compiler: Move double_inputs to gl_program::DualSlotInputs

Previously, we had two field in shader_info: double_inputs_read and
double_inputs.  Presumably, the one was for all double inputs that are
read and the other is all that exist.  However, because nir_gather_info
regenerates these two values, there is a possibility, if a variable gets
deleted, that the value of double_inputs could change over time.  This
is a problem because double_inputs is used to remap the input locations
to a two-slot-per-dvec3/4 scheme for i965.  If that mapping were to
change between glsl_to_nir and back-end state setup, we would fall over
when trying to map the NIR outputs back onto the GL location space.

This commit changes the way slot re-mapping works.  Instead of the
double_inputs field in shader_info, it adds a DualSlotInputs bitfield to
gl_program.  By having it in gl_program, we more easily guarantee that
NIR passes won't touch it after it's been set.  It also makes more sense
to put it in a GL data structure since it's really a mapping from GL
slots to back-end and/or NIR slots and not really a NIR shader thing.

Tested-by: Alejandro Piñeiro <apinheiro@igalia.com> (ARB_gl_spirv tests)
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

src/compiler/glsl/glsl_to_nir.cpp

@@ -149,8 +149,11 @@ glsl_to_nir(const struct gl_shader_program *shader_prog,
     * two locations. For instance, if we have in the IR code a dvec3 attr0 in
     * location 0 and vec4 attr1 in location 1, in NIR attr0 will use
     * locations/slots 0 and 1, and attr1 will use location/slot 2 */
-   if (shader->info.stage == MESA_SHADER_VERTEX)
-      nir_remap_attributes(shader, options);
+   if (shader->info.stage == MESA_SHADER_VERTEX) {
+      sh->Program->DualSlotInputs = nir_get_dual_slot_attributes(shader);
+      if (options->vs_inputs_dual_locations)
+         nir_remap_dual_slot_attributes(shader, sh->Program->DualSlotInputs);
+   }
 
    shader->info.name = ralloc_asprintf(shader, "GLSL%d", shader_prog->Name);
    if (shader_prog->Label)
@@ -344,15 +347,6 @@ nir_visitor::visit(ir_variable *ir)
             var->data.compact = ir->type->without_array()->is_scalar();
          }
       }
-
-      /* Mark all the locations that require two slots */
-      if (shader->info.stage == MESA_SHADER_VERTEX &&
-          glsl_type_is_dual_slot(glsl_without_array(var->type))) {
-         for (unsigned i = 0; i < glsl_count_attribute_slots(var->type, true); i++) {
-            uint64_t bitfield = BITFIELD64_BIT(var->data.location + i);
-            shader->info.vs.double_inputs |= bitfield;
-         }
-      }
       break;
 
    case ir_var_shader_out:

src/compiler/glsl/ir_set_program_inouts.cpp

@@ -118,7 +118,7 @@ mark(struct gl_program *prog, ir_variable *var, int offset, int len,
          /* double inputs read is only for vertex inputs */
          if (stage == MESA_SHADER_VERTEX &&
              var->type->without_array()->is_dual_slot())
-            prog->info.vs.double_inputs_read |= bitfield;
+            prog->DualSlotInputs |= bitfield;
 
          if (stage == MESA_SHADER_FRAGMENT) {
             prog->info.fs.uses_sample_qualifier |= var->data.sample;

src/compiler/glsl/serialize.cpp

@@ -1035,6 +1035,7 @@ write_shader_metadata(struct blob *metadata, gl_linked_shader *shader)
    struct gl_program *glprog = shader->Program;
    unsigned i;
 
+   blob_write_uint64(metadata, glprog->DualSlotInputs);
    blob_write_bytes(metadata, glprog->TexturesUsed,
                     sizeof(glprog->TexturesUsed));
    blob_write_uint64(metadata, glprog->SamplersUsed);
@@ -1088,6 +1089,7 @@ read_shader_metadata(struct blob_reader *metadata,
 {
    unsigned i;
 
+   glprog->DualSlotInputs = blob_read_uint64(metadata);
    blob_copy_bytes(metadata, (uint8_t *) glprog->TexturesUsed,
                    sizeof(glprog->TexturesUsed));
    glprog->SamplersUsed = blob_read_uint64(metadata);

src/compiler/nir/nir.c

@@ -1854,23 +1854,48 @@ nir_system_value_from_intrinsic(nir_intrinsic_op intrin)
    }
 }
 
+uint64_t
+nir_get_dual_slot_attributes(nir_shader *shader)
+{
+   assert(shader->info.stage == MESA_SHADER_VERTEX);
+
+   uint64_t dual_slot = 0;
+   nir_foreach_variable(var, &shader->inputs) {
+      if (glsl_type_is_dual_slot(glsl_without_array(var->type))) {
+         unsigned slots = glsl_count_attribute_slots(var->type, true);
+         dual_slot |= BITFIELD64_MASK(slots) << var->data.location;
+      }
+   }
+
+   return dual_slot;
+}
+
 /* OpenGL utility method that remaps the location attributes if they are
  * doubles. Not needed for vulkan due the differences on the input location
  * count for doubles on vulkan vs OpenGL
  */
 void
-nir_remap_attributes(nir_shader *shader,
-                     const nir_shader_compiler_options *options)
+nir_remap_dual_slot_attributes(nir_shader *shader, uint64_t dual_slot)
 {
-   if (options->vs_inputs_dual_locations) {
-      nir_foreach_variable(var, &shader->inputs) {
-         var->data.location +=
-            _mesa_bitcount_64(shader->info.vs.double_inputs &
-                              BITFIELD64_MASK(var->data.location));
-      }
-   }
+   assert(shader->info.stage == MESA_SHADER_VERTEX);
 
-   /* Once the remap is done, reset double_inputs_read, so later it will have
-    * which location/slots are doubles */
-   shader->info.vs.double_inputs = 0;
+   nir_foreach_variable(var, &shader->inputs) {
+      var->data.location +=
+         _mesa_bitcount_64(dual_slot & BITFIELD64_MASK(var->data.location));
+   }
+}
+
+/* Returns an attribute mask that has been re-compacted using the given
+ * dual_slot mask.
+ */
+uint64_t
+nir_get_single_slot_attribs_mask(uint64_t attribs, uint64_t dual_slot)
+{
+   while (dual_slot) {
+      unsigned loc = u_bit_scan64(&dual_slot);
+      /* mask of all bits up to and including loc */
+      uint64_t mask = BITFIELD64_MASK(loc + 1);
+      attribs = (attribs & mask) | ((attribs & ~mask) >> 1);
+   }
+   return attribs;
 }

src/compiler/nir/nir.h

@@ -3039,8 +3039,10 @@ bool nir_opt_conditional_discard(nir_shader *shader);
 
 void nir_sweep(nir_shader *shader);
 
-void nir_remap_attributes(nir_shader *shader,
-                          const nir_shader_compiler_options *options);
+uint64_t nir_get_dual_slot_attributes(nir_shader *shader);
+void nir_remap_dual_slot_attributes(nir_shader *shader,
+                                    uint64_t dual_slot);
+uint64_t nir_get_single_slot_attribs_mask(uint64_t attribs, uint64_t dual_slot);
 
 nir_intrinsic_op nir_intrinsic_from_system_value(gl_system_value val);
 gl_system_value nir_system_value_from_intrinsic(nir_intrinsic_op intrin);

src/compiler/nir/nir_gather_info.c

@@ -54,11 +54,6 @@ set_io_mask(nir_shader *shader, nir_variable *var, int offset, int len,
          else
             shader->info.inputs_read |= bitfield;
 
-         /* double inputs read is only for vertex inputs */
-         if (shader->info.stage == MESA_SHADER_VERTEX &&
-             glsl_type_is_dual_slot(glsl_without_array(var->type)))
-            shader->info.vs.double_inputs_read |= bitfield;
-
          if (shader->info.stage == MESA_SHADER_FRAGMENT) {
             shader->info.fs.uses_sample_qualifier |= var->data.sample;
          }
@@ -417,7 +412,6 @@ nir_shader_gather_info(nir_shader *shader, nir_function_impl *entrypoint)
    shader->info.system_values_read = 0;
    if (shader->info.stage == MESA_SHADER_VERTEX) {
       shader->info.vs.double_inputs = 0;
-      shader->info.vs.double_inputs_read = 0;
    }
    if (shader->info.stage == MESA_SHADER_FRAGMENT) {
       shader->info.fs.uses_sample_qualifier = false;

src/compiler/shader_info.h

@@ -134,9 +134,6 @@ typedef struct shader_info {
       struct {
          /* Which inputs are doubles */
          uint64_t double_inputs;
-
-         /* Which inputs are actually read and are double */
-         uint64_t double_inputs_read;
       } vs;
 
       struct {

src/mesa/drivers/dri/i965/brw_draw_upload.c

@@ -29,6 +29,7 @@
 #include "main/enums.h"
 #include "main/macros.h"
 #include "main/glformats.h"
+#include "nir.h"
 
 #include "brw_draw.h"
 #include "brw_defines.h"
@@ -454,10 +455,14 @@ brw_prepare_vertices(struct brw_context *brw)
 {
    const struct gen_device_info *devinfo = &brw->screen->devinfo;
    struct gl_context *ctx = &brw->ctx;
+   /* BRW_NEW_VERTEX_PROGRAM */
+   const struct gl_program *vp = brw->programs[MESA_SHADER_VERTEX];
    /* BRW_NEW_VS_PROG_DATA */
    const struct brw_vs_prog_data *vs_prog_data =
       brw_vs_prog_data(brw->vs.base.prog_data);
-   GLbitfield64 vs_inputs = vs_prog_data->inputs_read;
+   GLbitfield64 vs_inputs =
+      nir_get_single_slot_attribs_mask(vs_prog_data->inputs_read,
+                                       vp->DualSlotInputs);
    const unsigned char *ptr = NULL;
    GLuint interleaved = 0;
    unsigned int min_index = brw->vb.min_index + brw->basevertex;
@@ -486,16 +491,12 @@ brw_prepare_vertices(struct brw_context *brw)
    /* Accumulate the list of enabled arrays. */
    brw->vb.nr_enabled = 0;
    while (vs_inputs) {
-      GLuint first = ffsll(vs_inputs) - 1;
-      assert (first < 64);
-      GLuint index =
-         first - DIV_ROUND_UP(_mesa_bitcount_64(vs_prog_data->double_inputs_read &
-                                                BITFIELD64_MASK(first)), 2);
+      const unsigned index = ffsll(vs_inputs) - 1;
+      assert(index < 64);
+
       struct brw_vertex_element *input = &brw->vb.inputs[index];
-      input->is_dual_slot = (vs_prog_data->double_inputs_read & BITFIELD64_BIT(first)) != 0;
-      vs_inputs &= ~BITFIELD64_BIT(first);
-      if (input->is_dual_slot)
-         vs_inputs &= ~BITFIELD64_BIT(first + 1);
+      input->is_dual_slot = (vp->DualSlotInputs & BITFIELD64_BIT(index)) != 0;
+      vs_inputs &= ~BITFIELD64_BIT(index);
       brw->vb.enabled[brw->vb.nr_enabled++] = input;
    }
 

src/mesa/drivers/dri/i965/genX_state_upload.c

@@ -933,6 +933,7 @@ static const struct brw_tracked_state genX(vertices) = {
       .mesa = _NEW_POLYGON,
       .brw = BRW_NEW_BATCH |
              BRW_NEW_BLORP |
+             BRW_NEW_VERTEX_PROGRAM |
              BRW_NEW_VERTICES |
              BRW_NEW_VS_PROG_DATA,
    },

src/mesa/main/glspirv.c

@@ -182,20 +182,6 @@ _mesa_spirv_link_shaders(struct gl_context *ctx, struct gl_shader_program *prog)
       prog->last_vert_prog = prog->_LinkedShaders[last_vert_stage - 1]->Program;
 }
 
-static void
-nir_compute_double_inputs(nir_shader *shader,
-                          const nir_shader_compiler_options *options)
-{
-   nir_foreach_variable(var, &shader->inputs) {
-      if (glsl_type_is_dual_slot(glsl_without_array(var->type))) {
-         for (unsigned i = 0; i < glsl_count_attribute_slots(var->type, true); i++) {
-            uint64_t bitfield = BITFIELD64_BIT(var->data.location + i);
-            shader->info.vs.double_inputs |= bitfield;
-         }
-      }
-   }
-}
-
 nir_shader *
 _mesa_spirv_to_nir(struct gl_context *ctx,
                    const struct gl_shader_program *prog,
@@ -278,8 +264,10 @@ _mesa_spirv_to_nir(struct gl_context *ctx,
    NIR_PASS_V(nir, nir_split_per_member_structs);
 
    if (nir->info.stage == MESA_SHADER_VERTEX) {
-      nir_compute_double_inputs(nir, options);
-      nir_remap_attributes(nir, options);
+      uint64_t dual_slot_inputs = nir_get_dual_slot_attributes(nir);
+      if (options->vs_inputs_dual_locations)
+         nir_remap_dual_slot_attributes(nir, dual_slot_inputs);
+      linked_shader->Program->DualSlotInputs = dual_slot_inputs;
    }
 
    return nir;

src/mesa/main/mtypes.h

@@ -2066,6 +2066,21 @@ struct gl_program
    /** Is this program written to on disk shader cache */
    bool program_written_to_cache;
 
+   /** A bitfield indicating which vertex shader inputs consume two slots
+    *
+    * This is used for mapping from single-slot input locations in the GL API
+    * to dual-slot double input locations in the shader.  This field is set
+    * once as part of linking and never updated again to ensure the mapping
+    * remains consistent.
+    *
+    * Note: There may be dual-slot variables in the original shader source
+    * which do not appear in this bitfield due to having been eliminated by
+    * the compiler prior to DualSlotInputs being calculated.  There may also
+    * be bits set in this bitfield which are set but which the shader never
+    * reads due to compiler optimizations eliminating such variables after
+    * DualSlotInputs is calculated.
+    */
+   GLbitfield64 DualSlotInputs;
    /** Subset of OutputsWritten outputs written with non-zero index. */
    GLbitfield64 SecondaryOutputsWritten;
    /** TEXTURE_x_BIT bitmask */

src/mesa/state_tracker/st_glsl_to_nir.cpp

@@ -91,7 +91,7 @@ st_nir_assign_vs_in_locations(struct gl_program *prog, nir_shader *nir)
       if ((prog->info.inputs_read & BITFIELD64_BIT(attr)) != 0) {
          input_to_index[attr] = num_inputs;
          num_inputs++;
-         if ((prog->info.vs.double_inputs_read & BITFIELD64_BIT(attr)) != 0) {
+         if ((prog->DualSlotInputs & BITFIELD64_BIT(attr)) != 0) {
             /* add placeholder for second part of a double attribute */
             num_inputs++;
          }

src/mesa/state_tracker/st_glsl_to_tgsi.cpp

@@ -7129,7 +7129,7 @@ get_mesa_program_tgsi(struct gl_context *ctx,
    _mesa_copy_linked_program_data(shader_program, shader);
    shrink_array_declarations(v->inputs, v->num_inputs,
                              &prog->info.inputs_read,
-                             prog->info.vs.double_inputs_read,
+                             prog->DualSlotInputs,
                              &prog->info.patch_inputs_read);
    shrink_array_declarations(v->outputs, v->num_outputs,
                              &prog->info.outputs_written, 0ULL,

src/mesa/state_tracker/st_program.c

@@ -406,8 +406,7 @@ st_translate_vertex_program(struct st_context *st,
          stvp->input_to_index[attr] = stvp->num_inputs;
          stvp->index_to_input[stvp->num_inputs] = attr;
          stvp->num_inputs++;
-         if ((stvp->Base.info.vs.double_inputs_read &
-              BITFIELD64_BIT(attr)) != 0) {
+         if ((stvp->Base.DualSlotInputs & BITFIELD64_BIT(attr)) != 0) {
             /* add placeholder for second part of a double attribute */
             stvp->index_to_input[stvp->num_inputs] = ST_DOUBLE_ATTRIB_PLACEHOLDER;
             stvp->num_inputs++;