[Mesa-dev] [PATCH 27/30] i965: Wholesale replace the color resolve tracking code
Pohjolainen, Topi
topi.pohjolainen at gmail.com
Wed May 31 18:05:21 UTC 2017
On Wed, May 31, 2017 at 10:01:30AM -0700, Jason Ekstrand wrote:
> On Wed, May 31, 2017 at 8:43 AM, Jason Ekstrand <jason at jlekstrand.net>
> wrote:
>
> > On Wed, May 31, 2017 at 6:03 AM, Pohjolainen, Topi <
> > topi.pohjolainen at gmail.com> wrote:
> >
> >> On Fri, May 26, 2017 at 04:30:31PM -0700, Jason Ekstrand wrote:
> >> > This commit reworks the resolve tracking for CCS and MCS to use the new
> >> > isl_aux_state enum. This should provide much more accurate and easy to
> >> > reason about tracking. In order to understand, for instance, the
> >> > intel_miptree_prepare_ccs_access function, one only has to go look at
> >> > the giant comment for the isl_aux_state enum and follow the arrows.
> >> > Unfortunately, there's no good way to split this up without making a
> >> > real mess so there are a bunch of changes in here:
> >> >
> >> > 1) We now do partial resolves. I really have no idea how this ever
> >> > worked before. So far as I can tell, the only time the old code
> >> > ever did a partial resolve was when it was using CCS_D where a
> >> > partial resolve and a full resolve are the same thing.
> >> >
> >> > 2) We are now tracking 4 states instead of 3 for CCS_E. In particular,
> >> > we distinguish between compressed with clear and compressed without
> >> > clear. The end result is that you will never get two partial
> >> > resolves in a row.
> >> >
> >> > 3) The texture view rules are now more correct. Previously, we would
> >> > only bail if compression was not supported by the destination
> >> > format. However, this is not actually correct. Not all format
> >> > pairs are supported for texture views with CCS even if both support
> >> > CCS individually. Fortunately, ISL has a helper for this.
> >> >
> >> > 4) We are no longer using intel_resolve_map for tracking aux state but
> >> > are instead using a simple array of enum isl_aux_state indexed by
> >> > level and layer. This is because, now that we're tracking 4
> >> > different states, it's no longer clear which should be the "default"
> >> > and array lookups are faster than linked list searches.
> >> >
> >> > 5) The new code is very assert-happy. Incorrect transitions will now
> >> > get caught by assertions rather than by rendering corruption.
> >> > ---
> >> > src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 479
> >> +++++++++++++++++---------
> >> > src/mesa/drivers/dri/i965/intel_mipmap_tree.h | 39 +--
> >> > src/mesa/drivers/dri/i965/intel_resolve_map.h | 62 +---
> >> > 3 files changed, 320 insertions(+), 260 deletions(-)
> >> >
> >> > diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
> >> b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
> >> > index cae8358..3e30b2a 100644
> >> > --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
> >> > +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
> >> > @@ -326,7 +326,7 @@ intel_miptree_create_layout(struct brw_context
> >> *brw,
> >> > mt->aux_disable |= INTEL_AUX_DISABLE_CCS;
> >> > mt->is_scanout = (layout_flags & MIPTREE_LAYOUT_FOR_SCANOUT) != 0;
> >> > exec_list_make_empty(&mt->hiz_map);
> >> > - exec_list_make_empty(&mt->color_resolve_map);
> >> > + mt->aux_state = NULL;
> >> > mt->cpp = _mesa_get_format_bytes(format);
> >> > mt->num_samples = num_samples;
> >> > mt->compressed = _mesa_is_format_compressed(format);
> >> > @@ -576,6 +576,46 @@ intel_lower_compressed_format(struct brw_context
> >> *brw, mesa_format format)
> >> > }
> >> > }
> >> >
> >> > +static enum isl_aux_state **
> >> > +create_aux_state_map(struct intel_mipmap_tree *mt,
> >> > + enum isl_aux_state initial)
> >> > +{
> >> > + const uint32_t levels = mt->last_level + 1;
> >> > +
> >> > + uint32_t total_slices = 0;
> >> > + for (uint32_t level = 0; level < levels; level++)
> >> > + total_slices += mt->level[level].depth;
> >> > +
> >> > + const size_t per_level_array_size = levels * sizeof(enum
> >> isl_aux_state *);
> >> > +
> >> > + /* We're going to allocate a single chunk of data for both the
> >> per-level
> >> > + * reference array and the arrays of aux_state. This makes cleanup
> >> > + * significantly easier.
> >> > + */
> >> > + const size_t total_size = per_level_array_size +
> >> > + total_slices * sizeof(enum isl_aux_state);
> >> > + void *data = malloc(total_size);
> >> > + if (data == NULL)
> >> > + return NULL;
> >> > +
> >> > + enum isl_aux_state **per_level_arr = data;
> >> > + enum isl_aux_state *s = data + per_level_array_size;
> >> > + for (uint32_t level = 0; level < levels; level++) {
> >> > + per_level_arr[level] = s;
> >> > + for (uint32_t a = 0; a < mt->level[level].depth; a++)
> >> > + *(s++) = initial;
> >> > + }
> >> > + assert((void *)s == data + total_size);
> >> > +
> >> > + return per_level_arr;
> >> > +}
> >> > +
> >> > +static void
> >> > +free_aux_state_map(enum isl_aux_state **state)
> >> > +{
> >> > + free(state);
> >> > +}
> >> > +
> >> > static struct intel_mipmap_tree *
> >> > miptree_create(struct brw_context *brw,
> >> > GLenum target,
> >> > @@ -935,7 +975,7 @@ intel_miptree_release(struct intel_mipmap_tree **mt)
> >> > free((*mt)->mcs_buf);
> >> > }
> >> > intel_resolve_map_clear(&(*mt)->hiz_map);
> >> > - intel_resolve_map_clear(&(*mt)->color_resolve_map);
> >> > + free_aux_state_map((*mt)->aux_state);
> >> >
> >> > intel_miptree_release(&(*mt)->plane[0]);
> >> > intel_miptree_release(&(*mt)->plane[1]);
> >> > @@ -1482,27 +1522,31 @@ intel_miptree_alloc_mcs(struct brw_context *brw,
> >> > unreachable("Unrecognized sample count in
> >> intel_miptree_alloc_mcs");
> >> > };
> >> >
> >> > + /* Multisampled miptrees are only supported for single level. */
> >> > + assert(mt->first_level == 0);
> >> > + enum isl_aux_state **aux_state =
> >> > + create_aux_state_map(mt, ISL_AUX_STATE_CLEAR);
> >> > + if (!aux_state)
> >> > + return false;
> >> > +
> >> > mt->mcs_buf =
> >> > intel_mcs_miptree_buf_create(brw, mt,
> >> > format,
> >> > mt->logical_width0,
> >> > mt->logical_height0,
> >> > MIPTREE_LAYOUT_ACCELERATED_UPLOAD);
> >> > - if (!mt->mcs_buf)
> >> > + if (!mt->mcs_buf) {
> >> > + free(aux_state);
> >> > return false;
> >> > + }
> >> >
> >> > - intel_miptree_init_mcs(brw, mt, 0xFF);
> >> > + mt->aux_state = aux_state;
> >> >
> >> > - /* Multisampled miptrees are only supported for single level. */
> >> > - assert(mt->first_level == 0);
> >> > - intel_miptree_set_fast_clear_state(brw, mt, mt->first_level, 0,
> >> > - mt->logical_depth0,
> >> > - INTEL_FAST_CLEAR_STATE_CLEAR);
> >> > + intel_miptree_init_mcs(brw, mt, 0xFF);
> >> >
> >> > return true;
> >> > }
> >> >
> >> > -
> >> > bool
> >> > intel_miptree_alloc_non_msrt_mcs(struct brw_context *brw,
> >> > struct intel_mipmap_tree *mt,
> >> > @@ -1528,6 +1572,13 @@ intel_miptree_alloc_non_msrt_mcs(struct
> >> brw_context *brw,
> >> > if (!buf)
> >> > return false;
> >> >
> >> > + enum isl_aux_state **aux_state =
> >> > + create_aux_state_map(mt, ISL_AUX_STATE_PASS_THROUGH);
> >> > + if (!aux_state) {
> >> > + free(buf);
> >> > + return false;
> >> > + }
> >> > +
> >> > buf->size = temp_ccs_surf.size;
> >> > buf->pitch = temp_ccs_surf.row_pitch;
> >> > buf->qpitch = isl_surf_get_array_pitch_sa_rows(&temp_ccs_surf);
> >> > @@ -1549,10 +1600,12 @@ intel_miptree_alloc_non_msrt_mcs(struct
> >> brw_context *brw,
> >> > 1, I915_TILING_Y, &buf->pitch,
> >> alloc_flags);
> >> > if (!buf->bo) {
> >> > free(buf);
> >> > + free(aux_state);
> >> > return false;
> >> > }
> >> >
> >> > mt->mcs_buf = buf;
> >> > + mt->aux_state = aux_state;
> >> >
> >> > /* From Gen9 onwards single-sampled (non-msrt) auxiliary buffers are
> >> > * used for lossless compression which requires similar
> >> initialisation
> >> > @@ -1975,19 +2028,35 @@ intel_miptree_all_slices_resolve_depth(struct
> >> brw_context *brw,
> >> > BLORP_HIZ_OP_DEPTH_RESOLVE);
> >> > }
> >> >
> >> > -enum intel_fast_clear_state
> >> > -intel_miptree_get_fast_clear_state(const struct intel_mipmap_tree *mt,
> >> > - unsigned level, unsigned layer)
> >> > +bool
> >> > +intel_miptree_has_color_unresolved(const struct intel_mipmap_tree *mt,
> >> > + unsigned start_level, unsigned
> >> num_levels,
> >> > + unsigned start_layer, unsigned
> >> num_layers)
> >> > {
> >> > - intel_miptree_check_level_layer(mt, level, layer);
> >> > + assert(_mesa_is_format_color_format(mt->format));
> >> >
> >> > - const struct intel_resolve_map *item =
> >> > - intel_resolve_map_const_get(&mt->color_resolve_map, level,
> >> layer);
> >> > + if (!mt->mcs_buf)
> >> > + return false;
> >> >
> >> > - if (!item)
> >> > - return INTEL_FAST_CLEAR_STATE_RESOLVED;
> >> > + /* Clamp the level range to fit the miptree */
> >> > + assert(start_level + num_levels >= start_level);
> >>
> >> This assert looks odd, and did you mean to have it after adjusting?
> >>
> >
> > No, this is a "does the sum overflow" assertion that we do right before we
> > do a sum. There's a bunch of places where we pass [0, UINT32_MAX] as the
> > range and if someone decided to pass [5, UINT32_MAX], it would overflow.
> >
> > That said, I think my intention was to do the more direct "if (num_levels
> > == INTEL_REMAINING_LEVELS)" check. I switch it to that.
> >
>
> Ok, I think I have something better now. Take a look at
>
> https://cgit.freedesktop.org/~jekstrand/mesa/commit/?h=wip/i965-resolve-rework-v3&id=7ae758d95bee414d677df845d4781254f9f334f3
>
> If you like it, I'm happy to rebase a bit and scatter the changes through
> the series.
That looks nice. Just a small nit, you can pass miptree as const as it is only
used for reading.
>
>
> > > + const uint32_t last_level =
> >> > + MIN2(mt->last_level, start_level + num_levels - 1);
> >> > + start_level = MAX2(mt->first_level, start_level);
> >> > + num_levels = last_level - start_level + 1;
> >> > +
> >> > + for (uint32_t level = start_level; level <= last_level; level++) {
> >> > + const uint32_t level_layers = MIN2(num_layers,
> >> mt->level[level].depth);
> >> > + for (unsigned a = 0; a < level_layers; a++) {
> >> > + enum isl_aux_state aux_state =
> >> > + intel_miptree_get_aux_state(mt, level, start_layer + a);
> >> > + assert(aux_state != ISL_AUX_STATE_AUX_INVALID);
> >> > + if (aux_state != ISL_AUX_STATE_PASS_THROUGH)
> >> > + return true;
> >> > + }
> >> > + }
> >> >
> >> > - return item->fast_clear_state;
> >> > + return false;
> >> > }
> >> >
> >> > static void
> >> > @@ -2014,135 +2083,188 @@ intel_miptree_check_color_resolve(const
> >> struct brw_context *brw,
> >> > (void)layer;
> >> > }
> >> >
> >> > -void
> >> > -intel_miptree_set_fast_clear_state(const struct brw_context *brw,
> >> > - struct intel_mipmap_tree *mt,
> >> > - unsigned level,
> >> > - unsigned first_layer,
> >> > - unsigned num_layers,
> >> > - enum intel_fast_clear_state
> >> new_state)
> >> > -{
> >> > - /* Setting the state to resolved means removing the item from the
> >> list
> >> > - * altogether.
> >> > - */
> >> > - assert(new_state != INTEL_FAST_CLEAR_STATE_RESOLVED);
> >> > +static enum blorp_fast_clear_op
> >> > +get_ccs_d_resolve_op(enum isl_aux_state aux_state,
> >> > + bool ccs_supported, bool fast_clear_supported)
> >> > +{
> >> > + assert(ccs_supported == fast_clear_supported);
> >> >
> >> > - intel_miptree_check_color_resolve(brw, mt, level, first_layer);
> >> > + switch (aux_state) {
> >> > + case ISL_AUX_STATE_CLEAR:
> >> > + case ISL_AUX_STATE_COMPRESSED_CLEAR:
> >> > + if (!ccs_supported)
> >> > + return BLORP_FAST_CLEAR_OP_RESOLVE_FULL;
> >> > + else
> >> > + return BLORP_FAST_CLEAR_OP_NONE;
> >> >
> >> > - assert(first_layer + num_layers <= mt->physical_depth0);
> >> > + case ISL_AUX_STATE_PASS_THROUGH:
> >> > + return BLORP_FAST_CLEAR_OP_NONE;
> >> >
> >> > - for (unsigned i = 0; i < num_layers; i++)
> >> > - intel_resolve_map_set(&mt->color_resolve_map, level,
> >> > - first_layer + i, new_state);
> >> > -}
> >> > + case ISL_AUX_STATE_RESOLVED:
> >> > + case ISL_AUX_STATE_AUX_INVALID:
> >> > + case ISL_AUX_STATE_COMPRESSED_NO_CLEAR:
> >> > + break;
> >> > + }
> >> >
> >> > -bool
> >> > -intel_miptree_has_color_unresolved(const struct intel_mipmap_tree *mt,
> >> > - unsigned start_level, unsigned
> >> num_levels,
> >> > - unsigned start_layer, unsigned
> >> num_layers)
> >> > -{
> >> > - return intel_resolve_map_find_any(&mt->color_resolve_map,
> >> > - start_level, num_levels,
> >> > - start_layer, num_layers) != NULL;
> >> > + unreachable("Invalid aux state for CCS_D");
> >> > }
> >> >
> >> > -void
> >> > -intel_miptree_used_for_rendering(const struct brw_context *brw,
> >> > - struct intel_mipmap_tree *mt,
> >> unsigned level,
> >> > - unsigned start_layer, unsigned
> >> num_layers)
> >> > +static enum blorp_fast_clear_op
> >> > +get_ccs_e_resolve_op(enum isl_aux_state aux_state,
> >> > + bool ccs_supported, bool fast_clear_supported)
> >> > {
> >> > - const bool is_lossless_compressed =
> >> > - intel_miptree_is_lossless_compressed(brw, mt);
> >> > + switch (aux_state) {
> >> > + case ISL_AUX_STATE_CLEAR:
> >> > + case ISL_AUX_STATE_COMPRESSED_CLEAR:
> >> > + if (!ccs_supported)
> >> > + return BLORP_FAST_CLEAR_OP_RESOLVE_FULL;
> >> > + else if (!fast_clear_supported)
> >> > + return BLORP_FAST_CLEAR_OP_RESOLVE_PARTIAL;
> >> > + else
> >> > + return BLORP_FAST_CLEAR_OP_NONE;
> >> >
> >> > - for (unsigned i = 0; i < num_layers; ++i) {
> >> > - const enum intel_fast_clear_state fast_clear_state =
> >> > - intel_miptree_get_fast_clear_state(mt, level, start_layer +
> >> i);
> >> > + case ISL_AUX_STATE_COMPRESSED_NO_CLEAR:
> >> > + if (!ccs_supported)
> >> > + return BLORP_FAST_CLEAR_OP_RESOLVE_FULL;
> >> > + else
> >> > + return BLORP_FAST_CLEAR_OP_NONE;
> >> >
> >> > - /* If the buffer was previously in fast clear state, change it to
> >> > - * unresolved state, since it won't be guaranteed to be clear
> >> after
> >> > - * rendering occurs.
> >> > - */
> >> > - if (is_lossless_compressed ||
> >> > - fast_clear_state == INTEL_FAST_CLEAR_STATE_CLEAR) {
> >> > - intel_miptree_set_fast_clear_state(
> >> > - brw, mt, level, start_layer + i, 1,
> >> > - INTEL_FAST_CLEAR_STATE_UNRESOLVED);
> >> > - }
> >> > + case ISL_AUX_STATE_PASS_THROUGH:
> >> > + return BLORP_FAST_CLEAR_OP_NONE;
> >> > +
> >> > + case ISL_AUX_STATE_RESOLVED:
> >> > + case ISL_AUX_STATE_AUX_INVALID:
> >> > + break;
> >> > }
> >> > +
> >> > + unreachable("Invalid aux state for CCS_E");
> >> > }
> >> >
> >> > -static bool
> >> > -intel_miptree_needs_color_resolve(const struct brw_context *brw,
> >> > - const struct intel_mipmap_tree *mt,
> >> > - int flags)
> >> > +static void
> >> > +intel_miptree_prepare_ccs_access(struct brw_context *brw,
> >> > + struct intel_mipmap_tree *mt,
> >> > + uint32_t level, uint32_t layer,
> >> > + bool aux_supported,
> >> > + bool fast_clear_supported)
> >> > {
> >> > - if (mt->aux_disable & INTEL_AUX_DISABLE_CCS)
> >> > - return false;
> >> > + enum isl_aux_state aux_state = intel_miptree_get_aux_state(mt,
> >> level, layer);
> >> >
> >> > - const bool is_lossless_compressed =
> >> > - intel_miptree_is_lossless_compressed(brw, mt);
> >> > + enum blorp_fast_clear_op resolve_op;
> >> > + if (intel_miptree_is_lossless_compressed(brw, mt)) {
> >> > + resolve_op = get_ccs_e_resolve_op(aux_state, aux_supported,
> >> > + fast_clear_supported);
> >> > + } else {
> >> > + resolve_op = get_ccs_d_resolve_op(aux_state, aux_supported,
> >> > + fast_clear_supported);
> >> > + }
> >> >
> >> > - /* From gen9 onwards there is new compression scheme for single
> >> sampled
> >> > - * surfaces called "lossless compressed". These don't need to be
> >> always
> >> > - * resolved.
> >> > - */
> >> > - if ((flags & INTEL_MIPTREE_IGNORE_CCS_E) && is_lossless_compressed)
> >> > - return false;
> >> > + if (resolve_op != BLORP_FAST_CLEAR_OP_NONE) {
> >> > + intel_miptree_check_color_resolve(brw, mt, level, layer);
> >> > + brw_blorp_resolve_color(brw, mt, level, layer, resolve_op);
> >> >
> >> > - /* Fast color clear resolves only make sense for non-MSAA buffers.
> >> */
> >> > - if (mt->msaa_layout != INTEL_MSAA_LAYOUT_NONE &&
> >> !is_lossless_compressed)
> >> > - return false;
> >> > + switch (resolve_op) {
> >> > + case BLORP_FAST_CLEAR_OP_RESOLVE_FULL:
> >> > + /* The CCS full resolve operation destroys the CCS and sets
> >> it to the
> >> > + * pass-through state. (You can also think of this as being
> >> both a
> >> > + * resolve and an ambiguate in one operation.)
> >> > + */
> >> > + intel_miptree_set_aux_state(brw, mt, level, layer, 1,
> >> > + ISL_AUX_STATE_PASS_THROUGH);
> >> > + break;
> >> >
> >> > - return true;
> >> > + case BLORP_FAST_CLEAR_OP_RESOLVE_PARTIAL:
> >> > + intel_miptree_set_aux_state(brw, mt, level, layer, 1,
> >> > + ISL_AUX_STATE_COMPRESSED_NO_C
> >> LEAR);
> >> > + break;
> >> > +
> >> > + default:
> >> > + unreachable("Invalid resolve op");
> >> > + }
> >> > + }
> >> > }
> >> >
> >> > -static bool
> >> > -intel_miptree_resolve_color(struct brw_context *brw,
> >> > - struct intel_mipmap_tree *mt,
> >> > - uint32_t start_level, uint32_t num_levels,
> >> > - uint32_t start_layer, uint32_t num_layers,
> >> > - int flags)
> >> > +static void
> >> > +intel_miptree_finish_ccs_write(struct brw_context *brw,
> >> > + struct intel_mipmap_tree *mt,
> >> > + uint32_t level, uint32_t layer,
> >> > + bool written_with_ccs)
> >> > {
> >> > - intel_miptree_check_color_resolve(brw, mt, start_level,
> >> start_layer);
> >> > + enum isl_aux_state aux_state = intel_miptree_get_aux_state(mt,
> >> level, layer);
> >> >
> >> > - if (!intel_miptree_needs_color_resolve(brw, mt, flags))
> >> > - return false;
> >> > + if (intel_miptree_is_lossless_compressed(brw, mt)) {
> >> > + switch (aux_state) {
> >> > + case ISL_AUX_STATE_CLEAR:
> >> > + assert(written_with_ccs);
> >> > + intel_miptree_set_aux_state(brw, mt, level, layer, 1,
> >> > + ISL_AUX_STATE_COMPRESSED_CLEAR);
> >> > + break;
> >> >
> >> > - enum blorp_fast_clear_op resolve_op;
> >> > - if (brw->gen >= 9) {
> >> > - if (intel_miptree_is_lossless_compressed(brw, mt)) {
> >> > - resolve_op = BLORP_FAST_CLEAR_OP_RESOLVE_FULL;
> >> > - } else {
> >> > - resolve_op = BLORP_FAST_CLEAR_OP_RESOLVE_PARTIAL;
> >> > + case ISL_AUX_STATE_COMPRESSED_CLEAR:
> >> > + case ISL_AUX_STATE_COMPRESSED_NO_CLEAR:
> >> > + assert(written_with_ccs);
> >> > + break; /* Nothing to do */
> >> > +
> >> > + case ISL_AUX_STATE_PASS_THROUGH:
> >> > + if (written_with_ccs) {
> >> > + intel_miptree_set_aux_state(brw, mt, level, layer, 1,
> >> > + ISL_AUX_STATE_COMPRESSED_NO_CL
> >> EAR);
> >> > + } else {
> >> > + /* Nothing to do */
> >> > + }
> >> > + break;
> >> > +
> >> > + case ISL_AUX_STATE_RESOLVED:
> >> > + case ISL_AUX_STATE_AUX_INVALID:
> >> > + unreachable("Invalid aux state for CCS_E");
> >> > }
> >> > } else {
> >> > - /* Broadwell and earlier do not have a partial resolve */
> >> > - assert(!intel_miptree_is_lossless_compressed(brw, mt));
> >> > - resolve_op = BLORP_FAST_CLEAR_OP_RESOLVE_FULL;
> >> > - }
> >> > + /* CCS_D is a bit simpler */
> >> > + switch (aux_state) {
> >> > + case ISL_AUX_STATE_CLEAR:
> >> > + assert(written_with_ccs);
> >> > + intel_miptree_set_aux_state(brw, mt, level, layer, 1,
> >> > + ISL_AUX_STATE_COMPRESSED_CLEAR);
> >> > + break;
> >> >
> >> > - bool resolved = false;
> >> > - foreach_list_typed_safe(struct intel_resolve_map, map, link,
> >> > - &mt->color_resolve_map) {
> >> > - if (map->level < start_level ||
> >> > - map->level >= (start_level + num_levels) ||
> >> > - map->layer < start_layer ||
> >> > - map->layer >= (start_layer + num_layers))
> >> > - continue;
> >> > + case ISL_AUX_STATE_COMPRESSED_CLEAR:
> >> > + assert(written_with_ccs);
> >> > + break; /* Nothing to do */
> >> > +
> >> > + case ISL_AUX_STATE_PASS_THROUGH:
> >> > + /* Nothing to do */
> >> > + break;
> >> >
> >> > - /* Arrayed fast clear is only supported for gen8+. */
> >> > - assert(brw->gen >= 8 || map->level == 0);
> >> > + case ISL_AUX_STATE_COMPRESSED_NO_CLEAR:
> >> > + case ISL_AUX_STATE_RESOLVED:
> >> > + case ISL_AUX_STATE_AUX_INVALID:
> >> > + unreachable("Invalid aux state for CCS_D");
> >> > + }
> >> > + }
> >> > +}
> >> >
> >> > - intel_miptree_check_level_layer(mt, map->level, map->layer);
> >> > +static void
> >> > +intel_miptree_finish_mcs_write(struct brw_context *brw,
> >> > + struct intel_mipmap_tree *mt,
> >> > + uint32_t level, uint32_t layer,
> >> > + bool written_with_aux)
> >> > +{
> >> > + switch (intel_miptree_get_aux_state(mt, level, layer)) {
> >> > + case ISL_AUX_STATE_CLEAR:
> >> > + assert(written_with_aux);
> >> > + intel_miptree_set_aux_state(brw, mt, level, layer, 1,
> >> > + ISL_AUX_STATE_COMPRESSED_CLEAR);
> >> > + break;
> >> >
> >> > - assert(map->fast_clear_state != INTEL_FAST_CLEAR_STATE_RESOLVE
> >> D);
> >> > + case ISL_AUX_STATE_COMPRESSED_CLEAR:
> >> > + assert(written_with_aux);
> >> > + break; /* Nothing to do */
> >> >
> >> > - brw_blorp_resolve_color(brw, mt, map->level, map->layer,
> >> resolve_op);
> >> > - intel_resolve_map_remove(map);
> >> > - resolved = true;
> >> > + case ISL_AUX_STATE_COMPRESSED_NO_CLEAR:
> >> > + case ISL_AUX_STATE_RESOLVED:
> >> > + case ISL_AUX_STATE_PASS_THROUGH:
> >> > + case ISL_AUX_STATE_AUX_INVALID:
> >> > + unreachable("Invalid aux state for MCS");
> >> > }
> >> > -
> >> > - return resolved;
> >> > }
> >> >
> >> > void
> >> > @@ -2152,17 +2274,29 @@ intel_miptree_prepare_access(struct
> >> brw_context *brw,
> >> > uint32_t start_layer, uint32_t num_layers,
> >> > bool aux_supported, bool
> >> fast_clear_supported)
> >> > {
> >> > + /* Clamp the level range to fit the miptree */
> >> > + assert(start_level + num_levels >= start_level);
> >>
> >> This also looks odd.
> >>
> >> > + const uint32_t last_level =
> >> > + MIN2(mt->last_level, start_level + num_levels - 1);
> >> > + start_level = MAX2(mt->first_level, start_level);
> >> > + num_levels = last_level - start_level + 1;
> >> > +
> >> > if (_mesa_is_format_color_format(mt->format)) {
> >> > if (!mt->mcs_buf)
> >> > return;
> >> >
> >> > if (mt->num_samples > 1) {
> >> > /* Nothing to do for MSAA */
> >> > + assert(aux_supported && fast_clear_supported);
> >> > } else {
> >> > - /* TODO: This is fairly terrible. We can do better. */
> >> > - if (!aux_supported || !fast_clear_supported) {
> >> > - intel_miptree_resolve_color(brw, mt, start_level,
> >> num_levels,
> >> > - start_layer, num_layers, 0);
> >> > + for (uint32_t level = start_level; level <= last_level;
> >> level++) {
> >> > + const uint32_t level_layers =
> >> > + MIN2(num_layers, mt->level[level].depth);
> >> > + for (uint32_t a = 0; a < level_layers; a++) {
> >> > + intel_miptree_prepare_ccs_access(brw, mt, level,
> >> > + start_layer + a,
> >> aux_supported,
> >> > + fast_clear_supported);
> >> > + }
> >> > }
> >> > }
> >> > } else if (mt->format == MESA_FORMAT_S_UINT8) {
> >> > @@ -2197,12 +2331,18 @@ intel_miptree_finish_write(struct brw_context
> >> *brw,
> >> > assert(num_layers <= mt->level[level].depth - start_layer);
> >> >
> >> > if (_mesa_is_format_color_format(mt->format)) {
> >> > + if (!mt->mcs_buf)
> >> > + return;
> >> > +
> >> > if (mt->num_samples > 1) {
> >> > - /* Nothing to do for MSAA */
> >> > + for (uint32_t a = 0; a < num_layers; a++) {
> >> > + intel_miptree_finish_mcs_write(brw, mt, level,
> >> start_layer + a,
> >> > + written_with_aux);
> >> > + }
> >> > } else {
> >> > - if (written_with_aux) {
> >> > - intel_miptree_used_for_rendering(brw, mt, level,
> >> > - start_layer, num_layers);
> >> > + for (uint32_t a = 0; a < num_layers; a++) {
> >> > + intel_miptree_finish_ccs_write(brw, mt, level,
> >> start_layer + a,
> >> > + written_with_aux);
> >> > }
> >> > }
> >> > } else if (mt->format == MESA_FORMAT_S_UINT8) {
> >> > @@ -2231,22 +2371,12 @@ enum isl_aux_state
> >> > intel_miptree_get_aux_state(const struct intel_mipmap_tree *mt,
> >> > uint32_t level, uint32_t layer)
> >> > {
> >> > + intel_miptree_check_level_layer(mt, level, layer);
> >> > +
> >> > if (_mesa_is_format_color_format(mt->format)) {
> >> > assert(mt->mcs_buf != NULL);
> >> > - if (mt->num_samples > 1) {
> >> > - return ISL_AUX_STATE_COMPRESSED_CLEAR;
> >> > - } else {
> >> > - switch (intel_miptree_get_fast_clear_state(mt, level,
> >> layer)) {
> >> > - case INTEL_FAST_CLEAR_STATE_RESOLVED:
> >> > - return ISL_AUX_STATE_RESOLVED;
> >> > - case INTEL_FAST_CLEAR_STATE_UNRESOLVED:
> >> > - return ISL_AUX_STATE_COMPRESSED_CLEAR;
> >> > - case INTEL_FAST_CLEAR_STATE_CLEAR:
> >> > - return ISL_AUX_STATE_CLEAR;
> >> > - default:
> >> > - unreachable("Invalid fast clear state");
> >> > - }
> >> > - }
> >> > + assert(mt->num_samples <= 1 || mt->msaa_layout ==
> >> INTEL_MSAA_LAYOUT_CMS);
> >> > + return mt->aux_state[level][layer];
> >> > } else if (mt->format == MESA_FORMAT_S_UINT8) {
> >> > unreachable("Cannot get aux state for stencil");
> >> > } else {
> >> > @@ -2277,19 +2407,20 @@ intel_miptree_set_aux_state(struct brw_context
> >> *brw,
> >> > num_layers = mt->level[level].depth - start_layer;
> >> > assert(num_layers <= mt->level[level].depth - start_layer);
> >> >
> >> > - /* Right now, this only applies to clears. */
> >> > - assert(aux_state == ISL_AUX_STATE_CLEAR);
> >> > -
> >> > if (_mesa_is_format_color_format(mt->format)) {
> >> > - if (mt->num_samples > 1)
> >> > - assert(mt->msaa_layout == INTEL_MSAA_LAYOUT_CMS);
> >> > + assert(mt->mcs_buf != NULL);
> >> > + assert(mt->num_samples <= 1 || mt->msaa_layout ==
> >> INTEL_MSAA_LAYOUT_CMS);
> >> >
> >> > - assert(level == 0 && start_layer == 0 && num_layers == 1);
> >> > - intel_miptree_set_fast_clear_state(brw, mt, 0, 0, 1,
> >> > - INTEL_FAST_CLEAR_STATE_CLEAR)
> >> ;
> >> > + for (unsigned a = 0; a < num_layers; a++)
> >> > + mt->aux_state[level][start_layer + a] = aux_state;
> >> > } else if (mt->format == MESA_FORMAT_S_UINT8) {
> >> > - assert(!"Cannot set aux state for stencil");
> >> > + unreachable("Cannot get aux state for stencil");
> >> > } else {
> >> > + assert(mt->hiz_buf != NULL);
> >> > +
> >> > + /* Right now, this only applies to clears. */
> >> > + assert(aux_state == ISL_AUX_STATE_CLEAR);
> >> > +
> >> > for (unsigned a = 0; a < num_layers; a++) {
> >> > intel_miptree_check_level_layer(mt, level, start_layer);
> >> > intel_resolve_map_set(&mt->hiz_map, level, start_layer + a,
> >> > @@ -2310,22 +2441,23 @@ intel_miptree_set_aux_state(struct brw_context
> >> *brw,
> >> > * set).
> >> > */
> >> > static bool
> >> > -intel_texture_view_requires_resolve(struct brw_context *brw,
> >> > - struct intel_mipmap_tree *mt,
> >> > - mesa_format format)
> >> > +can_texture_with_ccs(struct brw_context *brw,
> >> > + struct intel_mipmap_tree *mt,
> >> > + mesa_format view_format)
> >> > {
> >> > - if (brw->gen < 9 ||
> >> > - !intel_miptree_is_lossless_compressed(brw, mt))
> >> > - return false;
> >> > + if (!intel_miptree_is_lossless_compressed(brw, mt))
> >> > + return false;
> >> >
> >> > - const enum isl_format isl_format = brw_isl_format_for_mesa_format
> >> (format);
> >> > + enum isl_format isl_mt_format = brw_isl_format_for_mesa_format
> >> (mt->format);
> >> > + enum isl_format isl_view_format = brw_isl_format_for_mesa_format
> >> (view_format);
> >> >
> >> > - if (isl_format_supports_ccs_e(&brw->screen->devinfo, isl_format))
> >> > + if (!isl_formats_are_ccs_e_compatible(&brw->screen->devinfo,
> >> > + isl_mt_format,
> >> isl_view_format)) {
> >> > + perf_debug("Incompatible sampling format (%s) for rbc (%s)\n",
> >> > + _mesa_get_format_name(view_format),
> >> > + _mesa_get_format_name(mt->format));
> >> > return false;
> >> > -
> >> > - perf_debug("Incompatible sampling format (%s) for rbc (%s)\n",
> >> > - _mesa_get_format_name(format),
> >> > - _mesa_get_format_name(mt->format));
> >> > + }
> >> >
> >> > return true;
> >> > }
> >> > @@ -2338,19 +2470,29 @@ intel_miptree_prepare_texture_slices(struct
> >> brw_context *brw,
> >> > uint32_t start_layer, uint32_t
> >> num_layers,
> >> > bool *aux_supported_out)
> >> > {
> >> > - bool aux_supported;
> >> > + bool aux_supported, clear_supported;
> >> > if (_mesa_is_format_color_format(mt->format)) {
> >> > - aux_supported = intel_miptree_is_lossless_compressed(brw, mt) &&
> >> > - !intel_texture_view_requires_resolve(brw, mt,
> >> view_format);
> >> > + if (mt->num_samples > 1) {
> >> > + aux_supported = clear_supported = true;
> >> > + } else {
> >> > + aux_supported = can_texture_with_ccs(brw, mt, view_format);
> >> > +
> >> > + /* Clear color is specified as ints or floats and the
> >> conversion is
> >> > + * done by the sampler. If we have a texture view, we would
> >> have to
> >> > + * perform the clear color conversion manually. Just disable
> >> clear
> >> > + * color.
> >> > + */
> >> > + clear_supported = aux_supported && (mt->format ==
> >> view_format);
> >> > + }
> >> > } else if (mt->format == MESA_FORMAT_S_UINT8) {
> >> > - aux_supported = false;
> >> > + aux_supported = clear_supported = false;
> >> > } else {
> >> > - aux_supported = intel_miptree_sample_with_hiz(brw, mt);
> >> > + aux_supported = clear_supported = intel_miptree_sample_with_hiz(brw,
> >> mt);
> >> > }
> >> >
> >> > intel_miptree_prepare_access(brw, mt, start_level, num_levels,
> >> > start_layer, num_layers,
> >> > - aux_supported, aux_supported);
> >> > + aux_supported, clear_supported);
> >> > if (aux_supported_out)
> >> > *aux_supported_out = aux_supported;
> >> > }
> >> > @@ -2488,7 +2630,8 @@ intel_miptree_make_shareable(struct brw_context
> >> *brw,
> >> > * execute any will likely crash due to the missing aux buffer.
> >> So let's
> >> > * delete all pending ops.
> >> > */
> >> > - exec_list_make_empty(&mt->color_resolve_map);
> >> > + free(mt->aux_state);
> >> > + mt->aux_state = NULL;
> >> > }
> >> >
> >> > if (mt->hiz_buf) {
> >> > diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
> >> b/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
> >> > index 3f3a78d..528c32c 100644
> >> > --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
> >> > +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.h
> >> > @@ -557,7 +557,14 @@ struct intel_mipmap_tree
> >> > * indicates state other than RESOLVED.
> >> > */
> >> > struct exec_list hiz_map; /* List of intel_resolve_map. */
> >> > - struct exec_list color_resolve_map; /* List of intel_resolve_map. */
> >> > +
> >> > + /**
> >> > + * \brief Maps miptree slices to their current aux state
> >> > + *
> >> > + * This two-dimensional array is indexed as [level][layer] and
> >> stores an
> >> > + * aux state for each slice.
> >> > + */
> >> > + enum isl_aux_state **aux_state;
> >> >
> >> > /**
> >> > * \brief Stencil miptree for depthstencil textures.
> >> > @@ -827,41 +834,11 @@ intel_miptree_all_slices_resolve_depth(struct
> >> brw_context *brw,
> >> >
> >> > /**\}*/
> >> >
> >> > -enum intel_fast_clear_state
> >> > -intel_miptree_get_fast_clear_state(const struct intel_mipmap_tree *mt,
> >> > - unsigned level, unsigned layer);
> >> > -
> >> > -void
> >> > -intel_miptree_set_fast_clear_state(const struct brw_context *brw,
> >> > - struct intel_mipmap_tree *mt,
> >> > - unsigned level,
> >> > - unsigned first_layer,
> >> > - unsigned num_layers,
> >> > - enum intel_fast_clear_state
> >> new_state);
> >> > -
> >> > bool
> >> > intel_miptree_has_color_unresolved(const struct intel_mipmap_tree *mt,
> >> > unsigned start_level, unsigned
> >> num_levels,
> >> > unsigned start_layer, unsigned
> >> num_layers);
> >> >
> >> > -/**
> >> > - * Update the fast clear state for a miptree to indicate that it has
> >> been used
> >> > - * for rendering.
> >> > - */
> >> > -void
> >> > -intel_miptree_used_for_rendering(const struct brw_context *brw,
> >> > - struct intel_mipmap_tree *mt,
> >> unsigned level,
> >> > - unsigned start_layer, unsigned
> >> num_layers);
> >> > -
> >> > -/**
> >> > - * Flag values telling color resolve pass which special types of
> >> buffers
> >> > - * can be ignored.
> >> > - *
> >> > - * INTEL_MIPTREE_IGNORE_CCS_E: Lossless compressed (single-sample
> >> > - * compression scheme since gen9)
> >> > - */
> >> > -#define INTEL_MIPTREE_IGNORE_CCS_E (1 << 0)
> >> > -
> >> >
> >> > #define INTEL_REMAINING_LAYERS UINT32_MAX
> >> > #define INTEL_REMAINING_LEVELS UINT32_MAX
> >> > diff --git a/src/mesa/drivers/dri/i965/intel_resolve_map.h
> >> b/src/mesa/drivers/dri/i965/intel_resolve_map.h
> >> > index 17d3983..771d855 100644
> >> > --- a/src/mesa/drivers/dri/i965/intel_resolve_map.h
> >> > +++ b/src/mesa/drivers/dri/i965/intel_resolve_map.h
> >> > @@ -33,63 +33,6 @@ extern "C" {
> >> > #endif
> >> >
> >> > /**
> >> > - * Enum for keeping track of the fast clear state of a buffer
> >> associated with
> >> > - * a miptree.
> >> > - *
> >> > - * Fast clear works by deferring the memory writes that would be used
> >> to clear
> >> > - * the buffer, so that instead of performing them at the time of the
> >> clear
> >> > - * operation, the hardware automatically performs them at the time
> >> that the
> >> > - * buffer is later accessed for rendering. The MCS buffer keeps track
> >> of
> >> > - * which regions of the buffer still have pending clear writes.
> >> > - *
> >> > - * This enum keeps track of the driver's knowledge of pending fast
> >> clears in
> >> > - * the MCS buffer.
> >> > - *
> >> > - * MCS buffers only exist on Gen7+.
> >> > - */
> >> > -enum intel_fast_clear_state
> >> > -{
> >> > - /**
> >> > - * No deferred clears are pending for this miptree, and the
> >> contents of the
> >> > - * color buffer are entirely correct. An MCS buffer may or may not
> >> exist
> >> > - * for this miptree. If it does exist, it is entirely in the "no
> >> deferred
> >> > - * clears pending" state. If it does not exist, it will be created
> >> the
> >> > - * first time a fast color clear is executed.
> >> > - *
> >> > - * In this state, the color buffer can be used for purposes other
> >> than
> >> > - * rendering without needing a render target resolve.
> >> > - *
> >> > - * Since there is no such thing as a "fast color clear resolve" for
> >> MSAA
> >> > - * buffers, an MSAA buffer will never be in this state.
> >> > - */
> >> > - INTEL_FAST_CLEAR_STATE_RESOLVED,
> >> > -
> >> > - /**
> >> > - * An MCS buffer exists for this miptree, and deferred clears are
> >> pending
> >> > - * for some regions of the color buffer, as indicated by the MCS
> >> buffer.
> >> > - * The contents of the color buffer are only correct for the
> >> regions where
> >> > - * the MCS buffer doesn't indicate a deferred clear.
> >> > - *
> >> > - * If a single-sample buffer is in this state, a render target
> >> resolve must
> >> > - * be performed before it can be used for purposes other than
> >> rendering.
> >> > - */
> >> > - INTEL_FAST_CLEAR_STATE_UNRESOLVED,
> >> > -
> >> > - /**
> >> > - * An MCS buffer exists for this miptree, and deferred clears are
> >> pending
> >> > - * for the entire color buffer, and the contents of the MCS buffer
> >> reflect
> >> > - * this. The contents of the color buffer are undefined.
> >> > - *
> >> > - * If a single-sample buffer is in this state, a render target
> >> resolve must
> >> > - * be performed before it can be used for purposes other than
> >> rendering.
> >> > - *
> >> > - * If the client attempts to clear a buffer which is already in
> >> this state,
> >> > - * the clear can be safely skipped, since the buffer is already
> >> clear.
> >> > - */
> >> > - INTEL_FAST_CLEAR_STATE_CLEAR,
> >> > -};
> >> > -
> >> > -/**
> >> > * \brief Map of miptree slices to needed resolves.
> >> > *
> >> > * The map is implemented as a linear doubly-linked list.
> >> > @@ -121,10 +64,7 @@ struct intel_resolve_map {
> >> > uint32_t level;
> >> > uint32_t layer;
> >> >
> >> > - union {
> >> > - enum blorp_hiz_op need;
> >> > - enum intel_fast_clear_state fast_clear_state;
> >> > - };
> >> > + enum blorp_hiz_op need;
> >> > };
> >> >
> >> > void
> >> > --
> >> > 2.5.0.400.gff86faf
> >> >
> >> > _______________________________________________
> >> > mesa-dev mailing list
> >> > mesa-dev at lists.freedesktop.org
> >> > https://lists.freedesktop.org/mailman/listinfo/mesa-dev
> >>
> >
> >
More information about the mesa-dev
mailing list