Mesa (master): panfrost: Set the MALI_WRITES_{Z,S} flags when needed

[GitLab Mirror]

Module: Mesa
Branch: master
Commit: 38c20696a5358d6898c4ee96fb127d603c1e1404
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=38c20696a5358d6898c4ee96fb127d603c1e1404

Author: Boris Brezillon <boris.brezillon at collabora.com>
Date:   Fri Jan 31 10:55:49 2020 +0100

panfrost: Set the MALI_WRITES_{Z,S} flags when needed

In order to make Z/S writes from fragment shaders effective, we need
to set the MALI_WRITES_{Z,S} flags when the shader has a
FRAG_RESULT_{DEPTH,STENCIL} output variable.

Now that shaders can change the S value, we can expose the
STENCIL_EXPORT cap.

Signed-off-by: Boris Brezillon <boris.brezillon at collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig at collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3697>

---

 src/gallium/drivers/panfrost/pan_assemble.c | 4 ++++
 src/gallium/drivers/panfrost/pan_context.c  | 9 ++++++++-
 src/gallium/drivers/panfrost/pan_context.h  | 2 ++
 src/gallium/drivers/panfrost/pan_screen.c   | 3 +++
 4 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/src/gallium/drivers/panfrost/pan_assemble.c b/src/gallium/drivers/panfrost/pan_assemble.c
index fb5cb4cb7e2..03ca63ccc32 100644
--- a/src/gallium/drivers/panfrost/pan_assemble.c
+++ b/src/gallium/drivers/panfrost/pan_assemble.c
@@ -114,6 +114,10 @@ panfrost_shader_compile(
         case MESA_SHADER_FRAGMENT:
                 meta->attribute_count = 0;
                 meta->varying_count = util_bitcount64(s->info.inputs_read);
+                if (s->info.outputs_written & BITFIELD64_BIT(FRAG_RESULT_DEPTH))
+                        state->writes_depth = true;
+                if (s->info.outputs_written & BITFIELD64_BIT(FRAG_RESULT_STENCIL))
+                        state->writes_stencil = true;
                 break;
         case MESA_SHADER_COMPUTE:
                 /* TODO: images */
diff --git a/src/gallium/drivers/panfrost/pan_context.c b/src/gallium/drivers/panfrost/pan_context.c
index faa56ee33ff..417c900c961 100644
--- a/src/gallium/drivers/panfrost/pan_context.c
+++ b/src/gallium/drivers/panfrost/pan_context.c
@@ -961,7 +961,14 @@ panfrost_emit_for_draw(struct panfrost_context *ctx, bool with_vertex_data)
                 /* Depending on whether it's legal to in the given shader, we
                  * try to enable early-z testing (or forward-pixel kill?) */
 
-                SET_BIT(ctx->fragment_shader_core.midgard1.flags_lo, MALI_EARLY_Z, !variant->can_discard);
+                SET_BIT(ctx->fragment_shader_core.midgard1.flags_lo, MALI_EARLY_Z,
+                        !variant->can_discard && !variant->writes_depth);
+
+                /* Add the writes Z/S flags if needed. */
+                SET_BIT(ctx->fragment_shader_core.midgard1.flags_lo,
+                        MALI_WRITES_Z, variant->writes_depth);
+                SET_BIT(ctx->fragment_shader_core.midgard1.flags_hi,
+                        MALI_WRITES_S, variant->writes_stencil);
 
                 /* Any time texturing is used, derivatives are implicitly
                  * calculated, so we need to enable helper invocations */
diff --git a/src/gallium/drivers/panfrost/pan_context.h b/src/gallium/drivers/panfrost/pan_context.h
index 98d327cfde4..79d082da11f 100644
--- a/src/gallium/drivers/panfrost/pan_context.h
+++ b/src/gallium/drivers/panfrost/pan_context.h
@@ -210,6 +210,8 @@ struct panfrost_shader_state {
         int uniform_count;
         bool can_discard;
         bool writes_point_size;
+        bool writes_depth;
+        bool writes_stencil;
         bool reads_point_coord;
         bool reads_face;
         bool reads_frag_coord;
diff --git a/src/gallium/drivers/panfrost/pan_screen.c b/src/gallium/drivers/panfrost/pan_screen.c
index 0fe6062df91..19d646bf7e7 100644
--- a/src/gallium/drivers/panfrost/pan_screen.c
+++ b/src/gallium/drivers/panfrost/pan_screen.c
@@ -245,6 +245,9 @@ panfrost_get_param(struct pipe_screen *screen, enum pipe_cap param)
                 return (int)(system_memory >> 20);
         }
 
+        case PIPE_CAP_SHADER_STENCIL_EXPORT:
+                return 1;
+
         case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
                 return 4;