panfrost: Add the MALI_WRITES_{Z,S} flags

We discovered 2 new shader flags used when a fragment shader updates
the depth/stencil value through a ZS writeout. If those flags are not
set, the depth/stencil value stored in the depth/stencil tilebuffer
remain unchanged.

While at it, rename unknown2 into flags_hi and rename flags into
flags_lo.

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

diff --git a/src/gallium/drivers/panfrost/pan_assemble.c b/src/gallium/drivers/panfrost/pan_assemble.c
index acb688651e6877049f416f12ecf2345afa26ca6e..fb5cb4cb7e2ac45c8b70fd2ae8bab5f8cf53c459 100644 (file)
--- a/src/gallium/drivers/panfrost/pan_assemble.c
+++ b/src/gallium/drivers/panfrost/pan_assemble.c
@@ -136,7 +136,7 @@ panfrost_shader_compile(
         /* Separate as primary uniform count is truncated */
         state->uniform_count = program.uniform_count;
 
-        meta->midgard1.unknown2 = 8; /* XXX */
+        meta->midgard1.flags_hi = 8; /* XXX */
 
         unsigned default_vec1_swizzle = panfrost_get_default_swizzle(1);
         unsigned default_vec2_swizzle = panfrost_get_default_swizzle(2);

diff --git a/src/gallium/drivers/panfrost/pan_context.c b/src/gallium/drivers/panfrost/pan_context.c
index 169e5956c2f3e5d937b5b30590cc133a468013e0..faa56ee33ffcba0c7cf560e1ae33aa16007f0e24 100644 (file)
--- a/src/gallium/drivers/panfrost/pan_context.c
+++ b/src/gallium/drivers/panfrost/pan_context.c
@@ -843,7 +843,7 @@ panfrost_patch_shader_state(struct panfrost_context *ctx,
         ss->tripipe->texture_count = ctx->sampler_view_count[stage];
         ss->tripipe->sampler_count = ctx->sampler_count[stage];
 
-        ss->tripipe->midgard1.flags = 0x220;
+        ss->tripipe->midgard1.flags_lo = 0x220;
 
         unsigned ubo_count = panfrost_ubo_count(ctx, stage);
         ss->tripipe->midgard1.uniform_buffer_count = ubo_count;
@@ -935,8 +935,8 @@ panfrost_emit_for_draw(struct panfrost_context *ctx, bool with_vertex_data)
                 COPY(midgard1.uniform_count);
                 COPY(midgard1.uniform_buffer_count);
                 COPY(midgard1.work_count);
-                COPY(midgard1.flags);
-                COPY(midgard1.unknown2);
+                COPY(midgard1.flags_lo);
+                COPY(midgard1.flags_hi);
 
 #undef COPY
 
@@ -961,12 +961,12 @@ 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, MALI_EARLY_Z, !variant->can_discard);
+                SET_BIT(ctx->fragment_shader_core.midgard1.flags_lo, MALI_EARLY_Z, !variant->can_discard);
 
                 /* Any time texturing is used, derivatives are implicitly
                  * calculated, so we need to enable helper invocations */
 
-                SET_BIT(ctx->fragment_shader_core.midgard1.flags, MALI_HELPER_INVOCATIONS, variant->helper_invocations);
+                SET_BIT(ctx->fragment_shader_core.midgard1.flags_lo, MALI_HELPER_INVOCATIONS, variant->helper_invocations);
 
                 /* Assign the stencil refs late */
 
@@ -985,7 +985,7 @@ panfrost_emit_for_draw(struct panfrost_context *ctx, bool with_vertex_data)
                  */
 
                 SET_BIT(ctx->fragment_shader_core.unknown2_3, MALI_CAN_DISCARD, variant->can_discard);
-                SET_BIT(ctx->fragment_shader_core.midgard1.flags, 0x400, variant->can_discard);
+                SET_BIT(ctx->fragment_shader_core.midgard1.flags_lo, 0x400, variant->can_discard);
 
                 /* Even on MFBD, the shader descriptor gets blend shaders. It's
                  * *also* copied to the blend_meta appended (by convention),

diff --git a/src/panfrost/include/panfrost-job.h b/src/panfrost/include/panfrost-job.h
index 0861d10e5096b4180a0449e50d95cc574efb5905..1a59f4c77c838b3b52ad400ae986ec25bc1afeb0 100644 (file)
--- a/src/panfrost/include/panfrost-job.h
+++ b/src/panfrost/include/panfrost-job.h
@@ -384,7 +384,10 @@ enum mali_format {
 #define MALI_ALPHA_COVERAGE(clampf) ((uint16_t) (int) (clampf * 15.0f))
 #define MALI_GET_ALPHA_COVERAGE(nibble) ((float) nibble / 15.0f)
 
-/* Applies to midgard1.flags */
+/* Applies to midgard1.flags_lo */
+
+/* Should be set when the fragment shader updates the depth value. */
+#define MALI_WRITES_Z (1 << 4)
 
 /* Should the hardware perform early-Z testing? Normally should be set
  * for performance reasons. Clear if you use: discard,
@@ -406,6 +409,11 @@ enum mali_format {
 #define MALI_READS_ZS (1 << 8)
 #define MALI_READS_TILEBUFFER (1 << 12)
 
+/* Applies to midgard1.flags_hi */
+
+/* Should be set when the fragment shader updates the stencil value. */
+#define MALI_WRITES_S (1 << 2)
+
 /* The raw Midgard blend payload can either be an equation or a shader
  * address, depending on the context */
 
@@ -534,12 +542,12 @@ struct mali_shader_meta {
                 } bifrost1;
                 struct {
                         unsigned uniform_buffer_count : 4;
-                        unsigned flags : 12;
+                        unsigned flags_lo : 12;
 
                         /* vec4 units */
                         unsigned work_count : 5;
                         unsigned uniform_count : 5;
-                        unsigned unknown2 : 6;
+                        unsigned flags_hi : 6;
                 } midgard1;
         };
 

diff --git a/src/panfrost/pandecode/decode.c b/src/panfrost/pandecode/decode.c
index 32a6ba42a7a8850cfd88407fd293f5d402569b64..dc755fcc3640b90fa534da87d5fb658cf686e53e 100644 (file)
--- a/src/panfrost/pandecode/decode.c
+++ b/src/panfrost/pandecode/decode.c
@@ -263,12 +263,18 @@ static const struct pandecode_flag_info mfbd_extra_flag_lo_info[] = {
 #undef FLAG_INFO
 
 #define FLAG_INFO(flag) { MALI_##flag, "MALI_" #flag }
-static const struct pandecode_flag_info shader_midgard1_flag_info [] = {
+static const struct pandecode_flag_info shader_midgard1_flag_lo_info [] = {
+        FLAG_INFO(WRITES_Z),
         FLAG_INFO(EARLY_Z),
         FLAG_INFO(READS_TILEBUFFER),
         FLAG_INFO(READS_ZS),
         {}
 };
+
+static const struct pandecode_flag_info shader_midgard1_flag_hi_info [] = {
+        FLAG_INFO(WRITES_S),
+        {}
+};
 #undef FLAG_INFO
 
 #define FLAG_INFO(flag) { MALI_MFBD_##flag, "MALI_MFBD_" #flag }
@@ -2213,19 +2219,21 @@ pandecode_vertex_tiler_postfix_pre(
                 if (is_bifrost) {
                         pandecode_prop("bifrost1.unk1 = 0x%" PRIx32, s->bifrost1.unk1);
                 } else {
-                        bool helpers = s->midgard1.flags & MALI_HELPER_INVOCATIONS;
-                        s->midgard1.flags &= ~MALI_HELPER_INVOCATIONS;
+                        bool helpers = s->midgard1.flags_lo & MALI_HELPER_INVOCATIONS;
+                        s->midgard1.flags_lo &= ~MALI_HELPER_INVOCATIONS;
 
                         if (helpers != info.helper_invocations) {
                                 pandecode_msg("XXX: expected helpers %u but got %u\n",
                                                 info.helper_invocations, helpers);
                         }
 
-                        pandecode_log(".midgard1.flags = ");
-                        pandecode_log_decoded_flags(shader_midgard1_flag_info, s->midgard1.flags);
+                        pandecode_log(".midgard1.flags_lo = ");
+                        pandecode_log_decoded_flags(shader_midgard1_flag_lo_info, s->midgard1.flags_lo);
                         pandecode_log_cont(",\n");
 
-                        pandecode_prop("midgard1.unknown2 = 0x%" PRIx32, s->midgard1.unknown2);
+                        pandecode_log(".midgard1.flags_hi = ");
+                        pandecode_log_decoded_flags(shader_midgard1_flag_hi_info, s->midgard1.flags_hi);
+                        pandecode_log_cont(",\n");
                 }
 
                 if (s->depth_units || s->depth_factor) {