task->bin = bin;
task->x = x * TILE_SIZE;
task->y = y * TILE_SIZE;
+ task->width = TILE_SIZE + x * TILE_SIZE > task->scene->fb.width ?
+ task->scene->fb.width - x * TILE_SIZE : TILE_SIZE;
+ task->height = TILE_SIZE + y * TILE_SIZE > task->scene->fb.height ?
+ task->scene->fb.height - y * TILE_SIZE : TILE_SIZE;
/* reset pointers to color and depth tile(s) */
memset(task->color_tiles, 0, sizeof(task->color_tiles));
/**
* Clear the rasterizer's current color tile.
* This is a bin command called during bin processing.
+ * Clear commands always clear all bound layers.
*/
static void
lp_rast_clear_color(struct lp_rasterizer_task *task,
util_format_write_4ui(format, arg.clear_color.ui, 0, &uc, 0, 0, 0, 1, 1);
}
- util_fill_rect(scene->cbufs[i].map,
- scene->fb.cbufs[i]->format,
- scene->cbufs[i].stride,
- task->x,
- task->y,
- TILE_SIZE,
- TILE_SIZE,
- &uc);
+ util_fill_box(scene->cbufs[i].map,
+ format,
+ scene->cbufs[i].stride,
+ scene->cbufs[i].layer_stride,
+ task->x,
+ task->y,
+ 0,
+ task->width,
+ task->height,
+ scene->fb_max_layer + 1,
+ &uc);
}
}
else {
clear_color[3]);
for (i = 0; i < scene->fb.nr_cbufs; i++) {
-
util_pack_color(arg.clear_color.f,
scene->fb.cbufs[i]->format, &uc);
- util_fill_rect(scene->cbufs[i].map,
- scene->fb.cbufs[i]->format,
- scene->cbufs[i].stride,
- task->x,
- task->y,
- TILE_SIZE,
- TILE_SIZE,
- &uc);
+ util_fill_box(scene->cbufs[i].map,
+ scene->fb.cbufs[i]->format,
+ scene->cbufs[i].stride,
+ scene->cbufs[i].layer_stride,
+ task->x,
+ task->y,
+ 0,
+ task->width,
+ task->height,
+ scene->fb_max_layer + 1,
+ &uc);
}
}
}
/**
* Clear the rasterizer's current z/stencil tile.
* This is a bin command called during bin processing.
+ * Clear commands always clear all bound layers.
*/
static void
lp_rast_clear_zstencil(struct lp_rasterizer_task *task,
uint64_t clear_mask64 = arg.clear_zstencil.mask;
uint32_t clear_value = (uint32_t) clear_value64;
uint32_t clear_mask = (uint32_t) clear_mask64;
- const unsigned height = TILE_SIZE;
- const unsigned width = TILE_SIZE;
- const unsigned block_size = scene->zsbuf.blocksize;
+ const unsigned height = task->height;
+ const unsigned width = task->width;
const unsigned dst_stride = scene->zsbuf.stride;
uint8_t *dst;
unsigned i, j;
+ unsigned block_size;
LP_DBG(DEBUG_RAST, "%s: value=0x%08x, mask=0x%08x\n",
__FUNCTION__, clear_value, clear_mask);
*/
if (scene->fb.zsbuf) {
-
- dst = lp_rast_get_unswizzled_depth_tile_pointer(task, LP_TEX_USAGE_READ_WRITE);
+ unsigned layer;
+ uint8_t *dst_layer = lp_rast_get_unswizzled_depth_tile_pointer(task, LP_TEX_USAGE_READ_WRITE);
+ block_size = util_format_get_blocksize(scene->fb.zsbuf->format);
clear_value &= clear_mask;
- switch (block_size) {
- case 1:
- assert(clear_mask == 0xff);
- memset(dst, (uint8_t) clear_value, height * width);
- break;
- case 2:
- if (clear_mask == 0xffff) {
- for (i = 0; i < height; i++) {
- uint16_t *row = (uint16_t *)dst;
- for (j = 0; j < width; j++)
- *row++ = (uint16_t) clear_value;
- dst += dst_stride;
+ for (layer = 0; layer <= scene->fb_max_layer; layer++) {
+ dst = dst_layer;
+
+ switch (block_size) {
+ case 1:
+ assert(clear_mask == 0xff);
+ memset(dst, (uint8_t) clear_value, height * width);
+ break;
+ case 2:
+ if (clear_mask == 0xffff) {
+ for (i = 0; i < height; i++) {
+ uint16_t *row = (uint16_t *)dst;
+ for (j = 0; j < width; j++)
+ *row++ = (uint16_t) clear_value;
+ dst += dst_stride;
+ }
}
- }
- else {
- for (i = 0; i < height; i++) {
- uint16_t *row = (uint16_t *)dst;
- for (j = 0; j < width; j++) {
- uint16_t tmp = ~clear_mask & *row;
- *row++ = clear_value | tmp;
+ else {
+ for (i = 0; i < height; i++) {
+ uint16_t *row = (uint16_t *)dst;
+ for (j = 0; j < width; j++) {
+ uint16_t tmp = ~clear_mask & *row;
+ *row++ = clear_value | tmp;
+ }
+ dst += dst_stride;
}
- dst += dst_stride;
}
- }
- break;
- case 4:
- if (clear_mask == 0xffffffff) {
- for (i = 0; i < height; i++) {
- uint32_t *row = (uint32_t *)dst;
- for (j = 0; j < width; j++)
- *row++ = clear_value;
- dst += dst_stride;
+ break;
+ case 4:
+ if (clear_mask == 0xffffffff) {
+ for (i = 0; i < height; i++) {
+ uint32_t *row = (uint32_t *)dst;
+ for (j = 0; j < width; j++)
+ *row++ = clear_value;
+ dst += dst_stride;
+ }
}
- }
- else {
- for (i = 0; i < height; i++) {
- uint32_t *row = (uint32_t *)dst;
- for (j = 0; j < width; j++) {
- uint32_t tmp = ~clear_mask & *row;
- *row++ = clear_value | tmp;
+ else {
+ for (i = 0; i < height; i++) {
+ uint32_t *row = (uint32_t *)dst;
+ for (j = 0; j < width; j++) {
+ uint32_t tmp = ~clear_mask & *row;
+ *row++ = clear_value | tmp;
+ }
+ dst += dst_stride;
}
- dst += dst_stride;
}
- }
- break;
- case 8:
- clear_value64 &= clear_mask64;
- if (clear_mask64 == 0xffffffffffULL) {
- for (i = 0; i < height; i++) {
- uint64_t *row = (uint64_t *)dst;
- for (j = 0; j < width; j++)
- *row++ = clear_value64;
- dst += dst_stride;
+ break;
+ case 8:
+ clear_value64 &= clear_mask64;
+ if (clear_mask64 == 0xffffffffffULL) {
+ for (i = 0; i < height; i++) {
+ uint64_t *row = (uint64_t *)dst;
+ for (j = 0; j < width; j++)
+ *row++ = clear_value64;
+ dst += dst_stride;
+ }
}
- }
- else {
- for (i = 0; i < height; i++) {
- uint64_t *row = (uint64_t *)dst;
- for (j = 0; j < width; j++) {
- uint64_t tmp = ~clear_mask64 & *row;
- *row++ = clear_value64 | tmp;
+ else {
+ for (i = 0; i < height; i++) {
+ uint64_t *row = (uint64_t *)dst;
+ for (j = 0; j < width; j++) {
+ uint64_t tmp = ~clear_mask64 & *row;
+ *row++ = clear_value64 | tmp;
+ }
+ dst += dst_stride;
}
- dst += dst_stride;
}
- }
- break;
+ break;
- default:
- assert(0);
- break;
+ default:
+ assert(0);
+ break;
+ }
+ dst_layer += scene->zsbuf.layer_stride;
}
}
}
variant = state->variant;
/* render the whole 64x64 tile in 4x4 chunks */
- for (y = 0; y < TILE_SIZE; y += 4){
- for (x = 0; x < TILE_SIZE; x += 4) {
+ for (y = 0; y < task->height; y += 4){
+ for (x = 0; x < task->width; x += 4) {
uint8_t *color[PIPE_MAX_COLOR_BUFS];
unsigned stride[PIPE_MAX_COLOR_BUFS];
uint8_t *depth = NULL;
/* color buffer */
for (i = 0; i < scene->fb.nr_cbufs; i++){
stride[i] = scene->cbufs[i].stride;
-
- color[i] = lp_rast_get_unswizzled_color_block_pointer(task, i, tile_x + x, tile_y + y);
+ color[i] = lp_rast_get_unswizzled_color_block_pointer(task, i, tile_x + x,
+ tile_y + y, inputs->layer);
}
/* depth buffer */
if (scene->zsbuf.map) {
- depth = lp_rast_get_unswizzled_depth_block_pointer(task, tile_x + x, tile_y + y);
+ depth = lp_rast_get_unswizzled_depth_block_pointer(task, tile_x + x,
+ tile_y + y, inputs->layer);
depth_stride = scene->zsbuf.stride;
}
-
/* run shader on 4x4 block */
BEGIN_JIT_CALL(state, task);
variant->jit_function[RAST_WHOLE]( &state->jit_context,
const struct lp_scene *scene = task->scene;
uint8_t *color[PIPE_MAX_COLOR_BUFS];
unsigned stride[PIPE_MAX_COLOR_BUFS];
- void *depth = NULL;
+ uint8_t *depth = NULL;
unsigned depth_stride = 0;
unsigned i;
/* color buffer */
for (i = 0; i < scene->fb.nr_cbufs; i++) {
stride[i] = scene->cbufs[i].stride;
-
- color[i] = lp_rast_get_unswizzled_color_block_pointer(task, i, x, y);
+ color[i] = lp_rast_get_unswizzled_color_block_pointer(task, i, x, y, inputs->layer);
}
/* depth buffer */
if (scene->zsbuf.map) {
depth_stride = scene->zsbuf.stride;
- depth = lp_rast_get_unswizzled_depth_block_pointer(task, x, y);
+ depth = lp_rast_get_unswizzled_depth_block_pointer(task, x, y, inputs->layer);
}
assert(lp_check_alignment(state->jit_context.u8_blend_color, 16));
- /* run shader on 4x4 block */
- BEGIN_JIT_CALL(state, task);
- variant->jit_function[RAST_EDGE_TEST](&state->jit_context,
- x, y,
- inputs->frontfacing,
- GET_A0(inputs),
- GET_DADX(inputs),
- GET_DADY(inputs),
- color,
- depth,
- mask,
- &task->thread_data,
- stride,
- depth_stride);
- END_JIT_CALL();
+ /*
+ * The rasterizer may produce fragments outside our
+ * allocated 4x4 blocks hence need to filter them out here.
+ */
+ if ((x % TILE_SIZE) < task->width && (y % TILE_SIZE) < task->height) {
+ if (task->query[PIPE_QUERY_PIPELINE_STATISTICS]) {
+ /* not very accurate would need a popcount on the mask */
+ task->ps_invocations++;
+ }
+ /* run shader on 4x4 block */
+ BEGIN_JIT_CALL(state, task);
+ variant->jit_function[RAST_EDGE_TEST](&state->jit_context,
+ x, y,
+ inputs->frontfacing,
+ GET_A0(inputs),
+ GET_DADX(inputs),
+ GET_DADY(inputs),
+ color,
+ depth,
+ mask,
+ &task->thread_data,
+ stride,
+ depth_stride);
+ END_JIT_CALL();
+ }
}
switch (pq->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
+ case PIPE_QUERY_OCCLUSION_PREDICATE:
task->thread_data.vis_counter = 0;
break;
+ case PIPE_QUERY_PIPELINE_STATISTICS:
+ task->ps_invocations = 0;
+ break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
case PIPE_QUERY_PRIMITIVES_EMITTED:
case PIPE_QUERY_SO_STATISTICS:
- case PIPE_QUERY_PIPELINE_STATISTICS:
case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
+ case PIPE_QUERY_TIMESTAMP_DISJOINT:
break;
default:
assert(0);
const union lp_rast_cmd_arg arg)
{
struct llvmpipe_query *pq = arg.query_obj;
- assert(task->query[pq->type] == pq || pq->type == PIPE_QUERY_TIMESTAMP);
+ assert(task->query[pq->type] == pq ||
+ pq->type == PIPE_QUERY_TIMESTAMP ||
+ pq->type == PIPE_QUERY_GPU_FINISHED);
switch (pq->type) {
case PIPE_QUERY_OCCLUSION_COUNTER:
+ case PIPE_QUERY_OCCLUSION_PREDICATE:
pq->count[task->thread_index] += task->thread_data.vis_counter;
break;
case PIPE_QUERY_TIMESTAMP:
pq->count[task->thread_index] = os_time_get_nano();
break;
+ case PIPE_QUERY_PIPELINE_STATISTICS:
+ pq->count[task->thread_index] += task->ps_invocations;
+ break;
case PIPE_QUERY_PRIMITIVES_GENERATED:
case PIPE_QUERY_PRIMITIVES_EMITTED:
case PIPE_QUERY_SO_STATISTICS:
- case PIPE_QUERY_PIPELINE_STATISTICS:
case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
+ case PIPE_QUERY_TIMESTAMP_DISJOINT:
+ case PIPE_QUERY_GPU_FINISHED:
break;
default:
assert(0);