etnaviv: Add sampler TS support

Sampler TS is an hardware optimization that can be used when rendering
to textures. After rendering to a resource with TS enabled, the
texture unit can use this to bypass lookups to empty tiles. This also
means a resolve-in-place can be avoided to flush the TS.

This commit is also an optimization when not using sampler TS, as
resolve-in-place will now be skipped if a resource has no (valid) TS.

Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>

src/gallium/drivers/etnaviv/etnaviv_emit.c

@@ -632,6 +632,32 @@ etna_emit_state(struct etna_context *ctx)
       /*01668*/ EMIT_STATE_RELOC(TS_DEPTH_SURFACE_BASE, &ctx->framebuffer.TS_DEPTH_SURFACE_BASE);
       /*0166C*/ EMIT_STATE(TS_DEPTH_CLEAR_VALUE, ctx->framebuffer.TS_DEPTH_CLEAR_VALUE);
    }
+   if (unlikely(dirty & ETNA_DIRTY_SAMPLER_VIEWS)) {
+      for (int x = 0; x < VIVS_TS_SAMPLER__LEN; ++x) {
+         if ((1 << x) & active_samplers) {
+            struct etna_sampler_view *sv = etna_sampler_view(ctx->sampler_view[x]);
+            /*01720*/ EMIT_STATE(TS_SAMPLER_CONFIG(x), sv->TS_SAMPLER_CONFIG);
+         }
+      }
+      for (int x = 0; x < VIVS_TS_SAMPLER__LEN; ++x) {
+         if ((1 << x) & active_samplers) {
+            struct etna_sampler_view *sv = etna_sampler_view(ctx->sampler_view[x]);
+            /*01740*/ EMIT_STATE_RELOC(TS_SAMPLER_STATUS_BASE(x), &sv->TS_SAMPLER_STATUS_BASE);
+         }
+      }
+      for (int x = 0; x < VIVS_TS_SAMPLER__LEN; ++x) {
+         if ((1 << x) & active_samplers) {
+            struct etna_sampler_view *sv = etna_sampler_view(ctx->sampler_view[x]);
+            /*01760*/ EMIT_STATE(TS_SAMPLER_CLEAR_VALUE(x), sv->TS_SAMPLER_CLEAR_VALUE);
+         }
+      }
+      for (int x = 0; x < VIVS_TS_SAMPLER__LEN; ++x) {
+         if ((1 << x) & active_samplers) {
+            struct etna_sampler_view *sv = etna_sampler_view(ctx->sampler_view[x]);
+            /*01780*/ EMIT_STATE(TS_SAMPLER_CLEAR_VALUE2(x), sv->TS_SAMPLER_CLEAR_VALUE2);
+         }
+      }
+   }
    if (unlikely(dirty & (ETNA_DIRTY_SAMPLER_VIEWS | ETNA_DIRTY_SAMPLERS))) {
       for (int x = 0; x < VIVS_TE_SAMPLER__LEN; ++x) {
          uint32_t val = 0; /* 0 == sampler inactive */

src/gallium/drivers/etnaviv/etnaviv_texture.c

@@ -112,11 +112,67 @@ etna_delete_sampler_state(struct pipe_context *pctx, void *ss)
    FREE(ss);
 }
 
+/* Return true if the GPU can use sampler TS with this sampler view.
+ * Sampler TS is an optimization used when rendering to textures, where
+ * a resolve-in-place can be avoided when rendering has left a (valid) TS.
+ */
+static bool
+etna_can_use_sampler_ts(struct pipe_sampler_view *view, int num)
+{
+    /* Can use sampler TS when:
+     * - the hardware supports sampler TS.
+     * - the sampler view will be bound to sampler <VIVS_TS_SAMPLER__LEN.
+     *   HALTI5 adds a mapping from sampler to sampler TS unit, but this is AFAIK
+     *   absent on earlier models.
+     * - it is a texture, not a buffer.
+     * - the sampler view has a supported format for sampler TS.
+     * - the sampler will have one LOD, and it happens to be level 0.
+     *   (it is not sure if the hw supports it for other levels, but available
+     *   state strongly suggests only one at a time).
+     * - the resource TS is valid for level 0.
+     */
+   struct etna_resource *rsc = etna_resource(view->texture);
+   struct etna_screen *screen = etna_screen(rsc->base.screen);
+   return VIV_FEATURE(screen, chipMinorFeatures2, TEXTURE_TILED_READ) &&
+      num < VIVS_TS_SAMPLER__LEN &&
+      rsc->base.target != PIPE_BUFFER &&
+      translate_ts_sampler_format(rsc->base.format) != ETNA_NO_MATCH &&
+      view->u.tex.first_level == 0 && MIN2(view->u.tex.last_level, rsc->base.last_level) == 0 &&
+      rsc->levels[0].ts_valid;
+}
+
 static void
-etna_update_sampler_source(struct pipe_sampler_view *view)
+etna_configure_sampler_ts(struct pipe_sampler_view *pview, bool enable)
+{
+   struct etna_sampler_view *view = etna_sampler_view(pview);
+   if (enable) {
+      struct etna_resource *rsc = etna_resource(view->base.texture);
+      struct etna_resource_level *lev = &rsc->levels[0];
+      assert(rsc->ts_bo && lev->ts_valid);
+
+      view->TE_SAMPLER_CONFIG1 |= VIVS_TE_SAMPLER_CONFIG1_USE_TS;
+      view->TS_SAMPLER_CONFIG =
+         VIVS_TS_SAMPLER_CONFIG_ENABLE(1) |
+         VIVS_TS_SAMPLER_CONFIG_FORMAT(translate_ts_sampler_format(rsc->base.format));
+      view->TS_SAMPLER_CLEAR_VALUE = lev->clear_value;
+      view->TS_SAMPLER_CLEAR_VALUE2 = lev->clear_value; /* To handle 64-bit formats this needs a different value */
+      view->TS_SAMPLER_STATUS_BASE.bo = rsc->ts_bo;
+      view->TS_SAMPLER_STATUS_BASE.offset = lev->ts_offset;
+      view->TS_SAMPLER_STATUS_BASE.flags = ETNA_RELOC_READ;
+   } else {
+      view->TE_SAMPLER_CONFIG1 &= ~VIVS_TE_SAMPLER_CONFIG1_USE_TS;
+      view->TS_SAMPLER_CONFIG = 0;
+      view->TS_SAMPLER_STATUS_BASE.bo = NULL;
+   }
+   /* n.b.: relies on caller to mark ETNA_DIRTY_SAMPLER_VIEWS */
+}
+
+static void
+etna_update_sampler_source(struct pipe_sampler_view *view, int num)
 {
    struct etna_resource *base = etna_resource(view->texture);
    struct etna_resource *to = base, *from = base;
+   bool enable_sampler_ts = false;
 
    if (base->external && etna_resource_newer(etna_resource(base->external), base))
       from = etna_resource(base->external);
@@ -129,11 +185,17 @@ etna_update_sampler_source(struct pipe_sampler_view *view)
                          view->texture->last_level);
       to->seqno = from->seqno;
    } else if ((to == from) && etna_resource_needs_flush(to)) {
-      /* Resolve TS if needed, remove when adding sampler TS */
-      etna_copy_resource(view->context, &to->base, &from->base, 0,
-                         view->texture->last_level);
-      to->flush_seqno = from->seqno;
+      if (etna_can_use_sampler_ts(view, num)) {
+         enable_sampler_ts = true;
+         /* Do not set flush_seqno because the resolve-to-self was bypassed */
+      } else {
+         /* Resolve TS if needed */
+         etna_copy_resource(view->context, &to->base, &from->base, 0,
+                            view->texture->last_level);
+         to->flush_seqno = from->seqno;
+      }
    }
+   etna_configure_sampler_ts(view, enable_sampler_ts);
 }
 
 static bool
@@ -332,7 +394,7 @@ etna_set_sampler_views(struct pipe_context *pctx, enum pipe_shader_type shader,
 
    for (unsigned idx = 0; idx < num_views; ++idx) {
       if (views[idx])
-         etna_update_sampler_source(views[idx]);
+         etna_update_sampler_source(views[idx], idx);
    }
 
    switch (shader) {

src/gallium/drivers/etnaviv/etnaviv_texture.h

@@ -62,6 +62,11 @@ struct etna_sampler_view {
    uint32_t TE_SAMPLER_ASTC0;
    struct etna_reloc TE_SAMPLER_LOD_ADDR[VIVS_TE_SAMPLER_LOD_ADDR__LEN];
    unsigned min_lod, max_lod; /* 5.5 fixp */
+
+   uint32_t TS_SAMPLER_CONFIG;
+   struct etna_reloc TS_SAMPLER_STATUS_BASE;
+   uint32_t TS_SAMPLER_CLEAR_VALUE;
+   uint32_t TS_SAMPLER_CLEAR_VALUE2;
 };
 
 static inline struct etna_sampler_view *