[PATCH 0/1] MR321: vkd3d-shader/hlsl: Implement intrinsic tan.

[Petrichor Park (＠petrathekat)]

Now that we can write HLSL intrinsics in HLSL, cleaning up missing
functions becomes much easier.

This commit also extends the trigonometry tests a little bit to make
sure that tan works right.

[PATCH 1/1] vkd3d-shader/hlsl: Implement intrinsic tan.

[Petrichor Park]

From: Petrichor Park

Now that we can write HLSL intrinsics in HLSL, cleaning up missing
functions becomes much easier.

This commit also extends the trigonometry tests a little bit to make
sure that tan works right.
---
 libs/vkd3d-shader/hlsl.y            | 28 ++++++++++++++++
 tests/hlsl/trigonometry.shader_test | 50 +++++++++++++++++++----------
 2 files changed, 61 insertions(+), 17 deletions(-)

diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y
index 161d1ab42..73f4bb6cc 100644
--- a/libs/vkd3d-shader/hlsl.y
+++ b/libs/vkd3d-shader/hlsl.y
@@ -3419,6 +3419,33 @@ static bool intrinsic_step(struct hlsl_ctx *ctx,
     return !!add_implicit_conversion(ctx, params->instrs, ge, type, loc);
 }

+static bool intrinsic_tan(struct hlsl_ctx *ctx,
+        const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+    struct hlsl_ir_function_decl *func;
+    struct hlsl_type *type;
+    char *body;
+
+    static const char template[] =
+            "%s tan(%s x)\n"
+            "{\n"
+            "    return sin(x) / cos(x);\n"
+            "}";
+
+    if (!(type = elementwise_intrinsic_get_common_type(ctx, params, loc)))
+        return false;
+    type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_FLOAT, type->dimx, type->dimy);
+
+    if (!(body = hlsl_sprintf_alloc(ctx, template, type->name, type->name)))
+        return false;
+    func = hlsl_compile_internal_function(ctx, "tan", body);
+    vkd3d_free(body);
+    if (!func)
+        return false;
+
+    return add_user_call(ctx, func, params, loc);
+}
+
 static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer *params,
         const struct vkd3d_shader_location *loc, const char *name, enum hlsl_sampler_dim dim)
 {
@@ -3652,6 +3679,7 @@ intrinsic_functions[] =
     {"smoothstep",                          3, true,  intrinsic_smoothstep},
     {"sqrt",                                1, true,  intrinsic_sqrt},
     {"step",                                2, true,  intrinsic_step},
+    {"tan",                                 1, true,  intrinsic_tan},
     {"tex2D",                              -1, false, intrinsic_tex2D},
     {"tex3D",                              -1, false, intrinsic_tex3D},
     {"texCUBE",                            -1, false, intrinsic_texCUBE},
diff --git a/tests/hlsl/trigonometry.shader_test b/tests/hlsl/trigonometry.shader_test
index 09933d3b7..d087044f3 100644
--- a/tests/hlsl/trigonometry.shader_test
+++ b/tests/hlsl/trigonometry.shader_test
@@ -8,27 +8,27 @@ void main(out float tex : texcoord, inout float4 pos : sv_position)
 float4 main(float tex : texcoord) : sv_target
 {
     tex = floor(tex + 0.25);
-    return float4(sin(tex), cos(tex), 0, 0);
+    return float4(sin(tex), cos(tex), tan(tex), 0);
 }

[test]
 draw quad
-probe ( 0, 0) rgba ( 0.00000000,  1.00000000, 0.0, 0.0)
-probe ( 1, 0) rgba ( 0.84147098,  0.54030231, 0.0, 0.0) 1024
-probe ( 2, 0) rgba ( 0.90929743, -0.41614684, 0.0, 0.0) 1024
-probe ( 3, 0) rgba ( 0.14112001, -0.98999250, 0.0, 0.0) 1024
-probe ( 4, 0) rgba (-0.75680250, -0.65364362, 0.0, 0.0) 1024
-probe ( 5, 0) rgba (-0.95892427,  0.28366219, 0.0, 0.0) 1024
-probe ( 6, 0) rgba (-0.27941550,  0.96017029, 0.0, 0.0) 1024
-probe ( 7, 0) rgba ( 0.65698660,  0.75390225, 0.0, 0.0) 1024
-probe ( 8, 0) rgba ( 0.98935825, -0.14550003, 0.0, 0.0) 1024
-probe ( 9, 0) rgba ( 0.41211849, -0.91113026, 0.0, 0.0) 1024
-probe (10, 0) rgba (-0.54402111, -0.83907153, 0.0, 0.0) 1024
-probe (11, 0) rgba (-0.99999021,  0.00442570, 0.0, 0.0) 2048
-probe (12, 0) rgba (-0.53657292,  0.84385396, 0.0, 0.0) 1024
-probe (13, 0) rgba ( 0.42016704,  0.90744678, 0.0, 0.0) 1024
-probe (14, 0) rgba ( 0.99060736,  0.13673722, 0.0, 0.0) 1024
-probe (15, 0) rgba ( 0.65028784, -0.75968791, 0.0, 0.0) 1024
+probe ( 0, 0) rgba ( 0.00000000,  1.00000000,  0.00000000, 0.0)
+probe ( 1, 0) rgba ( 0.84147098,  0.54030231,  1.55740772, 0.0) 1024
+probe ( 2, 0) rgba ( 0.90929743, -0.41614684, -2.18503986, 0.0) 1024
+probe ( 3, 0) rgba ( 0.14112001, -0.98999250, -0.14254654, 0.0) 1024
+probe ( 4, 0) rgba (-0.75680250, -0.65364362,  1.15782128, 0.0) 1024
+probe ( 5, 0) rgba (-0.95892427,  0.28366219, -3.38051501, 0.0) 1024
+probe ( 6, 0) rgba (-0.27941550,  0.96017029, -0.29100619, 0.0) 1024
+probe ( 7, 0) rgba ( 0.65698660,  0.75390225,  0.87144798, 0.0) 1024
+probe ( 8, 0) rgba ( 0.98935825, -0.14550003, -6.79971146, 0.0) 1024
+probe ( 9, 0) rgba ( 0.41211849, -0.91113026, -0.45231566, 0.0) 1024
+probe (10, 0) rgba (-0.54402111, -0.83907153,  0.64836083, 0.0) 1024
+probe (11, 0) rgba (-0.99999021,  0.00442570, -225.95084645, 0.0) 2048
+probe (12, 0) rgba (-0.53657292,  0.84385396, -0.63585993, 0.0) 1024
+probe (13, 0) rgba ( 0.42016704,  0.90744678,  0.46302113, 0.0) 1024
+probe (14, 0) rgba ( 0.99060736,  0.13673722,  7.24460662, 0.0) 1024
+probe (15, 0) rgba ( 0.65028784, -0.75968791, -0.85599340, 0.0) 1024

[pixel shader]
@@ -57,3 +57,19 @@ float4 main() : sv_target
 uniform 0 float4 0.0 0.78539816 1.57079632 2.35619449
 draw quad
 probe all rgba (1000.0, 707.0, -0.0, -707.0)
+
+
+[pixel shader]
+uniform float4 a;
+
+float4 main() : sv_target
+{
+    return round(1000 * tan(a));
+}
+
+[test]
+% tan(pi/2) is infinity, which the test harness can't recognize
+% so query 0, pi/4, 3pi/4, pi
+uniform 0 float4 0.0 0.78539816 2.35619449 3.14159265
+draw quad
+probe all rgba (0, 1000, -1000.0, 0)

[Zebediah Figura (＠zfigura)]

tan is a simple divide, so I think in this case it's probably better to define it in C.

sscanf() can actually scan INF, but it needs POSIX to work in the crosstests. You'll need something like [1].

[1] https://gitlab.winehq.org/zfigura/vkd3d/-/commit/ec79b5aa5682dd6f486738bd50a...