Francisco Casas : vkd3d-shader/hlsl: Emit fixmes on non-constant vector addressing.

[Alexandre Julliard]

Module: vkd3d
Branch: master
Commit: 43ff28b00b2be949cb9f5cb31d28a73a4648dde2
URL:    https://gitlab.winehq.org/wine/vkd3d/-/commit/43ff28b00b2be949cb9f5cb31d28a7...

Author: Francisco Casas 
Date:   Thu Jan 11 16:54:20 2024 -0300

vkd3d-shader/hlsl: Emit fixmes on non-constant vector addressing.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=56162

Storing to a vector component using a non-constant index is not allowed
on profiles lower than 6.0. Unless this happens inside a loop that can be
unrolled, which we are not doing yet.

For this reason, a validate_nonconstant_vector_store_derefs pass is
added to detect these cases.

Ideally we would want to emit an hlsl_error on this pass, but before
implementing loop unrolling, we could reach this point on valid HLSL.

Also, as pointed out by Nikolay in the mentioned bug, currently
new_offset_from_path_index() fails an assertion when this happens,
because it expects an hlsl_ir_constant, so a check is added.
It also felt correct to emit an hlsl_fixme there, despite the
redundancy.

---

libs/vkd3d-shader/hlsl_codegen.c       | 39 ++++++++++++++++++++++++++++++++++
 tests/hlsl/vector-indexing.shader_test | 13 ++++++++++++
 2 files changed, 52 insertions(+)

diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c
index acc19dd8..d37bef15 100644
--- a/libs/vkd3d-shader/hlsl_codegen.c
+++ b/libs/vkd3d-shader/hlsl_codegen.c
@@ -32,6 +32,11 @@ static struct hlsl_ir_node *new_offset_from_path_index(struct hlsl_ctx *ctx, str
     switch (type->class)
     {
         case HLSL_CLASS_VECTOR:
+            if (idx->type != HLSL_IR_CONSTANT)
+            {
+                hlsl_fixme(ctx, &idx->loc, "Non-constant vector addressing.");
+                break;
+            }
             *offset_component += hlsl_ir_constant(idx)->value.u[0].u;
             break;

@@ -2478,6 +2483,38 @@ static bool lower_nonconstant_vector_derefs(struct hlsl_ctx *ctx, struct hlsl_ir
     return false;
 }

+static bool validate_nonconstant_vector_store_derefs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, struct hlsl_block *block)
+{
+    struct hlsl_ir_node *idx;
+    struct hlsl_deref *deref;
+    struct hlsl_type *type;
+    unsigned int i;
+
+    if (instr->type != HLSL_IR_STORE)
+        return false;
+
+    deref = &hlsl_ir_store(instr)->lhs;
+    assert(deref->var);
+
+    if (deref->path_len == 0)
+        return false;
+
+    type = deref->var->data_type;
+    for (i = 0; i < deref->path_len - 1; ++i)
+        type = hlsl_get_element_type_from_path_index(ctx, type, deref->path[i].node);
+
+    idx = deref->path[deref->path_len - 1].node;
+
+    if (type->class == HLSL_CLASS_VECTOR && idx->type != HLSL_IR_CONSTANT)
+    {
+        /* We should turn this into an hlsl_error after we implement unrolling, because if we get
+         * here after that, it means that the HLSL is invalid. */
+        hlsl_fixme(ctx, &instr->loc, "Non-constant vector addressing on store. Unrolling may be missing.");
+    }
+
+    return false;
+}
+
 /* Lower combined samples and sampler variables to synthesized separated textures and samplers.
  * That is, translate SM1-style samples in the source to SM4-style samples in the bytecode. */
 static bool lower_combined_samples(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
@@ -4969,6 +5006,8 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry
         lower_ir(ctx, lower_abs, body);
     }

+    lower_ir(ctx, validate_nonconstant_vector_store_derefs, body);
+
     /* TODO: move forward, remove when no longer needed */
     transform_derefs(ctx, replace_deref_path_with_offset, body);
     while (hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, body, NULL));
diff --git a/tests/hlsl/vector-indexing.shader_test b/tests/hlsl/vector-indexing.shader_test
index 9f6f9a2c..446b2d0e 100644
--- a/tests/hlsl/vector-indexing.shader_test
+++ b/tests/hlsl/vector-indexing.shader_test
@@ -49,3 +49,16 @@ float4 main() : SV_TARGET
 [test]
 draw quad
 probe all rgba (2.0, 2.0, 2.0, 2.0)
+
+
+[pixel shader fail(sm<6) todo]
+float4 v;
+float i;
+
+float4 main() : sv_target
+{
+    float4 p = v;
+    p[i] = 2.0;
+
+    return 0;
+}