[PATCH 0/1] MR80: vkd3d-shader/hlsl: Support reflect() intrinsic.

[PATCH 1/1] vkd3d-shader/hlsl: Support reflect() intrinsic.

[Nikolay Sivov]

From: Nikolay Sivov

---
 Makefile.am               |  1 +
 libs/vkd3d-shader/hlsl.y  | 25 +++++++++++++++++++++++++
 tests/reflect.shader_test | 11 +++++++++++
 3 files changed, 37 insertions(+)
 create mode 100644 tests/reflect.shader_test

diff --git a/Makefile.am b/Makefile.am
index f9199472..8822e9eb 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -130,6 +130,7 @@ vkd3d_shader_tests = \
    tests/preproc-invalid.shader_test \
    tests/preproc-macro.shader_test \
    tests/preproc-misc.shader_test \
+	tests/reflect.shader_test \
    tests/return.shader_test \
    tests/round.shader_test \
    tests/sampler.shader_test \
diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y
index a4de0edd..db7bbde6 100644
--- a/libs/vkd3d-shader/hlsl.y
+++ b/libs/vkd3d-shader/hlsl.y
@@ -2767,6 +2767,30 @@ static bool intrinsic_pow(struct hlsl_ctx *ctx,
     return true;
 }

+static bool intrinsic_reflect(struct hlsl_ctx *ctx,
+        const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+    struct hlsl_ir_node *i = params->args[0], *n = params->args[1];
+    struct hlsl_ir_node *dot, *mul_n, *two_dot, *neg;
+
+    if (!elementwise_intrinsic_float_convert_args(ctx, params, loc))
+        return false;
+
+    if (!(dot = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_DOT, i, n, loc)))
+        return false;
+
+    if (!(two_dot = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_ADD, dot, dot, loc)))
+        return false;
+
+    if (!(mul_n = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, n, two_dot, loc)))
+        return false;
+
+    if (!(neg = add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_NEG, mul_n, loc)))
+        return false;
+
+    return !!add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_ADD, i, neg, loc);
+}
+
 static bool intrinsic_round(struct hlsl_ctx *ctx,
         const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
 {
@@ -2975,6 +2999,7 @@ intrinsic_functions[] =
     {"mul",                                 2, true,  intrinsic_mul},
     {"normalize",                           1, true,  intrinsic_normalize},
     {"pow",                                 2, true,  intrinsic_pow},
+    {"reflect",                             2, true,  intrinsic_reflect},
     {"round",                               1, true,  intrinsic_round},
     {"saturate",                            1, true,  intrinsic_saturate},
     {"sin",                                 1, true,  intrinsic_sin},
diff --git a/tests/reflect.shader_test b/tests/reflect.shader_test
new file mode 100644
index 00000000..a5746599
--- /dev/null
+++ b/tests/reflect.shader_test
@@ -0,0 +1,11 @@
+[pixel shader]
+float4 main(uniform float4 i, uniform float4 n) : sv_target
+{
+    return reflect(i, n);
+}
+
+[test]
+uniform 0 float4 0.5 -0.1 0.2 0.3
+uniform 1 float4 0.6 0.4 -0.3 1.0
+draw quad
+probe all rgba (-0.5, 0.6, 0.4, -0.3)

[Giovanni Mascellani (＠giomasce)]

Giovanni Mascellani (@giomasce) commented about tests/reflect.shader_test:
> +[pixel shader]
> +float4 main(uniform float4 i, uniform float4 n) : sv_target
> +{
> - return reflect(i, n);
>
> +}
> -
>
> +[test]
> +uniform 0 float4 0.5 -0.1 0.2 0.3
> +uniform 1 float4 0.6 0.4 -0.3 1.0

This should be `uniform 4`.

[Giovanni Mascellani (＠giomasce)]

Giovanni Mascellani (@giomasce) commented about libs/vkd3d-shader/hlsl.y:
>  return true;
>
> }
>
> +static bool intrinsic_reflect(struct hlsl_ctx *ctx,
> -
>    const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
>
> +{
> - struct hlsl_ir_node *i = params->args[0], *n = params->args[1];
> - struct hlsl_ir_node *dot, *mul_n, *two_dot, *neg;
> -
> - if (!elementwise_intrinsic_float_convert_args(ctx, params, loc))
> -
>    return false;
> -
> - if (!(dot = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_DOT, i, n, loc)))

Please use `add_binary_dot_expr()`, which does things properly: it generates a scalar rather than a vector, it errors out on matrix inputs, it generates `HLSL_OP2_MUL` instead of `HLSL_OP2_DOT` when necessary.

[Giovanni Mascellani (＠giomasce)]

On Mon Feb  6 14:41:36 2023 +0000, Giovanni Mascellani wrote:
> Please use `add_binary_dot_expr()`, which does things properly: it
> generates a scalar rather than a vector, it errors out on matrix inputs,
> it generates `HLSL_OP2_MUL` instead of `HLSL_OP2_DOT` when necessary.

Thinking again, this is probably another instance of https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/76#note_22500: this is not an elementwise operation, so you shouldn't use `elementwise_intrinsic_float_convert_args()`. But please check what happens when providing matrices, or vectors of different size, or vectors and scalars, or other funny combinations.