #ifdef __cplusplus
 }
 #endif
-
-#endif // ALIGNED_VECTOR_H
diff --git a/include/gl.h b/include/GL/gl.h
similarity index 100%
rename from include/gl.h
rename to include/GL/gl.h
diff --git a/include/glext.h b/include/GL/glext.h
similarity index 100%
rename from include/glext.h
rename to include/GL/glext.h
diff --git a/include/glkos.h b/include/GL/glkos.h
similarity index 99%
rename from include/glkos.h
rename to include/GL/glkos.h
index dd290bb..184ecaf 100644
--- a/include/glkos.h
+++ b/include/GL/glkos.h
@@ -4,6 +4,9 @@
 
 __BEGIN_DECLS
 
+extern const char* GLDC_VERSION;
+
+
 /*
  * Dreamcast specific compressed + twiddled formats.
  * We use constants from the range 0xEEE0 onwards
diff --git a/include/glu.h b/include/GL/glu.h
similarity index 98%
rename from include/glu.h
rename to include/GL/glu.h
index ee9717d..0c8420c 100644
--- a/include/glu.h
+++ b/include/GL/glu.h
@@ -17,7 +17,7 @@
 __BEGIN_DECLS
 
 #ifndef BUILD_LIBGL
-#include <gl.h>
+#include "gl.h"
 #endif
 
 #define GLU_FALSE 0
diff --git a/samples/blend_test/main.c b/samples/blend_test/main.c
index 2635e9b..191bbe9 100644
--- a/samples/blend_test/main.c
+++ b/samples/blend_test/main.c
@@ -4,9 +4,9 @@
  * output and incorrect depth testing
  */
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/depth_funcs/main.c b/samples/depth_funcs/main.c
index aab0184..fe90f35 100644
--- a/samples/depth_funcs/main.c
+++ b/samples/depth_funcs/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/depth_funcs_alpha_testing/gl_png.c b/samples/depth_funcs_alpha_testing/gl_png.c
index 05dedc4..78298b3 100644
--- a/samples/depth_funcs_alpha_testing/gl_png.c
+++ b/samples/depth_funcs_alpha_testing/gl_png.c
@@ -1,4 +1,3 @@
-#include <kos.h>
 #include <string.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -188,7 +187,7 @@ int dtex_to_gl_texture(texture *tex, char* filename) {
 		printf("Texture %s loaded\n", tex->path);
 
 		return(1);
-};
+}
 
 void draw_textured_quad(texture *tex) {
 	if(glIsTexture(tex->id)) {
diff --git a/samples/depth_funcs_alpha_testing/gl_png.h b/samples/depth_funcs_alpha_testing/gl_png.h
index 21070b7..29b312f 100644
--- a/samples/depth_funcs_alpha_testing/gl_png.h
+++ b/samples/depth_funcs_alpha_testing/gl_png.h
@@ -2,7 +2,7 @@
 #define __GL_PNG_H__
 
 #include <stdio.h>
-#include <kos.h>
+#include <stdint.h>
 #include <GL/gl.h>
 
 typedef struct _texture {
diff --git a/samples/depth_funcs_alpha_testing/main.c b/samples/depth_funcs_alpha_testing/main.c
index 5bdbaad..c0e7e81 100644
--- a/samples/depth_funcs_alpha_testing/main.c
+++ b/samples/depth_funcs_alpha_testing/main.c
@@ -1,11 +1,11 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 #include  "gl_png.h"
 
 //$KOS_BASE/utils/texconv/texconv  --in disk.png --format ARGB4444 --preview disk_preview.png --out disk.dtex
 
-extern uint8 romdisk[];
+extern uint8_t romdisk[];
 KOS_INIT_ROMDISK(romdisk);
 texture t;
 int 	blendActive = -1;
diff --git a/samples/depth_funcs_ortho/main.c b/samples/depth_funcs_ortho/main.c
index 01a3639..837ede0 100644
--- a/samples/depth_funcs_ortho/main.c
+++ b/samples/depth_funcs_ortho/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/lerabot01/main.c b/samples/lerabot01/main.c
index 4ce79b6..dfcb696 100644
--- a/samples/lerabot01/main.c
+++ b/samples/lerabot01/main.c
@@ -1,11 +1,16 @@
 #include <stdio.h>
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
 #include <time.h>
+#include <stdlib.h>
+#include <stdio.h>
 
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
+
+#ifdef __DREAMCAST__
 extern uint8 romdisk[];
 KOS_INIT_ROMDISK(romdisk);
+#endif
 
 /* floats for x rotation, y rotation, z rotation */
 float xrot, yrot, zrot;
diff --git a/samples/lerabot_blend_test/main.c b/samples/lerabot_blend_test/main.c
index 65e58ec..24722ef 100644
--- a/samples/lerabot_blend_test/main.c
+++ b/samples/lerabot_blend_test/main.c
@@ -4,9 +4,9 @@
  * output and incorrect depth testing
  */
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/lights/main.c b/samples/lights/main.c
index 72777eb..15088d7 100644
--- a/samples/lights/main.c
+++ b/samples/lights/main.c
@@ -1,12 +1,23 @@
 #include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glext.h"
+#include "GL/glkos.h"
 
-extern uint8 romdisk[];
+#ifdef __DREAMCAST__
+
+#include <kos.h>
+extern uint8_t romdisk[];
 KOS_INIT_ROMDISK(romdisk);
 
+#define IMAGE_FILENAME "/rd/NeHe.bmp"
+
+#else
+#define IMAGE_FILENAME "samples/lights/romdisk/NeHe.bmp"
+#endif
 
 float xrot, yrot, zrot;
 
@@ -89,7 +100,7 @@ int ImageLoad(char *filename, Image *image) {
     }
 
     if ((i = fread(image->data, size, 1, file)) != 1) {
-        printf(stderr, "Error reading image data from %s.\n", filename);
+        fprintf(stderr, "Error reading image data from %s.\n", filename);
         return 0;
     }
 
@@ -115,7 +126,7 @@ void LoadGLTextures() {
         exit(0);
     }
 
-    if (!ImageLoad("/rd/NeHe.bmp", image1)) {
+    if (!ImageLoad(IMAGE_FILENAME, image1)) {
         exit(1);
     }
 
diff --git a/samples/mipmap/main.c b/samples/mipmap/main.c
index 15397d3..0ab8447 100644
--- a/samples/mipmap/main.c
+++ b/samples/mipmap/main.c
@@ -1,13 +1,21 @@
 #include <stdio.h>
 #include <stdlib.h>
+#include <stdint.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glext.h"
+#include "GL/glkos.h"
 
-extern uint8 romdisk[];
+extern uint8_t romdisk[];
 KOS_INIT_ROMDISK(romdisk);
 
+#ifdef __DREAMCAST__
+#define IMAGE_FILENAME "/rd/NeHe.bmp"
+#else
+#define IMAGE_FILENAME "samples/mipmap/romdisk/NeHe.bmp"
+#endif
+
 /* storage for one texture  */
 int texture[1];
 
@@ -87,7 +95,7 @@ int ImageLoad(char *filename, Image *image) {
     }
 
     if ((i = fread(image->data, size, 1, file)) != 1) {
-        printf(stderr, "Error reading image data from %s.\n", filename);
+        fprintf(stderr, "Error reading image data from %s.\n", filename);
         return 0;
     }
 
@@ -113,7 +121,7 @@ void LoadGLTextures() {
         exit(0);
     }
 
-    if (!ImageLoad("/rd/NeHe.bmp", image1)) {
+    if (!ImageLoad(IMAGE_FILENAME, image1)) {
         exit(1);
     }
 
diff --git a/samples/multitexture_arrays/main.c b/samples/multitexture_arrays/main.c
index 58bf860..0923c6d 100644
--- a/samples/multitexture_arrays/main.c
+++ b/samples/multitexture_arrays/main.c
@@ -8,11 +8,21 @@
 */
 
 #include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
-#include "glext.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
+#include "GL/glext.h"
+
+#ifdef __DREAMCAST__
+#define IMAGE1_FILENAME "/rd/wp001vq.pvr"
+#define IMAGE2_FILENAME "/rd/FlareWS_256.pvr"
+#else
+#define IMAGE1_FILENAME "samples/multitexture_arrays/romdisk/wp001vq.pvr"
+#define IMAGE2_FILENAME "samples/multitexture_arrays/romdisk/FlareWS_256.pvr"
+#endif
 
 /* Load a PVR texture - located in pvr-texture.c */
 extern GLuint glTextureLoadPVR(char *fname, unsigned char isMipMapped, unsigned char glMipMap);
@@ -29,7 +39,7 @@ GLfloat TEXCOORD_ARRAY[4 * 2] = { 0, 0,
                                   0, 1
                                 };
 
-GLuint ARGB_ARRAY[4 * 1] = { 0xFFFF0000, 0xFF0000FF, 0xFF00FF00, 0xFFFFFF00 };
+GLuint ARGB_ARRAY[4] = { 0xFFFF0000, 0xFF0000FF, 0xFF00FF00, 0xFFFFFF00 };
 
 
 /* Multi-Texture Example using Open GL Vertex Buffer Submission. */
@@ -91,7 +101,7 @@ void RenderCallback(GLuint texID0, GLuint texID1) {
     glDisableClientState(GL_VERTEX_ARRAY);
 }
 
-extern uint8 romdisk[];
+extern uint8_t romdisk[];
 KOS_INIT_ROMDISK(romdisk);
 
 int main(int argc, char **argv) {
@@ -105,8 +115,8 @@ int main(int argc, char **argv) {
     glLoadIdentity();
 
     /* Load two PVR textures to OpenGL */
-    GLuint texID0 = glTextureLoadPVR("/rd/wp001vq.pvr", 0, 0);
-    GLuint texID1 = glTextureLoadPVR("/rd/FlareWS_256.pvr", 0, 0);
+    GLuint texID0 = glTextureLoadPVR(IMAGE1_FILENAME, 0, 0);
+    GLuint texID1 = glTextureLoadPVR(IMAGE2_FILENAME, 0, 0);
 
     while(1) {
         /* Draw the "scene" */
diff --git a/samples/multitexture_arrays/pvr-texture.c b/samples/multitexture_arrays/pvr-texture.c
index 3732803..c9d5f92 100644
--- a/samples/multitexture_arrays/pvr-texture.c
+++ b/samples/multitexture_arrays/pvr-texture.c
@@ -6,13 +6,14 @@
 
    Load A PVR Texture to the PVR using Open GL
 */
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdint.h>
 
-#include <kos.h>
-
-#include "../include/gl.h"
-#include "../include/glu.h"
-#include "../include/glkos.h"
-#include "../include/glext.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
+#include "GL/glext.h"
 
 #define PVR_HDR_SIZE 0x20
 #define MAX(x, y) ((x > y) ? x : y)
@@ -51,8 +52,8 @@ static GLuint _glGetMipmapDataSize(GLuint width, GLuint height) {
    glMipMap should be passed as 1 if Open GL should calculate the Mipmap levels, 0 otherwise */
 GLuint glTextureLoadPVR(char *fname, unsigned char isMipMapped, unsigned char glMipMap) {
     FILE *tex = NULL;
-    uint16 *TEX0 = NULL;
-    uint8 HDR[PVR_HDR_SIZE];
+    uint16_t *TEX0 = NULL;
+    uint8_t HDR[PVR_HDR_SIZE];
     GLuint texID, texSize, texW, texH, texFormat;
 
     /* Open the PVR texture file, and get its file size */
diff --git a/samples/nehe02/main.c b/samples/nehe02/main.c
index b3782e6..ae6a0d0 100644
--- a/samples/nehe02/main.c
+++ b/samples/nehe02/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
@@ -16,7 +16,7 @@ void InitGL(int Width, int Height)	        // We call this right after our OpenG
 
     gluPerspective(45.0f,(GLfloat)Width/(GLfloat)Height,0.1f,100.0f);	// Calculate The Aspect Ratio Of The Window
 
-    glMatrixMode(GL_MODELVIEW);
+    glMatrixMode(GL_MODELVIEW);    
 }
 
 /* The function called when our window is resized (which shouldn't happen, because we're fullscreen) */
diff --git a/samples/nehe02de/main.c b/samples/nehe02de/main.c
index eba7fca..f84775c 100644
--- a/samples/nehe02de/main.c
+++ b/samples/nehe02de/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/nehe02va/main.c b/samples/nehe02va/main.c
index 5256cc7..ebaf30a 100644
--- a/samples/nehe02va/main.c
+++ b/samples/nehe02va/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/nehe03/main.c b/samples/nehe03/main.c
index 61361ee..a4e7c35 100644
--- a/samples/nehe03/main.c
+++ b/samples/nehe03/main.c
@@ -1,6 +1,7 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/nehe04/main.c b/samples/nehe04/main.c
index d4cb75a..a84e94e 100644
--- a/samples/nehe04/main.c
+++ b/samples/nehe04/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 GLfloat rtri;
 GLfloat rquad;
diff --git a/samples/nehe05/main.c b/samples/nehe05/main.c
index 77b7e52..6bffe61 100644
--- a/samples/nehe05/main.c
+++ b/samples/nehe05/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 GLfloat rtri;
 GLfloat rquad;
diff --git a/samples/nehe06/main.c b/samples/nehe06/main.c
index cd3e9f1..c712ba3 100644
--- a/samples/nehe06/main.c
+++ b/samples/nehe06/main.c
@@ -1,17 +1,26 @@
 #include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glext.h"
+#include "GL/glkos.h"
 
-extern uint8 romdisk[];
+#ifdef __DREAMCAST__
+#define IMAGE_FILENAME "/rd/NeHe.bmp"
+extern uint8_t romdisk[];
 KOS_INIT_ROMDISK(romdisk);
+#else
+#define IMAGE_FILENAME "samples/nehe06/romdisk/NeHe.bmp"
+#endif
 
 /* floats for x rotation, y rotation, z rotation */
 float xrot, yrot, zrot;
 
 /* storage for one texture  */
-int texture[1];
+GLuint texture[1];
 
 /* Image type - contains height, width, and data */
 struct Image {
@@ -89,7 +98,7 @@ int ImageLoad(char *filename, Image *image) {
     }
 
     if ((i = fread(image->data, size, 1, file)) != 1) {
-        printf(stderr, "Error reading image data from %s.\n", filename);
+        fprintf(stderr, "Error reading image data from %s.\n", filename);
         return 0;
     }
 
@@ -115,7 +124,7 @@ void LoadGLTextures() {
         exit(0);
     }
 
-    if (!ImageLoad("/rd/NeHe.bmp", image1)) {
+    if (!ImageLoad(IMAGE_FILENAME, image1)) {
         exit(1);
     }
 
diff --git a/samples/nehe06_vq/main.c b/samples/nehe06_vq/main.c
index ad2cc25..c4501e2 100644
--- a/samples/nehe06_vq/main.c
+++ b/samples/nehe06_vq/main.c
@@ -1,12 +1,15 @@
 #include <stdio.h>
+#include <stdlib.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
-#include "glext.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
+#include "GL/glext.h"
 
+#ifdef __DREAMCAST__
 extern uint8 romdisk[];
 KOS_INIT_ROMDISK(romdisk);
+#endif
 
 /* floats for x rotation, y rotation, z rotation */
 float xrot, yrot, zrot;
diff --git a/samples/ortho2d/main.c b/samples/ortho2d/main.c
index e2173da..99354c7 100644
--- a/samples/ortho2d/main.c
+++ b/samples/ortho2d/main.c
@@ -1,6 +1,6 @@
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height)	        // We call this right after our OpenGL window is created.
diff --git a/samples/paletted/main.c b/samples/paletted/main.c
index 5def756..69826da 100644
--- a/samples/paletted/main.c
+++ b/samples/paletted/main.c
@@ -1,13 +1,14 @@
 #include <stdio.h>
 #include <string.h>
 #include <stdlib.h>
+#include <stdint.h>
 
-#include "gl.h"
-#include "glext.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glext.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
-extern uint8 romdisk[];
+extern uint8_t romdisk[];
 KOS_INIT_ROMDISK(romdisk);
 
 /* floats for x rotation, y rotation, z rotation */
diff --git a/samples/paletted_pcx/main.c b/samples/paletted_pcx/main.c
index b82a54a..bb8a0e0 100644
--- a/samples/paletted_pcx/main.c
+++ b/samples/paletted_pcx/main.c
@@ -4,13 +4,16 @@
 #include <stdint.h>
 #include <assert.h>
 
-#include "gl.h"
-#include "glext.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glext.h"
+#include "GL/glkos.h"
 
-extern uint8 romdisk[];
-KOS_INIT_ROMDISK(romdisk);
+#ifdef __DREAMCAST__
+    #include <kos.h>
+    extern uint8 romdisk[];
+    KOS_INIT_ROMDISK(romdisk);
+#endif
 
 /* floats for x rotation, y rotation, z rotation */
 float xrot, yrot, zrot;
diff --git a/samples/polygon_offset/main.c b/samples/polygon_offset/main.c
index 101f4d6..a00278d 100644
--- a/samples/polygon_offset/main.c
+++ b/samples/polygon_offset/main.c
@@ -1,12 +1,17 @@
-#include <kos.h>
+
 typedef enum
 {
     false,
     true
 } bool;
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+
+#ifdef __DREAMCAST__
+#include <kos.h>
+#endif
+
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
 void InitGL(int Width, int Height) // We call this right after our OpenGL window is created.
@@ -96,6 +101,7 @@ void drawPolygon()
 int frames = 0;
 void check_input()
 {
+#ifdef __DREAMCAST__
     maple_device_t *cont;
     cont_state_t *state;
 
@@ -117,6 +123,7 @@ void check_input()
             }
         }
     }
+#endif
 }
 
 /* The main drawing function. */
diff --git a/samples/polymark/main.c b/samples/polymark/main.c
index 2a671fc..d0d6378 100644
--- a/samples/polymark/main.c
+++ b/samples/polymark/main.c
@@ -10,6 +10,7 @@
 #include <kos.h>
 
 #include <GL/gl.h>
+#include <GL/glkos.h>
 
 #include <stdlib.h>
 #include <time.h>
diff --git a/samples/quadmark/main.c b/samples/quadmark/main.c
index 3788875..59da143 100644
--- a/samples/quadmark/main.c
+++ b/samples/quadmark/main.c
@@ -10,6 +10,7 @@
 #include <kos.h>
 
 #include <GL/gl.h>
+#include <GL/glkos.h>
 
 #include <stdlib.h>
 #include <time.h>
diff --git a/samples/terrain/main.c b/samples/terrain/main.c
index 4b4adb2..3ab9163 100644
--- a/samples/terrain/main.c
+++ b/samples/terrain/main.c
@@ -1,9 +1,9 @@
 #include <stdio.h>
 #include <stdlib.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 #define TERRAIN_SIZE 100
 #define TERRAIN_SCALE 1.0f
diff --git a/samples/zclip/main.c b/samples/zclip/main.c
index cbe2148..2da470b 100644
--- a/samples/zclip/main.c
+++ b/samples/zclip/main.c
@@ -1,15 +1,18 @@
 #include <stdio.h>
 #include <string.h>
 #include <stdlib.h>
+#include <stdint.h>
 
-#include "gl.h"
-#include "glext.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glext.h"
+#include "GL/glkos.h"
 
+#ifdef __DREAMCAST__
+#include <kos.h>
 extern uint8 romdisk[];
 KOS_INIT_ROMDISK(romdisk);
-
+#endif
 
 /* Image type - contains height, width, and data */
 typedef struct                                      // Create A Structure
@@ -29,85 +32,85 @@ TextureImage lightmap;
 
 GLboolean LoadTGA(TextureImage *texture, const char *filename)			// Loads A TGA File Into Memory
 {
-	GLubyte		TGAheader[12]={0,0,2,0,0,0,0,0,0,0,0,0};	// Uncompressed TGA Header
-	GLubyte		TGAcompare[12];								// Used To Compare TGA Header
-	GLubyte		header[6];									// First 6 Useful Bytes From The Header
-	GLuint		bytesPerPixel;								// Holds Number Of Bytes Per Pixel Used In The TGA File
-	GLuint		imageSize;									// Used To Store The Image Size When Setting Aside Ram
-	GLuint		temp;										// Temporary Variable
-	GLuint		type=GL_RGBA;								// Set The Default GL Mode To RBGA (32 BPP)
+    GLubyte		TGAheader[12]={0,0,2,0,0,0,0,0,0,0,0,0};	// Uncompressed TGA Header
+    GLubyte		TGAcompare[12];								// Used To Compare TGA Header
+    GLubyte		header[6];									// First 6 Useful Bytes From The Header
+    GLuint		bytesPerPixel;								// Holds Number Of Bytes Per Pixel Used In The TGA File
+    GLuint		imageSize;									// Used To Store The Image Size When Setting Aside Ram
+    GLuint		temp;										// Temporary Variable
+    GLuint		type=GL_RGBA;								// Set The Default GL Mode To RBGA (32 BPP)
 
-	FILE *file = fopen(filename, "rb");						// Open The TGA File
+    FILE *file = fopen(filename, "rb");						// Open The TGA File
 
-	if(	file==NULL ||										// Does File Even Exist?
-		fread(TGAcompare,1,sizeof(TGAcompare),file)!=sizeof(TGAcompare) ||	// Are There 12 Bytes To Read?
-		memcmp(TGAheader,TGAcompare,sizeof(TGAheader))!=0				||	// Does The Header Match What We Want?
-		fread(header,1,sizeof(header),file)!=sizeof(header))				// If So Read Next 6 Header Bytes
-	{
-		if (file == NULL) {								// Did The File Even Exist? *Added Jim Strong*
+    if(	file==NULL ||										// Does File Even Exist?
+        fread(TGAcompare,1,sizeof(TGAcompare),file)!=sizeof(TGAcompare) ||	// Are There 12 Bytes To Read?
+        memcmp(TGAheader,TGAcompare,sizeof(TGAheader))!=0				||	// Does The Header Match What We Want?
+        fread(header,1,sizeof(header),file)!=sizeof(header))				// If So Read Next 6 Header Bytes
+    {
+        if (file == NULL) {								// Did The File Even Exist? *Added Jim Strong*
             fprintf(stderr, "Missing file\n");
-			return GL_FALSE;									// Return False
+            return GL_FALSE;									// Return False
         } else
-		{
+        {
             fprintf(stderr, "Invalid format\n");
-			fclose(file);									// If Anything Failed, Close The File
-			return GL_FALSE;									// Return False
-		}
-	}
+            fclose(file);									// If Anything Failed, Close The File
+            return GL_FALSE;									// Return False
+        }
+    }
 
-	texture->width  = header[1] * 256 + header[0];			// Determine The TGA Width	(highbyte*256+lowbyte)
-	texture->height = header[3] * 256 + header[2];			// Determine The TGA Height	(highbyte*256+lowbyte)
+    texture->width  = header[1] * 256 + header[0];			// Determine The TGA Width	(highbyte*256+lowbyte)
+    texture->height = header[3] * 256 + header[2];			// Determine The TGA Height	(highbyte*256+lowbyte)
 
- 	if(	texture->width	<=0	||								// Is The Width Less Than Or Equal To Zero
-		texture->height	<=0	||								// Is The Height Less Than Or Equal To Zero
-		(header[4]!=24 && header[4]!=32))					// Is The TGA 24 or 32 Bit?
-	{
-		fprintf(stderr, "Wrong format\n");
+    if(	texture->width	<=0	||								// Is The Width Less Than Or Equal To Zero
+        texture->height	<=0	||								// Is The Height Less Than Or Equal To Zero
+        (header[4]!=24 && header[4]!=32))					// Is The TGA 24 or 32 Bit?
+    {
+        fprintf(stderr, "Wrong format\n");
         fclose(file);										// If Anything Failed, Close The File
-		return GL_FALSE;										// Return False
-	}
+        return GL_FALSE;										// Return False
+    }
 
-	texture->bpp	= header[4];							// Grab The TGA's Bits Per Pixel (24 or 32)
-	bytesPerPixel	= texture->bpp/8;						// Divide By 8 To Get The Bytes Per Pixel
-	imageSize		= texture->width*texture->height*bytesPerPixel;	// Calculate The Memory Required For The TGA Data
+    texture->bpp	= header[4];							// Grab The TGA's Bits Per Pixel (24 or 32)
+    bytesPerPixel	= texture->bpp/8;						// Divide By 8 To Get The Bytes Per Pixel
+    imageSize		= texture->width*texture->height*bytesPerPixel;	// Calculate The Memory Required For The TGA Data
 
-	texture->imageData=(GLubyte *)malloc(imageSize);		// Reserve Memory To Hold The TGA Data
+    texture->imageData=(GLubyte *)malloc(imageSize);		// Reserve Memory To Hold The TGA Data
 
-	if(	texture->imageData==NULL ||							// Does The Storage Memory Exist?
-		fread(texture->imageData, 1, imageSize, file)!=imageSize)	// Does The Image Size Match The Memory Reserved?
-	{
-		if(texture->imageData!=NULL)						// Was Image Data Loaded
-			free(texture->imageData);						// If So, Release The Image Data
+    if(	texture->imageData==NULL ||							// Does The Storage Memory Exist?
+        fread(texture->imageData, 1, imageSize, file)!=imageSize)	// Does The Image Size Match The Memory Reserved?
+    {
+        if(texture->imageData!=NULL)						// Was Image Data Loaded
+            free(texture->imageData);						// If So, Release The Image Data
 
-		fclose(file);										// Close The File
-		return GL_FALSE;										// Return False
-	}
+        fclose(file);										// Close The File
+        return GL_FALSE;										// Return False
+    }
 
     GLuint i;
-	for(i = 0; i < (int)imageSize; i += bytesPerPixel)		// Loop Through The Image Data
-	{														// Swaps The 1st And 3rd Bytes ('R'ed and 'B'lue)
-		temp=texture->imageData[i];							// Temporarily Store The Value At Image Data 'i'
-		texture->imageData[i] = texture->imageData[i + 2];	// Set The 1st Byte To The Value Of The 3rd Byte
-		texture->imageData[i + 2] = temp;					// Set The 3rd Byte To The Value In 'temp' (1st Byte Value)
-	}
+    for(i = 0; i < (int)imageSize; i += bytesPerPixel)		// Loop Through The Image Data
+    {														// Swaps The 1st And 3rd Bytes ('R'ed and 'B'lue)
+        temp=texture->imageData[i];							// Temporarily Store The Value At Image Data 'i'
+        texture->imageData[i] = texture->imageData[i + 2];	// Set The 1st Byte To The Value Of The 3rd Byte
+        texture->imageData[i + 2] = temp;					// Set The 3rd Byte To The Value In 'temp' (1st Byte Value)
+    }
 
-	fclose (file);											// Close The File
+    fclose (file);											// Close The File
 
-	// Build A Texture From The Data
-	glGenTextures(1, &texture[0].texID);					// Generate OpenGL texture IDs
+    // Build A Texture From The Data
+    glGenTextures(1, &texture[0].texID);					// Generate OpenGL texture IDs
 
-	glBindTexture(GL_TEXTURE_2D, texture[0].texID);			// Bind Our Texture
-	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);	// Linear Filtered
-	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);	// Linear Filtered
+    glBindTexture(GL_TEXTURE_2D, texture[0].texID);			// Bind Our Texture
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);	// Linear Filtered
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);	// Linear Filtered
 
-	if (texture[0].bpp==24)									// Was The TGA 24 Bits
-	{
-		type=GL_RGB;										// If So Set The 'type' To GL_RGB
-	}
+    if (texture[0].bpp==24)									// Was The TGA 24 Bits
+    {
+        type=GL_RGB;										// If So Set The 'type' To GL_RGB
+    }
 
-	glTexImage2D(GL_TEXTURE_2D, 0, type, texture[0].width, texture[0].height, 0, type, GL_UNSIGNED_BYTE, texture[0].imageData);
+    glTexImage2D(GL_TEXTURE_2D, 0, type, texture[0].width, texture[0].height, 0, type, GL_UNSIGNED_BYTE, texture[0].imageData);
 
-	return GL_TRUE;											// Texture Building Went Ok, Return True
+    return GL_TRUE;											// Texture Building Went Ok, Return True
 }
 // Load Bitmaps And Convert To Textures
 void LoadGLTextures() {
diff --git a/samples/zclip_triangle/main.c b/samples/zclip_triangle/main.c
index 46ea274..0400f0b 100644
--- a/samples/zclip_triangle/main.c
+++ b/samples/zclip_triangle/main.c
@@ -2,9 +2,9 @@
 #include <string.h>
 #include <stdlib.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
diff --git a/samples/zclip_trianglestrip/main.c b/samples/zclip_trianglestrip/main.c
index f6b8e80..332ec3e 100644
--- a/samples/zclip_trianglestrip/main.c
+++ b/samples/zclip_trianglestrip/main.c
@@ -2,9 +2,9 @@
 #include <string.h>
 #include <stdlib.h>
 
-#include "gl.h"
-#include "glu.h"
-#include "glkos.h"
+#include "GL/gl.h"
+#include "GL/glu.h"
+#include "GL/glkos.h"
 
 
 /* A general OpenGL initialization function.  Sets all of the initial parameters. */
diff --git a/toolchains/Dreamcast.cmake b/toolchains/Dreamcast.cmake
new file mode 100644
index 0000000..9a10474
--- /dev/null
+++ b/toolchains/Dreamcast.cmake
@@ -0,0 +1,59 @@
+SET(CMAKE_SYSTEM_NAME Generic)
+
+SET(PLATFORM_DREAMCAST TRUE)
+
+SET(CMAKE_SYSTEM_VERSION 1)
+
+set(CMAKE_CROSSCOMPILING TRUE)
+
+set(CMAKE_C_COMPILER $ENV{KOS_CC_BASE}/bin/sh-elf-gcc)
+set(CMAKE_CXX_COMPILER $ENV{KOS_CC_BASE}/bin/sh-elf-g++)
+
+add_compile_options(-ml -m4-single-only -ffunction-sections -fdata-sections)
+
+set(CMAKE_EXE_LINKER_FLAGS " -ml -m4-single-only -Wl,-Ttext=0x8c010000 -Wl,--gc-sections -T$ENV{KOS_BASE}/utils/ldscripts/shlelf.xc -nodefaultlibs" CACHE INTERNAL "" FORCE)
+
+set(CMAKE_FIND_LIBRARY_SUFFIXES ".a")
+set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
+
+set(CMAKE_SYSTEM_INCLUDE_PATH "${CMAKE_SYSTEM_INCLUDE_PATH} $ENV{KOS_BASE}/include $ENV{KOS_BASE}/kernel/arch/dreamcast/include $ENV{KOS_BASE}/addons/include $ENV{KOS_BASE}/../kos-ports/include")
+
+INCLUDE_DIRECTORIES(
+    $ENV{KOS_BASE}/include
+    $ENV{KOS_BASE}/kernel/arch/dreamcast/include
+    $ENV{KOS_BASE}/addons/include
+    $ENV{KOS_BASE}/../kos-ports/include
+)
+
+LINK_DIRECTORIES(
+    $ENV{KOS_BASE}/addons/lib/dreamcast
+    $ENV{KOS_PORTS}/lib
+)
+
+IF(${CMAKE_BUILD_TYPE} MATCHES Debug)
+LINK_DIRECTORIES($ENV{KOS_BASE}/lib/dreamcast/debug)
+ELSE()
+LINK_DIRECTORIES($ENV{KOS_BASE}/lib/dreamcast)
+ENDIF()
+
+
+add_link_options(-L$ENV{KOS_BASE}/lib/dreamcast)
+link_libraries(-Wl,--start-group -lstdc++ -lkallisti -lc -lgcc -Wl,--end-group m)
+
+SET(CMAKE_EXECUTABLE_SUFFIX ".elf")
+SET(CMAKE_EXECUTABLE_SUFFIX_CXX ".elf")
+
+ADD_DEFINITIONS(
+    -D__DREAMCAST__
+    -DDREAMCAST
+    -D_arch_dreamcast
+    -D__arch_dreamcast
+    -D_arch_sub_pristine
+)
+
+if (NOT CMAKE_BUILD_TYPE MATCHES Debug)
+    ADD_DEFINITIONS(-DNDEBUG)
+endif()
+
+SET(CMAKE_ASM_FLAGS "")
+SET(CMAKE_ASM_FLAGS_RELEASE "")