MVP for higher order functions (#2849)

* MVP for higher order functions

* Add shader subgroup partitioned glsl intrinsics

* Implement parsing and checking for tuple types

Currently there is no way to do anything useful with them from the source language however

* neaten

* Correct precedence of function type parsing

* neaten

* higher order function tests

* function types of any arity

* Inference for higher order functions

* Add second test for unsynchronized params

* regenerate vs projects

* dx11 -> dx12 for saturated cooperations tests

* Disable saturated cooperation tests on vulkan

They fail on release builds in CI, not essential for the higher order function work however

* remove saturated-cooperation tests

* Remove unnecessary assert and clarify control flow in AddDeclRefOverloadCandidates

* Add Tuple type name mangling

* Use functype keyword to introduce function types

* Add more inference tests for hof

---------

Co-authored-by: Yong He <yonghe@outlook.com>

8 files changed, 141 insertions, 0 deletions

diff --git a/tests/language-feature/higher-order-functions/fix.slang b/tests/language-feature/higher-order-functions/fix.slang
new file mode 100644
index 000000000..812353477
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/fix.slang
@@ -0,0 +1,26 @@
+//DISABLE_TEST(compute, vulkan):COMPARE_COMPUTE_EX():-vk -compute -shaderobj -output-using-type
+//DISABLE_TEST(compute):COMPARE_COMPUTE_EX():-slang -compute -shaderobj -output-using-type
+// Disabled due to https://github.com/shader-slang/slang/issues/2198
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<uint> outputBuffer;
+
+func fix(float x) -> float
+{
+    let epsilon = 0.01;
+    let x_ = go(x);
+    if(abs(x_ - x) < epsilon)
+        return x_;
+    return fix(x_);
+}
+
+func go(x : float) -> float
+{
+    return (x + 121 / x) / 2;
+}
+
+[numthreads(4, 1, 1)]
+void computeMain(uint tig : SV_GroupIndex)
+{
+    outputBuffer[tig] = uint(fix(11));
+}
diff --git a/tests/language-feature/higher-order-functions/generic.slang b/tests/language-feature/higher-order-functions/generic.slang
new file mode 100644
index 000000000..e7d63eb27
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/generic.slang
@@ -0,0 +1,14 @@
+//TEST:SIMPLE:
+
+func foo(f : functype (float) -> int) -> int{
+    return f(0);
+}
+
+int bap<T>(float) {
+    return 1;
+}
+
+int zap() {
+    // We should be able to specify which foo we want
+    return foo(bap<bool>);
+}
diff --git a/tests/language-feature/higher-order-functions/inference-generic.slang b/tests/language-feature/higher-order-functions/inference-generic.slang
new file mode 100644
index 000000000..52e65d0d9
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/inference-generic.slang
@@ -0,0 +1,18 @@
+//DISABLE_TEST:SIMPLE:
+// Disabled because we don't actually perform this inference yet
+
+func foo(f : functype (float) -> int) -> int
+{
+    return f(0);
+}
+
+int boo<T>(T)
+{
+    return 1;
+}
+
+int zoo()
+{
+    // We should infer that we want boo<float>
+    return foo(boo);
+}
diff --git a/tests/language-feature/higher-order-functions/inference.slang b/tests/language-feature/higher-order-functions/inference.slang
new file mode 100644
index 000000000..c01f5b19e
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/inference.slang
@@ -0,0 +1,21 @@
+//TEST:SIMPLE:
+
+func x<A, B>(f : functype (A) -> B)
+{ }
+
+float f(int)
+{
+    return 0;
+}
+
+func y<A, B, C>(g : functype (A, B) -> C)
+{ }
+
+void g<A>(A, bool)
+{ }
+
+void main()
+{
+    x(f);
+    y(g<int>);
+}
diff --git a/tests/language-feature/higher-order-functions/overloaded.slang b/tests/language-feature/higher-order-functions/overloaded.slang
new file mode 100644
index 000000000..ebbc6ed19
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/overloaded.slang
@@ -0,0 +1,18 @@
+//TEST:SIMPLE:
+
+func foo(f : functype (float) -> int) -> int{
+    return f(0);
+}
+
+int bit<T>(T) {
+    return 1;
+}
+
+int bit<T, let N : int>(vector<T, N>) {
+    return 1;
+}
+
+int zit() {
+    // In an ideal world in this case we could infer that we want bit<T>
+    return foo(bit<float>);
+}
diff --git a/tests/language-feature/higher-order-functions/overloaded.slang.expected b/tests/language-feature/higher-order-functions/overloaded.slang.expected
new file mode 100644
index 000000000..3d02bf06f
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/overloaded.slang.expected
@@ -0,0 +1,8 @@
+result code = -1
+standard error = {
+tests/language-feature/higher-order-functions/overloaded.slang(17): error 39999: passing overloaded functions to higher order functions is not supported
+    return foo(bit<float>);
+              ^
+}
+standard output = {
+}
diff --git a/tests/language-feature/higher-order-functions/simple.slang b/tests/language-feature/higher-order-functions/simple.slang
new file mode 100644
index 000000000..7c6bf5a26
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/simple.slang
@@ -0,0 +1,31 @@
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX():-vk -compute -shaderobj -output-using-type
+//TEST(compute):COMPARE_COMPUTE_EX():-slang -compute -shaderobj -output-using-type
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<uint> outputBuffer;
+
+func twice<A>(f : functype (A) -> A, a : A) -> A
+{
+    return f(f(a));
+}
+
+func apply2<A, B, C>(f : functype (A, B) -> C, a : A, b : B) -> C
+{
+    return f(a, b);
+}
+
+func multiply(x : float, y : float) -> float
+{
+    return x * y;
+}
+
+func square(x : float) -> float
+{
+    return apply2<float, float, float>(multiply, x, x);
+}
+
+[numthreads(4, 1, 1)]
+void computeMain(uint tig : SV_GroupIndex)
+{
+    outputBuffer[tig] = uint(twice(square, -2.f));
+}
diff --git a/tests/language-feature/higher-order-functions/simple.slang.expected.txt b/tests/language-feature/higher-order-functions/simple.slang.expected.txt
new file mode 100644
index 000000000..d9d031441
--- /dev/null
+++ b/tests/language-feature/higher-order-functions/simple.slang.expected.txt
@@ -0,0 +1,5 @@
+type: uint32_t
+16
+16
+16
+16