Cleanup: replace attrib w/ attr

Also rename GPUVertexAttribs to GPUVertAttrLayers,
avoids confusion with GPUVertAttr which isn't closely related.
This commit is contained in:
2019-01-29 07:46:25 +11:00
parent a9c2971658
commit 68cab3aff6
76 changed files with 667 additions and 629 deletions

View File

@@ -66,15 +66,15 @@ typedef struct {
/* current vertex */
uint vertex_idx;
GLubyte *vertex_data;
uint16_t unassigned_attrib_bits; /* which attributes of current vertex have not been given values? */
uint16_t unassigned_attr_bits; /* which attributes of current vertex have not been given values? */
GLuint vbo_id;
GLuint vao_id;
GLuint bound_program;
const GPUShaderInterface *shader_interface;
GPUAttrBinding attrib_binding;
uint16_t prev_enabled_attrib_bits; /* <-- only affects this VAO, so we're ok */
GPUAttrBinding attr_binding;
uint16_t prev_enabled_attr_bits; /* <-- only affects this VAO, so we're ok */
} Immediate;
/* size of internal buffer -- make this adjustable? */
@@ -121,7 +121,7 @@ void immDeactivate(void)
#endif
GPU_vao_free(imm.vao_id, imm.context);
imm.vao_id = 0;
imm.prev_enabled_attrib_bits = 0;
imm.prev_enabled_attr_bits = 0;
}
void immDestroy(void)
@@ -150,7 +150,7 @@ void immBindProgram(GLuint program, const GPUShaderInterface *shaderface)
VertexFormat_pack(&imm.vertex_format);
glUseProgram(program);
get_attrib_locations(&imm.vertex_format, &imm.attrib_binding, shaderface);
get_attr_locations(&imm.vertex_format, &imm.attr_binding, shaderface);
GPU_matrix_bind(shaderface);
}
@@ -210,7 +210,7 @@ void immBegin(GPUPrimType prim_type, uint vertex_len)
imm.prim_type = prim_type;
imm.vertex_len = vertex_len;
imm.vertex_idx = 0;
imm.unassigned_attrib_bits = imm.attrib_binding.enabled_bits;
imm.unassigned_attr_bits = imm.attr_binding.enabled_bits;
/* how many bytes do we need for this draw call? */
const uint bytes_needed = vertex_buffer_size(&imm.vertex_format, vertex_len);
@@ -270,7 +270,7 @@ GPUBatch *immBeginBatch(GPUPrimType prim_type, uint vertex_len)
imm.prim_type = prim_type;
imm.vertex_len = vertex_len;
imm.vertex_idx = 0;
imm.unassigned_attrib_bits = imm.attrib_binding.enabled_bits;
imm.unassigned_attr_bits = imm.attr_binding.enabled_bits;
GPUVertBuf *verts = GPU_vertbuf_create_with_format(&imm.vertex_format);
GPU_vertbuf_data_alloc(verts, vertex_len);
@@ -295,11 +295,11 @@ static void immDrawSetup(void)
/* set up VAO -- can be done during Begin or End really */
glBindVertexArray(imm.vao_id);
/* enable/disable vertex attribs as needed */
if (imm.attrib_binding.enabled_bits != imm.prev_enabled_attrib_bits) {
/* Enable/Disable vertex attributes as needed. */
if (imm.attr_binding.enabled_bits != imm.prev_enabled_attr_bits) {
for (uint loc = 0; loc < GPU_VERT_ATTR_MAX_LEN; ++loc) {
bool is_enabled = imm.attrib_binding.enabled_bits & (1 << loc);
bool was_enabled = imm.prev_enabled_attrib_bits & (1 << loc);
bool is_enabled = imm.attr_binding.enabled_bits & (1 << loc);
bool was_enabled = imm.prev_enabled_attr_bits & (1 << loc);
if (is_enabled && !was_enabled) {
glEnableVertexAttribArray(loc);
@@ -309,18 +309,18 @@ static void immDrawSetup(void)
}
}
imm.prev_enabled_attrib_bits = imm.attrib_binding.enabled_bits;
imm.prev_enabled_attr_bits = imm.attr_binding.enabled_bits;
}
const uint stride = imm.vertex_format.stride;
for (uint a_idx = 0; a_idx < imm.vertex_format.attr_len; ++a_idx) {
const GPUVertAttr *a = imm.vertex_format.attribs + a_idx;
const GPUVertAttr *a = &imm.vertex_format.attrs[a_idx];
const uint offset = imm.buffer_offset + a->offset;
const GLvoid *pointer = (const GLubyte *)0 + offset;
const uint loc = read_attrib_location(&imm.attrib_binding, a_idx);
const uint loc = read_attr_location(&imm.attr_binding, a_idx);
switch (a->fetch_mode) {
case GPU_FETCH_FLOAT:
@@ -400,29 +400,29 @@ void immEnd(void)
imm.strict_vertex_len = true;
}
static void setAttribValueBit(uint attrib_id)
static void setAttrValueBit(uint attr_id)
{
uint16_t mask = 1 << attrib_id;
uint16_t mask = 1 << attr_id;
#if TRUST_NO_ONE
assert(imm.unassigned_attrib_bits & mask); /* not already set */
assert(imm.unassigned_attr_bits & mask); /* not already set */
#endif
imm.unassigned_attrib_bits &= ~mask;
imm.unassigned_attr_bits &= ~mask;
}
/* --- generic attribute functions --- */
void immAttr1f(uint attrib_id, float x)
void immAttr1f(uint attr_id, float x)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_F32);
assert(attr->comp_len == 1);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
float *data = (float *)(imm.vertex_data + attr->offset);
/* printf("%s %td %p\n", __FUNCTION__, (GLubyte*)data - imm.buffer_data, data); */
@@ -430,17 +430,17 @@ void immAttr1f(uint attrib_id, float x)
data[0] = x;
}
void immAttr2f(uint attrib_id, float x, float y)
void immAttr2f(uint attr_id, float x, float y)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_F32);
assert(attr->comp_len == 2);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
float *data = (float *)(imm.vertex_data + attr->offset);
/* printf("%s %td %p\n", __FUNCTION__, (GLubyte*)data - imm.buffer_data, data); */
@@ -449,17 +449,17 @@ void immAttr2f(uint attrib_id, float x, float y)
data[1] = y;
}
void immAttr3f(uint attrib_id, float x, float y, float z)
void immAttr3f(uint attr_id, float x, float y, float z)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_F32);
assert(attr->comp_len == 3);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
float *data = (float *)(imm.vertex_data + attr->offset);
/* printf("%s %td %p\n", __FUNCTION__, (GLubyte*)data - imm.buffer_data, data); */
@@ -469,17 +469,17 @@ void immAttr3f(uint attrib_id, float x, float y, float z)
data[2] = z;
}
void immAttr4f(uint attrib_id, float x, float y, float z, float w)
void immAttr4f(uint attr_id, float x, float y, float z, float w)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_F32);
assert(attr->comp_len == 4);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
float *data = (float *)(imm.vertex_data + attr->offset);
/* printf("%s %td %p\n", __FUNCTION__, (GLubyte*)data - imm.buffer_data, data); */
@@ -490,34 +490,34 @@ void immAttr4f(uint attrib_id, float x, float y, float z, float w)
data[3] = w;
}
void immAttr1u(uint attrib_id, uint x)
void immAttr1u(uint attr_id, uint x)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_U32);
assert(attr->comp_len == 1);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
uint *data = (uint *)(imm.vertex_data + attr->offset);
data[0] = x;
}
void immAttr2i(uint attrib_id, int x, int y)
void immAttr2i(uint attr_id, int x, int y)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_I32);
assert(attr->comp_len == 2);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
int *data = (int *)(imm.vertex_data + attr->offset);
@@ -525,17 +525,17 @@ void immAttr2i(uint attrib_id, int x, int y)
data[1] = y;
}
void immAttr2s(uint attrib_id, short x, short y)
void immAttr2s(uint attr_id, short x, short y)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_I16);
assert(attr->comp_len == 2);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
short *data = (short *)(imm.vertex_data + attr->offset);
@@ -543,32 +543,32 @@ void immAttr2s(uint attrib_id, short x, short y)
data[1] = y;
}
void immAttr2fv(uint attrib_id, const float data[2])
void immAttr2fv(uint attr_id, const float data[2])
{
immAttr2f(attrib_id, data[0], data[1]);
immAttr2f(attr_id, data[0], data[1]);
}
void immAttr3fv(uint attrib_id, const float data[3])
void immAttr3fv(uint attr_id, const float data[3])
{
immAttr3f(attrib_id, data[0], data[1], data[2]);
immAttr3f(attr_id, data[0], data[1], data[2]);
}
void immAttr4fv(uint attrib_id, const float data[4])
void immAttr4fv(uint attr_id, const float data[4])
{
immAttr4f(attrib_id, data[0], data[1], data[2], data[3]);
immAttr4f(attr_id, data[0], data[1], data[2], data[3]);
}
void immAttr3ub(uint attrib_id, uchar r, uchar g, uchar b)
void immAttr3ub(uint attr_id, uchar r, uchar g, uchar b)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_U8);
assert(attr->comp_len == 3);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
GLubyte *data = imm.vertex_data + attr->offset;
/* printf("%s %td %p\n", __FUNCTION__, data - imm.buffer_data, data); */
@@ -578,17 +578,17 @@ void immAttr3ub(uint attrib_id, uchar r, uchar g, uchar b)
data[2] = b;
}
void immAttr4ub(uint attrib_id, uchar r, uchar g, uchar b, uchar a)
void immAttr4ub(uint attr_id, uchar r, uchar g, uchar b, uchar a)
{
GPUVertAttr *attr = imm.vertex_format.attribs + attrib_id;
GPUVertAttr *attr = &imm.vertex_format.attrs[attr_id];
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(attr->comp_type == GPU_COMP_U8);
assert(attr->comp_len == 4);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
GLubyte *data = imm.vertex_data + attr->offset;
/* printf("%s %td %p\n", __FUNCTION__, data - imm.buffer_data, data); */
@@ -599,24 +599,24 @@ void immAttr4ub(uint attrib_id, uchar r, uchar g, uchar b, uchar a)
data[3] = a;
}
void immAttr3ubv(uint attrib_id, const uchar data[3])
void immAttr3ubv(uint attr_id, const uchar data[3])
{
immAttr3ub(attrib_id, data[0], data[1], data[2]);
immAttr3ub(attr_id, data[0], data[1], data[2]);
}
void immAttr4ubv(uint attrib_id, const uchar data[4])
void immAttr4ubv(uint attr_id, const uchar data[4])
{
immAttr4ub(attrib_id, data[0], data[1], data[2], data[3]);
immAttr4ub(attr_id, data[0], data[1], data[2], data[3]);
}
void immAttrSkip(uint attrib_id)
void immAttrSkip(uint attr_id)
{
#if TRUST_NO_ONE
assert(attrib_id < imm.vertex_format.attr_len);
assert(attr_id < imm.vertex_format.attr_len);
assert(imm.vertex_idx < imm.vertex_len);
assert(imm.prim_type != GPU_PRIM_NONE); /* make sure we're between a Begin/End pair */
#endif
setAttribValueBit(attrib_id);
setAttrValueBit(attr_id);
}
static void immEndVertex(void) /* and move on to the next vertex */
@@ -626,15 +626,15 @@ static void immEndVertex(void) /* and move on to the next vertex */
assert(imm.vertex_idx < imm.vertex_len);
#endif
/* have all attribs been assigned values?
* if not, copy value from previous vertex */
if (imm.unassigned_attrib_bits) {
/* Have all attributes been assigned values?
* If not, copy value from previous vertex. */
if (imm.unassigned_attr_bits) {
#if TRUST_NO_ONE
assert(imm.vertex_idx > 0); /* first vertex must have all attribs specified */
assert(imm.vertex_idx > 0); /* first vertex must have all attributes specified */
#endif
for (uint a_idx = 0; a_idx < imm.vertex_format.attr_len; ++a_idx) {
if ((imm.unassigned_attrib_bits >> a_idx) & 1) {
const GPUVertAttr *a = imm.vertex_format.attribs + a_idx;
if ((imm.unassigned_attr_bits >> a_idx) & 1) {
const GPUVertAttr *a = &imm.vertex_format.attrs[a_idx];
/* printf("copying %s from vertex %u to %u\n", a->name, imm.vertex_idx - 1, imm.vertex_idx); */
@@ -647,54 +647,54 @@ static void immEndVertex(void) /* and move on to the next vertex */
imm.vertex_idx++;
imm.vertex_data += imm.vertex_format.stride;
imm.unassigned_attrib_bits = imm.attrib_binding.enabled_bits;
imm.unassigned_attr_bits = imm.attr_binding.enabled_bits;
}
void immVertex2f(uint attrib_id, float x, float y)
void immVertex2f(uint attr_id, float x, float y)
{
immAttr2f(attrib_id, x, y);
immAttr2f(attr_id, x, y);
immEndVertex();
}
void immVertex3f(uint attrib_id, float x, float y, float z)
void immVertex3f(uint attr_id, float x, float y, float z)
{
immAttr3f(attrib_id, x, y, z);
immAttr3f(attr_id, x, y, z);
immEndVertex();
}
void immVertex4f(uint attrib_id, float x, float y, float z, float w)
void immVertex4f(uint attr_id, float x, float y, float z, float w)
{
immAttr4f(attrib_id, x, y, z, w);
immAttr4f(attr_id, x, y, z, w);
immEndVertex();
}
void immVertex2i(uint attrib_id, int x, int y)
void immVertex2i(uint attr_id, int x, int y)
{
immAttr2i(attrib_id, x, y);
immAttr2i(attr_id, x, y);
immEndVertex();
}
void immVertex2s(uint attrib_id, short x, short y)
void immVertex2s(uint attr_id, short x, short y)
{
immAttr2s(attrib_id, x, y);
immAttr2s(attr_id, x, y);
immEndVertex();
}
void immVertex2fv(uint attrib_id, const float data[2])
void immVertex2fv(uint attr_id, const float data[2])
{
immAttr2f(attrib_id, data[0], data[1]);
immAttr2f(attr_id, data[0], data[1]);
immEndVertex();
}
void immVertex3fv(uint attrib_id, const float data[3])
void immVertex3fv(uint attr_id, const float data[3])
{
immAttr3f(attrib_id, data[0], data[1], data[2]);
immAttr3f(attr_id, data[0], data[1], data[2]);
immEndVertex();
}
void immVertex2iv(uint attrib_id, const int data[2])
void immVertex2iv(uint attr_id, const int data[2])
{
immAttr2i(attrib_id, data[0], data[1]);
immAttr2i(attr_id, data[0], data[1]);
immEndVertex();
}