Re: [Mesa-dev] [PATCH v2 2/9] intel/blorp: Add indirect clear color support to mcs_partial_resolve

Mon, 26 Feb 2018 07:15:32 -0800

On February 26, 2018 05:33:15 "Pohjolainen, Topi" <topi.pohjolai...@gmail.com> wrote: 


On Fri, Feb 23, 2018 at 10:22:57PM -0800, Jason Ekstrand wrote:

This is a bit complicated because we have to get the indirect clear
color in there somehow.  In order to not do any more work in the shader
than needed, we set it up as it's own vertex binding which points
directly at the clear color address specified by the client.
---
 src/intel/blorp/blorp_clear.c     | 25 +++++++++++++++++-
 src/intel/blorp/blorp_genX_exec.h | 54 ++++++++++++++++++++++++++++++++-------
 src/intel/blorp/blorp_priv.h      |  1 +
 3 files changed, 70 insertions(+), 10 deletions(-)

diff --git a/src/intel/blorp/blorp_clear.c b/src/intel/blorp/blorp_clear.c
index dde116f..832e8ee 100644
--- a/src/intel/blorp/blorp_clear.c
+++ b/src/intel/blorp/blorp_clear.c
@@ -833,9 +833,18 @@ blorp_ccs_resolve(struct blorp_batch *batch,
    batch->blorp->exec(batch, &params);
 }

+static nir_ssa_def *
+blorp_nir_bit(nir_builder *b, nir_ssa_def *src, unsigned bit)
+{
+   return nir_iand(b, nir_ushr(b, src, nir_imm_int(b, bit)),
+                      nir_imm_int(b, 1));
+}
+
 struct blorp_mcs_partial_resolve_key
 {
    enum blorp_shader_type shader_type;
+   bool indirect_clear_color;
+   bool int_format;
    uint32_t num_samples;
 };
@@ -845,6 +854,8 @@ blorp_params_get_mcs_partial_resolve_kernel(struct blorp_context *blorp, 
 {
    const struct blorp_mcs_partial_resolve_key blorp_key = {
       .shader_type = BLORP_SHADER_TYPE_MCS_PARTIAL_RESOLVE,
+      .indirect_clear_color = params->dst.clear_color_addr.buffer != NULL,
+      .int_format = isl_format_has_int_channel(params->dst.view.format),
       .num_samples = params->num_samples,
    };
@@ -879,7 +890,18 @@ blorp_params_get_mcs_partial_resolve_kernel(struct blorp_context *blorp, 
    discard->src[0] = nir_src_for_ssa(nir_inot(&b, is_clear));
    nir_builder_instr_insert(&b, &discard->instr);

-   nir_copy_var(&b, frag_color, v_color);
+   nir_ssa_def *clear_color = nir_load_var(&b, v_color);
+   if (blorp_key.indirect_clear_color && blorp->isl_dev->info->gen <= 8) {
+      /* Gen7-8 clear colors are stored as single 0/1 bits */
+      clear_color = nir_vec4(&b, blorp_nir_bit(&b, clear_color, 31),
+                                 blorp_nir_bit(&b, clear_color, 30),
+                                 blorp_nir_bit(&b, clear_color, 29),
+                                 blorp_nir_bit(&b, clear_color, 28));


I was expecting to see right hand side in the form:
blorp_nir_bit(&b nir_channel(&b, clear_color, 0), 31) and so on. So omitting
the channel selection defaults to first or am I missing something else?
More or less, yes. In this case, I think it's actually doing a vectorised thing in blorp_nir_bit and then the vec4 at the end is only taking the first channel of reach argument. The computer then scalarizes the whole thing and throw away the unnerved components. In any case, it's perfectly safe.
In addition I was suprised to see that we can treat it as integer in expressions 
even though it is always defined as float - it gets its type after v_color,
right?
NIR doesn't really have types in the classic sense. The only you're information a SSA value has in NIR is its number of components and bits per component. Whether those boots are interpreted as an int or a float are up to the individual instruction. 


And that in turn is unconditionally declared as:

   nir_variable *v_color =
      BLORP_CREATE_NIR_INPUT(b.shader, clear_color, glsl_vec4_type());
We could have just as easily declared it as a uvec4. The important part is that it's declared to be flat. 


+
+      if (!blorp_key.int_format)
+         clear_color = nir_i2f32(&b, clear_color);
+   }
+   nir_store_var(&b, frag_color, clear_color, 0xf);

    struct brw_wm_prog_key wm_key;
    brw_blorp_init_wm_prog_key(&wm_key);
@@ -925,6 +947,7 @@ blorp_mcs_partial_resolve(struct blorp_batch *batch,

    params.num_samples = params.dst.surf.samples;
    params.num_layers = num_layers;
+   params.dst_clear_color_as_input = surf->clear_color_addr.buffer != NULL;

    memcpy(&params.wm_inputs.clear_color,
           surf->clear_color.f32, sizeof(float) * 4);
diff --git a/src/intel/blorp/blorp_genX_exec.h b/src/intel/blorp/blorp_genX_exec.h 
index cea514e..cc408ca 100644
--- a/src/intel/blorp/blorp_genX_exec.h
+++ b/src/intel/blorp/blorp_genX_exec.h
@@ -297,7 +297,7 @@ static void
 blorp_emit_vertex_buffers(struct blorp_batch *batch,
                           const struct blorp_params *params)
 {
-   struct GENX(VERTEX_BUFFER_STATE) vb[2];
+   struct GENX(VERTEX_BUFFER_STATE) vb[3];
    memset(vb, 0, sizeof(vb));

    struct blorp_address addr;
@@ -308,12 +308,20 @@ 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);

-   const unsigned num_dwords = 1 + GENX(VERTEX_BUFFER_STATE_length) * 2;
+   uint32_t num_vbs = 2;
+   if (params->dst_clear_color_as_input) {
+      blorp_fill_vertex_buffer_state(batch, vb, num_vbs++,
+                                     params->dst.clear_color_addr,
+ batch->blorp->isl_dev->ss.clear_value_size, 
+                                     0);
+   }
+
+   const unsigned num_dwords = 1 + num_vbs * GENX(VERTEX_BUFFER_STATE_length);
    uint32_t *dw = blorp_emitn(batch, GENX(3DSTATE_VERTEX_BUFFERS), num_dwords);
    if (!dw)
       return;

-   for (unsigned i = 0; i < 2; i++) {
+   for (unsigned i = 0; i < num_vbs; i++) {
       GENX(VERTEX_BUFFER_STATE_pack)(batch, dw, &vb[i]);
       dw += GENX(VERTEX_BUFFER_STATE_length);
    }
@@ -440,21 +448,49 @@ blorp_emit_vertex_elements(struct blorp_batch *batch,
    };
    slot++;

-   for (unsigned i = 0; i < num_varyings; ++i) {
+   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);
       ve[slot] = (struct GENX(VERTEX_ELEMENT_STATE)) {
-         .VertexBufferIndex = 1,
+         .VertexBufferIndex = 2,
          .Valid = true,
- .SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32G32B32A32_FLOAT, 
-         .SourceElementOffset = 16 + i * 4 * sizeof(float),
+         .SourceElementOffset = 0,
          .Component0Control = VFCOMP_STORE_SRC,
+#if GEN_GEN >= 9
+ .SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32G32B32A32_FLOAT, 
          .Component1Control = VFCOMP_STORE_SRC,
          .Component2Control = VFCOMP_STORE_SRC,
          .Component3Control = VFCOMP_STORE_SRC,
-#if GEN_GEN <= 5
-         .DestinationElementOffset = slot * 4,
+#else
+         /* Clear colors on gen7-8 are for bits out of one dword */
+ .SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32_FLOAT, 
+         .Component1Control = VFCOMP_STORE_0,
+         .Component2Control = VFCOMP_STORE_0,
+         .Component3Control = VFCOMP_STORE_0,
 #endif
       };
       slot++;
+   } else {
+      for (unsigned i = 0; i < num_varyings; ++i) {
+         ve[slot] = (struct GENX(VERTEX_ELEMENT_STATE)) {
+            .VertexBufferIndex = 1,
+            .Valid = true,
+ .SourceElementFormat = (enum GENX(SURFACE_FORMAT)) 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 =
diff --git a/src/intel/blorp/blorp_priv.h b/src/intel/blorp/blorp_priv.h
index 3fd2203..3811ec6 100644
--- a/src/intel/blorp/blorp_priv.h
+++ b/src/intel/blorp/blorp_priv.h
@@ -196,6 +196,7 @@ struct blorp_params
    bool color_write_disable[4];
    struct brw_blorp_wm_inputs wm_inputs;
    struct blorp_vs_inputs vs_inputs;
+   bool dst_clear_color_as_input;
    unsigned num_samples;
    unsigned num_draw_buffers;
    unsigned num_layers;
--
2.5.0.400.gff86faf

_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev



_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to