#ifndef LP_RAST_PRIV_H
#define LP_RAST_PRIV_H
-#include "os/os_thread.h"
-#include "util/u_format.h"
+#include "util/format/u_format.h"
+#include "util/u_thread.h"
#include "gallivm/lp_bld_debug.h"
#include "lp_memory.h"
#include "lp_rast.h"
/** Non-interpolated passthru state and occlude counter for visible pixels */
struct lp_jit_thread_data thread_data;
- uint64_t ps_invocations;
- uint8_t ps_inv_multiplier;
pipe_semaphore work_ready;
pipe_semaphore work_done;
struct lp_rasterizer_task tasks[LP_MAX_THREADS];
unsigned num_threads;
- pipe_thread threads[LP_MAX_THREADS];
+ thrd_t threads[LP_MAX_THREADS];
/** For synchronizing the rasterization threads */
- pipe_barrier barrier;
+ util_barrier barrier;
};
-
+void
+lp_rast_shade_quads_mask_sample(struct lp_rasterizer_task *task,
+ const struct lp_rast_shader_inputs *inputs,
+ unsigned x, unsigned y,
+ uint64_t mask);
void
lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
const struct lp_rast_shader_inputs *inputs,
unsigned mask);
-
-/**
- * Get pointer to the unswizzled color tile
- */
-static INLINE uint8_t *
-lp_rast_get_unswizzled_color_tile_pointer(struct lp_rasterizer_task *task,
- unsigned buf, enum lp_texture_usage usage)
-{
- const struct lp_scene *scene = task->scene;
- unsigned format_bytes;
-
- assert(task->x < scene->tiles_x * TILE_SIZE);
- assert(task->y < scene->tiles_y * TILE_SIZE);
- assert(task->x % TILE_SIZE == 0);
- assert(task->y % TILE_SIZE == 0);
- assert(buf < scene->fb.nr_cbufs);
-
- if (!task->color_tiles[buf]) {
- struct pipe_surface *cbuf = scene->fb.cbufs[buf];
- assert(cbuf);
-
- format_bytes = util_format_get_blocksize(cbuf->format);
- task->color_tiles[buf] = scene->cbufs[buf].map + scene->cbufs[buf].stride * task->y + format_bytes * task->x;
- }
-
- return task->color_tiles[buf];
-}
-
-
/**
- * Get pointer to the unswizzled depth tile
- */
-static INLINE uint8_t *
-lp_rast_get_unswizzled_depth_tile_pointer(struct lp_rasterizer_task *task,
- enum lp_texture_usage usage)
-{
- const struct lp_scene *scene = task->scene;
- unsigned format_bytes;
-
- assert(task->x < scene->tiles_x * TILE_SIZE);
- assert(task->y < scene->tiles_y * TILE_SIZE);
- assert(task->x % TILE_SIZE == 0);
- assert(task->y % TILE_SIZE == 0);
-
- if (!task->depth_tile) {
- struct pipe_surface *dbuf = scene->fb.zsbuf;
- assert(dbuf);
-
- format_bytes = util_format_get_blocksize(dbuf->format);
- task->depth_tile = scene->zsbuf.map + scene->zsbuf.stride * task->y + format_bytes * task->x;
- }
-
- return task->depth_tile;
-}
-
-
-/**
- * Get the pointer to an unswizzled 4x4 color block (within an unswizzled 64x64 tile).
+ * Get the pointer to a 4x4 color block (within a 64x64 tile).
* \param x, y location of 4x4 block in window coords
*/
-static INLINE uint8_t *
-lp_rast_get_unswizzled_color_block_pointer(struct lp_rasterizer_task *task,
- unsigned buf, unsigned x, unsigned y,
- unsigned layer)
+static inline uint8_t *
+lp_rast_get_color_block_pointer(struct lp_rasterizer_task *task,
+ unsigned buf, unsigned x, unsigned y,
+ unsigned layer)
{
- unsigned px, py, pixel_offset, format_bytes;
+ unsigned px, py, pixel_offset;
uint8_t *color;
assert(x < task->scene->tiles_x * TILE_SIZE);
assert((y % TILE_VECTOR_HEIGHT) == 0);
assert(buf < task->scene->fb.nr_cbufs);
- format_bytes = util_format_get_blocksize(task->scene->fb.cbufs[buf]->format);
-
- color = lp_rast_get_unswizzled_color_tile_pointer(task, buf, LP_TEX_USAGE_READ_WRITE);
- assert(color);
+ assert(task->color_tiles[buf]);
+ /*
+ * We don't actually benefit from having per tile cbuf/zsbuf pointers,
+ * it's just extra work - the mul/add would be exactly the same anyway.
+ * Fortunately the extra work (modulo) here is very cheap at least...
+ */
px = x % TILE_SIZE;
py = y % TILE_SIZE;
- pixel_offset = px * format_bytes + py * task->scene->cbufs[buf].stride;
- color = color + pixel_offset;
+ pixel_offset = px * task->scene->cbufs[buf].format_bytes +
+ py * task->scene->cbufs[buf].stride;
+ color = task->color_tiles[buf] + pixel_offset;
if (layer) {
color += layer * task->scene->cbufs[buf].layer_stride;
/**
- * Get the pointer to an unswizzled 4x4 depth block (within an unswizzled 64x64 tile).
+ * Get the pointer to a 4x4 depth block (within a 64x64 tile).
* \param x, y location of 4x4 block in window coords
*/
-static INLINE uint8_t *
-lp_rast_get_unswizzled_depth_block_pointer(struct lp_rasterizer_task *task,
- unsigned x, unsigned y, unsigned layer)
+static inline uint8_t *
+lp_rast_get_depth_block_pointer(struct lp_rasterizer_task *task,
+ unsigned x, unsigned y, unsigned layer)
{
- unsigned px, py, pixel_offset, format_bytes;
+ unsigned px, py, pixel_offset;
uint8_t *depth;
assert(x < task->scene->tiles_x * TILE_SIZE);
assert((x % TILE_VECTOR_WIDTH) == 0);
assert((y % TILE_VECTOR_HEIGHT) == 0);
- format_bytes = util_format_get_blocksize(task->scene->fb.zsbuf->format);
-
- depth = lp_rast_get_unswizzled_depth_tile_pointer(task, LP_TEX_USAGE_READ_WRITE);
- assert(depth);
+ assert(task->depth_tile);
px = x % TILE_SIZE;
py = y % TILE_SIZE;
- pixel_offset = px * format_bytes + py * task->scene->zsbuf.stride;
- depth = depth + pixel_offset;
+ pixel_offset = px * task->scene->zsbuf.format_bytes +
+ py * task->scene->zsbuf.stride;
+ depth = task->depth_tile + pixel_offset;
if (layer) {
depth += layer * task->scene->zsbuf.layer_stride;
* triangle in/out tests.
* \param x, y location of 4x4 block in window coords
*/
-static INLINE void
+static inline void
lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
const struct lp_rast_shader_inputs *inputs,
unsigned x, unsigned y )
struct lp_fragment_shader_variant *variant = state->variant;
uint8_t *color[PIPE_MAX_COLOR_BUFS];
unsigned stride[PIPE_MAX_COLOR_BUFS];
+ unsigned sample_stride[PIPE_MAX_COLOR_BUFS];
uint8_t *depth = NULL;
unsigned depth_stride = 0;
+ unsigned depth_sample_stride = 0;
unsigned i;
/* color buffer */
for (i = 0; i < scene->fb.nr_cbufs; i++) {
if (scene->fb.cbufs[i]) {
stride[i] = scene->cbufs[i].stride;
- color[i] = lp_rast_get_unswizzled_color_block_pointer(task, i, x, y,
- inputs->layer);
+ sample_stride[i] = scene->cbufs[i].sample_stride;
+ color[i] = lp_rast_get_color_block_pointer(task, i, x, y,
+ inputs->layer);
}
else {
stride[i] = 0;
+ sample_stride[i] = 0;
color[i] = NULL;
}
}
if (scene->zsbuf.map) {
- depth = lp_rast_get_unswizzled_depth_block_pointer(task, x, y, inputs->layer);
+ depth = lp_rast_get_depth_block_pointer(task, x, y, inputs->layer);
+ depth_sample_stride = scene->zsbuf.sample_stride;
depth_stride = scene->zsbuf.stride;
}
+ uint64_t mask = 0;
+ for (unsigned i = 0; i < scene->fb_max_samples; i++)
+ mask |= (uint64_t)0xffff << (16 * i);
+
/*
* 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) {
- /* not very accurate would need a popcount on the mask */
- /* always count this not worth bothering? */
- task->ps_invocations += 1 * variant->ps_inv_multiplier;
-
/* Propagate non-interpolated raster state. */
task->thread_data.raster_state.viewport_index = inputs->viewport_index;
GET_DADY(inputs),
color,
depth,
- 0xffff,
+ mask,
&task->thread_data,
stride,
- depth_stride);
+ depth_stride,
+ sample_stride,
+ depth_sample_stride);
END_JIT_CALL();
}
}
void lp_rast_triangle_32_4_16( struct lp_rasterizer_task *,
const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_1( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_2( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_3( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_4( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_5( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_6( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_7( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_8( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_3_4(struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_3_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_4_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_32_1( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_2( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_3( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_4( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_5( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_6( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_7( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+void lp_rast_triangle_ms_32_8( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_32_3_4(struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_32_3_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
+void lp_rast_triangle_ms_32_4_16( struct lp_rasterizer_task *,
+ const union lp_rast_cmd_arg );
+
void
lp_rast_set_state(struct lp_rasterizer_task *task,
const union lp_rast_cmd_arg arg);