/**************************************************************************
*
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2003 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
**************************************************************************/
+#include <stdio.h>
+#include "main/arrayobj.h"
#include "main/glheader.h"
#include "main/context.h"
#include "pipe/p_defines.h"
#include "st_context.h"
#include "st_atom.h"
-#include "st_cb_bitmap.h"
#include "st_program.h"
#include "st_manager.h"
+#include "st_util.h"
-/**
- * This is used to initialize st->atoms[].
- */
-static const struct st_tracked_state *atoms[] =
+typedef void (*update_func_t)(struct st_context *st);
+
+/* The list state update functions. */
+static const update_func_t update_functions[] =
{
- &st_update_depth_stencil_alpha,
- &st_update_clip,
-
- &st_finalize_textures,
- &st_update_fp,
- &st_update_gp,
- &st_update_vp,
-
- &st_update_rasterizer,
- &st_update_polygon_stipple,
- &st_update_viewport,
- &st_update_scissor,
- &st_update_blend,
- &st_update_vertex_texture,
- &st_update_fragment_texture,
- &st_update_geometry_texture,
- &st_update_sampler, /* depends on update_*_texture for swizzle */
- &st_update_framebuffer,
- &st_update_msaa,
- &st_update_vs_constants,
- &st_update_gs_constants,
- &st_update_fs_constants,
- &st_bind_vs_ubos,
- &st_bind_fs_ubos,
- &st_update_pixel_transfer,
-
- /* this must be done after the vertex program update */
- &st_update_array
+#define ST_STATE(FLAG, st_update) st_update,
+#include "st_atom_list.h"
+#undef ST_STATE
};
void st_init_atoms( struct st_context *st )
{
- /* no-op */
+ STATIC_ASSERT(ARRAY_SIZE(update_functions) <= 64);
}
}
-/***********************************************************************
- */
-
-static GLboolean check_state( const struct st_state_flags *a,
- const struct st_state_flags *b )
-{
- return ((a->mesa & b->mesa) ||
- (a->st & b->st));
-}
-
-static void accumulate_state( struct st_state_flags *a,
- const struct st_state_flags *b )
-{
- a->mesa |= b->mesa;
- a->st |= b->st;
-}
-
-
-static void xor_states( struct st_state_flags *result,
- const struct st_state_flags *a,
- const struct st_state_flags *b )
-{
- result->mesa = a->mesa ^ b->mesa;
- result->st = a->st ^ b->st;
-}
-
-
/* Too complex to figure out, just check every time:
*/
static void check_program_state( struct st_context *st )
{
struct gl_context *ctx = st->ctx;
+ struct st_program *old_vp = st->vp;
+ struct st_program *old_tcp = st->tcp;
+ struct st_program *old_tep = st->tep;
+ struct st_program *old_gp = st->gp;
+ struct st_program *old_fp = st->fp;
+
+ struct gl_program *new_vp = ctx->VertexProgram._Current;
+ struct gl_program *new_tcp = ctx->TessCtrlProgram._Current;
+ struct gl_program *new_tep = ctx->TessEvalProgram._Current;
+ struct gl_program *new_gp = ctx->GeometryProgram._Current;
+ struct gl_program *new_fp = ctx->FragmentProgram._Current;
+ uint64_t dirty = 0;
+ unsigned num_viewports = 1;
+
+ /* Flag states used by both new and old shaders to unbind shader resources
+ * properly when transitioning to shaders that don't use them.
+ */
+ if (unlikely(new_vp != (old_vp ? &old_vp->Base : NULL))) {
+ if (old_vp)
+ dirty |= old_vp->affected_states;
+ if (new_vp)
+ dirty |= ST_NEW_VERTEX_PROGRAM(st, st_program(new_vp));
+ }
+
+ if (unlikely(new_tcp != &old_tcp->Base)) {
+ if (old_tcp)
+ dirty |= old_tcp->affected_states;
+ if (new_tcp)
+ dirty |= st_program(new_tcp)->affected_states;
+ }
+
+ if (unlikely(new_tep != &old_tep->Base)) {
+ if (old_tep)
+ dirty |= old_tep->affected_states;
+ if (new_tep)
+ dirty |= st_program(new_tep)->affected_states;
+ }
+
+ if (unlikely(new_gp != &old_gp->Base)) {
+ if (old_gp)
+ dirty |= old_gp->affected_states;
+ if (new_gp)
+ dirty |= st_program(new_gp)->affected_states;
+ }
- if (ctx->VertexProgram._Current != &st->vp->Base)
- st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+ if (unlikely(new_fp != &old_fp->Base)) {
+ if (old_fp)
+ dirty |= old_fp->affected_states;
+ if (new_fp)
+ dirty |= st_program(new_fp)->affected_states;
+ }
- if (ctx->FragmentProgram._Current != &st->fp->Base)
- st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+ /* Find out the number of viewports. This determines how many scissors
+ * and viewport states we need to update.
+ */
+ struct gl_program *last_prim_shader = new_gp ? new_gp :
+ new_tep ? new_tep : new_vp;
+ if (last_prim_shader &&
+ last_prim_shader->info.outputs_written & (
+ VARYING_BIT_VIEWPORT | VARYING_BIT_VIEWPORT_MASK))
+ num_viewports = ctx->Const.MaxViewports;
+
+ if (st->state.num_viewports != num_viewports) {
+ st->state.num_viewports = num_viewports;
+ dirty |= ST_NEW_VIEWPORT;
+
+ if (ctx->Scissor.EnableFlags & u_bit_consecutive(0, num_viewports))
+ dirty |= ST_NEW_SCISSOR;
+ }
- if (ctx->GeometryProgram._Current != &st->gp->Base)
- st->dirty.st |= ST_NEW_GEOMETRY_PROGRAM;
+ st->dirty |= dirty;
}
static void check_attrib_edgeflag(struct st_context *st)
{
- const struct gl_client_array **arrays = st->ctx->Array._DrawArrays;
- GLboolean vertDataEdgeFlags;
+ GLboolean vertdata_edgeflags, edgeflag_culls_prims, edgeflags_enabled;
+ struct gl_program *vp = st->ctx->VertexProgram._Current;
- if (!arrays)
- return;
+ edgeflags_enabled = st->ctx->Polygon.FrontMode != GL_FILL ||
+ st->ctx->Polygon.BackMode != GL_FILL;
- vertDataEdgeFlags = arrays[VERT_ATTRIB_EDGEFLAG]->BufferObj &&
- arrays[VERT_ATTRIB_EDGEFLAG]->BufferObj->Name;
- if (vertDataEdgeFlags != st->vertdata_edgeflags) {
- st->vertdata_edgeflags = vertDataEdgeFlags;
- st->dirty.st |= ST_NEW_EDGEFLAGS_DATA;
+ vertdata_edgeflags = edgeflags_enabled &&
+ _mesa_draw_edge_flag_array_enabled(st->ctx);
+
+ if (vertdata_edgeflags != st->vertdata_edgeflags) {
+ st->vertdata_edgeflags = vertdata_edgeflags;
+ if (vp)
+ st->dirty |= ST_NEW_VERTEX_PROGRAM(st, st_program(vp));
+ }
+
+ edgeflag_culls_prims = edgeflags_enabled && !vertdata_edgeflags &&
+ !st->ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0];
+ if (edgeflag_culls_prims != st->edgeflag_culls_prims) {
+ st->edgeflag_culls_prims = edgeflag_culls_prims;
+ st->dirty |= ST_NEW_RASTERIZER;
}
}
* Update all derived state:
*/
-void st_validate_state( struct st_context *st )
+void st_validate_state( struct st_context *st, enum st_pipeline pipeline )
{
- struct st_state_flags *state = &st->dirty;
- GLuint i;
+ struct gl_context *ctx = st->ctx;
+ uint64_t dirty, pipeline_mask;
+ uint32_t dirty_lo, dirty_hi;
+
+ /* Get Mesa driver state.
+ *
+ * Inactive states are shader states not used by shaders at the moment.
+ */
+ st->dirty |= ctx->NewDriverState & st->active_states & ST_ALL_STATES_MASK;
+ ctx->NewDriverState &= ~st->dirty;
+
+ /* Get pipeline state. */
+ switch (pipeline) {
+ case ST_PIPELINE_RENDER:
+ if (st->ctx->API == API_OPENGL_COMPAT)
+ check_attrib_edgeflag(st);
+
+ if (st->gfx_shaders_may_be_dirty) {
+ check_program_state(st);
+ st->gfx_shaders_may_be_dirty = false;
+ }
- /* Get Mesa driver state. */
- st->dirty.st |= st->ctx->NewDriverState;
- st->ctx->NewDriverState = 0;
+ st_manager_validate_framebuffers(st);
- check_attrib_edgeflag(st);
+ pipeline_mask = ST_PIPELINE_RENDER_STATE_MASK;
+ break;
- if (state->mesa)
- st_flush_bitmap_cache(st);
+ case ST_PIPELINE_CLEAR:
+ st_manager_validate_framebuffers(st);
+ pipeline_mask = ST_PIPELINE_CLEAR_STATE_MASK;
+ break;
- check_program_state( st );
+ case ST_PIPELINE_META:
+ if (st->gfx_shaders_may_be_dirty) {
+ check_program_state(st);
+ st->gfx_shaders_may_be_dirty = false;
+ }
- st_manager_validate_framebuffers(st);
+ st_manager_validate_framebuffers(st);
+ pipeline_mask = ST_PIPELINE_META_STATE_MASK;
+ break;
- if (state->st == 0)
- return;
+ case ST_PIPELINE_UPDATE_FRAMEBUFFER:
+ st_manager_validate_framebuffers(st);
+ pipeline_mask = ST_PIPELINE_UPDATE_FB_STATE_MASK;
+ break;
- /*printf("%s %x/%x\n", __FUNCTION__, state->mesa, state->st);*/
+ case ST_PIPELINE_COMPUTE: {
+ struct st_program *old_cp = st->cp;
+ struct gl_program *new_cp = ctx->ComputeProgram._Current;
-#ifdef DEBUG
- if (1) {
-#else
- if (0) {
-#endif
- /* Debug version which enforces various sanity checks on the
- * state flags which are generated and checked to help ensure
- * state atoms are ordered correctly in the list.
- */
- struct st_state_flags examined, prev;
- memset(&examined, 0, sizeof(examined));
- prev = *state;
-
- for (i = 0; i < Elements(atoms); i++) {
- const struct st_tracked_state *atom = atoms[i];
- struct st_state_flags generated;
-
- /*printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st);*/
-
- if (!(atom->dirty.mesa || atom->dirty.st) ||
- !atom->update) {
- printf("malformed atom %s\n", atom->name);
- assert(0);
- }
-
- if (check_state(state, &atom->dirty)) {
- atoms[i]->update( st );
- /*printf("after: %x\n", atom->dirty.mesa);*/
- }
-
- accumulate_state(&examined, &atom->dirty);
-
- /* generated = (prev ^ state)
- * if (examined & generated)
- * fail;
- */
- xor_states(&generated, &prev, state);
- assert(!check_state(&examined, &generated));
- prev = *state;
+ if (new_cp != &old_cp->Base) {
+ if (old_cp)
+ st->dirty |= old_cp->affected_states;
+ assert(new_cp);
+ st->dirty |= st_program(new_cp)->affected_states;
}
- /*printf("\n");*/
+ st->compute_shader_may_be_dirty = false;
+
+ /*
+ * We add the ST_NEW_FB_STATE bit here as well, because glBindFramebuffer
+ * acts as a barrier that breaks feedback loops between the framebuffer
+ * and textures bound to the framebuffer, even when those textures are
+ * accessed by compute shaders; so we must inform the driver of new
+ * framebuffer state.
+ */
+ pipeline_mask = ST_PIPELINE_COMPUTE_STATE_MASK | ST_NEW_FB_STATE;
+ break;
}
- else {
- for (i = 0; i < Elements(atoms); i++) {
- if (check_state(state, &atoms[i]->dirty))
- atoms[i]->update( st );
- }
+
+ default:
+ unreachable("Invalid pipeline specified");
}
- memset(state, 0, sizeof(*state));
-}
+ dirty = st->dirty & pipeline_mask;
+ if (!dirty)
+ return;
+ dirty_lo = dirty;
+ dirty_hi = dirty >> 32;
+ /* Update states.
+ *
+ * Don't use u_bit_scan64, it may be slower on 32-bit.
+ */
+ while (dirty_lo)
+ update_functions[u_bit_scan(&dirty_lo)](st);
+ while (dirty_hi)
+ update_functions[32 + u_bit_scan(&dirty_hi)](st);
+ /* Clear the render or compute state bits. */
+ st->dirty &= ~pipeline_mask;
+}