static void brw_emit_vertices(struct brw_context *brw)
{
struct gl_context *ctx = &brw->intel.ctx;
struct intel_context *intel = intel_context(ctx);
GLuint i, nr_elements;
brw_prepare_vertices(brw);
brw_emit_query_begin(brw);
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
* VERTEX_ELEMENT struct and bail.
*
* The stale VB state stays in place, but they don't do anything unless
* a VE loads from them.
*/
if (brw->vb.nr_enabled == 0) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | 1);
if (intel->gen >= 6) {
OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
}
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
CACHED_BATCH();
return;
}
/* Now emit VB and VEP state packets.
*/
if (brw->vb.nr_buffers) {
if (intel->gen >= 6) {
assert(brw->vb.nr_buffers <= 33);
} else {
assert(brw->vb.nr_buffers <= 17);
}
BEGIN_BATCH(1 + 4*brw->vb.nr_buffers);
OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4*brw->vb.nr_buffers - 1));
for (i = 0; i < brw->vb.nr_buffers; i++) {
struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
uint32_t dw0;
if (intel->gen >= 6) {
dw0 = buffer->step_rate
? GEN6_VB0_ACCESS_INSTANCEDATA
: GEN6_VB0_ACCESS_VERTEXDATA;
dw0 |= i << GEN6_VB0_INDEX_SHIFT;
} else {
dw0 = buffer->step_rate
? BRW_VB0_ACCESS_INSTANCEDATA
: BRW_VB0_ACCESS_VERTEXDATA;
dw0 |= i << BRW_VB0_INDEX_SHIFT;
}
if (intel->gen >= 7)
dw0 |= GEN7_VB0_ADDRESS_MODIFYENABLE;
OUT_BATCH(dw0 | (buffer->stride << BRW_VB0_PITCH_SHIFT));
OUT_RELOC(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->offset);
if (intel->gen >= 5) {
OUT_RELOC(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->bo->size - 1);
} else
OUT_BATCH(0);
OUT_BATCH(buffer->step_rate);
brw->vb.current_buffers[i].handle = buffer->bo->handle;
brw->vb.current_buffers[i].offset = buffer->offset;
brw->vb.current_buffers[i].stride = buffer->stride;
brw->vb.current_buffers[i].step_rate = buffer->step_rate;
}
brw->vb.nr_current_buffers = i;
ADVANCE_BATCH();
}
nr_elements = brw->vb.nr_enabled + brw->vs.prog_data->uses_vertexid;
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS, presumably
* for VertexID/InstanceID.
*/
if (intel->gen >= 6) {
assert(nr_elements <= 34);
} else {
assert(nr_elements <= 18);
}
BEGIN_BATCH(1 + nr_elements * 2);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
for (i = 0; i < brw->vb.nr_enabled; i++) {
struct brw_vertex_element *input = brw->vb.enabled[i];
uint32_t format = get_surface_type(input->glarray->Type,
input->glarray->Size,
input->glarray->Format,
input->glarray->Normalized,
input->glarray->Integer);
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
switch (input->glarray->Size) {
case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
: BRW_VE1_COMPONENT_STORE_1_FLT;
break;
}
if (intel->gen >= 6) {
OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((input->buffer << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
}
if (intel->gen >= 5)
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT));
else
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
}
if (brw->vs.prog_data->uses_vertexid) {
uint32_t dw0 = 0, dw1 = 0;
dw1 = ((BRW_VE1_COMPONENT_STORE_VID << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_IID << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
if (intel->gen >= 6) {
dw0 |= GEN6_VE0_VALID;
} else {
dw0 |= BRW_VE0_VALID;
dw1 |= (i * 4) << BRW_VE1_DST_OFFSET_SHIFT;
}
/* Note that for gl_VertexID, gl_InstanceID, and gl_PrimitiveID values,
* the format is ignored and the value is always int.
*/
OUT_BATCH(dw0);
OUT_BATCH(dw1);
}
CACHED_BATCH();
}
static void brw_emit_vertices(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
GLuint i, nr_elements;
brw_prepare_vertices(brw);
brw_emit_query_begin(brw);
nr_elements = brw->vb.nr_enabled + brw->vs.prog_data->uses_vertexid;
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
* VERTEX_ELEMENT struct and bail.
*
* The stale VB state stays in place, but they don't do anything unless
* a VE loads from them.
*/
if (nr_elements == 0) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | 1);
if (brw->gen >= 6) {
OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
}
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
ADVANCE_BATCH();
return;
}
/* Now emit VB and VEP state packets.
*/
if (brw->vb.nr_buffers) {
if (brw->gen >= 6) {
assert(brw->vb.nr_buffers <= 33);
} else {
assert(brw->vb.nr_buffers <= 17);
}
BEGIN_BATCH(1 + 4*brw->vb.nr_buffers);
OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4*brw->vb.nr_buffers - 1));
for (i = 0; i < brw->vb.nr_buffers; i++) {
struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
uint32_t dw0;
if (brw->gen >= 6) {
dw0 = buffer->step_rate
? GEN6_VB0_ACCESS_INSTANCEDATA
: GEN6_VB0_ACCESS_VERTEXDATA;
dw0 |= i << GEN6_VB0_INDEX_SHIFT;
} else {
dw0 = buffer->step_rate
? BRW_VB0_ACCESS_INSTANCEDATA
: BRW_VB0_ACCESS_VERTEXDATA;
dw0 |= i << BRW_VB0_INDEX_SHIFT;
}
if (brw->gen >= 7)
dw0 |= GEN7_VB0_ADDRESS_MODIFYENABLE;
if (brw->gen == 7)
dw0 |= GEN7_MOCS_L3 << 16;
WARN_ONCE(buffer->stride >= (brw->gen >= 5 ? 2048 : 2047),
"VBO stride %d too large, bad rendering may occur\n",
buffer->stride);
OUT_BATCH(dw0 | (buffer->stride << BRW_VB0_PITCH_SHIFT));
OUT_RELOC(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->offset);
if (brw->gen >= 5) {
OUT_RELOC(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->bo->size - 1);
} else
OUT_BATCH(0);
OUT_BATCH(buffer->step_rate);
}
ADVANCE_BATCH();
}
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS, presumably
* for VertexID/InstanceID.
*/
if (brw->gen >= 6) {
assert(nr_elements <= 34);
} else {
assert(nr_elements <= 18);
}
struct brw_vertex_element *gen6_edgeflag_input = NULL;
BEGIN_BATCH(1 + nr_elements * 2);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
for (i = 0; i < brw->vb.nr_enabled; i++) {
struct brw_vertex_element *input = brw->vb.enabled[i];
uint32_t format = brw_get_vertex_surface_type(brw, input->glarray);
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
if (input == &brw->vb.inputs[VERT_ATTRIB_EDGEFLAG]) {
/* Gen6+ passes edgeflag as sideband along with the vertex, instead
* of in the VUE. We have to upload it sideband as the last vertex
* element according to the B-Spec.
*/
if (brw->gen >= 6) {
gen6_edgeflag_input = input;
continue;
}
}
switch (input->glarray->Size) {
case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
: BRW_VE1_COMPONENT_STORE_1_FLT;
break;
}
if (brw->gen >= 6) {
OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((input->buffer << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
}
if (brw->gen >= 5)
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT));
else
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
}
if (brw->gen >= 6 && gen6_edgeflag_input) {
uint32_t format =
brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
GEN6_VE0_EDGE_FLAG_ENABLE |
(format << BRW_VE0_FORMAT_SHIFT) |
(gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
if (brw->vs.prog_data->uses_vertexid) {
uint32_t dw0 = 0, dw1 = 0;
dw1 = ((BRW_VE1_COMPONENT_STORE_VID << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_IID << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
if (brw->gen >= 6) {
dw0 |= GEN6_VE0_VALID;
} else {
dw0 |= BRW_VE0_VALID;
dw1 |= (i * 4) << BRW_VE1_DST_OFFSET_SHIFT;
}
/* Note that for gl_VertexID, gl_InstanceID, and gl_PrimitiveID values,
* the format is ignored and the value is always int.
*/
OUT_BATCH(dw0);
OUT_BATCH(dw1);
}
ADVANCE_BATCH();
}
static void brw_emit_vertices(struct brw_context *brw)
{
GLuint i;
brw_prepare_vertices(brw);
brw_prepare_shader_draw_parameters(brw);
brw_emit_query_begin(brw);
unsigned nr_elements = brw->vb.nr_enabled;
if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid)
++nr_elements;
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
* VERTEX_ELEMENT struct and bail.
*
* The stale VB state stays in place, but they don't do anything unless
* a VE loads from them.
*/
if (nr_elements == 0) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | 1);
if (brw->gen >= 6) {
OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
}
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
ADVANCE_BATCH();
return;
}
/* Now emit VB and VEP state packets.
*/
unsigned nr_buffers =
brw->vb.nr_buffers + brw->vs.prog_data->uses_vertexid;
if (nr_buffers) {
if (brw->gen >= 6) {
assert(nr_buffers <= 33);
} else {
assert(nr_buffers <= 17);
}
BEGIN_BATCH(1 + 4 * nr_buffers);
OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4 * nr_buffers - 1));
for (i = 0; i < brw->vb.nr_buffers; i++) {
struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
emit_vertex_buffer_state(brw, i, buffer->bo, buffer->bo->size - 1,
buffer->offset, buffer->stride,
buffer->step_rate);
}
if (brw->vs.prog_data->uses_vertexid) {
emit_vertex_buffer_state(brw, brw->vb.nr_buffers,
brw->draw.draw_params_bo,
brw->draw.draw_params_bo->size - 1,
brw->draw.draw_params_offset,
0, /* stride */
0); /* step rate */
}
ADVANCE_BATCH();
}
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS, presumably
* for VertexID/InstanceID.
*/
if (brw->gen >= 6) {
assert(nr_elements <= 34);
} else {
assert(nr_elements <= 18);
}
struct brw_vertex_element *gen6_edgeflag_input = NULL;
BEGIN_BATCH(1 + nr_elements * 2);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
for (i = 0; i < brw->vb.nr_enabled; i++) {
struct brw_vertex_element *input = brw->vb.enabled[i];
uint32_t format = brw_get_vertex_surface_type(brw, input->glarray);
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
if (input == &brw->vb.inputs[VERT_ATTRIB_EDGEFLAG]) {
/* Gen6+ passes edgeflag as sideband along with the vertex, instead
* of in the VUE. We have to upload it sideband as the last vertex
* element according to the B-Spec.
*/
if (brw->gen >= 6) {
gen6_edgeflag_input = input;
continue;
}
}
switch (input->glarray->Size) {
case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
: BRW_VE1_COMPONENT_STORE_1_FLT;
break;
}
if (brw->gen >= 6) {
OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((input->buffer << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
}
if (brw->gen >= 5)
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT));
else
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
}
if (brw->gen >= 6 && gen6_edgeflag_input) {
uint32_t format =
brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
GEN6_VE0_EDGE_FLAG_ENABLE |
(format << BRW_VE0_FORMAT_SHIFT) |
(gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid) {
uint32_t dw0 = 0, dw1 = 0;
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_0;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_0;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_0;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_0;
if (brw->vs.prog_data->uses_vertexid) {
comp0 = BRW_VE1_COMPONENT_STORE_SRC;
comp2 = BRW_VE1_COMPONENT_STORE_VID;
}
if (brw->vs.prog_data->uses_instanceid) {
comp3 = BRW_VE1_COMPONENT_STORE_IID;
}
dw1 = (comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT);
if (brw->gen >= 6) {
dw0 |= GEN6_VE0_VALID |
brw->vb.nr_buffers << GEN6_VE0_INDEX_SHIFT |
BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT;
} else {
dw0 |= BRW_VE0_VALID |
brw->vb.nr_buffers << BRW_VE0_INDEX_SHIFT |
BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT;
dw1 |= (i * 4) << BRW_VE1_DST_OFFSET_SHIFT;
}
/* Note that for gl_VertexID, gl_InstanceID, and gl_PrimitiveID values,
* the format is ignored and the value is always int.
*/
OUT_BATCH(dw0);
OUT_BATCH(dw1);
}
ADVANCE_BATCH();
}
static void brw_emit_vertices(struct brw_context *brw)
{
struct gl_context *ctx = &brw->intel.ctx;
struct intel_context *intel = intel_context(ctx);
GLuint i, nr_elements;
brw_prepare_vertices(brw);
brw_emit_query_begin(brw);
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
* VERTEX_ELEMENT struct and bail.
*
* The stale VB state stays in place, but they don't do anything unless
* a VE loads from them.
*/
if (brw->vb.nr_enabled == 0) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | 1);
if (intel->gen >= 6) {
OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((0 << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
}
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
CACHED_BATCH();
return;
}
/* Now emit VB and VEP state packets.
*/
if (brw->vb.nr_buffers) {
if (intel->gen >= 6) {
assert(brw->vb.nr_buffers <= 33);
} else {
assert(brw->vb.nr_buffers <= 17);
}
BEGIN_BATCH(1 + 4*brw->vb.nr_buffers);
OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4*brw->vb.nr_buffers - 1));
for (i = 0; i < brw->vb.nr_buffers; i++) {
struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
uint32_t dw0;
if (intel->gen >= 6) {
dw0 = buffer->step_rate
? GEN6_VB0_ACCESS_INSTANCEDATA
: GEN6_VB0_ACCESS_VERTEXDATA;
dw0 |= i << GEN6_VB0_INDEX_SHIFT;
} else {
dw0 = buffer->step_rate
? BRW_VB0_ACCESS_INSTANCEDATA
: BRW_VB0_ACCESS_VERTEXDATA;
dw0 |= i << BRW_VB0_INDEX_SHIFT;
}
if (intel->gen >= 7)
dw0 |= GEN7_VB0_ADDRESS_MODIFYENABLE;
OUT_BATCH(dw0 | (buffer->stride << BRW_VB0_PITCH_SHIFT));
OUT_RELOC(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->offset);
if (intel->gen >= 5) {
OUT_RELOC(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->bo->size - 1);
} else
OUT_BATCH(0);
OUT_BATCH(buffer->step_rate);
brw->vb.current_buffers[i].handle = buffer->bo->handle;
brw->vb.current_buffers[i].offset = buffer->offset;
brw->vb.current_buffers[i].stride = buffer->stride;
brw->vb.current_buffers[i].step_rate = buffer->step_rate;
}
brw->vb.nr_current_buffers = i;
ADVANCE_BATCH();
}
nr_elements = brw->vb.nr_enabled + brw->vs.prog_data->uses_vertexid;
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS, presumably
* for VertexID/InstanceID.
*/
if (intel->gen >= 6) {
assert(nr_elements <= 34);
} else {
assert(nr_elements <= 18);
}
struct brw_vertex_element *gen6_edgeflag_input = NULL;
BEGIN_BATCH(1 + nr_elements * 2);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
for (i = 0; i < brw->vb.nr_enabled; i++) {
struct brw_vertex_element *input = brw->vb.enabled[i];
uint32_t format = get_surface_type(input->glarray->Type,
input->glarray->Size,
input->glarray->Format,
input->glarray->Normalized,
input->glarray->Integer);
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
/* The gen4 driver expects edgeflag to come in as a float, and passes
* that float on to the tests in the clipper. Mesa's current vertex
* attribute value for EdgeFlag is stored as a float, which works out.
* glEdgeFlagPointer, on the other hand, gives us an unnormalized
* integer ubyte. Just rewrite that to convert to a float.
*/
if (input->attrib == VERT_ATTRIB_EDGEFLAG) {
/* Gen6+ passes edgeflag as sideband along with the vertex, instead
* of in the VUE. We have to upload it sideband as the last vertex
* element according to the B-Spec.
*/
if (intel->gen >= 6) {
gen6_edgeflag_input = input;
continue;
}
if (format == BRW_SURFACEFORMAT_R8_UINT)
format = BRW_SURFACEFORMAT_R8_SSCALED;
}
switch (input->glarray->Size) {
case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
: BRW_VE1_COMPONENT_STORE_1_FLT;
break;
}
if (intel->gen >= 6) {
OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
} else {
OUT_BATCH((input->buffer << BRW_VE0_INDEX_SHIFT) |
BRW_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
}
if (intel->gen >= 5)
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT));
else
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT) |
((i * 4) << BRW_VE1_DST_OFFSET_SHIFT));
}
if (intel->gen >= 6 && gen6_edgeflag_input) {
uint32_t format = get_surface_type(gen6_edgeflag_input->glarray->Type,
gen6_edgeflag_input->glarray->Size,
gen6_edgeflag_input->glarray->Format,
gen6_edgeflag_input->glarray->Normalized,
gen6_edgeflag_input->glarray->Integer);
OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
GEN6_VE0_EDGE_FLAG_ENABLE |
(format << BRW_VE0_FORMAT_SHIFT) |
(gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
if (brw->vs.prog_data->uses_vertexid) {
uint32_t dw0 = 0, dw1 = 0;
dw1 = ((BRW_VE1_COMPONENT_STORE_VID << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_IID << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
if (intel->gen >= 6) {
dw0 |= GEN6_VE0_VALID;
} else {
dw0 |= BRW_VE0_VALID;
dw1 |= (i * 4) << BRW_VE1_DST_OFFSET_SHIFT;
}
/* Note that for gl_VertexID, gl_InstanceID, and gl_PrimitiveID values,
* the format is ignored and the value is always int.
*/
OUT_BATCH(dw0);
OUT_BATCH(dw1);
}
CACHED_BATCH();
}
/* May fail if out of video memory for texture or vbo upload, or on
* fallback conditions.
*/
static GLboolean brw_try_draw_prims( GLcontext *ctx,
const struct gl_client_array *arrays[],
const struct _mesa_prim *prim,
GLuint nr_prims,
const struct _mesa_index_buffer *ib,
GLuint min_index,
GLuint max_index )
{
struct intel_context *intel = intel_context(ctx);
struct brw_context *brw = brw_context(ctx);
GLboolean retval = GL_FALSE;
GLuint i;
GLuint ib_offset;
dri_bo *ib_bo;
GLboolean force_flush = GL_FALSE;
int ret;
if (ctx->NewState)
_mesa_update_state( ctx );
brw_validate_textures( brw );
/* Bind all inputs, derive varying and size information:
*/
brw_merge_inputs( brw, arrays );
/* Have to validate state quite late. Will rebuild tnl_program,
* which depends on varying information.
*
* Note this is where brw->vs->prog_data.inputs_read is calculated,
* so can't access it earlier.
*/
LOCK_HARDWARE(intel);
if (brw->intel.numClipRects == 0) {
UNLOCK_HARDWARE(intel);
return GL_TRUE;
}
{
/* Flush the batch if it's approaching full, so that we don't wrap while
* we've got validated state that needs to be in the same batch as the
* primitives. This fraction is just a guess (minimal full state plus
* a primitive is around 512 bytes), and would be better if we had
* an upper bound of how much we might emit in a single
* brw_try_draw_prims().
*/
flush:
if (force_flush)
brw->no_batch_wrap = GL_FALSE;
if (intel->batch->ptr - intel->batch->map > intel->batch->size * 3 / 4
/* brw_emit_prim may change the cliprect_mode to LOOP_CLIPRECTS */
|| intel->batch->cliprect_mode != LOOP_CLIPRECTS || (force_flush == GL_TRUE))
intel_batchbuffer_flush(intel->batch);
force_flush = GL_FALSE;
brw->no_batch_wrap = GL_TRUE;
/* Set the first primitive early, ahead of validate_state:
*/
brw_set_prim(brw, prim[0].mode, &force_flush);
/* XXX: Need to separate validate and upload of state.
*/
ret = brw_validate_state( brw );
if (ret) {
force_flush = GL_TRUE;
goto flush;
}
/* Various fallback checks:
*/
if (brw->intel.Fallback)
goto out;
if (check_fallbacks( brw, prim, nr_prims ))
goto out;
/* need to account for index buffer and vertex buffer */
if (ib) {
ret = brw_prepare_indices( brw, ib , &ib_bo, &ib_offset);
if (ret) {
force_flush = GL_TRUE;
goto flush;
}
}
ret = brw_prepare_vertices( brw, min_index, max_index);
if (ret < 0)
goto out;
if (ret > 0) {
force_flush = GL_TRUE;
goto flush;
}
/* Upload index, vertex data:
*/
if (ib)
brw_emit_indices( brw, ib, ib_bo, ib_offset);
brw_emit_vertices( brw, min_index, max_index);
for (i = 0; i < nr_prims; i++) {
brw_emit_prim(brw, &prim[i]);
}
retval = GL_TRUE;
}
out:
brw->no_batch_wrap = GL_FALSE;
UNLOCK_HARDWARE(intel);
if (!retval)
DBG("%s failed\n", __FUNCTION__);
return retval;
}
static void
gen8_emit_vertices(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
uint32_t mocs_wb = brw->gen >= 9 ? SKL_MOCS_WB : BDW_MOCS_WB;
bool uses_edge_flag;
brw_prepare_vertices(brw);
brw_prepare_shader_draw_parameters(brw);
uses_edge_flag = (ctx->Polygon.FrontMode != GL_FILL ||
ctx->Polygon.BackMode != GL_FILL);
if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid) {
unsigned vue = brw->vb.nr_enabled;
/* The element for the edge flags must always be last, so we have to
* insert the SGVS before it in that case.
*/
if (uses_edge_flag) {
assert(vue > 0);
vue--;
}
WARN_ONCE(vue >= 33,
"Trying to insert VID/IID past 33rd vertex element, "
"need to reorder the vertex attrbutes.");
unsigned dw1 = 0;
if (brw->vs.prog_data->uses_vertexid) {
dw1 |= GEN8_SGVS_ENABLE_VERTEX_ID |
(2 << GEN8_SGVS_VERTEX_ID_COMPONENT_SHIFT) | /* .z channel */
(vue << GEN8_SGVS_VERTEX_ID_ELEMENT_OFFSET_SHIFT);
}
if (brw->vs.prog_data->uses_instanceid) {
dw1 |= GEN8_SGVS_ENABLE_INSTANCE_ID |
(3 << GEN8_SGVS_INSTANCE_ID_COMPONENT_SHIFT) | /* .w channel */
(vue << GEN8_SGVS_INSTANCE_ID_ELEMENT_OFFSET_SHIFT);
}
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
OUT_BATCH(dw1);
ADVANCE_BATCH();
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
OUT_BATCH(vue | GEN8_VF_INSTANCING_ENABLE);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
* VERTEX_ELEMENT struct and bail.
*
* The stale VB state stays in place, but they don't do anything unless
* a VE loads from them.
*/
if (brw->vb.nr_enabled == 0) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (3 - 2));
OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
ADVANCE_BATCH();
return;
}
/* Now emit 3DSTATE_VERTEX_BUFFERS and 3DSTATE_VERTEX_ELEMENTS packets. */
unsigned nr_buffers = brw->vb.nr_buffers + brw->vs.prog_data->uses_vertexid;
if (nr_buffers) {
assert(nr_buffers <= 33);
BEGIN_BATCH(1 + 4 * nr_buffers);
OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4 * nr_buffers - 1));
for (unsigned i = 0; i < brw->vb.nr_buffers; i++) {
struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
uint32_t dw0 = 0;
dw0 |= i << GEN6_VB0_INDEX_SHIFT;
dw0 |= GEN7_VB0_ADDRESS_MODIFYENABLE;
dw0 |= buffer->stride << BRW_VB0_PITCH_SHIFT;
dw0 |= mocs_wb << 16;
OUT_BATCH(dw0);
OUT_RELOC64(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->offset);
OUT_BATCH(buffer->bo->size);
}
if (brw->vs.prog_data->uses_vertexid) {
OUT_BATCH(brw->vb.nr_buffers << GEN6_VB0_INDEX_SHIFT |
GEN7_VB0_ADDRESS_MODIFYENABLE |
mocs_wb << 16);
OUT_RELOC64(brw->draw.draw_params_bo, I915_GEM_DOMAIN_VERTEX, 0,
brw->draw.draw_params_offset);
OUT_BATCH(brw->draw.draw_params_bo->size);
}
ADVANCE_BATCH();
}
/* Normally we don't need an element for the SGVS attribute because the
* 3DSTATE_VF_SGVS instruction lets you store the generated attribute in an
* element that is past the list in 3DSTATE_VERTEX_ELEMENTS. However if the
* vertex ID is used then it needs an element for the base vertex buffer.
* Additionally if there is an edge flag element then the SGVS can't be
* inserted past that so we need a dummy element to ensure that the edge
* flag is the last one.
*/
bool needs_sgvs_element = (brw->vs.prog_data->uses_vertexid ||
(brw->vs.prog_data->uses_instanceid &&
uses_edge_flag));
unsigned nr_elements = brw->vb.nr_enabled + needs_sgvs_element;
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS,
* presumably for VertexID/InstanceID.
*/
assert(nr_elements <= 34);
struct brw_vertex_element *gen6_edgeflag_input = NULL;
BEGIN_BATCH(1 + nr_elements * 2);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
for (unsigned i = 0; i < brw->vb.nr_enabled; i++) {
struct brw_vertex_element *input = brw->vb.enabled[i];
uint32_t format = brw_get_vertex_surface_type(brw, input->glarray);
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
/* The gen4 driver expects edgeflag to come in as a float, and passes
* that float on to the tests in the clipper. Mesa's current vertex
* attribute value for EdgeFlag is stored as a float, which works out.
* glEdgeFlagPointer, on the other hand, gives us an unnormalized
* integer ubyte. Just rewrite that to convert to a float.
*/
if (input == &brw->vb.inputs[VERT_ATTRIB_EDGEFLAG]) {
/* Gen6+ passes edgeflag as sideband along with the vertex, instead
* of in the VUE. We have to upload it sideband as the last vertex
* element according to the B-Spec.
*/
gen6_edgeflag_input = input;
continue;
}
switch (input->glarray->Size) {
case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
: BRW_VE1_COMPONENT_STORE_1_FLT;
break;
}
OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT));
}
if (needs_sgvs_element) {
if (brw->vs.prog_data->uses_vertexid) {
OUT_BATCH(GEN6_VE0_VALID |
brw->vb.nr_buffers << GEN6_VE0_INDEX_SHIFT |
BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT);
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
} else {
OUT_BATCH(GEN6_VE0_VALID);
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
}
if (gen6_edgeflag_input) {
uint32_t format =
brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
GEN6_VE0_EDGE_FLAG_ENABLE |
(format << BRW_VE0_FORMAT_SHIFT) |
(gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
ADVANCE_BATCH();
for (unsigned i = 0, j = 0; i < brw->vb.nr_enabled; i++) {
const struct brw_vertex_element *input = brw->vb.enabled[i];
const struct brw_vertex_buffer *buffer = &brw->vb.buffers[input->buffer];
unsigned element_index;
/* The edge flag element is reordered to be the last one in the code
* above so we need to compensate for that in the element indices used
* below.
*/
if (input == gen6_edgeflag_input)
element_index = nr_elements - 1;
else
element_index = j++;
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
OUT_BATCH(element_index |
(buffer->step_rate ? GEN8_VF_INSTANCING_ENABLE : 0));
OUT_BATCH(buffer->step_rate);
ADVANCE_BATCH();
}
}
static void
gen8_emit_vertices(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
bool uses_edge_flag;
brw_prepare_vertices(brw);
brw_prepare_shader_draw_parameters(brw);
uses_edge_flag = (ctx->Polygon.FrontMode != GL_FILL ||
ctx->Polygon.BackMode != GL_FILL);
if (brw->vs.prog_data->uses_vertexid || brw->vs.prog_data->uses_instanceid) {
unsigned vue = brw->vb.nr_enabled;
/* The element for the edge flags must always be last, so we have to
* insert the SGVS before it in that case.
*/
if (uses_edge_flag) {
assert(vue > 0);
vue--;
}
WARN_ONCE(vue >= 33,
"Trying to insert VID/IID past 33rd vertex element, "
"need to reorder the vertex attrbutes.");
unsigned dw1 = 0;
if (brw->vs.prog_data->uses_vertexid) {
dw1 |= GEN8_SGVS_ENABLE_VERTEX_ID |
(2 << GEN8_SGVS_VERTEX_ID_COMPONENT_SHIFT) | /* .z channel */
(vue << GEN8_SGVS_VERTEX_ID_ELEMENT_OFFSET_SHIFT);
}
if (brw->vs.prog_data->uses_instanceid) {
dw1 |= GEN8_SGVS_ENABLE_INSTANCE_ID |
(3 << GEN8_SGVS_INSTANCE_ID_COMPONENT_SHIFT) | /* .w channel */
(vue << GEN8_SGVS_INSTANCE_ID_ELEMENT_OFFSET_SHIFT);
}
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
OUT_BATCH(dw1);
ADVANCE_BATCH();
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
OUT_BATCH(vue | GEN8_VF_INSTANCING_ENABLE);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(2);
OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
OUT_BATCH(0);
ADVANCE_BATCH();
}
/* If the VS doesn't read any inputs (calculating vertex position from
* a state variable for some reason, for example), emit a single pad
* VERTEX_ELEMENT struct and bail.
*
* The stale VB state stays in place, but they don't do anything unless
* a VE loads from them.
*/
if (brw->vb.nr_enabled == 0) {
BEGIN_BATCH(3);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (3 - 2));
OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
(0 << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
ADVANCE_BATCH();
return;
}
/* Now emit 3DSTATE_VERTEX_BUFFERS and 3DSTATE_VERTEX_ELEMENTS packets. */
const bool uses_draw_params =
brw->vs.prog_data->uses_basevertex ||
brw->vs.prog_data->uses_baseinstance;
const unsigned nr_buffers = brw->vb.nr_buffers +
uses_draw_params + brw->vs.prog_data->uses_drawid;
if (nr_buffers) {
assert(nr_buffers <= 33);
BEGIN_BATCH(1 + 4 * nr_buffers);
OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4 * nr_buffers - 1));
for (unsigned i = 0; i < brw->vb.nr_buffers; i++) {
const struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
EMIT_VERTEX_BUFFER_STATE(brw, i, buffer->bo,
buffer->offset,
buffer->offset + buffer->size,
buffer->stride, 0 /* unused */);
}
if (uses_draw_params) {
EMIT_VERTEX_BUFFER_STATE(brw, brw->vb.nr_buffers,
brw->draw.draw_params_bo,
brw->draw.draw_params_offset,
brw->draw.draw_params_bo->size,
0 /* stride */,
0 /* unused */);
}
if (brw->vs.prog_data->uses_drawid) {
EMIT_VERTEX_BUFFER_STATE(brw, brw->vb.nr_buffers + 1,
brw->draw.draw_id_bo,
brw->draw.draw_id_offset,
brw->draw.draw_id_bo->size,
0 /* stride */,
0 /* unused */);
}
ADVANCE_BATCH();
}
/* Normally we don't need an element for the SGVS attribute because the
* 3DSTATE_VF_SGVS instruction lets you store the generated attribute in an
* element that is past the list in 3DSTATE_VERTEX_ELEMENTS. However if
* we're using draw parameters then we need an element for the those
* values. Additionally if there is an edge flag element then the SGVS
* can't be inserted past that so we need a dummy element to ensure that
* the edge flag is the last one.
*/
const bool needs_sgvs_element = (brw->vs.prog_data->uses_basevertex ||
brw->vs.prog_data->uses_baseinstance ||
((brw->vs.prog_data->uses_instanceid ||
brw->vs.prog_data->uses_vertexid) &&
uses_edge_flag));
const unsigned nr_elements =
brw->vb.nr_enabled + needs_sgvs_element + brw->vs.prog_data->uses_drawid;
/* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS,
* presumably for VertexID/InstanceID.
*/
assert(nr_elements <= 34);
struct brw_vertex_element *gen6_edgeflag_input = NULL;
BEGIN_BATCH(1 + nr_elements * 2);
OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
for (unsigned i = 0; i < brw->vb.nr_enabled; i++) {
struct brw_vertex_element *input = brw->vb.enabled[i];
uint32_t format = brw_get_vertex_surface_type(brw, input->glarray);
uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
/* From the BDW PRM, Volume 2d, page 588 (VERTEX_ELEMENT_STATE):
* "Any SourceElementFormat of *64*_PASSTHRU cannot be used with an
* element which has edge flag enabled."
*/
assert(!(is_passthru_format(format) && uses_edge_flag));
/* The gen4 driver expects edgeflag to come in as a float, and passes
* that float on to the tests in the clipper. Mesa's current vertex
* attribute value for EdgeFlag is stored as a float, which works out.
* glEdgeFlagPointer, on the other hand, gives us an unnormalized
* integer ubyte. Just rewrite that to convert to a float.
*/
if (input == &brw->vb.inputs[VERT_ATTRIB_EDGEFLAG]) {
/* Gen6+ passes edgeflag as sideband along with the vertex, instead
* of in the VUE. We have to upload it sideband as the last vertex
* element according to the B-Spec.
*/
gen6_edgeflag_input = input;
continue;
}
switch (input->glarray->Size) {
case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
: BRW_VE1_COMPONENT_STORE_1_FLT;
break;
}
/* From the BDW PRM, Volume 2d, page 586 (VERTEX_ELEMENT_STATE):
*
* "When SourceElementFormat is set to one of the *64*_PASSTHRU
* formats, 64-bit components are stored in the URB without any
* conversion. In this case, vertex elements must be written as 128
* or 256 bits, with VFCOMP_STORE_0 being used to pad the output
* as required. E.g., if R64_PASSTHRU is used to copy a 64-bit Red
* component into the URB, Component 1 must be specified as
* VFCOMP_STORE_0 (with Components 2,3 set to VFCOMP_NOSTORE)
* in order to output a 128-bit vertex element, or Components 1-3 must
* be specified as VFCOMP_STORE_0 in order to output a 256-bit vertex
* element. Likewise, use of R64G64B64_PASSTHRU requires Component 3
* to be specified as VFCOMP_STORE_0 in order to output a 256-bit vertex
* element."
*/
if (input->glarray->Doubles) {
switch (input->glarray->Size) {
case 0:
case 1:
case 2:
/* Use 128-bits instead of 256-bits to write double and dvec2
* vertex elements.
*/
comp2 = BRW_VE1_COMPONENT_NOSTORE;
comp3 = BRW_VE1_COMPONENT_NOSTORE;
break;
case 3:
/* Pad the output using VFCOMP_STORE_0 as suggested
* by the BDW PRM.
*/
comp3 = BRW_VE1_COMPONENT_STORE_0;
break;
}
}
OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
(format << BRW_VE0_FORMAT_SHIFT) |
(input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
(comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
(comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
(comp3 << BRW_VE1_COMPONENT_3_SHIFT));
}
if (needs_sgvs_element) {
if (brw->vs.prog_data->uses_basevertex ||
brw->vs.prog_data->uses_baseinstance) {
OUT_BATCH(GEN6_VE0_VALID |
brw->vb.nr_buffers << GEN6_VE0_INDEX_SHIFT |
BRW_SURFACEFORMAT_R32G32_UINT << BRW_VE0_FORMAT_SHIFT);
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
} else {
OUT_BATCH(GEN6_VE0_VALID);
OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
}
if (brw->vs.prog_data->uses_drawid) {
OUT_BATCH(GEN6_VE0_VALID |
((brw->vb.nr_buffers + 1) << GEN6_VE0_INDEX_SHIFT) |
(BRW_SURFACEFORMAT_R32_UINT << BRW_VE0_FORMAT_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
if (gen6_edgeflag_input) {
uint32_t format =
brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
GEN6_VE0_VALID |
GEN6_VE0_EDGE_FLAG_ENABLE |
(format << BRW_VE0_FORMAT_SHIFT) |
(gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
(BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
}
ADVANCE_BATCH();
for (unsigned i = 0, j = 0; i < brw->vb.nr_enabled; i++) {
const struct brw_vertex_element *input = brw->vb.enabled[i];
const struct brw_vertex_buffer *buffer = &brw->vb.buffers[input->buffer];
unsigned element_index;
/* The edge flag element is reordered to be the last one in the code
* above so we need to compensate for that in the element indices used
* below.
*/
if (input == gen6_edgeflag_input)
element_index = nr_elements - 1;
else
element_index = j++;
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
OUT_BATCH(element_index |
(buffer->step_rate ? GEN8_VF_INSTANCING_ENABLE : 0));
OUT_BATCH(buffer->step_rate);
ADVANCE_BATCH();
}
if (brw->vs.prog_data->uses_drawid) {
const unsigned element = brw->vb.nr_enabled + needs_sgvs_element;
BEGIN_BATCH(3);
OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
OUT_BATCH(element);
OUT_BATCH(0);
ADVANCE_BATCH();
}
}
