nvc0: invalidate textures/samplers on GK104+

Like Fermi, textures and samplers are aliased between 3D and compute,
especially the TIC_FLUSH/TSC_FLUSH methods and we have to re-validate
these resources when switching between the two pipelines.

This fixes a GPU hang with Elemental (and most likely with other UE4 demos).

Tested on GK107 and GM107.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
CC: <mesa-stable@lists.freedesktop.org>

diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c b/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c
index 4fa262195f3bde0faa56ba2cf4ca4ed76084eee4..cbc270d9f19f2d7b09e6b45ba9ecd5e0c0f58a00 100644 (file)
--- a/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c
+++ b/src/gallium/drivers/nouveau/nvc0/nvc0_tex.c
@@ -605,13 +605,11 @@ void nvc0_validate_textures(struct nvc0_context *nvc0)
       PUSH_DATA (nvc0->base.pushbuf, 0);
    }
 
-   if (nvc0->screen->base.class_3d < NVE4_3D_CLASS) {
-      /* Invalidate all CP textures because they are aliased. */
-      for (int i = 0; i < nvc0->num_textures[5]; i++)
-         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_CP_TEX(i));
-      nvc0->textures_dirty[5] = ~0;
-      nvc0->dirty_cp |= NVC0_NEW_CP_TEXTURES;
-   }
+   /* Invalidate all CP textures because they are aliased. */
+   for (int i = 0; i < nvc0->num_textures[5]; i++)
+      nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_CP_TEX(i));
+   nvc0->textures_dirty[5] = ~0;
+   nvc0->dirty_cp |= NVC0_NEW_CP_TEXTURES;
 }
 
 bool
@@ -716,11 +714,9 @@ void nvc0_validate_samplers(struct nvc0_context *nvc0)
       PUSH_DATA (nvc0->base.pushbuf, 0);
    }
 
-   if (nvc0->screen->base.class_3d < NVE4_3D_CLASS) {
-      /* Invalidate all CP samplers because they are aliased. */
-      nvc0->samplers_dirty[5] = ~0;
-      nvc0->dirty_cp |= NVC0_NEW_CP_SAMPLERS;
-   }
+   /* Invalidate all CP samplers because they are aliased. */
+   nvc0->samplers_dirty[5] = ~0;
+   nvc0->dirty_cp |= NVC0_NEW_CP_SAMPLERS;
 }
 
 /* Upload the "diagonal" entries for the possible texture sources ($t == $s).

diff --git a/src/gallium/drivers/nouveau/nvc0/nve4_compute.c b/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
index f1ea4d95b91c46d8564cd5c64da75b835bf94f5b..e85e9b48b7f8ec1e88488627967374c71d04e6a7 100644 (file)
--- a/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
+++ b/src/gallium/drivers/nouveau/nvc0/nve4_compute.c
@@ -304,7 +304,13 @@ nve4_compute_validate_samplers(struct nvc0_context *nvc0)
       BEGIN_NVC0(nvc0->base.pushbuf, NVE4_CP(TSC_FLUSH), 1);
       PUSH_DATA (nvc0->base.pushbuf, 0);
    }
+
+   /* Invalidate all 3D samplers because they are aliased. */
+   for (int s = 0; s < 5; s++)
+      nvc0->samplers_dirty[s] = ~0;
+   nvc0->dirty_3d |= NVC0_NEW_3D_SAMPLERS;
 }
+
 /* (Code duplicated at bottom for various non-convincing reasons.
  *  E.g. we might want to use the COMPUTE subchannel to upload TIC/TSC
  *  entries to avoid a subchannel switch.
@@ -754,6 +760,14 @@ nve4_compute_validate_textures(struct nvc0_context *nvc0)
    }
 
    nvc0->state.num_textures[s] = nvc0->num_textures[s];
+
+   /* Invalidate all 3D textures because they are aliased. */
+   for (int s = 0; s < 5; s++) {
+      for (int i = 0; i < nvc0->num_textures[s]; i++)
+         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_TEX(s, i));
+      nvc0->textures_dirty[s] = ~0;
+   }
+   nvc0->dirty_3d |= NVC0_NEW_3D_TEXTURES;
 }