Mesa (master): i965/fs: Rework uniform handling

[Jason Ekstrand]

Module: Mesa
Branch: master
Commit: 259f7291de2387aa3ac5f856b39b7b934a1d8e7d
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=259f7291de2387aa3ac5f856b39b7b934a1d8e7d

Author: Jason Ekstrand <jason.ekstrand at intel.com>
Date:   Tue Aug 18 12:00:15 2015 -0700

i965/fs: Rework uniform handling

Previously, we treated the entire UNIFORM file as if it had two elements:
One for direct things and one for indirect.  This is substantially
different from how the old visitor code handled it where each element was
effectively its own uniform.  This commit makes the NIR path more like the
old ir_visitor path where each uniform is separate.  This should allow us
to more easily make decisions about what to push.

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

---

 src/mesa/drivers/dri/i965/brw_fs.h       |    3 ---
 src/mesa/drivers/dri/i965/brw_fs_nir.cpp |   30 ++++++++----------------------
 src/mesa/drivers/dri/i965/brw_nir.c      |    7 +++----
 3 files changed, 11 insertions(+), 29 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h
index 6d18929..6bca762 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.h
+++ b/src/mesa/drivers/dri/i965/brw_fs.h
@@ -318,9 +318,6 @@ public:
    /** Number of uniform variable components visited. */
    unsigned uniforms;
 
-   /** Total number of direct uniforms we can get from NIR */
-   unsigned num_direct_uniforms;
-
    /** Byte-offset for the next available spot in the scratch space buffer. */
    unsigned last_scratch;
 
diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
index ad51d80..a62dbb8 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
@@ -175,19 +175,9 @@ fs_visitor::nir_setup_outputs(nir_shader *shader)
 void
 fs_visitor::nir_setup_uniforms(nir_shader *shader)
 {
-   num_direct_uniforms = shader->num_direct_uniforms;
-
    if (dispatch_width != 8)
       return;
 
-   /* We split the uniform register file in half.  The first half is
-    * entirely direct uniforms.  The second half is indirect.
-    */
-   if (num_direct_uniforms > 0)
-      param_size[0] = num_direct_uniforms;
-   if (shader->num_uniforms > num_direct_uniforms)
-      param_size[num_direct_uniforms] = shader->num_uniforms - num_direct_uniforms;
-
    uniforms = shader->num_uniforms;
 
    if (shader_prog) {
@@ -200,15 +190,19 @@ fs_visitor::nir_setup_uniforms(nir_shader *shader)
             nir_setup_builtin_uniform(var);
          else
             nir_setup_uniform(var);
+
+         param_size[var->data.driver_location] = type_size_scalar(var->type);
       }
    } else {
-      /* prog_to_nir doesn't create uniform variables; set param up directly. */
+      /* prog_to_nir only creates a single giant uniform variable so we can
+       * just set param up directly. */
       for (unsigned p = 0; p < prog->Parameters->NumParameters; p++) {
          for (unsigned int i = 0; i < 4; i++) {
             stage_prog_data->param[4 * p + i] =
                &prog->Parameters->ParameterValues[p][i];
          }
       }
+      param_size[0] = prog->Parameters->NumParameters * 4;
    }
 }
 
@@ -1504,21 +1498,13 @@ fs_visitor::nir_emit_intrinsic(const fs_builder &bld, nir_intrinsic_instr *instr
       has_indirect = true;
       /* fallthrough */
    case nir_intrinsic_load_uniform: {
-      unsigned index = instr->const_index[0] + instr->const_index[1];
-
-      fs_reg uniform_reg;
-      if (index < num_direct_uniforms) {
-         uniform_reg = fs_reg(UNIFORM, 0);
-      } else {
-         uniform_reg = fs_reg(UNIFORM, num_direct_uniforms);
-         index -= num_direct_uniforms;
-      }
+      fs_reg uniform_reg(UNIFORM, instr->const_index[0]);
+      uniform_reg.reg_offset = instr->const_index[1];
 
       for (unsigned j = 0; j < instr->num_components; j++) {
-         fs_reg src = offset(retype(uniform_reg, dest.type), bld, index);
+         fs_reg src = offset(retype(uniform_reg, dest.type), bld, j);
          if (has_indirect)
             src.reladdr = new(mem_ctx) fs_reg(get_nir_src(instr->src[0]));
-         index++;
 
          bld.MOV(dest, src);
          dest = offset(dest, bld, 1);
diff --git a/src/mesa/drivers/dri/i965/brw_nir.c b/src/mesa/drivers/dri/i965/brw_nir.c
index 3d04363..dfac44f 100644
--- a/src/mesa/drivers/dri/i965/brw_nir.c
+++ b/src/mesa/drivers/dri/i965/brw_nir.c
@@ -111,10 +111,9 @@ brw_create_nir(struct brw_context *brw,
    nir_optimize(nir, is_scalar);
 
    if (is_scalar) {
-      nir_assign_var_locations_direct_first(nir, &nir->uniforms,
-                                            &nir->num_direct_uniforms,
-                                            &nir->num_uniforms,
-                                            type_size_scalar);
+      nir_assign_var_locations(&nir->uniforms,
+                               &nir->num_uniforms,
+                               type_size_scalar);
       nir_assign_var_locations(&nir->inputs, &nir->num_inputs, type_size_scalar);
       nir_assign_var_locations(&nir->outputs, &nir->num_outputs, type_size_scalar);
       nir_lower_io(nir, type_size_scalar);