Merge branch 'resize_coldpath' into 'master'

Move aligned_vector_reserve onto cold path since it should rarely be called

See merge request simulant/GLdc!132

containers/aligned_vector.c

@@ -14,7 +14,6 @@
 
 extern inline void* aligned_vector_resize(AlignedVector* vector, const uint32_t element_count);
 extern inline void* aligned_vector_extend(AlignedVector* vector, const uint32_t additional_count);
-extern inline void* aligned_vector_reserve(AlignedVector* vector, uint32_t element_count);
 extern inline void* aligned_vector_push_back(AlignedVector* vector, const void* objs, uint32_t count);
 
 void aligned_vector_init(AlignedVector* vector, uint32_t element_size) {
@@ -31,6 +30,31 @@ void aligned_vector_init(AlignedVector* vector, uint32_t element_size) {
     (void) ptr;
 }
 
+void* aligned_vector_reserve(AlignedVector* vector, uint32_t element_count) {
+    AlignedVectorHeader* hdr = &vector->hdr;
+
+    if(element_count < hdr->capacity) {
+        return aligned_vector_at(vector, element_count);
+    }
+
+    uint32_t original_byte_size = (hdr->size * hdr->element_size);
+
+    /* We overallocate so that we don't make small allocations during push backs */
+    element_count = ROUND_TO_CHUNK_SIZE(element_count);
+
+    uint32_t new_byte_size = (element_count * hdr->element_size);
+    uint8_t* original_data = vector->data;
+
+    vector->data = (uint8_t*) memalign(0x20, new_byte_size);
+    assert(vector->data);
+
+    AV_MEMCPY4(vector->data, original_data, original_byte_size);
+    free(original_data);
+
+    hdr->capacity = element_count;
+    return vector->data + original_byte_size;
+}
+
 void aligned_vector_shrink_to_fit(AlignedVector* vector) {
     AlignedVectorHeader* const hdr = &vector->hdr;
     if(hdr->size == 0) {

containers/aligned_vector.h

@@ -27,6 +27,7 @@ static inline void* memalign(size_t alignment, size_t size) {
 #define AV_INLINE_DEBUG AV_NO_INSTRUMENT __attribute__((always_inline))
 #define AV_FORCE_INLINE static AV_INLINE_DEBUG
 #endif
+#define AV_NO_INLINE __attribute__((noinline))
 
 
 #ifdef __DREAMCAST__
@@ -86,37 +87,16 @@ typedef struct {
 
 void aligned_vector_init(AlignedVector* vector, uint32_t element_size);
 
+/* Resizes the backing array if necessary so that element_count elements can fit in the array */
+/* (note that only capacity is potentially changed, size is never changed) */
+AV_NO_INLINE void* aligned_vector_reserve(AlignedVector* vector, uint32_t element_count);
+
 AV_FORCE_INLINE void* aligned_vector_at(const AlignedVector* vector, const uint32_t index) {
     const AlignedVectorHeader* hdr = &vector->hdr;
     assert(index < hdr->size);
     return vector->data + (index * hdr->element_size);
 }
 
-AV_FORCE_INLINE void* aligned_vector_reserve(AlignedVector* vector, uint32_t element_count) {
-    AlignedVectorHeader* hdr = &vector->hdr;
-
-    if(element_count < hdr->capacity) {
-        return aligned_vector_at(vector, element_count);
-    }
-
-    uint32_t original_byte_size = (hdr->size * hdr->element_size);
-
-    /* We overallocate so that we don't make small allocations during push backs */
-    element_count = ROUND_TO_CHUNK_SIZE(element_count);
-
-    uint32_t new_byte_size = (element_count * hdr->element_size);
-    uint8_t* original_data = vector->data;
-
-    vector->data = (uint8_t*) memalign(0x20, new_byte_size);
-    assert(vector->data);
-
-    AV_MEMCPY4(vector->data, original_data, original_byte_size);
-    free(original_data);
-
-    hdr->capacity = element_count;
-    return vector->data + original_byte_size;
-}
-
 AV_FORCE_INLINE AlignedVectorHeader* aligned_vector_header(const AlignedVector* vector) {
     return (AlignedVectorHeader*) &vector->hdr;
 }

samples/primitive_modes/main.c

@@ -2,10 +2,11 @@
 #include <math.h>
 #include <stdlib.h>
 
-#ifdef SDL2_BUILD
+#ifndef __DREAMCAST__
 #include <SDL2/SDL.h>
 static SDL_Window* win_handle;
 #else
+#include <kos.h>
 #include <GL/glkos.h>
 #endif
 
@@ -116,15 +117,36 @@ static void DrawTriFan(float y) {
 	glEnd();
 }
 
-int main(int argc, char *argv[]) {
-#ifdef SDL2_BUILD
+static void sample_init() {
+#ifndef __DREAMCAST__
 	SDL_Init(SDL_INIT_EVERYTHING);
 	win_handle = SDL_CreateWindow("Shapes", 0, 0, 640, 480, SDL_WINDOW_OPENGL);
 	SDL_GL_CreateContext(win_handle);
 #else
     glKosInit();
 #endif
+}
 
+static int sample_should_exit() {
+#ifndef __DREAMCAST__
+    SDL_Event event;
+    while (SDL_PollEvent(&event)) {
+        if(event.type == SDL_QUIT) return 1;
+    }
+    return 0;
+#else
+    maple_device_t *cont = maple_enum_type(0, MAPLE_FUNC_CONTROLLER);
+    if (!cont) return 0;
+
+    cont_state_t *state = (cont_state_t *)maple_dev_status(cont);
+    if (!state) return 0;
+
+    return state->buttons & CONT_START;
+#endif
+}
+
+int main(int argc, char *argv[]) {
+    sample_init();
 	glClearColor(0.5f, 0.5f, 0.5f, 1);
 	glViewport(0, 0, 640, 480);
 
@@ -139,18 +161,9 @@ int main(int argc, char *argv[]) {
 	mat[3][3] = 1;
 	glLoadMatrixf(mat);
 
-	int running = 1;
-
-	while (running) {
+	while (!sample_should_exit()) {
 		glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
 
-#ifdef SDL2_BUILD
-		SDL_Event event;
-		while (SDL_PollEvent(&event)) {
-			if(event.type == SDL_QUIT) running = 0;
-		}
-#endif
-
 		DrawTriStrip(300);
 		DrawTriFan(300);
 		DrawTriList(300);
@@ -160,9 +173,9 @@ int main(int argc, char *argv[]) {
 
 		glPointSize(5);
 		DrawPoint(100);
-		glLineWidth(0.5f);
-		DrawLineLoop(100);
 		glLineWidth(2);
+		DrawLineLoop(100);
+		glLineWidth(4);
 		DrawLineStrip(100);
 		glLineWidth(10);
 		DrawLine(100);