#include "util/u_pack_color.h"
#include "tgsi/tgsi_parse.h"
#include "radeonsi_pipe.h"
+#include "radeonsi_shader.h"
#include "si_state.h"
#include "sid.h"
*/
static void si_update_fb_blend_state(struct r600_context *rctx)
{
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_pm4_state *pm4;
struct si_state_blend *blend = rctx->queued.named.blend;
uint32_t mask;
- if (pm4 == NULL || blend == NULL)
+ if (blend == NULL)
+ return;
+
+ pm4 = CALLOC_STRUCT(si_pm4_state);
+ if (pm4 == NULL)
return;
mask = (1ULL << ((unsigned)rctx->framebuffer.nr_cbufs * 4)) - 1;
static void si_update_fb_rs_state(struct r600_context *rctx)
{
struct si_state_rasterizer *rs = rctx->queued.named.rasterizer;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_pm4_state *pm4;
unsigned offset_db_fmt_cntl = 0, depth;
float offset_units;
- if (!rs || !rctx->framebuffer.zsbuf) {
- FREE(pm4);
+ if (!rs || !rctx->framebuffer.zsbuf)
return;
- }
offset_units = rctx->queued.named.rasterizer->offset_units;
switch (rctx->framebuffer.zsbuf->texture->format) {
return;
}
+ pm4 = CALLOC_STRUCT(si_pm4_state);
/* FIXME some of those reg can be computed with cso */
offset_db_fmt_cntl |= S_028B78_POLY_OFFSET_NEG_NUM_DB_BITS(depth);
si_pm4_set_reg(pm4, R_028B80_PA_SU_POLY_OFFSET_FRONT_SCALE,
* DSA
*/
-/* transnates straight */
-static uint32_t si_translate_ds_func(int func)
+static uint32_t si_translate_stencil_op(int s_op)
{
- return func;
+ switch (s_op) {
+ case PIPE_STENCIL_OP_KEEP:
+ return V_02842C_STENCIL_KEEP;
+ case PIPE_STENCIL_OP_ZERO:
+ return V_02842C_STENCIL_ZERO;
+ case PIPE_STENCIL_OP_REPLACE:
+ return V_02842C_STENCIL_REPLACE_TEST;
+ case PIPE_STENCIL_OP_INCR:
+ return V_02842C_STENCIL_ADD_CLAMP;
+ case PIPE_STENCIL_OP_DECR:
+ return V_02842C_STENCIL_SUB_CLAMP;
+ case PIPE_STENCIL_OP_INCR_WRAP:
+ return V_02842C_STENCIL_ADD_WRAP;
+ case PIPE_STENCIL_OP_DECR_WRAP:
+ return V_02842C_STENCIL_SUB_WRAP;
+ case PIPE_STENCIL_OP_INVERT:
+ return V_02842C_STENCIL_INVERT;
+ default:
+ R600_ERR("Unknown stencil op %d", s_op);
+ assert(0);
+ break;
+ }
+ return 0;
}
static void *si_create_dsa_state(struct pipe_context *ctx,
struct si_pm4_state *pm4 = &dsa->pm4;
unsigned db_depth_control, /* alpha_test_control, */ alpha_ref;
unsigned db_render_override, db_render_control;
+ uint32_t db_stencil_control = 0;
if (dsa == NULL) {
return NULL;
/* stencil */
if (state->stencil[0].enabled) {
db_depth_control |= S_028800_STENCIL_ENABLE(1);
- db_depth_control |= S_028800_STENCILFUNC(si_translate_ds_func(state->stencil[0].func));
- //db_depth_control |= S_028800_STENCILFAIL(r600_translate_stencil_op(state->stencil[0].fail_op));
- //db_depth_control |= S_028800_STENCILZPASS(r600_translate_stencil_op(state->stencil[0].zpass_op));
- //db_depth_control |= S_028800_STENCILZFAIL(r600_translate_stencil_op(state->stencil[0].zfail_op));
+ db_depth_control |= S_028800_STENCILFUNC(state->stencil[0].func);
+ db_stencil_control |= S_02842C_STENCILFAIL(si_translate_stencil_op(state->stencil[0].fail_op));
+ db_stencil_control |= S_02842C_STENCILZPASS(si_translate_stencil_op(state->stencil[0].zpass_op));
+ db_stencil_control |= S_02842C_STENCILZFAIL(si_translate_stencil_op(state->stencil[0].zfail_op));
if (state->stencil[1].enabled) {
db_depth_control |= S_028800_BACKFACE_ENABLE(1);
- db_depth_control |= S_028800_STENCILFUNC_BF(si_translate_ds_func(state->stencil[1].func));
- //db_depth_control |= S_028800_STENCILFAIL_BF(r600_translate_stencil_op(state->stencil[1].fail_op));
- //db_depth_control |= S_028800_STENCILZPASS_BF(r600_translate_stencil_op(state->stencil[1].zpass_op));
- //db_depth_control |= S_028800_STENCILZFAIL_BF(r600_translate_stencil_op(state->stencil[1].zfail_op));
+ db_depth_control |= S_028800_STENCILFUNC_BF(state->stencil[1].func);
+ db_stencil_control |= S_02842C_STENCILFAIL_BF(si_translate_stencil_op(state->stencil[1].fail_op));
+ db_stencil_control |= S_02842C_STENCILZPASS_BF(si_translate_stencil_op(state->stencil[1].zpass_op));
+ db_stencil_control |= S_02842C_STENCILZFAIL_BF(si_translate_stencil_op(state->stencil[1].zfail_op));
}
}
si_pm4_set_reg(pm4, R_028800_DB_DEPTH_CONTROL, db_depth_control);
si_pm4_set_reg(pm4, R_028000_DB_RENDER_CONTROL, db_render_control);
si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE, db_render_override);
+ si_pm4_set_reg(pm4, R_02842C_DB_STENCIL_CONTROL, db_stencil_control);
si_pm4_set_reg(pm4, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0);
si_pm4_set_reg(pm4, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0);
si_pm4_set_reg(pm4, R_028AC8_DB_PRELOAD_CONTROL, 0x0);
}
}
+/* Returns the size in bits of the widest component of a CB format */
+static unsigned si_colorformat_max_comp_size(uint32_t colorformat)
+{
+ switch(colorformat) {
+ case V_028C70_COLOR_4_4_4_4:
+ return 4;
+
+ case V_028C70_COLOR_1_5_5_5:
+ case V_028C70_COLOR_5_5_5_1:
+ return 5;
+
+ case V_028C70_COLOR_5_6_5:
+ return 6;
+
+ case V_028C70_COLOR_8:
+ case V_028C70_COLOR_8_8:
+ case V_028C70_COLOR_8_8_8_8:
+ return 8;
+
+ case V_028C70_COLOR_10_10_10_2:
+ case V_028C70_COLOR_2_10_10_10:
+ return 10;
+
+ case V_028C70_COLOR_10_11_11:
+ case V_028C70_COLOR_11_11_10:
+ return 11;
+
+ case V_028C70_COLOR_16:
+ case V_028C70_COLOR_16_16:
+ case V_028C70_COLOR_16_16_16_16:
+ return 16;
+
+ case V_028C70_COLOR_8_24:
+ case V_028C70_COLOR_24_8:
+ return 24;
+
+ case V_028C70_COLOR_32:
+ case V_028C70_COLOR_32_32:
+ case V_028C70_COLOR_32_32_32_32:
+ case V_028C70_COLOR_X24_8_32_FLOAT:
+ return 32;
+ }
+
+ assert(!"Unknown maximum component size");
+ return 0;
+}
+
static uint32_t si_translate_dbformat(enum pipe_format format)
{
switch (format) {
const struct util_format_description *desc,
int first_non_void)
{
- uint32_t result;
+ unsigned type = desc->channel[first_non_void].type;
+ int i;
- if (desc->channel[first_non_void].type == UTIL_FORMAT_TYPE_FIXED)
- return ~0;
+ if (type == UTIL_FORMAT_TYPE_FIXED)
+ return V_008F0C_BUF_DATA_FORMAT_INVALID;
+
+ /* See whether the components are of the same size. */
+ for (i = 0; i < desc->nr_channels; i++) {
+ if (desc->channel[first_non_void].size != desc->channel[i].size)
+ return V_008F0C_BUF_DATA_FORMAT_INVALID;
+ }
+
+ switch (desc->channel[first_non_void].size) {
+ case 8:
+ switch (desc->nr_channels) {
+ case 1:
+ return V_008F0C_BUF_DATA_FORMAT_8;
+ case 2:
+ return V_008F0C_BUF_DATA_FORMAT_8_8;
+ case 3:
+ case 4:
+ return V_008F0C_BUF_DATA_FORMAT_8_8_8_8;
+ }
+ break;
+ case 16:
+ switch (desc->nr_channels) {
+ case 1:
+ return V_008F0C_BUF_DATA_FORMAT_16;
+ case 2:
+ return V_008F0C_BUF_DATA_FORMAT_16_16;
+ case 3:
+ case 4:
+ return V_008F0C_BUF_DATA_FORMAT_16_16_16_16;
+ }
+ break;
+ case 32:
+ if (type != UTIL_FORMAT_TYPE_FLOAT)
+ return V_008F0C_BUF_DATA_FORMAT_INVALID;
- result = si_translate_texformat(screen, format, desc, first_non_void);
- if (result == V_008F0C_BUF_DATA_FORMAT_INVALID ||
- result > V_008F0C_BUF_DATA_FORMAT_32_32_32_32)
- result = ~0;
+ switch (desc->nr_channels) {
+ case 1:
+ return V_008F0C_BUF_DATA_FORMAT_32;
+ case 2:
+ return V_008F0C_BUF_DATA_FORMAT_32_32;
+ case 3:
+ return V_008F0C_BUF_DATA_FORMAT_32_32_32;
+ case 4:
+ return V_008F0C_BUF_DATA_FORMAT_32_32_32_32;
+ }
+ break;
+ }
- return result;
+ return V_008F0C_BUF_DATA_FORMAT_INVALID;
}
static bool si_is_vertex_format_supported(struct pipe_screen *screen, enum pipe_format format)
{
- return si_translate_vertexformat(screen, format, util_format_description(format),
- util_format_get_first_non_void_channel(format)) != ~0U;
+ const struct util_format_description *desc;
+ int first_non_void;
+ unsigned data_format;
+
+ desc = util_format_description(format);
+ first_non_void = util_format_get_first_non_void_channel(format);
+ data_format = si_translate_vertexformat(screen, format, desc, first_non_void);
+ return data_format != V_008F0C_BUF_DATA_FORMAT_INVALID;
}
static bool si_is_colorbuffer_format_supported(enum pipe_format format)
const struct util_format_description *desc;
int i;
unsigned blend_clamp = 0, blend_bypass = 0;
+ unsigned max_comp_size;
surf = (struct r600_surface *)state->cbufs[cb];
rtex = (struct r600_resource_texture*)state->cbufs[cb]->texture;
}
si_pm4_set_reg(pm4, R_028C70_CB_COLOR0_INFO + cb * 0x3C, color_info);
si_pm4_set_reg(pm4, R_028C74_CB_COLOR0_ATTRIB + cb * 0x3C, color_attrib);
+
+ /* Determine pixel shader export format */
+ max_comp_size = si_colorformat_max_comp_size(format);
+ if (ntype == V_028C70_NUMBER_SRGB ||
+ ((ntype == V_028C70_NUMBER_UNORM || ntype == V_028C70_NUMBER_SNORM) &&
+ max_comp_size <= 10) ||
+ (ntype == V_028C70_NUMBER_FLOAT && max_comp_size <= 16)) {
+ rctx->export_16bpc |= 1 << cb;
+ rctx->spi_shader_col_format |= V_028714_SPI_SHADER_FP16_ABGR << (4 * cb);
+ } else
+ rctx->spi_shader_col_format |= V_028714_SPI_SHADER_32_ABGR << (4 * cb);
}
static void si_db(struct r600_context *rctx, struct si_pm4_state *pm4,
{
struct r600_resource_texture *rtex;
struct r600_surface *surf;
- unsigned level, first_layer, pitch, slice, format;
- uint32_t db_z_info, stencil_info;
- uint64_t offset;
+ unsigned level, pitch, slice, format;
+ uint32_t z_info, s_info;
+ uint64_t z_offs, s_offs;
if (state->zsbuf == NULL) {
si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, 0);
level = surf->base.u.tex.level;
rtex = (struct r600_resource_texture*)surf->base.texture;
- first_layer = surf->base.u.tex.first_layer;
format = si_translate_dbformat(rtex->real_format);
- offset = r600_resource_va(rctx->context.screen, surf->base.texture);
- offset += rtex->surface.level[level].offset;
+ z_offs = r600_resource_va(rctx->context.screen, surf->base.texture);
+ z_offs += rtex->surface.level[level].offset;
+
+ s_offs = r600_resource_va(rctx->context.screen, surf->base.texture);
+ s_offs += rtex->surface.stencil_offset;
+ z_offs += rtex->surface.level[level].offset / 4;
+
+ z_offs >>= 8;
+ s_offs >>= 8;
+
pitch = (rtex->surface.level[level].nblk_x / 8) - 1;
slice = (rtex->surface.level[level].nblk_x * rtex->surface.level[level].nblk_y) / 64;
if (slice) {
slice = slice - 1;
}
- offset >>= 8;
- si_pm4_add_bo(pm4, &rtex->resource, RADEON_USAGE_READWRITE);
- si_pm4_set_reg(pm4, R_028048_DB_Z_READ_BASE, offset);
- si_pm4_set_reg(pm4, R_028050_DB_Z_WRITE_BASE, offset);
- si_pm4_set_reg(pm4, R_028008_DB_DEPTH_VIEW,
- S_028008_SLICE_START(state->zsbuf->u.tex.first_layer) |
- S_028008_SLICE_MAX(state->zsbuf->u.tex.last_layer));
+ z_info = S_028040_FORMAT(format);
+ s_info = S_028044_FORMAT(1);
- db_z_info = S_028040_FORMAT(format);
- stencil_info = S_028044_FORMAT(rtex->stencil != 0);
+ if (rtex->surface.level[level].mode == RADEON_SURF_MODE_1D) {
+ z_info |= S_028040_TILE_MODE_INDEX(4);
+ s_info |= S_028044_TILE_MODE_INDEX(4);
- switch (format) {
- case V_028040_Z_16:
- db_z_info |= S_028040_TILE_MODE_INDEX(5);
- stencil_info |= S_028044_TILE_MODE_INDEX(5);
- break;
- case V_028040_Z_24:
- case V_028040_Z_32_FLOAT:
- db_z_info |= S_028040_TILE_MODE_INDEX(6);
- stencil_info |= S_028044_TILE_MODE_INDEX(6);
- break;
- default:
- db_z_info |= S_028040_TILE_MODE_INDEX(7);
- stencil_info |= S_028044_TILE_MODE_INDEX(7);
- }
-
- if (rtex->stencil) {
- uint64_t stencil_offset =
- r600_texture_get_offset(rtex->stencil, level, first_layer);
-
- stencil_offset += r600_resource_va(rctx->context.screen, (void*)rtex->stencil);
- stencil_offset >>= 8;
+ } else if (rtex->surface.level[level].mode == RADEON_SURF_MODE_2D) {
+ switch (format) {
+ case V_028040_Z_16:
+ z_info |= S_028040_TILE_MODE_INDEX(5);
+ s_info |= S_028044_TILE_MODE_INDEX(5);
+ break;
+ case V_028040_Z_24:
+ case V_028040_Z_32_FLOAT:
+ z_info |= S_028040_TILE_MODE_INDEX(6);
+ s_info |= S_028044_TILE_MODE_INDEX(6);
+ break;
+ default:
+ z_info |= S_028040_TILE_MODE_INDEX(7);
+ s_info |= S_028044_TILE_MODE_INDEX(7);
+ }
- si_pm4_add_bo(pm4, &rtex->stencil->resource, RADEON_USAGE_READWRITE);
- si_pm4_set_reg(pm4, R_02804C_DB_STENCIL_READ_BASE, stencil_offset);
- si_pm4_set_reg(pm4, R_028054_DB_STENCIL_WRITE_BASE, stencil_offset);
- si_pm4_set_reg(pm4, R_028044_DB_STENCIL_INFO, stencil_info);
} else {
+ R600_ERR("Invalid DB tiling mode %d!\n",
+ rtex->surface.level[level].mode);
+ si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, 0);
si_pm4_set_reg(pm4, R_028044_DB_STENCIL_INFO, 0);
+ return;
}
+ si_pm4_set_reg(pm4, R_028008_DB_DEPTH_VIEW,
+ S_028008_SLICE_START(state->zsbuf->u.tex.first_layer) |
+ S_028008_SLICE_MAX(state->zsbuf->u.tex.last_layer));
+
+ si_pm4_set_reg(pm4, R_02803C_DB_DEPTH_INFO, 0x1);
if (format != ~0U) {
- si_pm4_set_reg(pm4, R_02803C_DB_DEPTH_INFO, 0x1);
- si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, db_z_info);
- si_pm4_set_reg(pm4, R_028058_DB_DEPTH_SIZE, S_028058_PITCH_TILE_MAX(pitch));
- si_pm4_set_reg(pm4, R_02805C_DB_DEPTH_SLICE, S_02805C_SLICE_TILE_MAX(slice));
+ si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, z_info);
} else {
si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, 0);
}
+
+ if (rtex->surface.flags & RADEON_SURF_SBUFFER) {
+ si_pm4_set_reg(pm4, R_028044_DB_STENCIL_INFO, s_info);
+ } else {
+ si_pm4_set_reg(pm4, R_028044_DB_STENCIL_INFO, 0);
+ }
+
+ si_pm4_add_bo(pm4, &rtex->resource, RADEON_USAGE_READWRITE);
+ si_pm4_set_reg(pm4, R_028048_DB_Z_READ_BASE, z_offs);
+ si_pm4_set_reg(pm4, R_02804C_DB_STENCIL_READ_BASE, s_offs);
+ si_pm4_set_reg(pm4, R_028050_DB_Z_WRITE_BASE, z_offs);
+ si_pm4_set_reg(pm4, R_028054_DB_STENCIL_WRITE_BASE, s_offs);
+
+ si_pm4_set_reg(pm4, R_028058_DB_DEPTH_SIZE, S_028058_PITCH_TILE_MAX(pitch));
+ si_pm4_set_reg(pm4, R_02805C_DB_DEPTH_SLICE, S_02805C_SLICE_TILE_MAX(slice));
}
static void si_set_framebuffer_state(struct pipe_context *ctx,
/* build states */
rctx->have_depth_fb = 0;
+ rctx->export_16bpc = 0;
+ rctx->spi_shader_col_format = 0;
for (int i = 0; i < state->nr_cbufs; i++) {
si_cb(rctx, pm4, state, i);
}
+ assert(!(rctx->export_16bpc & ~0xff));
si_db(rctx, pm4, state);
shader_mask = 0;
si_pm4_set_reg(pm4, R_028200_PA_SC_WINDOW_OFFSET, 0x00000000);
si_pm4_set_reg(pm4, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA);
si_pm4_set_reg(pm4, R_02823C_CB_SHADER_MASK, shader_mask);
+ si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT,
+ rctx->spi_shader_col_format);
si_pm4_set_reg(pm4, R_028BE0_PA_SC_AA_CONFIG, 0x00000000);
si_pm4_set_state(rctx, framebuffer, pm4);
* shaders
*/
+/* Compute the key for the hw shader variant */
+static INLINE unsigned si_shader_selector_key(struct pipe_context *ctx,
+ struct si_pipe_shader_selector *sel)
+{
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ unsigned key = 0;
+
+ if (sel->type == PIPE_SHADER_FRAGMENT) {
+ if (sel->fs_write_all)
+ key |= rctx->framebuffer.nr_cbufs;
+ key |= rctx->export_16bpc << 4;
+ /*if (rctx->queued.named.rasterizer)
+ key |= rctx->queued.named.rasterizer->flatshade << 12;*/
+ /*key |== rctx->two_side << 13;*/
+ }
+
+ return key;
+}
+
+/* Select the hw shader variant depending on the current state.
+ * (*dirty) is set to 1 if current variant was changed */
+int si_shader_select(struct pipe_context *ctx,
+ struct si_pipe_shader_selector *sel,
+ unsigned *dirty)
+{
+ unsigned key;
+ struct si_pipe_shader * shader = NULL;
+ int r;
+
+ key = si_shader_selector_key(ctx, sel);
+
+ /* Check if we don't need to change anything.
+ * This path is also used for most shaders that don't need multiple
+ * variants, it will cost just a computation of the key and this
+ * test. */
+ if (likely(sel->current && sel->current->key == key)) {
+ return 0;
+ }
+
+ /* lookup if we have other variants in the list */
+ if (sel->num_shaders > 1) {
+ struct si_pipe_shader *p = sel->current, *c = p->next_variant;
+
+ while (c && c->key != key) {
+ p = c;
+ c = c->next_variant;
+ }
+
+ if (c) {
+ p->next_variant = c->next_variant;
+ shader = c;
+ }
+ }
+
+ if (unlikely(!shader)) {
+ shader = CALLOC(1, sizeof(struct si_pipe_shader));
+ shader->selector = sel;
+
+ r = si_pipe_shader_create(ctx, shader);
+ if (unlikely(r)) {
+ R600_ERR("Failed to build shader variant (type=%u, key=%u) %d\n",
+ sel->type, key, r);
+ sel->current = NULL;
+ return r;
+ }
+
+ /* We don't know the value of fs_write_all property until we built
+ * at least one variant, so we may need to recompute the key (include
+ * rctx->framebuffer.nr_cbufs) after building first variant. */
+ if (sel->type == PIPE_SHADER_FRAGMENT &&
+ sel->num_shaders == 0 &&
+ shader->shader.fs_write_all) {
+ sel->fs_write_all = 1;
+ key = si_shader_selector_key(ctx, sel);
+ }
+
+ shader->key = key;
+ sel->num_shaders++;
+ }
+
+ if (dirty)
+ *dirty = 1;
+
+ shader->next_variant = sel->current;
+ sel->current = shader;
+
+ return 0;
+}
+
static void *si_create_shader_state(struct pipe_context *ctx,
- const struct pipe_shader_state *state)
+ const struct pipe_shader_state *state,
+ unsigned pipe_shader_type)
{
- struct si_pipe_shader *shader = CALLOC_STRUCT(si_pipe_shader);
+ struct si_pipe_shader_selector *sel = CALLOC_STRUCT(si_pipe_shader_selector);
+ int r;
- shader->tokens = tgsi_dup_tokens(state->tokens);
- shader->so = state->stream_output;
+ sel->type = pipe_shader_type;
+ sel->tokens = tgsi_dup_tokens(state->tokens);
+ sel->so = state->stream_output;
- return shader;
+ r = si_shader_select(ctx, sel, NULL);
+ if (r) {
+ free(sel);
+ return NULL;
+ }
+
+ return sel;
+}
+
+static void *si_create_fs_state(struct pipe_context *ctx,
+ const struct pipe_shader_state *state)
+{
+ return si_create_shader_state(ctx, state, PIPE_SHADER_FRAGMENT);
+}
+
+static void *si_create_vs_state(struct pipe_context *ctx,
+ const struct pipe_shader_state *state)
+{
+ return si_create_shader_state(ctx, state, PIPE_SHADER_VERTEX);
}
static void si_bind_vs_shader(struct pipe_context *ctx, void *state)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pipe_shader *shader = state;
+ struct si_pipe_shader_selector *sel = state;
- if (rctx->vs_shader == state)
+ if (rctx->vs_shader == sel)
return;
rctx->shader_dirty = true;
- rctx->vs_shader = shader;
+ rctx->vs_shader = sel;
- if (shader) {
- si_pm4_bind_state(rctx, vs, shader->pm4);
- }
+ if (sel && sel->current)
+ si_pm4_bind_state(rctx, vs, sel->current->pm4);
+ else
+ si_pm4_bind_state(rctx, vs, rctx->dummy_pixel_shader->pm4);
}
static void si_bind_ps_shader(struct pipe_context *ctx, void *state)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pipe_shader *shader = state;
+ struct si_pipe_shader_selector *sel = state;
- if (rctx->ps_shader == state)
+ if (rctx->ps_shader == sel)
return;
rctx->shader_dirty = true;
- rctx->ps_shader = shader;
+ rctx->ps_shader = sel;
- if (shader) {
- si_pm4_bind_state(rctx, ps, shader->pm4);
- }
+ if (sel && sel->current)
+ si_pm4_bind_state(rctx, ps, sel->current->pm4);
+ else
+ si_pm4_bind_state(rctx, ps, rctx->dummy_pixel_shader->pm4);
}
+static void si_delete_shader_selector(struct pipe_context *ctx,
+ struct si_pipe_shader_selector *sel)
+{
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_pipe_shader *p = sel->current, *c;
+
+ while (p) {
+ c = p->next_variant;
+ si_pm4_delete_state(rctx, vs, p->pm4);
+ si_pipe_shader_destroy(ctx, p);
+ free(p);
+ p = c;
+ }
+
+ free(sel->tokens);
+ free(sel);
+ }
+
static void si_delete_vs_shader(struct pipe_context *ctx, void *state)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pipe_shader *shader = (struct si_pipe_shader *)state;
+ struct si_pipe_shader_selector *sel = (struct si_pipe_shader_selector *)state;
- if (rctx->vs_shader == shader) {
+ if (rctx->vs_shader == sel) {
rctx->vs_shader = NULL;
}
- si_pm4_delete_state(rctx, vs, shader->pm4);
- free(shader->tokens);
- si_pipe_shader_destroy(ctx, shader);
- free(shader);
+ si_delete_shader_selector(ctx, sel);
}
static void si_delete_ps_shader(struct pipe_context *ctx, void *state)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pipe_shader *shader = (struct si_pipe_shader *)state;
+ struct si_pipe_shader_selector *sel = (struct si_pipe_shader_selector *)state;
- if (rctx->ps_shader == shader) {
+ if (rctx->ps_shader == sel) {
rctx->ps_shader = NULL;
}
- si_pm4_delete_state(rctx, ps, shader->pm4);
- free(shader->tokens);
- si_pipe_shader_destroy(ctx, shader);
- free(shader);
+ si_delete_shader_selector(ctx, sel);
}
/*
unsigned format, num_format, /*endian,*/ tiling_index;
uint32_t pitch = 0;
unsigned char state_swizzle[4], swizzle[4];
- unsigned height, depth, width;
+ unsigned height, depth, width, offset_level, last_level;
int first_non_void;
uint64_t va;
/* not supported any more */
//endian = si_colorformat_endian_swap(format);
- height = texture->height0;
- depth = texture->depth0;
- width = texture->width0;
- pitch = align(tmp->pitch_in_blocks[0] *
- util_format_get_blockwidth(state->format), 8);
+ offset_level = state->u.tex.first_level;
+ last_level = state->u.tex.last_level - offset_level;
+ width = tmp->surface.level[offset_level].npix_x;
+ height = tmp->surface.level[offset_level].npix_y;
+ depth = tmp->surface.level[offset_level].npix_z;
+ pitch = tmp->surface.level[offset_level].nblk_x * util_format_get_blockwidth(state->format);
if (texture->target == PIPE_TEXTURE_1D_ARRAY) {
height = 1;
}
va = r600_resource_va(ctx->screen, texture);
- if (state->u.tex.last_level) {
- view->state[0] = (va + tmp->offset[1]) >> 8;
- } else {
- view->state[0] = (va + tmp->offset[0]) >> 8;
- }
- view->state[1] = (S_008F14_BASE_ADDRESS_HI((va + tmp->offset[0]) >> 40) |
+ va += tmp->surface.level[offset_level].offset;
+ view->state[0] = va >> 8;
+ view->state[1] = (S_008F14_BASE_ADDRESS_HI(va >> 40) |
S_008F14_DATA_FORMAT(format) |
S_008F14_NUM_FORMAT(num_format));
view->state[2] = (S_008F18_WIDTH(width - 1) |
S_008F1C_DST_SEL_Y(si_map_swizzle(swizzle[1])) |
S_008F1C_DST_SEL_Z(si_map_swizzle(swizzle[2])) |
S_008F1C_DST_SEL_W(si_map_swizzle(swizzle[3])) |
- S_008F1C_BASE_LEVEL(state->u.tex.first_level) |
- S_008F1C_LAST_LEVEL(state->u.tex.last_level) |
+ S_008F1C_BASE_LEVEL(offset_level) |
+ S_008F1C_LAST_LEVEL(last_level) |
S_008F1C_TILING_INDEX(tiling_index) |
S_008F1C_TYPE(si_tex_dim(texture->target)));
view->state[4] = (S_008F20_DEPTH(depth - 1) | S_008F20_PITCH(pitch - 1));
struct r600_context *rctx = (struct r600_context *)ctx;
struct si_pipe_sampler_view **resource = (struct si_pipe_sampler_view **)views;
struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
- struct si_resource *bo;
- int i;
+ int i, j;
int has_depth = 0;
- uint64_t va;
- char *ptr;
if (!count)
goto out;
si_pm4_inval_texture_cache(pm4);
- bo = si_resource_create_custom(ctx->screen, PIPE_USAGE_IMMUTABLE,
- count * sizeof(resource[0]->state));
- ptr = rctx->ws->buffer_map(bo->cs_buf, rctx->cs, PIPE_TRANSFER_WRITE);
-
- for (i = 0; i < count; i++, ptr += sizeof(resource[0]->state)) {
- struct r600_resource_texture *tex = (void *)resource[i]->base.texture;
-
+ si_pm4_sh_data_begin(pm4);
+ for (i = 0; i < count; i++) {
pipe_sampler_view_reference(
(struct pipe_sampler_view **)&rctx->ps_samplers.views[i],
views[i]);
- si_pm4_add_bo(pm4, &tex->resource, RADEON_USAGE_READ);
+ if (views[i]) {
+ struct r600_resource_texture *tex = (void *)resource[i]->base.texture;
- if (resource[i]) {
- if (tex->depth)
- has_depth = 1;
+ si_pm4_add_bo(pm4, &tex->resource, RADEON_USAGE_READ);
- memcpy(ptr, resource[i]->state, sizeof(resource[0]->state));
- } else
- memset(ptr, 0, sizeof(resource[0]->state));
+ for (j = 0; j < Elements(resource[i]->state); ++j) {
+ si_pm4_sh_data_add(pm4, resource[i]->state[j]);
+ }
+ }
}
- rctx->ws->buffer_unmap(bo->cs_buf);
-
for (i = count; i < NUM_TEX_UNITS; i++) {
if (rctx->ps_samplers.views[i])
pipe_sampler_view_reference((struct pipe_sampler_view **)&rctx->ps_samplers.views[i], NULL);
}
- va = r600_resource_va(ctx->screen, (void *)bo);
- si_pm4_add_bo(pm4, bo, RADEON_USAGE_READ);
- si_pm4_set_reg(pm4, R_00B040_SPI_SHADER_USER_DATA_PS_4, va);
- si_pm4_set_reg(pm4, R_00B044_SPI_SHADER_USER_DATA_PS_5, va >> 32);
+ si_pm4_sh_data_end(pm4, R_00B040_SPI_SHADER_USER_DATA_PS_4);
out:
si_pm4_set_state(rctx, ps_sampler_views, pm4);
static void si_bind_vs_sampler(struct pipe_context *ctx, unsigned count, void **states)
{
- assert(0);
+ assert(count == 0);
}
static void si_bind_ps_sampler(struct pipe_context *ctx, unsigned count, void **states)
struct r600_context *rctx = (struct r600_context *)ctx;
struct si_pipe_sampler_state **rstates = (struct si_pipe_sampler_state **)states;
struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
- struct si_resource *bo;
- uint64_t va;
- char *ptr;
- int i;
+ int i, j;
if (!count)
goto out;
si_pm4_inval_texture_cache(pm4);
- bo = si_resource_create_custom(ctx->screen, PIPE_USAGE_IMMUTABLE,
- count * sizeof(rstates[0]->val));
- ptr = rctx->ws->buffer_map(bo->cs_buf, rctx->cs, PIPE_TRANSFER_WRITE);
-
- for (i = 0; i < count; i++, ptr += sizeof(rstates[0]->val)) {
- memcpy(ptr, rstates[i]->val, sizeof(rstates[0]->val));
+ si_pm4_sh_data_begin(pm4);
+ for (i = 0; i < count; i++) {
+ for (j = 0; j < Elements(rstates[i]->val); ++j) {
+ si_pm4_sh_data_add(pm4, rstates[i]->val[j]);
+ }
}
-
- rctx->ws->buffer_unmap(bo->cs_buf);
+ si_pm4_sh_data_end(pm4, R_00B038_SPI_SHADER_USER_DATA_PS_2);
memcpy(rctx->ps_samplers.samplers, states, sizeof(void*) * count);
- va = r600_resource_va(ctx->screen, (void *)bo);
- si_pm4_add_bo(pm4, bo, RADEON_USAGE_READ);
- si_pm4_set_reg(pm4, R_00B038_SPI_SHADER_USER_DATA_PS_2, va);
- si_pm4_set_reg(pm4, R_00B03C_SPI_SHADER_USER_DATA_PS_3, va >> 32);
-
out:
si_pm4_set_state(rctx, ps_sampler, pm4);
rctx->ps_samplers.n_samplers = count;
{
struct r600_context *rctx = (struct r600_context *)ctx;
struct si_resource *rbuffer = cb ? si_resource(cb->buffer) : NULL;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_pm4_state *pm4;
uint64_t va_offset;
uint32_t offset;
/* Note that the state tracker can unbind constant buffers by
* passing NULL here.
*/
- if (cb == NULL) {
- FREE(pm4);
+ if (cb == NULL)
return;
- }
+ pm4 = CALLOC_STRUCT(si_pm4_state);
si_pm4_inval_shader_cache(pm4);
if (cb->user_buffer)
default:
R600_ERR("unsupported %d\n", shader);
- return;
}
if (cb->buffer != &rbuffer->b.b)
}
}
-/*
- * Stream out
- */
-
-static struct pipe_stream_output_target *
-si_create_so_target(struct pipe_context *ctx,
- struct pipe_resource *buffer,
- unsigned buffer_offset,
- unsigned buffer_size)
-{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct r600_so_target *t;
- void *ptr;
-
- t = CALLOC_STRUCT(r600_so_target);
- if (!t) {
- return NULL;
- }
-
- t->b.reference.count = 1;
- t->b.context = ctx;
- pipe_resource_reference(&t->b.buffer, buffer);
- t->b.buffer_offset = buffer_offset;
- t->b.buffer_size = buffer_size;
-
- t->filled_size = si_resource_create_custom(ctx->screen, PIPE_USAGE_STATIC, 4);
- ptr = rctx->ws->buffer_map(t->filled_size->cs_buf, rctx->cs, PIPE_TRANSFER_WRITE);
- memset(ptr, 0, t->filled_size->buf->size);
- rctx->ws->buffer_unmap(t->filled_size->cs_buf);
-
- return &t->b;
-}
-
-static void si_so_target_destroy(struct pipe_context *ctx,
- struct pipe_stream_output_target *target)
-{
- struct r600_so_target *t = (struct r600_so_target*)target;
- pipe_resource_reference(&t->b.buffer, NULL);
- si_resource_reference(&t->filled_size, NULL);
- FREE(t);
-}
-
-static void si_set_so_targets(struct pipe_context *ctx,
- unsigned num_targets,
- struct pipe_stream_output_target **targets,
- unsigned append_bitmask)
-{
- struct r600_context *rctx = (struct r600_context *)ctx;
- unsigned i;
-
- /* Stop streamout. */
- if (rctx->num_so_targets) {
- r600_context_streamout_end(rctx);
- }
-
- /* Set the new targets. */
- for (i = 0; i < num_targets; i++) {
- pipe_so_target_reference((struct pipe_stream_output_target**)&rctx->so_targets[i], targets[i]);
- }
- for (; i < rctx->num_so_targets; i++) {
- pipe_so_target_reference((struct pipe_stream_output_target**)&rctx->so_targets[i], NULL);
- }
-
- rctx->num_so_targets = num_targets;
- rctx->streamout_start = num_targets != 0;
- rctx->streamout_append_bitmask = append_bitmask;
-}
-
/*
* Misc
*/
-#if 0
-static uint32_t r600_translate_stencil_op(int s_op)
-{
- switch (s_op) {
- case PIPE_STENCIL_OP_KEEP:
- return V_028800_STENCIL_KEEP;
- case PIPE_STENCIL_OP_ZERO:
- return V_028800_STENCIL_ZERO;
- case PIPE_STENCIL_OP_REPLACE:
- return V_028800_STENCIL_REPLACE;
- case PIPE_STENCIL_OP_INCR:
- return V_028800_STENCIL_INCR;
- case PIPE_STENCIL_OP_DECR:
- return V_028800_STENCIL_DECR;
- case PIPE_STENCIL_OP_INCR_WRAP:
- return V_028800_STENCIL_INCR_WRAP;
- case PIPE_STENCIL_OP_DECR_WRAP:
- return V_028800_STENCIL_DECR_WRAP;
- case PIPE_STENCIL_OP_INVERT:
- return V_028800_STENCIL_INVERT;
- default:
- R600_ERR("Unknown stencil op %d", s_op);
- assert(0);
- break;
- }
- return 0;
-}
-#endif
-
static void si_set_polygon_stipple(struct pipe_context *ctx,
const struct pipe_poly_stipple *state)
{
rctx->context.set_framebuffer_state = si_set_framebuffer_state;
- rctx->context.create_vs_state = si_create_shader_state;
- rctx->context.create_fs_state = si_create_shader_state;
+ rctx->context.create_vs_state = si_create_vs_state;
+ rctx->context.create_fs_state = si_create_fs_state;
rctx->context.bind_vs_state = si_bind_vs_shader;
rctx->context.bind_fs_state = si_bind_ps_shader;
rctx->context.delete_vs_state = si_delete_vs_shader;
{
struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ si_pm4_cmd_begin(pm4, PKT3_CONTEXT_CONTROL);
+ si_pm4_cmd_add(pm4, 0x80000000);
+ si_pm4_cmd_add(pm4, 0x80000000);
+ si_pm4_cmd_end(pm4, false);
+
si_pm4_set_reg(pm4, R_028A4C_PA_SC_MODE_CNTL_1, 0x0);
si_pm4_set_reg(pm4, R_028A10_VGT_OUTPUT_PATH_CNTL, 0x0);