[Mesa-stable] [PATCH] intel/blorp: Don't vertex fetch directly from clear values

[Jason Ekstrand]

On gen8+, we have to VF cache flush whenever a vertex binding aliases a
previous binding at the same index modulo 4GiB.  We deal with this in
Vulkan by ensuring that vertex buffers and the dynamic state (from which
BLORP pulls its vertex buffers) are in the same 4GiB region of the
address space.  That doesn't work if we're reading clear colors with the
VF unit.  In order to work around this we switch to using MI commands to
copy the clear value into the vertex buffer we allocate for the normal
constant data.

Cc: mesa-stable at lists.freedesktop.org
---
 src/intel/blorp/blorp_genX_exec.h | 66 +++++++++++++------------------
 1 file changed, 27 insertions(+), 39 deletions(-)

diff --git a/src/intel/blorp/blorp_genX_exec.h b/src/intel/blorp/blorp_genX_exec.h
index 446743b5910..323ca9de1e5 100644
--- a/src/intel/blorp/blorp_genX_exec.h
+++ b/src/intel/blorp/blorp_genX_exec.h
@@ -200,6 +200,14 @@ emit_urb_config(struct blorp_batch *batch,
    blorp_emit_urb_config(batch, vs_entry_size, sf_entry_size);
 }
 
+#if GEN_GEN >= 7
+static void
+blorp_emit_memcpy(struct blorp_batch *batch,
+                  struct blorp_address dst,
+                  struct blorp_address src,
+                  uint32_t size);
+#endif
+
 static void
 blorp_emit_vertex_data(struct blorp_batch *batch,
                        const struct blorp_params *params,
@@ -298,6 +306,7 @@ blorp_emit_vertex_buffers(struct blorp_batch *batch,
                           const struct blorp_params *params)
 {
    struct GENX(VERTEX_BUFFER_STATE) vb[3];
+   uint32_t num_vbs = 2;
    memset(vb, 0, sizeof(vb));
 
    struct blorp_address addr;
@@ -308,13 +317,20 @@ blorp_emit_vertex_buffers(struct blorp_batch *batch,
    blorp_emit_input_varying_data(batch, params, &addr, &size);
    blorp_fill_vertex_buffer_state(batch, vb, 1, addr, size, 0);
 
-   uint32_t num_vbs = 2;
+   /* The clear color value starts 16B into the input VB */
+   struct blorp_address clear_color_input_addr = addr;
+   clear_color_input_addr.offset += 16;
+
    if (params->dst_clear_color_as_input) {
+#if GEN_GEN >= 7
       const unsigned clear_color_size =
          GEN_GEN < 10 ? batch->blorp->isl_dev->ss.clear_value_size : 4 * 4;
-      blorp_fill_vertex_buffer_state(batch, vb, num_vbs++,
-                                     params->dst.clear_color_addr,
-                                     clear_color_size, 0);
+      blorp_emit_memcpy(batch, clear_color_input_addr,
+                        params->dst.clear_color_addr,
+                        clear_color_size);
+#else
+      unreachable("MCS partial resolve is not a thing on SNB and earlier");
+#endif
    }
 
    const unsigned num_dwords = 1 + num_vbs * GENX(VERTEX_BUFFER_STATE_length);
@@ -449,49 +465,21 @@ blorp_emit_vertex_elements(struct blorp_batch *batch,
    };
    slot++;
 
-   if (params->dst_clear_color_as_input) {
-      /* If the caller wants the destination indirect clear color, redirect
-       * to vertex buffer 2 where we stored it earlier.  The only users of
-       * an indirect clear color source have that as their only vertex
-       * attribute.
-       */
-      assert(num_varyings == 1);
+   for (unsigned i = 0; i < num_varyings; ++i) {
       ve[slot] = (struct GENX(VERTEX_ELEMENT_STATE)) {
-         .VertexBufferIndex = 2,
+         .VertexBufferIndex = 1,
          .Valid = true,
-         .SourceElementOffset = 0,
-         .Component0Control = VFCOMP_STORE_SRC,
-#if GEN_GEN >= 9
          .SourceElementFormat = ISL_FORMAT_R32G32B32A32_FLOAT,
+         .SourceElementOffset = 16 + i * 4 * sizeof(float),
+         .Component0Control = VFCOMP_STORE_SRC,
          .Component1Control = VFCOMP_STORE_SRC,
          .Component2Control = VFCOMP_STORE_SRC,
          .Component3Control = VFCOMP_STORE_SRC,
-#else
-         /* Clear colors on gen7-8 are for bits out of one dword */
-         .SourceElementFormat = ISL_FORMAT_R32_FLOAT,
-         .Component1Control = VFCOMP_STORE_0,
-         .Component2Control = VFCOMP_STORE_0,
-         .Component3Control = VFCOMP_STORE_0,
+#if GEN_GEN <= 5
+         .DestinationElementOffset = slot * 4,
 #endif
       };
       slot++;
-   } else {
-      for (unsigned i = 0; i < num_varyings; ++i) {
-         ve[slot] = (struct GENX(VERTEX_ELEMENT_STATE)) {
-            .VertexBufferIndex = 1,
-            .Valid = true,
-            .SourceElementFormat = ISL_FORMAT_R32G32B32A32_FLOAT,
-            .SourceElementOffset = 16 + i * 4 * sizeof(float),
-            .Component0Control = VFCOMP_STORE_SRC,
-            .Component1Control = VFCOMP_STORE_SRC,
-            .Component2Control = VFCOMP_STORE_SRC,
-            .Component3Control = VFCOMP_STORE_SRC,
-#if GEN_GEN <= 5
-            .DestinationElementOffset = slot * 4,
-#endif
-         };
-         slot++;
-      }
    }
 
    const unsigned num_dwords =
@@ -1244,7 +1232,7 @@ blorp_emit_pipeline(struct blorp_batch *batch,
 
 #endif /* GEN_GEN >= 6 */
 
-#if GEN_GEN >= 7 && GEN_GEN < 10
+#if GEN_GEN >= 7
 static void
 blorp_emit_memcpy(struct blorp_batch *batch,
                   struct blorp_address dst,
-- 
2.17.0