spirv: parse unstructured CFG
[mesa.git] / src / compiler / spirv / spirv_to_nir.c
index 799926eb037b8195dfc583f9de5baee3b10466b2..747d1a0cdcce5e929984a34ce794f406e718ce1b 100644 (file)
@@ -775,6 +775,33 @@ wrap_type_in_array(const struct glsl_type *type,
                           glsl_get_explicit_stride(array_type));
 }
 
+static bool
+vtn_type_needs_explicit_layout(struct vtn_builder *b, enum vtn_variable_mode mode)
+{
+   /* For OpenCL we never want to strip the info from the types, and it makes
+    * type comparisons easier in later stages.
+    */
+   if (b->options->environment == NIR_SPIRV_OPENCL)
+      return true;
+
+   switch (mode) {
+   case vtn_variable_mode_input:
+   case vtn_variable_mode_output:
+      /* Layout decorations kept because we need offsets for XFB arrays of
+       * blocks.
+       */
+      return b->shader->info.has_transform_feedback_varyings;
+
+   case vtn_variable_mode_ssbo:
+   case vtn_variable_mode_phys_ssbo:
+   case vtn_variable_mode_ubo:
+      return true;
+
+   default:
+      return false;
+   }
+}
+
 const struct glsl_type *
 vtn_type_get_nir_type(struct vtn_builder *b, struct vtn_type *type,
                       enum vtn_variable_mode mode)
@@ -787,16 +814,65 @@ vtn_type_get_nir_type(struct vtn_builder *b, struct vtn_type *type,
    }
 
    if (mode == vtn_variable_mode_uniform) {
-      struct vtn_type *tail = vtn_type_without_array(type);
-      if (tail->base_type == vtn_base_type_image) {
-         return wrap_type_in_array(tail->glsl_image, type->type);
-      } else if (tail->base_type == vtn_base_type_sampler) {
-         return wrap_type_in_array(glsl_bare_sampler_type(), type->type);
-      } else if (tail->base_type == vtn_base_type_sampled_image) {
-         return wrap_type_in_array(tail->image->glsl_image, type->type);
+      switch (type->base_type) {
+      case vtn_base_type_array: {
+         const struct glsl_type *elem_type =
+            vtn_type_get_nir_type(b, type->array_element, mode);
+
+         return glsl_array_type(elem_type, type->length,
+                                glsl_get_explicit_stride(type->type));
+      }
+
+      case vtn_base_type_struct: {
+         bool need_new_struct = false;
+         const uint32_t num_fields = type->length;
+         NIR_VLA(struct glsl_struct_field, fields, num_fields);
+         for (unsigned i = 0; i < num_fields; i++) {
+            fields[i] = *glsl_get_struct_field_data(type->type, i);
+            const struct glsl_type *field_nir_type =
+               vtn_type_get_nir_type(b, type->members[i], mode);
+            if (fields[i].type != field_nir_type) {
+               fields[i].type = field_nir_type;
+               need_new_struct = true;
+            }
+         }
+         if (need_new_struct) {
+            if (glsl_type_is_interface(type->type)) {
+               return glsl_interface_type(fields, num_fields,
+                                          /* packing */ 0, false,
+                                          glsl_get_type_name(type->type));
+            } else {
+               return glsl_struct_type(fields, num_fields,
+                                       glsl_get_type_name(type->type),
+                                       glsl_struct_type_is_packed(type->type));
+            }
+         } else {
+            /* No changes, just pass it on */
+            return type->type;
+         }
+      }
+
+      case vtn_base_type_image:
+         return type->glsl_image;
+
+      case vtn_base_type_sampler:
+         return glsl_bare_sampler_type();
+
+      case vtn_base_type_sampled_image:
+         return type->image->glsl_image;
+
+      default:
+         return type->type;
       }
    }
 
+   /* Layout decorations are allowed but ignored in certain conditions,
+    * to allow SPIR-V generators perform type deduplication.  Discard
+    * unnecessary ones when passing to NIR.
+    */
+   if (!vtn_type_needs_explicit_layout(b, mode))
+      return glsl_get_bare_type(type->type);
+
    return type->type;
 }
 
@@ -2857,6 +2933,8 @@ vtn_handle_image(struct vtn_builder *b, SpvOp opcode,
    SpvScope scope = SpvScopeInvocation;
    SpvMemorySemanticsMask semantics = 0;
 
+   enum gl_access_qualifier access = 0;
+
    struct vtn_value *res_val;
    switch (opcode) {
    case SpvOpAtomicExchange:
@@ -2879,6 +2957,7 @@ vtn_handle_image(struct vtn_builder *b, SpvOp opcode,
       image = *res_val->image;
       scope = vtn_constant_uint(b, w[4]);
       semantics = vtn_constant_uint(b, w[5]);
+      access |= ACCESS_COHERENT;
       break;
 
    case SpvOpAtomicStore:
@@ -2886,6 +2965,7 @@ vtn_handle_image(struct vtn_builder *b, SpvOp opcode,
       image = *res_val->image;
       scope = vtn_constant_uint(b, w[2]);
       semantics = vtn_constant_uint(b, w[3]);
+      access |= ACCESS_COHERENT;
       break;
 
    case SpvOpImageQuerySize:
@@ -3032,7 +3112,6 @@ vtn_handle_image(struct vtn_builder *b, SpvOp opcode,
     * chains to find the NonUniform decoration.  It's either right there or we
     * can assume it doesn't exist.
     */
-   enum gl_access_qualifier access = 0;
    vtn_foreach_decoration(b, res_val, non_uniform_decoration_cb, &access);
    nir_intrinsic_set_access(intrin, access);
 
@@ -3300,6 +3379,8 @@ vtn_handle_atomics(struct vtn_builder *b, SpvOp opcode,
       nir_intrinsic_op op  = get_ssbo_nir_atomic_op(b, opcode);
       atomic = nir_intrinsic_instr_create(b->nb.shader, op);
 
+      nir_intrinsic_set_access(atomic, ACCESS_COHERENT);
+
       int src = 0;
       switch (opcode) {
       case SpvOpAtomicLoad:
@@ -3351,6 +3432,9 @@ vtn_handle_atomics(struct vtn_builder *b, SpvOp opcode,
       atomic = nir_intrinsic_instr_create(b->nb.shader, op);
       atomic->src[0] = nir_src_for_ssa(&deref->dest.ssa);
 
+      if (ptr->mode != vtn_variable_mode_workgroup)
+         nir_intrinsic_set_access(atomic, ACCESS_COHERENT);
+
       switch (opcode) {
       case SpvOpAtomicLoad:
          atomic->num_components = glsl_get_vector_elements(deref_type);
@@ -4441,14 +4525,7 @@ vtn_handle_execution_mode(struct vtn_builder *b, struct vtn_value *entry_point,
       b->shader->info.cs.local_size[2] = mode->operands[2];
       break;
 
-   case SpvExecutionModeLocalSizeId:
-      b->shader->info.cs.local_size[0] = vtn_constant_uint(b, mode->operands[0]);
-      b->shader->info.cs.local_size[1] = vtn_constant_uint(b, mode->operands[1]);
-      b->shader->info.cs.local_size[2] = vtn_constant_uint(b, mode->operands[2]);
-      break;
-
    case SpvExecutionModeLocalSizeHint:
-   case SpvExecutionModeLocalSizeHintId:
       break; /* Nothing to do with this */
 
    case SpvExecutionModeOutputVertices:
@@ -4578,8 +4655,60 @@ vtn_handle_execution_mode(struct vtn_builder *b, struct vtn_value *entry_point,
    case SpvExecutionModeDenormFlushToZero:
    case SpvExecutionModeSignedZeroInfNanPreserve:
    case SpvExecutionModeRoundingModeRTE:
-   case SpvExecutionModeRoundingModeRTZ:
-      /* Already handled in vtn_handle_rounding_mode_in_execution_mode() */
+   case SpvExecutionModeRoundingModeRTZ: {
+      unsigned execution_mode = 0;
+      switch (mode->exec_mode) {
+      case SpvExecutionModeDenormPreserve:
+         switch (mode->operands[0]) {
+         case 16: execution_mode = FLOAT_CONTROLS_DENORM_PRESERVE_FP16; break;
+         case 32: execution_mode = FLOAT_CONTROLS_DENORM_PRESERVE_FP32; break;
+         case 64: execution_mode = FLOAT_CONTROLS_DENORM_PRESERVE_FP64; break;
+         default: vtn_fail("Floating point type not supported");
+         }
+         break;
+      case SpvExecutionModeDenormFlushToZero:
+         switch (mode->operands[0]) {
+         case 16: execution_mode = FLOAT_CONTROLS_DENORM_FLUSH_TO_ZERO_FP16; break;
+         case 32: execution_mode = FLOAT_CONTROLS_DENORM_FLUSH_TO_ZERO_FP32; break;
+         case 64: execution_mode = FLOAT_CONTROLS_DENORM_FLUSH_TO_ZERO_FP64; break;
+         default: vtn_fail("Floating point type not supported");
+         }
+         break;
+      case SpvExecutionModeSignedZeroInfNanPreserve:
+         switch (mode->operands[0]) {
+         case 16: execution_mode = FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP16; break;
+         case 32: execution_mode = FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP32; break;
+         case 64: execution_mode = FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP64; break;
+         default: vtn_fail("Floating point type not supported");
+         }
+         break;
+      case SpvExecutionModeRoundingModeRTE:
+         switch (mode->operands[0]) {
+         case 16: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP16; break;
+         case 32: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP32; break;
+         case 64: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP64; break;
+         default: vtn_fail("Floating point type not supported");
+         }
+         break;
+      case SpvExecutionModeRoundingModeRTZ:
+         switch (mode->operands[0]) {
+         case 16: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP16; break;
+         case 32: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP32; break;
+         case 64: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP64; break;
+         default: vtn_fail("Floating point type not supported");
+         }
+         break;
+      default:
+         break;
+      }
+
+      b->shader->info.float_controls_execution_mode |= execution_mode;
+      break;
+   }
+
+   case SpvExecutionModeLocalSizeId:
+   case SpvExecutionModeLocalSizeHintId:
+      /* Handled later by vtn_handle_execution_mode_id(). */
       break;
 
    default:
@@ -4590,60 +4719,28 @@ vtn_handle_execution_mode(struct vtn_builder *b, struct vtn_value *entry_point,
 }
 
 static void
-vtn_handle_rounding_mode_in_execution_mode(struct vtn_builder *b, struct vtn_value *entry_point,
-                                           const struct vtn_decoration *mode, void *data)
+vtn_handle_execution_mode_id(struct vtn_builder *b, struct vtn_value *entry_point,
+                             const struct vtn_decoration *mode, UNUSED void *data)
 {
-   vtn_assert(b->entry_point == entry_point);
 
-   unsigned execution_mode = 0;
+   vtn_assert(b->entry_point == entry_point);
 
-   switch(mode->exec_mode) {
-   case SpvExecutionModeDenormPreserve:
-      switch (mode->operands[0]) {
-      case 16: execution_mode = FLOAT_CONTROLS_DENORM_PRESERVE_FP16; break;
-      case 32: execution_mode = FLOAT_CONTROLS_DENORM_PRESERVE_FP32; break;
-      case 64: execution_mode = FLOAT_CONTROLS_DENORM_PRESERVE_FP64; break;
-      default: vtn_fail("Floating point type not supported");
-      }
-      break;
-   case SpvExecutionModeDenormFlushToZero:
-      switch (mode->operands[0]) {
-      case 16: execution_mode = FLOAT_CONTROLS_DENORM_FLUSH_TO_ZERO_FP16; break;
-      case 32: execution_mode = FLOAT_CONTROLS_DENORM_FLUSH_TO_ZERO_FP32; break;
-      case 64: execution_mode = FLOAT_CONTROLS_DENORM_FLUSH_TO_ZERO_FP64; break;
-      default: vtn_fail("Floating point type not supported");
-      }
-       break;
-   case SpvExecutionModeSignedZeroInfNanPreserve:
-      switch (mode->operands[0]) {
-      case 16: execution_mode = FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP16; break;
-      case 32: execution_mode = FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP32; break;
-      case 64: execution_mode = FLOAT_CONTROLS_SIGNED_ZERO_INF_NAN_PRESERVE_FP64; break;
-      default: vtn_fail("Floating point type not supported");
-      }
-      break;
-   case SpvExecutionModeRoundingModeRTE:
-      switch (mode->operands[0]) {
-      case 16: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP16; break;
-      case 32: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP32; break;
-      case 64: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP64; break;
-      default: vtn_fail("Floating point type not supported");
-      }
+   switch (mode->exec_mode) {
+   case SpvExecutionModeLocalSizeId:
+      b->shader->info.cs.local_size[0] = vtn_constant_uint(b, mode->operands[0]);
+      b->shader->info.cs.local_size[1] = vtn_constant_uint(b, mode->operands[1]);
+      b->shader->info.cs.local_size[2] = vtn_constant_uint(b, mode->operands[2]);
       break;
-   case SpvExecutionModeRoundingModeRTZ:
-      switch (mode->operands[0]) {
-      case 16: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP16; break;
-      case 32: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP32; break;
-      case 64: execution_mode = FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP64; break;
-      default: vtn_fail("Floating point type not supported");
-      }
+
+   case SpvExecutionModeLocalSizeHintId:
+      /* Nothing to do with this hint. */
       break;
 
    default:
+      /* Nothing to do.  Literal execution modes already handled by
+       * vtn_handle_execution_mode(). */
       break;
    }
-
-   b->shader->info.float_controls_execution_mode |= execution_mode;
 }
 
 static bool
@@ -5438,12 +5535,9 @@ spirv_to_nir(const uint32_t *words, size_t word_count,
    if (stage == MESA_SHADER_GEOMETRY)
       b->shader->info.gs.invocations = 1;
 
-   /* Parse rounding mode execution modes. This has to happen earlier than
-    * other changes in the execution modes since they can affect, for example,
-    * the result of the floating point constants.
-    */
+   /* Parse execution modes. */
    vtn_foreach_execution_mode(b, b->entry_point,
-                              vtn_handle_rounding_mode_in_execution_mode, NULL);
+                              vtn_handle_execution_mode, NULL);
 
    b->specializations = spec;
    b->num_specializations = num_spec;
@@ -5452,9 +5546,11 @@ spirv_to_nir(const uint32_t *words, size_t word_count,
    words = vtn_foreach_instruction(b, words, word_end,
                                    vtn_handle_variable_or_type_instruction);
 
-   /* Parse execution modes */
+   /* Parse execution modes that depend on IDs. Must happen after we have
+    * constants parsed.
+    */
    vtn_foreach_execution_mode(b, b->entry_point,
-                              vtn_handle_execution_mode, NULL);
+                              vtn_handle_execution_mode_id, NULL);
 
    if (b->workgroup_size_builtin) {
       vtn_assert(b->workgroup_size_builtin->type->type ==
@@ -5498,6 +5594,9 @@ spirv_to_nir(const uint32_t *words, size_t word_count,
    if (entry_point->num_params && b->shader->info.stage == MESA_SHADER_KERNEL)
       entry_point = vtn_emit_kernel_entry_point_wrapper(b, entry_point);
 
+   /* structurize the CFG */
+   nir_lower_goto_ifs(b->shader);
+
    entry_point->is_entrypoint = true;
 
    /* When multiple shader stages exist in the same SPIR-V module, we