iris: Use 3DSTATE_CONSTANT_ALL when possible.

Use this new instruction introduced in Gen12. The instruction itself is
smaller, and it also allows us to emit a single instruction to all
stages that have the same push constant buffers (e.g. when they don't
have constant buffers).

There's one restriction to use this instruction, though: the length
field is only 5 bits long, so we need to check whether we can use it,
and fallback to the old 3DSTATE_CONSTANT_XS if that field is >= 32.

v2 (Suggestions from Caio):
 - use max_length instead of large_buffers.
 - remove UNUSED and use #if GEN_GEN >= 12 instead.
 - inline "buffers" and drop BITSET_RANGE() usage.
 - add assert(n <= max_pointers)
 - move emit to outside of the loop.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>

diff --git a/src/gallium/drivers/iris/iris_genx_macros.h b/src/gallium/drivers/iris/iris_genx_macros.h
index 58680341a50d9a0b823ffea51a787c8b9f9ab760..74fe9139add6b7ec88b08f716c4f9d3a41e098d8 100644 (file)
--- a/src/gallium/drivers/iris/iris_genx_macros.h
+++ b/src/gallium/drivers/iris/iris_genx_macros.h
@@ -88,12 +88,15 @@ __gen_combine_address(struct iris_batch *batch, void *location,
 #define iris_pack_command(cmd, dst, name) \
    _iris_pack_command(NULL, cmd, dst, name)
 
-#define iris_pack_state(cmd, dst, name)                           \
+#define _iris_pack_state(batch, cmd, dst, name)                   \
    for (struct cmd name = {},                                     \
         *_dst = (void *)(dst); __builtin_expect(_dst != NULL, 1); \
-        __genxml_cmd_pack(cmd)(NULL, (void *)_dst, &name),        \
+        __genxml_cmd_pack(cmd)(batch, (void *)_dst, &name),       \
         _dst = NULL)
 
+#define iris_pack_state(cmd, dst, name)                           \
+   _iris_pack_state(NULL, cmd, dst, name)
+
 #define iris_emit_cmd(batch, cmd, name) \
    _iris_pack_command(batch, cmd, __gen_get_batch_dwords(batch, __genxml_cmd_length(cmd)), name)
 

diff --git a/src/gallium/drivers/iris/iris_state.c b/src/gallium/drivers/iris/iris_state.c
index 1eea996ab3779d8458d89800e8920f53e3a36b77..8ea79f722a3e6811ec0bfe448b9222ed702740cc 100644 (file)
--- a/src/gallium/drivers/iris/iris_state.c
+++ b/src/gallium/drivers/iris/iris_state.c
@@ -5095,6 +5095,7 @@ struct push_bos {
       uint32_t length;
    } buffers[4];
    int buffer_count;
+   uint32_t max_length;
 };
 
 static void
@@ -5114,6 +5115,9 @@ setup_constant_buffers(struct iris_context *ice,
       if (range->length == 0)
          continue;
 
+      if (range->length > push_bos->max_length)
+         push_bos->max_length = range->length;
+
       /* Range block is a binding table index, map back to UBO index. */
       unsigned block_index = iris_bti_to_group_index(
          &shader->bt, IRIS_SURFACE_GROUP_UBO, range->block);
@@ -5169,6 +5173,45 @@ emit_push_constant_packets(struct iris_context *ice,
    }
 }
 
+#if GEN_GEN >= 12
+static void
+emit_push_constant_packet_all(struct iris_context *ice,
+                              struct iris_batch *batch,
+                              uint32_t shader_mask,
+                              const struct push_bos *push_bos)
+{
+   if (!push_bos) {
+      iris_emit_cmd(batch, GENX(3DSTATE_CONSTANT_ALL), pc) {
+         pc.ShaderUpdateEnable = shader_mask;
+      }
+      return;
+   }
+
+   const uint32_t n = push_bos->buffer_count;
+   const uint32_t max_pointers = 4;
+   const uint32_t num_dwords = 2 + 2 * n;
+   uint32_t const_all[2 + 2 * max_pointers];
+   uint32_t *dw = &const_all[0];
+
+   assert(n <= max_pointers);
+   iris_pack_command(GENX(3DSTATE_CONSTANT_ALL), dw, all) {
+      all.DWordLength = num_dwords - 2;
+      all.ShaderUpdateEnable = shader_mask;
+      all.PointerBufferMask = (1 << n) - 1;
+   }
+   dw += 2;
+
+   for (int i = 0; i < n; i++) {
+      _iris_pack_state(batch, GENX(3DSTATE_CONSTANT_ALL_DATA),
+                       dw + i * 2, data) {
+         data.PointerToConstantBuffer = push_bos->buffers[i].addr;
+         data.ConstantBufferReadLength = push_bos->buffers[i].length;
+      }
+   }
+   iris_batch_emit(batch, const_all, sizeof(uint32_t) * num_dwords);
+}
+#endif
+
 static void
 iris_upload_dirty_render_state(struct iris_context *ice,
                                struct iris_batch *batch,
@@ -5347,6 +5390,10 @@ iris_upload_dirty_render_state(struct iris_context *ice,
       }
    }
 
+#if GEN_GEN >= 12
+   uint32_t nobuffer_stages = 0;
+#endif
+
    for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
       if (!(dirty & (IRIS_DIRTY_CONSTANTS_VS << stage)))
          continue;
@@ -5362,9 +5409,33 @@ iris_upload_dirty_render_state(struct iris_context *ice,
 
       struct push_bos push_bos = {};
       setup_constant_buffers(ice, batch, stage, &push_bos);
+
+#if GEN_GEN >= 12
+      /* If this stage doesn't have any push constants, emit it later in a
+       * single CONSTANT_ALL packet with all the other stages.
+       */
+      if (push_bos.buffer_count == 0) {
+         nobuffer_stages |= 1 << stage;
+         continue;
+      }
+
+      /* The Constant Buffer Read Length field from 3DSTATE_CONSTANT_ALL
+       * contains only 5 bits, so we can only use it for buffers smaller than
+       * 32.
+       */
+      if (push_bos.max_length < 32) {
+         emit_push_constant_packet_all(ice, batch, 1 << stage, &push_bos);
+         continue;
+      }
+#endif
       emit_push_constant_packets(ice, batch, stage, &push_bos);
    }
 
+#if GEN_GEN >= 12
+   if (nobuffer_stages)
+      emit_push_constant_packet_all(ice, batch, nobuffer_stages, NULL);
+#endif
+
    for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
       /* Gen9 requires 3DSTATE_BINDING_TABLE_POINTERS_XS to be re-emitted
        * in order to commit constants.  TODO: Investigate "Disable Gather