tint: Add f16 support for parts of float built-in, part 1

This patch add f16 support for a major part of numeric built-in, and
implement corresponding unittests for resolver and backends. This patch
also enable f16 constant evaluation for unary minus operator, `atan2`
and `clamp`.

The following numeric built-ins are not supported yet:
* frexp
* modf

The end-to-end tests for f16 built-in are not added yet.

Bug: tint:1473, tint:1502
Change-Id: If807185617b21c510a1a9c371179a60800c4f875
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/96722
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>

src/tint/intrinsics.def

@@ -165,6 +165,7 @@ match scalar_no_f16: f32 | i32 | u32 | bool
 match scalar_no_i32: f32 | f16 | u32 | bool
 match scalar_no_u32: f32 | f16 | i32 | bool
 match scalar_no_bool: f32 | f16 | i32 | u32
+match fia_fiu32_f16: fa | ia | f32 | i32 | u32 | f16
 match fia_fi32_f16: fa | ia | f32 | i32 | f16
 match fia_fiu32: fa | ia | f32 | i32 | u32
 match fa_f32: fa | f32
@@ -386,48 +387,48 @@ match storage
 ////////////////////////////////////////////////////////////////////////////////
 
 // https://gpuweb.github.io/gpuweb/wgsl/#builtin-functions
-fn abs<T: fiu32>(T) -> T
-fn abs<N: num, T: fiu32>(vec<N, T>) -> vec<N, T>
-fn acos(f32) -> f32
-fn acos<N: num>(vec<N, f32>) -> vec<N, f32>
-fn acosh(f32) -> f32
-fn acosh<N: num>(vec<N, f32>) -> vec<N, f32>
+fn abs<T: fiu32_f16>(T) -> T
+fn abs<N: num, T: fiu32_f16>(vec<N, T>) -> vec<N, T>
+fn acos<T: f32_f16>(T) -> T
+fn acos<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn acosh<T: f32_f16>(T) -> T
+fn acosh<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
 fn all(bool) -> bool
 fn all<N: num>(vec<N, bool>) -> bool
 fn any(bool) -> bool
 fn any<N: num>(vec<N, bool>) -> bool
 fn arrayLength<T, A: access>(ptr<storage, array<T>, A>) -> u32
-fn asin(f32) -> f32
-fn asin<N: num>(vec<N, f32>) -> vec<N, f32>
-fn asinh(f32) -> f32
-fn asinh<N: num>(vec<N, f32>) -> vec<N, f32>
-fn atan(f32) -> f32
-fn atan<N: num>(vec<N, f32>) -> vec<N, f32>
-@const fn atan2<T: fa_f32>(T, T) -> T
-@const fn atan2<N: num, T: fa_f32>(vec<N, T>, vec<N, T>) -> vec<N, T>
-fn atanh(f32) -> f32
-fn atanh<N: num>(vec<N, f32>) -> vec<N, f32>
-fn ceil(f32) -> f32
-fn ceil<N: num>(vec<N, f32>) -> vec<N, f32>
-@const fn clamp<T: fia_fiu32>(T, T, T) -> T
-@const fn clamp<N: num, T: fia_fiu32>(vec<N, T>, vec<N, T>, vec<N, T>) -> vec<N, T>
-fn cos(f32) -> f32
-fn cos<N: num>(vec<N, f32>) -> vec<N, f32>
-fn cosh(f32) -> f32
-fn cosh<N: num>(vec<N, f32>) -> vec<N, f32>
+fn asin<T: f32_f16>(T) -> T
+fn asin<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn asinh<T: f32_f16>(T) -> T
+fn asinh<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn atan<T: f32_f16>(T) -> T
+fn atan<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+@const fn atan2<T: fa_f32_f16>(T, T) -> T
+@const fn atan2<T: fa_f32_f16, N: num>(vec<N, T>, vec<N, T>) -> vec<N, T>
+fn atanh<T: f32_f16>(T) -> T
+fn atanh<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn ceil<T: f32_f16>(T) -> T
+fn ceil<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+@const fn clamp<T: fia_fiu32_f16>(T, T, T) -> T
+@const fn clamp<T: fia_fiu32_f16, N: num>(vec<N, T>, vec<N, T>, vec<N, T>) -> vec<N, T>
+fn cos<T: f32_f16>(T) -> T
+fn cos<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn cosh<T: f32_f16>(T) -> T
+fn cosh<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
 fn countLeadingZeros<T: iu32>(T) -> T
 fn countLeadingZeros<N: num, T: iu32>(vec<N, T>) -> vec<N, T>
 fn countOneBits<T: iu32>(T) -> T
 fn countOneBits<N: num, T: iu32>(vec<N, T>) -> vec<N, T>
 fn countTrailingZeros<T: iu32>(T) -> T
 fn countTrailingZeros<N: num, T: iu32>(vec<N, T>) -> vec<N, T>
-fn cross(vec3<f32>, vec3<f32>) -> vec3<f32>
-fn degrees(f32) -> f32
-fn degrees<N: num>(vec<N, f32>) -> vec<N, f32>
-fn determinant<N: num>(mat<N, N, f32>) -> f32
-fn distance(f32, f32) -> f32
-fn distance<N: num>(vec<N, f32>, vec<N, f32>) -> f32
-fn dot<N: num, T: fiu32>(vec<N, T>, vec<N, T>) -> T
+fn cross<T: f32_f16>(vec3<T>, vec3<T>) -> vec3<T>
+fn degrees<T: f32_f16>(T) -> T
+fn degrees<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn determinant<N: num, T: f32_f16>(mat<N, N, T>) -> T
+fn distance<T: f32_f16>(T, T) -> T
+fn distance<N: num, T: f32_f16>(vec<N, T>, vec<N, T>) -> T
+fn dot<N: num, T: fiu32_f16>(vec<N, T>, vec<N, T>) -> T
 fn dot4I8Packed(u32, u32) -> i32
 fn dot4U8Packed(u32, u32) -> u32
 @stage("fragment") fn dpdx(f32) -> f32
@@ -442,23 +443,23 @@ fn dot4U8Packed(u32, u32) -> u32
 @stage("fragment") fn dpdyCoarse<N: num>(vec<N, f32>) -> vec<N, f32>
 @stage("fragment") fn dpdyFine(f32) -> f32
 @stage("fragment") fn dpdyFine<N: num>(vec<N, f32>) -> vec<N, f32>
-fn exp(f32) -> f32
-fn exp<N: num>(vec<N, f32>) -> vec<N, f32>
-fn exp2(f32) -> f32
-fn exp2<N: num>(vec<N, f32>) -> vec<N, f32>
+fn exp<T: f32_f16>(T) -> T
+fn exp<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn exp2<T: f32_f16>(T) -> T
+fn exp2<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
 fn extractBits<T: iu32>(T, u32, u32) -> T
 fn extractBits<N: num, T: iu32>(vec<N, T>, u32, u32) -> vec<N, T>
-fn faceForward<N: num>(vec<N, f32>, vec<N, f32>, vec<N, f32>) -> vec<N, f32>
+fn faceForward<N: num, T: f32_f16>(vec<N, T>, vec<N, T>, vec<N, T>) -> vec<N, T>
 fn firstLeadingBit<T: iu32>(T) -> T
 fn firstLeadingBit<N: num, T: iu32>(vec<N, T>) -> vec<N, T>
 fn firstTrailingBit<T: iu32>(T) -> T
 fn firstTrailingBit<N: num, T: iu32>(vec<N, T>) -> vec<N, T>
-fn floor(f32) -> f32
-fn floor<N: num>(vec<N, f32>) -> vec<N, f32>
-fn fma(f32, f32, f32) -> f32
-fn fma<N: num>(vec<N, f32>, vec<N, f32>, vec<N, f32>) -> vec<N, f32>
-fn fract(f32) -> f32
-fn fract<N: num>(vec<N, f32>) -> vec<N, f32>
+fn floor<T: f32_f16>(T) -> T
+fn floor<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn fma<T: f32_f16>(T, T, T) -> T
+fn fma<N: num, T: f32_f16>(vec<N, T>, vec<N, T>, vec<N, T>) -> vec<N, T>
+fn fract<T: f32_f16>(T) -> T
+fn fract<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
 fn frexp(f32) -> __frexp_result
 fn frexp<N: num>(vec<N, f32>) -> __frexp_result_vec<N>
 @stage("fragment") fn fwidth(f32) -> f32
@@ -469,64 +470,64 @@ fn frexp<N: num>(vec<N, f32>) -> __frexp_result_vec<N>
 @stage("fragment") fn fwidthFine<N: num>(vec<N, f32>) -> vec<N, f32>
 fn insertBits<T: iu32>(T, T, u32, u32) -> T
 fn insertBits<N: num, T: iu32>(vec<N, T>, vec<N, T>, u32, u32) -> vec<N, T>
-fn inverseSqrt(f32) -> f32
-fn inverseSqrt<N: num>(vec<N, f32>) -> vec<N, f32>
-fn ldexp(f32, i32) -> f32
-fn ldexp<N: num>(vec<N, f32>, vec<N, i32>) -> vec<N, f32>
-fn length(f32) -> f32
-fn length<N: num>(vec<N, f32>) -> f32
-fn log(f32) -> f32
-fn log<N: num>(vec<N, f32>) -> vec<N, f32>
-fn log2(f32) -> f32
-fn log2<N: num>(vec<N, f32>) -> vec<N, f32>
-fn max<T: fiu32>(T, T) -> T
-fn max<N: num, T: fiu32>(vec<N, T>, vec<N, T>) -> vec<N, T>
-fn min<T: fiu32>(T, T) -> T
-fn min<N: num, T: fiu32>(vec<N, T>, vec<N, T>) -> vec<N, T>
-fn mix(f32, f32, f32) -> f32
-fn mix<N: num>(vec<N, f32>, vec<N, f32>, vec<N, f32>) -> vec<N, f32>
-fn mix<N: num>(vec<N, f32>, vec<N, f32>, f32) -> vec<N, f32>
+fn inverseSqrt<T: f32_f16>(T) -> T
+fn inverseSqrt<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn ldexp<T: f32_f16>(T, i32) -> T
+fn ldexp<N: num, T: f32_f16>(vec<N, T>, vec<N, i32>) -> vec<N, T>
+fn length<T: f32_f16>(T) -> T
+fn length<N: num, T: f32_f16>(vec<N, T>) -> T
+fn log<T: f32_f16>(T) -> T
+fn log<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn log2<T: f32_f16>(T) -> T
+fn log2<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn max<T: fiu32_f16>(T, T) -> T
+fn max<N: num, T: fiu32_f16>(vec<N, T>, vec<N, T>) -> vec<N, T>
+fn min<T: fiu32_f16>(T, T) -> T
+fn min<N: num, T: fiu32_f16>(vec<N, T>, vec<N, T>) -> vec<N, T>
+fn mix<T: f32_f16>(T, T, T) -> T
+fn mix<N: num, T: f32_f16>(vec<N, T>, vec<N, T>, vec<N, T>) -> vec<N, T>
+fn mix<N: num, T: f32_f16>(vec<N, T>, vec<N, T>, T) -> vec<N, T>
 fn modf(f32) -> __modf_result
 fn modf<N: num>(vec<N, f32>) -> __modf_result_vec<N>
-fn normalize<N: num>(vec<N, f32>) -> vec<N, f32>
+fn normalize<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
 fn pack2x16float(vec2<f32>) -> u32
 fn pack2x16snorm(vec2<f32>) -> u32
 fn pack2x16unorm(vec2<f32>) -> u32
 fn pack4x8snorm(vec4<f32>) -> u32
 fn pack4x8unorm(vec4<f32>) -> u32
-fn pow(f32, f32) -> f32
-fn pow<N: num>(vec<N, f32>, vec<N, f32>) -> vec<N, f32>
-fn radians(f32) -> f32
-fn radians<N: num>(vec<N, f32>) -> vec<N, f32>
-fn reflect<N: num>(vec<N, f32>, vec<N, f32>) -> vec<N, f32>
-fn refract<N: num>(vec<N, f32>, vec<N, f32>, f32) -> vec<N, f32>
+fn pow<T: f32_f16>(T, T) -> T
+fn pow<N: num, T: f32_f16>(vec<N, T>, vec<N, T>) -> vec<N, T>
+fn radians<T: f32_f16>(T) -> T
+fn radians<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn reflect<N: num, T: f32_f16>(vec<N, T>, vec<N, T>) -> vec<N, T>
+fn refract<N: num, T: f32_f16>(vec<N, T>, vec<N, T>, T) -> vec<N, T>
 fn reverseBits<T: iu32>(T) -> T
 fn reverseBits<N: num, T: iu32>(vec<N, T>) -> vec<N, T>
-fn round(f32) -> f32
-fn round<N: num>(vec<N, f32>) -> vec<N, f32>
-fn select<T: scalar_no_f16>(T, T, bool) -> T
-fn select<T: scalar_no_f16, N: num>(vec<N, T>, vec<N, T>, bool) -> vec<N, T>
-fn select<N: num, T: scalar_no_f16>(vec<N, T>, vec<N, T>, vec<N, bool>) -> vec<N, T>
-fn sign(f32) -> f32
-fn sign<N: num>(vec<N, f32>) -> vec<N, f32>
-fn sin(f32) -> f32
-fn sin<N: num>(vec<N, f32>) -> vec<N, f32>
-fn sinh(f32) -> f32
-fn sinh<N: num>(vec<N, f32>) -> vec<N, f32>
-fn smoothstep(f32, f32, f32) -> f32
-fn smoothstep<N: num>(vec<N, f32>, vec<N, f32>, vec<N, f32>) -> vec<N, f32>
-fn sqrt(f32) -> f32
-fn sqrt<N: num>(vec<N, f32>) -> vec<N, f32>
-fn step(f32, f32) -> f32
-fn step<N: num>(vec<N, f32>, vec<N, f32>) -> vec<N, f32>
+fn round<T: f32_f16>(T) -> T
+fn round<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn select<T: scalar>(T, T, bool) -> T
+fn select<T: scalar, N: num>(vec<N, T>, vec<N, T>, bool) -> vec<N, T>
+fn select<N: num, T: scalar>(vec<N, T>, vec<N, T>, vec<N, bool>) -> vec<N, T>
+fn sign<T: f32_f16>(T) -> T
+fn sign<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn sin<T: f32_f16>(T) -> T
+fn sin<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn sinh<T: f32_f16>(T) -> T
+fn sinh<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn smoothstep<T: f32_f16>(T, T, T) -> T
+fn smoothstep<N: num, T: f32_f16>(vec<N, T>, vec<N, T>, vec<N, T>) -> vec<N, T>
+fn sqrt<T: f32_f16>(T) -> T
+fn sqrt<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn step<T: f32_f16>(T, T) -> T
+fn step<N: num, T: f32_f16>(vec<N, T>, vec<N, T>) -> vec<N, T>
 @stage("compute") fn storageBarrier()
-fn tan(f32) -> f32
-fn tan<N: num>(vec<N, f32>) -> vec<N, f32>
-fn tanh(f32) -> f32
-fn tanh<N: num>(vec<N, f32>) -> vec<N, f32>
-fn transpose<M: num, N: num>(mat<M, N, f32>) -> mat<N, M, f32>
-fn trunc(f32) -> f32
-fn trunc<N: num>(vec<N, f32>) -> vec<N, f32>
+fn tan<T: f32_f16>(T) -> T
+fn tan<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn tanh<T: f32_f16>(T) -> T
+fn tanh<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
+fn transpose<M: num, N: num, T: f32_f16>(mat<M, N, T>) -> mat<N, M, T>
+fn trunc<T: f32_f16>(T) -> T
+fn trunc<N: num, T: f32_f16>(vec<N, T>) -> vec<N, T>
 fn unpack2x16float(u32) -> vec2<f32>
 fn unpack2x16snorm(u32) -> vec2<f32>
 fn unpack2x16unorm(u32) -> vec2<f32>

src/tint/resolver/builtin_test.cc

@@ -129,9 +129,9 @@ TEST_F(ResolverBuiltinTest, Select_Error_NoParams) {
               R"(error: no matching call to select()
 
 3 candidate functions:
-  select(T, T, bool) -> T  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, i32, u32 or bool
+  select(T, T, bool) -> T  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
 )");
 }
 
@@ -145,9 +145,9 @@ TEST_F(ResolverBuiltinTest, Select_Error_SelectorInt) {
               R"(error: no matching call to select(i32, i32, i32)
 
 3 candidate functions:
-  select(T, T, bool) -> T  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, i32, u32 or bool
+  select(T, T, bool) -> T  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
 )");
 }
 
@@ -162,9 +162,9 @@ TEST_F(ResolverBuiltinTest, Select_Error_Matrix) {
               R"(error: no matching call to select(mat2x2<f32>, mat2x2<f32>, bool)
 
 3 candidate functions:
-  select(T, T, bool) -> T  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, i32, u32 or bool
+  select(T, T, bool) -> T  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
 )");
 }
 
@@ -178,9 +178,9 @@ TEST_F(ResolverBuiltinTest, Select_Error_MismatchTypes) {
               R"(error: no matching call to select(f32, vec2<f32>, bool)
 
 3 candidate functions:
-  select(T, T, bool) -> T  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, i32, u32 or bool
+  select(T, T, bool) -> T  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
 )");
 }
 
@@ -194,9 +194,9 @@ TEST_F(ResolverBuiltinTest, Select_Error_MismatchVectorSize) {
               R"(error: no matching call to select(vec2<f32>, vec3<f32>, bool)
 
 3 candidate functions:
-  select(T, T, bool) -> T  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, i32, u32 or bool
-  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, i32, u32 or bool
+  select(T, T, bool) -> T  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, bool) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
+  select(vecN<T>, vecN<T>, vecN<bool>) -> vecN<T>  where: T is f32, f16, i32, u32 or bool
 )");
 }
 
@@ -458,6 +458,206 @@ TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, FourParams_Vector_f32) {
     }
 }
 
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, OneParam_Scalar_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, 1_h);
+    WrapInFunction(call);
+
+    if (param.args_number == 1u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(),
+                    HasSubstr("error: no matching call to " + std::string(param.name) + "(f16)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, OneParam_Vector_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    if (param.args_number == 1u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->is_float_vector());
+        EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+        ASSERT_NE(TypeOf(call)->As<sem::Vector>()->type(), nullptr);
+        EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(), HasSubstr("error: no matching call to " +
+                                            std::string(param.name) + "(vec3<f16>)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, TwoParams_Scalar_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, 1_h, 1_h);
+    WrapInFunction(call);
+
+    if (param.args_number == 2u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(), HasSubstr("error: no matching call to " +
+                                            std::string(param.name) + "(f16, f16)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, TwoParams_Vector_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, vec3<f16>(1_h, 1_h, 3_h), vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    if (param.args_number == 2u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->is_float_vector());
+        EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+        ASSERT_NE(TypeOf(call)->As<sem::Vector>()->type(), nullptr);
+        EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(), HasSubstr("error: no matching call to " +
+                                            std::string(param.name) + "(vec3<f16>, vec3<f16>)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, ThreeParams_Scalar_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, 1_h, 1_h, 1_h);
+    WrapInFunction(call);
+
+    if (param.args_number == 3u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(), HasSubstr("error: no matching call to " +
+                                            std::string(param.name) + "(f16, f16, f16)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, ThreeParams_Vector_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, vec3<f16>(1_h, 1_h, 3_h), vec3<f16>(1_h, 1_h, 3_h),
+                      vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    if (param.args_number == 3u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->is_float_vector());
+        EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+        ASSERT_NE(TypeOf(call)->As<sem::Vector>()->type(), nullptr);
+        EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(),
+                    HasSubstr("error: no matching call to " + std::string(param.name) +
+                              "(vec3<f16>, vec3<f16>, vec3<f16>)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, FourParams_Scalar_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, 1_h, 1_h, 1_h, 1_h);
+    WrapInFunction(call);
+
+    if (param.args_number == 4u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(), HasSubstr("error: no matching call to " +
+                                            std::string(param.name) + "(f16, f16, f16, f16)"));
+    }
+}
+
+TEST_P(ResolverBuiltinTest_FloatBuiltin_IdenticalType, FourParams_Vector_f16) {
+    auto param = GetParam();
+
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call(param.name, vec3<f16>(1_h, 1_h, 3_h), vec3<f16>(1_h, 1_h, 3_h),
+                      vec3<f16>(1_h, 1_h, 3_h), vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    if (param.args_number == 4u) {
+        // Parameter count matched.
+        EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+        ASSERT_NE(TypeOf(call), nullptr);
+        EXPECT_TRUE(TypeOf(call)->is_float_vector());
+        EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+        ASSERT_NE(TypeOf(call)->As<sem::Vector>()->type(), nullptr);
+        EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+    } else {
+        // Invalid parameter count.
+        EXPECT_FALSE(r()->Resolve());
+
+        EXPECT_THAT(r()->error(),
+                    HasSubstr("error: no matching call to " + std::string(param.name) +
+                              "(vec3<f16>, vec3<f16>, vec3<f16>, vec3<f16>)"));
+    }
+}
+
 INSTANTIATE_TEST_SUITE_P(
     ResolverTest,
     ResolverBuiltinTest_FloatBuiltin_IdenticalType,
@@ -526,6 +726,20 @@ TEST_F(ResolverBuiltinFloatTest, Cross_f32) {
     EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Cross_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("cross", vec3<f16>(1_h, 2_h, 3_h), vec3<f16>(1_h, 2_h, 3_h));
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->is_float_vector());
+    EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+    EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinFloatTest, Cross_Error_NoArgs) {
     auto* call = Call("cross");
     WrapInFunction(call);
@@ -535,7 +749,7 @@ TEST_F(ResolverBuiltinFloatTest, Cross_Error_NoArgs) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to cross()
 
 1 candidate function:
-  cross(vec3<f32>, vec3<f32>) -> vec3<f32>
+  cross(vec3<T>, vec3<T>) -> vec3<T>  where: T is f32 or f16
 )");
 }
 
@@ -548,7 +762,7 @@ TEST_F(ResolverBuiltinFloatTest, Cross_Error_Scalar) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to cross(f32, f32)
 
 1 candidate function:
-  cross(vec3<f32>, vec3<f32>) -> vec3<f32>
+  cross(vec3<T>, vec3<T>) -> vec3<T>  where: T is f32 or f16
 )");
 }
 
@@ -562,7 +776,7 @@ TEST_F(ResolverBuiltinFloatTest, Cross_Error_Vec3Int) {
               R"(error: no matching call to cross(vec3<i32>, vec3<i32>)
 
 1 candidate function:
-  cross(vec3<f32>, vec3<f32>) -> vec3<f32>
+  cross(vec3<T>, vec3<T>) -> vec3<T>  where: T is f32 or f16
 )");
 }
 
@@ -577,7 +791,7 @@ TEST_F(ResolverBuiltinFloatTest, Cross_Error_Vec4) {
               R"(error: no matching call to cross(vec4<f32>, vec4<f32>)
 
 1 candidate function:
-  cross(vec3<f32>, vec3<f32>) -> vec3<f32>
+  cross(vec3<T>, vec3<T>) -> vec3<T>  where: T is f32 or f16
 )");
 }
 
@@ -593,7 +807,7 @@ TEST_F(ResolverBuiltinFloatTest, Cross_Error_TooManyParams) {
               R"(error: no matching call to cross(vec3<f32>, vec3<f32>, vec3<f32>)
 
 1 candidate function:
-  cross(vec3<f32>, vec3<f32>) -> vec3<f32>
+  cross(vec3<T>, vec3<T>) -> vec3<T>  where: T is f32 or f16
 )");
 }
 
@@ -608,6 +822,18 @@ TEST_F(ResolverBuiltinFloatTest, Distance_Scalar_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Distance_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("distance", 1_h, 1_h);
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinFloatTest, Distance_Vector_f32) {
     auto* call = Call("distance", vec3<f32>(1_f, 1_f, 3_f), vec3<f32>(1_f, 1_f, 3_f));
     WrapInFunction(call);
@@ -618,31 +844,44 @@ TEST_F(ResolverBuiltinFloatTest, Distance_Vector_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Distance_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("distance", vec3<f16>(1_h, 1_h, 3_h), vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinFloatTest, Distance_TooManyParams) {
-    auto* call = Call("distance", 1_f, 1_f, 3_f);
+    auto* call = Call("distance", vec3<f32>(1_f, 1_f, 3_f), vec3<f32>(1_f, 1_f, 3_f),
+                      vec3<f32>(1_f, 1_f, 3_f));
     WrapInFunction(call);
 
     EXPECT_FALSE(r()->Resolve());
 
-    EXPECT_EQ(r()->error(), R"(error: no matching call to distance(f32, f32, f32)
+    EXPECT_EQ(r()->error(), R"(error: no matching call to distance(vec3<f32>, vec3<f32>, vec3<f32>)
 
 2 candidate functions:
-  distance(f32, f32) -> f32
-  distance(vecN<f32>, vecN<f32>) -> f32
+  distance(T, T) -> T  where: T is f32 or f16
+  distance(vecN<T>, vecN<T>) -> T  where: T is f32 or f16
 )");
 }
 
 TEST_F(ResolverBuiltinFloatTest, Distance_TooFewParams) {
-    auto* call = Call("distance", 1_f);
+    auto* call = Call("distance", vec3<f32>(1_f, 1_f, 3_f));
     WrapInFunction(call);
 
     EXPECT_FALSE(r()->Resolve());
 
-    EXPECT_EQ(r()->error(), R"(error: no matching call to distance(f32)
+    EXPECT_EQ(r()->error(), R"(error: no matching call to distance(vec3<f32>)
 
 2 candidate functions:
-  distance(f32, f32) -> f32
-  distance(vecN<f32>, vecN<f32>) -> f32
+  distance(T, T) -> T  where: T is f32 or f16
+  distance(vecN<T>, vecN<T>) -> T  where: T is f32 or f16
 )");
 }
 
@@ -655,8 +894,8 @@ TEST_F(ResolverBuiltinFloatTest, Distance_NoParams) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to distance()
 
 2 candidate functions:
-  distance(f32, f32) -> f32
-  distance(vecN<f32>, vecN<f32>) -> f32
+  distance(T, T) -> T  where: T is f32 or f16
+  distance(vecN<T>, vecN<T>) -> T  where: T is f32 or f16
 )");
 }
 
@@ -796,6 +1035,18 @@ TEST_F(ResolverBuiltinFloatTest, Length_Scalar_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Length_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("length", 1_h);
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinFloatTest, Length_FloatVector_f32) {
     auto* call = Call("length", vec3<f32>(1_f, 1_f, 3_f));
     WrapInFunction(call);
@@ -806,6 +1057,18 @@ TEST_F(ResolverBuiltinFloatTest, Length_FloatVector_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Length_FloatVector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("length", vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinFloatTest, Length_NoParams) {
     auto* call = Call("length");
     WrapInFunction(call);
@@ -815,8 +1078,8 @@ TEST_F(ResolverBuiltinFloatTest, Length_NoParams) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to length()
 
 2 candidate functions:
-  length(f32) -> f32
-  length(vecN<f32>) -> f32
+  length(T) -> T  where: T is f32 or f16
+  length(vecN<T>) -> T  where: T is f32 or f16
 )");
 }
 
@@ -829,8 +1092,8 @@ TEST_F(ResolverBuiltinFloatTest, Length_TooManyParams) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to length(f32, f32)
 
 2 candidate functions:
-  length(f32) -> f32
-  length(vecN<f32>) -> f32
+  length(T) -> T  where: T is f32 or f16
+  length(vecN<T>) -> T  where: T is f32 or f16
 )");
 }
 
@@ -849,6 +1112,21 @@ TEST_F(ResolverBuiltinFloatTest, Mix_VectorScalar_f32) {
     EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Mix_VectorScalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("mix", vec3<f16>(1_h, 1_h, 1_h), vec3<f16>(1_h, 1_h, 1_h), 4_h);
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->is_float_vector());
+    EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+    ASSERT_NE(TypeOf(call)->As<sem::Vector>()->type(), nullptr);
+    EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+}
+
 // modf: (f32) -> __modf_result, (vecN<f32>) -> __modf_result_vecN
 TEST_F(ResolverBuiltinFloatTest, ModfScalar) {
     auto* call = Call("modf", 1_f);
@@ -987,6 +1265,20 @@ TEST_F(ResolverBuiltinFloatTest, Normalize_Vector_f32) {
     EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinFloatTest, Normalize_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* call = Call("normalize", vec3<f16>(1_h, 1_h, 3_h));
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->is_float_vector());
+    EXPECT_EQ(TypeOf(call)->As<sem::Vector>()->Width(), 3u);
+    EXPECT_TRUE(TypeOf(call)->As<sem::Vector>()->type()->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinFloatTest, Normalize_Error_NoParams) {
     auto* call = Call("normalize");
     WrapInFunction(call);
@@ -996,7 +1288,7 @@ TEST_F(ResolverBuiltinFloatTest, Normalize_Error_NoParams) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to normalize()
 
 1 candidate function:
-  normalize(vecN<f32>) -> vecN<f32>
+  normalize(vecN<T>) -> vecN<T>  where: T is f32 or f16
 )");
 }
 
@@ -1436,6 +1728,20 @@ TEST_F(ResolverBuiltinTest, Determinant_2x2_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinTest, Determinant_2x2_f16) {
+    Enable(ast::Extension::kF16);
+
+    GlobalVar("var", ty.mat2x2<f16>(), ast::StorageClass::kPrivate);
+
+    auto* call = Call("determinant", "var");
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinTest, Determinant_3x3_f32) {
     GlobalVar("var", ty.mat3x3<f32>(), ast::StorageClass::kPrivate);
 
@@ -1448,6 +1754,20 @@ TEST_F(ResolverBuiltinTest, Determinant_3x3_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinTest, Determinant_3x3_f16) {
+    Enable(ast::Extension::kF16);
+
+    GlobalVar("var", ty.mat3x3<f16>(), ast::StorageClass::kPrivate);
+
+    auto* call = Call("determinant", "var");
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinTest, Determinant_4x4_f32) {
     GlobalVar("var", ty.mat4x4<f32>(), ast::StorageClass::kPrivate);
 
@@ -1460,6 +1780,20 @@ TEST_F(ResolverBuiltinTest, Determinant_4x4_f32) {
     EXPECT_TRUE(TypeOf(call)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinTest, Determinant_4x4_f16) {
+    Enable(ast::Extension::kF16);
+
+    GlobalVar("var", ty.mat4x4<f16>(), ast::StorageClass::kPrivate);
+
+    auto* call = Call("determinant", "var");
+    WrapInFunction(call);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(call), nullptr);
+    EXPECT_TRUE(TypeOf(call)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinTest, Determinant_NotSquare) {
     GlobalVar("var", ty.mat2x3<f32>(), ast::StorageClass::kPrivate);
 
@@ -1471,7 +1805,7 @@ TEST_F(ResolverBuiltinTest, Determinant_NotSquare) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to determinant(mat2x3<f32>)
 
 1 candidate function:
-  determinant(matNxN<f32>) -> f32
+  determinant(matNxN<T>) -> T  where: T is f32 or f16
 )");
 }
 
@@ -1486,7 +1820,7 @@ TEST_F(ResolverBuiltinTest, Determinant_NotMatrix) {
     EXPECT_EQ(r()->error(), R"(error: no matching call to determinant(f32)
 
 1 candidate function:
-  determinant(matNxN<f32>) -> f32
+  determinant(matNxN<T>) -> T  where: T is f32 or f16
 )");
 }
 
@@ -1507,6 +1841,20 @@ TEST_F(ResolverBuiltinTest, Dot_Vec2_f32) {
     EXPECT_TRUE(TypeOf(expr)->Is<sem::F32>());
 }
 
+TEST_F(ResolverBuiltinTest, Dot_Vec2_f16) {
+    Enable(ast::Extension::kF16);
+
+    GlobalVar("my_var", ty.vec2<f16>(), ast::StorageClass::kPrivate);
+
+    auto* expr = Call("dot", "my_var", "my_var");
+    WrapInFunction(expr);
+
+    EXPECT_TRUE(r()->Resolve()) << r()->error();
+
+    ASSERT_NE(TypeOf(expr), nullptr);
+    EXPECT_TRUE(TypeOf(expr)->Is<sem::F16>());
+}
+
 TEST_F(ResolverBuiltinTest, Dot_Vec3_i32) {
     GlobalVar("my_var", ty.vec3<i32>(), ast::StorageClass::kPrivate);
 
@@ -1541,7 +1889,7 @@ TEST_F(ResolverBuiltinTest, Dot_Error_Scalar) {
               R"(error: no matching call to dot(f32, f32)
 
 1 candidate function:
-  dot(vecN<T>, vecN<T>) -> T  where: T is f32, i32 or u32
+  dot(vecN<T>, vecN<T>) -> T  where: T is f32, i32, u32 or f16
 )");
 }
 

src/tint/resolver/const_eval.cc

@@ -73,33 +73,32 @@ auto Dispatch_fia_fi32_f16(F&& f, CONSTANTS&&... cs) {
         [&](const sem::AbstractFloat*) { return f(cs->template As<AFloat>()...); },
         [&](const sem::F32*) { return f(cs->template As<f32>()...); },
         [&](const sem::I32*) { return f(cs->template As<i32>()...); },
-        [&](const sem::F16*) {
-            // TODO(crbug.com/tint/1502): Support const eval for f16
-            return nullptr;
-        });
+        [&](const sem::F16*) { return f(cs->template As<f16>()...); });
 }
 
 /// Helper that calls `f` passing in the value of all `cs`.
 /// Assumes all `cs` are of the same type.
 template <typename F, typename... CONSTANTS>
-auto Dispatch_fia_fiu32(F&& f, CONSTANTS&&... cs) {
+auto Dispatch_fia_fiu32_f16(F&& f, CONSTANTS&&... cs) {
     return Switch(
         First(cs...)->Type(),  //
         [&](const sem::AbstractInt*) { return f(cs->template As<AInt>()...); },
         [&](const sem::AbstractFloat*) { return f(cs->template As<AFloat>()...); },
         [&](const sem::F32*) { return f(cs->template As<f32>()...); },
         [&](const sem::I32*) { return f(cs->template As<i32>()...); },
-        [&](const sem::U32*) { return f(cs->template As<u32>()...); });
+        [&](const sem::U32*) { return f(cs->template As<u32>()...); },
+        [&](const sem::F16*) { return f(cs->template As<f16>()...); });
 }
 
 /// Helper that calls `f` passing in the value of all `cs`.
 /// Assumes all `cs` are of the same type.
 template <typename F, typename... CONSTANTS>
-auto Dispatch_fa_f32(F&& f, CONSTANTS&&... cs) {
+auto Dispatch_fa_f32_f16(F&& f, CONSTANTS&&... cs) {
     return Switch(
         First(cs...)->Type(),  //
         [&](const sem::AbstractFloat*) { return f(cs->template As<AFloat>()...); },
-        [&](const sem::F32*) { return f(cs->template As<f32>()...); });
+        [&](const sem::F32*) { return f(cs->template As<f32>()...); },
+        [&](const sem::F16*) { return f(cs->template As<f16>()...); });
 }
 
 /// ZeroTypeDispatch is a helper for calling the function `f`, passing a single zero-value argument
@@ -730,7 +729,7 @@ const sem::Constant* ConstEval::atan2(const sem::Type*,
         auto create = [&](auto i, auto j) {
             return CreateElement(builder, c0->Type(), decltype(i)(std::atan2(i.value, j.value)));
         };
-        return Dispatch_fa_f32(create, c0, c1);
+        return Dispatch_fa_f32_f16(create, c0, c1);
     };
     return TransformElements(builder, transform, args[0]->ConstantValue(),
                              args[1]->ConstantValue());
@@ -744,7 +743,7 @@ const sem::Constant* ConstEval::clamp(const sem::Type*,
             return CreateElement(builder, c0->Type(),
                                  decltype(e)(std::min(std::max(e, low), high)));
         };
-        return Dispatch_fia_fiu32(create, c0, c1, c2);
+        return Dispatch_fia_fiu32_f16(create, c0, c1, c2);
     };
     return TransformElements(builder, transform, args[0]->ConstantValue(), args[1]->ConstantValue(),
                              args[2]->ConstantValue());

src/tint/resolver/const_eval_test.cc

@@ -2994,7 +2994,8 @@ struct Values {
 };
 
 struct Case {
-    std::variant<Values<AInt>, Values<AFloat>, Values<u32>, Values<i32>, Values<f32>> values;
+    std::variant<Values<AInt>, Values<AFloat>, Values<u32>, Values<i32>, Values<f32>, Values<f16>>
+        values;
 };
 
 static std::ostream& operator<<(std::ostream& o, const Case& c) {
@@ -3101,6 +3102,15 @@ INSTANTIATE_TEST_SUITE_P(Negation,
                                               C(-kHighest<f32>, kHighest<f32>),
                                               C(kLowest<f32>, Negate(kLowest<f32>)),
                                               C(Negate(kLowest<f32>), kLowest<f32>),
+                                              // f16
+                                              C(0.0_h, -0.0_h),
+                                              C(-0.0_h, 0.0_h),
+                                              C(1.0_h, -1.0_h),
+                                              C(-1.0_h, 1.0_h),
+                                              C(kHighest<f16>, -kHighest<f16>),
+                                              C(-kHighest<f16>, kHighest<f16>),
+                                              C(kLowest<f16>, Negate(kLowest<f16>)),
+                                              C(Negate(kLowest<f16>), kLowest<f16>),
                                           })));
 
 // Make sure UBSan doesn't trip on C++'s undefined behaviour of negating the smallest negative
@@ -3252,7 +3262,8 @@ INSTANTIATE_TEST_SUITE_P(  //
     ResolverConstEvalBuiltinTest,
     testing::Combine(testing::Values(sem::BuiltinType::kAtan2),
                      testing::ValuesIn(Concat(Atan2Cases<AFloat, true>(),  //
-                                              Atan2Cases<f32, false>()))));
+                                              Atan2Cases<f32, false>(),
+                                              Atan2Cases<f16, false>()))));
 
 template <typename T>
 std::vector<Case> ClampCases() {
@@ -3277,7 +3288,8 @@ INSTANTIATE_TEST_SUITE_P(  //
                                               ClampCases<i32>(),
                                               ClampCases<u32>(),
                                               ClampCases<AFloat>(),
-                                              ClampCases<f32>()))));
+                                              ClampCases<f32>(),
+                                              ClampCases<f16>()))));
 
 }  // namespace builtin
 

src/tint/writer/glsl/generator_impl.cc

@@ -1302,10 +1302,12 @@ bool GeneratorImpl::EmitFrexpCall(std::ostream& out,
 bool GeneratorImpl::EmitDegreesCall(std::ostream& out,
                                     const ast::CallExpression* expr,
                                     const sem::Builtin* builtin) {
+    auto* return_elem_type = sem::Type::DeepestElementOf(builtin->ReturnType());
+    const std::string suffix = Is<sem::F16>(return_elem_type) ? "hf" : "f";
     return CallBuiltinHelper(out, expr, builtin,
                              [&](TextBuffer* b, const std::vector<std::string>& params) {
                                  line(b) << "return " << params[0] << " * " << std::setprecision(20)
-                                         << sem::kRadToDeg << ";";
+                                         << sem::kRadToDeg << suffix << ";";
                                  return true;
                              });
 }
@@ -1313,10 +1315,12 @@ bool GeneratorImpl::EmitDegreesCall(std::ostream& out,
 bool GeneratorImpl::EmitRadiansCall(std::ostream& out,
                                     const ast::CallExpression* expr,
                                     const sem::Builtin* builtin) {
+    auto* return_elem_type = sem::Type::DeepestElementOf(builtin->ReturnType());
+    const std::string suffix = Is<sem::F16>(return_elem_type) ? "hf" : "f";
     return CallBuiltinHelper(out, expr, builtin,
                              [&](TextBuffer* b, const std::vector<std::string>& params) {
                                  line(b) << "return " << params[0] << " * " << std::setprecision(20)
-                                         << sem::kDegToRad << ";";
+                                         << sem::kDegToRad << suffix << ";";
                                  return true;
                              });
 }

src/tint/writer/glsl/generator_impl_builtin_test.cc

@@ -29,36 +29,40 @@ using BuiltinType = sem::BuiltinType;
 
 using GlslGeneratorImplTest_Builtin = TestHelper;
 
-enum class ParamType {
+enum class CallParamType {
     kF32,
     kU32,
     kBool,
+    kF16,
 };
 
 struct BuiltinData {
     BuiltinType builtin;
-    ParamType type;
+    CallParamType type;
     const char* glsl_name;
 };
 inline std::ostream& operator<<(std::ostream& out, BuiltinData data) {
-    out << data.glsl_name;
+    out << data.glsl_name << "<";
     switch (data.type) {
-        case ParamType::kF32:
+        case CallParamType::kF32:
             out << "f32";
             break;
-        case ParamType::kU32:
+        case CallParamType::kU32:
             out << "u32";
             break;
-        case ParamType::kBool:
+        case CallParamType::kBool:
             out << "bool";
             break;
+        case CallParamType::kF16:
+            out << "f16";
+            break;
     }
     out << ">";
     return out;
 }
 
 const ast::CallExpression* GenerateCall(BuiltinType builtin,
-                                        ParamType type,
+                                        CallParamType type,
                                         ProgramBuilder* builder) {
     std::string name;
     std::ostringstream str(name);
@@ -96,29 +100,51 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
         case BuiltinType::kTanh:
         case BuiltinType::kTrunc:
         case BuiltinType::kSign:
-            return builder->Call(str.str(), "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2");
+            } else {
+                return builder->Call(str.str(), "f2");
+            }
         case BuiltinType::kLdexp:
-            return builder->Call(str.str(), "f2", "i2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "i2");
+            } else {
+                return builder->Call(str.str(), "f2", "i2");
+            }
         case BuiltinType::kAtan2:
         case BuiltinType::kDot:
         case BuiltinType::kDistance:
         case BuiltinType::kPow:
         case BuiltinType::kReflect:
         case BuiltinType::kStep:
-            return builder->Call(str.str(), "f2", "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2");
+            }
         case BuiltinType::kCross:
-            return builder->Call(str.str(), "f3", "f3");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h3", "h3");
+            } else {
+                return builder->Call(str.str(), "f3", "f3");
+            }
         case BuiltinType::kFma:
         case BuiltinType::kMix:
         case BuiltinType::kFaceForward:
         case BuiltinType::kSmoothstep:
-            return builder->Call(str.str(), "f2", "f2", "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "h2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2", "f2");
+            }
         case BuiltinType::kAll:
         case BuiltinType::kAny:
             return builder->Call(str.str(), "b2");
         case BuiltinType::kAbs:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2");
             } else {
                 return builder->Call(str.str(), "u2");
             }
@@ -127,23 +153,39 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
             return builder->Call(str.str(), "u2");
         case BuiltinType::kMax:
         case BuiltinType::kMin:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2", "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2");
             } else {
                 return builder->Call(str.str(), "u2", "u2");
             }
         case BuiltinType::kClamp:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2", "f2", "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "h2");
             } else {
                 return builder->Call(str.str(), "u2", "u2", "u2");
             }
         case BuiltinType::kSelect:
-            return builder->Call(str.str(), "f2", "f2", "b2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "b2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2", "b2");
+            }
         case BuiltinType::kDeterminant:
-            return builder->Call(str.str(), "m2x2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "hm2x2");
+            } else {
+                return builder->Call(str.str(), "m2x2");
+            }
         case BuiltinType::kTranspose:
-            return builder->Call(str.str(), "m3x2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "hm3x2");
+            } else {
+                return builder->Call(str.str(), "m3x2");
+            }
         default:
             break;
     }
@@ -153,6 +195,15 @@ using GlslBuiltinTest = TestParamHelper<BuiltinData>;
 TEST_P(GlslBuiltinTest, Emit) {
     auto param = GetParam();
 
+    if (param.type == CallParamType::kF16) {
+        Enable(ast::Extension::kF16);
+
+        GlobalVar("h2", ty.vec2<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("h3", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("hm2x2", ty.mat2x2<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("hm3x2", ty.mat3x2<f16>(), ast::StorageClass::kPrivate);
+    }
+
     GlobalVar("f2", ty.vec2<f32>(), ast::StorageClass::kPrivate);
     GlobalVar("f3", ty.vec3<f32>(), ast::StorageClass::kPrivate);
     GlobalVar("u2", ty.vec2<u32>(), ast::StorageClass::kPrivate);
@@ -180,64 +231,110 @@ TEST_P(GlslBuiltinTest, Emit) {
 INSTANTIATE_TEST_SUITE_P(
     GlslGeneratorImplTest_Builtin,
     GlslBuiltinTest,
-    testing::Values(BuiltinData{BuiltinType::kAbs, ParamType::kF32, "abs"},
-                    BuiltinData{BuiltinType::kAbs, ParamType::kU32, "abs"},
-                    BuiltinData{BuiltinType::kAcos, ParamType::kF32, "acos"},
-                    BuiltinData{BuiltinType::kAll, ParamType::kBool, "all"},
-                    BuiltinData{BuiltinType::kAny, ParamType::kBool, "any"},
-                    BuiltinData{BuiltinType::kAsin, ParamType::kF32, "asin"},
-                    BuiltinData{BuiltinType::kAtan, ParamType::kF32, "atan"},
-                    BuiltinData{BuiltinType::kAtan2, ParamType::kF32, "atan"},
-                    BuiltinData{BuiltinType::kCeil, ParamType::kF32, "ceil"},
-                    BuiltinData{BuiltinType::kClamp, ParamType::kF32, "clamp"},
-                    BuiltinData{BuiltinType::kClamp, ParamType::kU32, "clamp"},
-                    BuiltinData{BuiltinType::kCos, ParamType::kF32, "cos"},
-                    BuiltinData{BuiltinType::kCosh, ParamType::kF32, "cosh"},
-                    BuiltinData{BuiltinType::kCountOneBits, ParamType::kU32, "bitCount"},
-                    BuiltinData{BuiltinType::kCross, ParamType::kF32, "cross"},
-                    BuiltinData{BuiltinType::kDeterminant, ParamType::kF32, "determinant"},
-                    BuiltinData{BuiltinType::kDistance, ParamType::kF32, "distance"},
-                    BuiltinData{BuiltinType::kDot, ParamType::kF32, "dot"},
-                    BuiltinData{BuiltinType::kDpdx, ParamType::kF32, "dFdx"},
-                    BuiltinData{BuiltinType::kDpdxCoarse, ParamType::kF32, "dFdx"},
-                    BuiltinData{BuiltinType::kDpdxFine, ParamType::kF32, "dFdx"},
-                    BuiltinData{BuiltinType::kDpdy, ParamType::kF32, "dFdy"},
-                    BuiltinData{BuiltinType::kDpdyCoarse, ParamType::kF32, "dFdy"},
-                    BuiltinData{BuiltinType::kDpdyFine, ParamType::kF32, "dFdy"},
-                    BuiltinData{BuiltinType::kExp, ParamType::kF32, "exp"},
-                    BuiltinData{BuiltinType::kExp2, ParamType::kF32, "exp2"},
-                    BuiltinData{BuiltinType::kFaceForward, ParamType::kF32, "faceforward"},
-                    BuiltinData{BuiltinType::kFloor, ParamType::kF32, "floor"},
-                    BuiltinData{BuiltinType::kFma, ParamType::kF32, "fma"},
-                    BuiltinData{BuiltinType::kFract, ParamType::kF32, "fract"},
-                    BuiltinData{BuiltinType::kFwidth, ParamType::kF32, "fwidth"},
-                    BuiltinData{BuiltinType::kFwidthCoarse, ParamType::kF32, "fwidth"},
-                    BuiltinData{BuiltinType::kFwidthFine, ParamType::kF32, "fwidth"},
-                    BuiltinData{BuiltinType::kInverseSqrt, ParamType::kF32, "inversesqrt"},
-                    BuiltinData{BuiltinType::kLdexp, ParamType::kF32, "ldexp"},
-                    BuiltinData{BuiltinType::kLength, ParamType::kF32, "length"},
-                    BuiltinData{BuiltinType::kLog, ParamType::kF32, "log"},
-                    BuiltinData{BuiltinType::kLog2, ParamType::kF32, "log2"},
-                    BuiltinData{BuiltinType::kMax, ParamType::kF32, "max"},
-                    BuiltinData{BuiltinType::kMax, ParamType::kU32, "max"},
-                    BuiltinData{BuiltinType::kMin, ParamType::kF32, "min"},
-                    BuiltinData{BuiltinType::kMin, ParamType::kU32, "min"},
-                    BuiltinData{BuiltinType::kMix, ParamType::kF32, "mix"},
-                    BuiltinData{BuiltinType::kNormalize, ParamType::kF32, "normalize"},
-                    BuiltinData{BuiltinType::kPow, ParamType::kF32, "pow"},
-                    BuiltinData{BuiltinType::kReflect, ParamType::kF32, "reflect"},
-                    BuiltinData{BuiltinType::kReverseBits, ParamType::kU32, "bitfieldReverse"},
-                    BuiltinData{BuiltinType::kRound, ParamType::kU32, "round"},
-                    BuiltinData{BuiltinType::kSign, ParamType::kF32, "sign"},
-                    BuiltinData{BuiltinType::kSin, ParamType::kF32, "sin"},
-                    BuiltinData{BuiltinType::kSinh, ParamType::kF32, "sinh"},
-                    BuiltinData{BuiltinType::kSmoothstep, ParamType::kF32, "smoothstep"},
-                    BuiltinData{BuiltinType::kSqrt, ParamType::kF32, "sqrt"},
-                    BuiltinData{BuiltinType::kStep, ParamType::kF32, "step"},
-                    BuiltinData{BuiltinType::kTan, ParamType::kF32, "tan"},
-                    BuiltinData{BuiltinType::kTanh, ParamType::kF32, "tanh"},
-                    BuiltinData{BuiltinType::kTranspose, ParamType::kF32, "transpose"},
-                    BuiltinData{BuiltinType::kTrunc, ParamType::kF32, "trunc"}));
+    testing::Values(/* Logical built-in */
+                    BuiltinData{BuiltinType::kAll, CallParamType::kBool, "all"},
+                    BuiltinData{BuiltinType::kAny, CallParamType::kBool, "any"},
+                    /* Float built-in */
+                    BuiltinData{BuiltinType::kAbs, CallParamType::kF32, "abs"},
+                    BuiltinData{BuiltinType::kAbs, CallParamType::kF16, "abs"},
+                    BuiltinData{BuiltinType::kAcos, CallParamType::kF32, "acos"},
+                    BuiltinData{BuiltinType::kAcos, CallParamType::kF16, "acos"},
+                    BuiltinData{BuiltinType::kAsin, CallParamType::kF32, "asin"},
+                    BuiltinData{BuiltinType::kAsin, CallParamType::kF16, "asin"},
+                    BuiltinData{BuiltinType::kAtan, CallParamType::kF32, "atan"},
+                    BuiltinData{BuiltinType::kAtan, CallParamType::kF16, "atan"},
+                    BuiltinData{BuiltinType::kAtan2, CallParamType::kF32, "atan"},
+                    BuiltinData{BuiltinType::kAtan2, CallParamType::kF16, "atan"},
+                    BuiltinData{BuiltinType::kCeil, CallParamType::kF32, "ceil"},
+                    BuiltinData{BuiltinType::kCeil, CallParamType::kF16, "ceil"},
+                    BuiltinData{BuiltinType::kClamp, CallParamType::kF32, "clamp"},
+                    BuiltinData{BuiltinType::kClamp, CallParamType::kF16, "clamp"},
+                    BuiltinData{BuiltinType::kCos, CallParamType::kF32, "cos"},
+                    BuiltinData{BuiltinType::kCos, CallParamType::kF16, "cos"},
+                    BuiltinData{BuiltinType::kCosh, CallParamType::kF32, "cosh"},
+                    BuiltinData{BuiltinType::kCosh, CallParamType::kF16, "cosh"},
+                    BuiltinData{BuiltinType::kCross, CallParamType::kF32, "cross"},
+                    BuiltinData{BuiltinType::kCross, CallParamType::kF16, "cross"},
+                    BuiltinData{BuiltinType::kDistance, CallParamType::kF32, "distance"},
+                    BuiltinData{BuiltinType::kDistance, CallParamType::kF16, "distance"},
+                    BuiltinData{BuiltinType::kExp, CallParamType::kF32, "exp"},
+                    BuiltinData{BuiltinType::kExp, CallParamType::kF16, "exp"},
+                    BuiltinData{BuiltinType::kExp2, CallParamType::kF32, "exp2"},
+                    BuiltinData{BuiltinType::kExp2, CallParamType::kF16, "exp2"},
+                    BuiltinData{BuiltinType::kFaceForward, CallParamType::kF32, "faceforward"},
+                    BuiltinData{BuiltinType::kFaceForward, CallParamType::kF16, "faceforward"},
+                    BuiltinData{BuiltinType::kFloor, CallParamType::kF32, "floor"},
+                    BuiltinData{BuiltinType::kFloor, CallParamType::kF16, "floor"},
+                    BuiltinData{BuiltinType::kFma, CallParamType::kF32, "fma"},
+                    BuiltinData{BuiltinType::kFma, CallParamType::kF16, "fma"},
+                    BuiltinData{BuiltinType::kFract, CallParamType::kF32, "fract"},
+                    BuiltinData{BuiltinType::kFract, CallParamType::kF16, "fract"},
+                    BuiltinData{BuiltinType::kInverseSqrt, CallParamType::kF32, "inversesqrt"},
+                    BuiltinData{BuiltinType::kInverseSqrt, CallParamType::kF16, "inversesqrt"},
+                    BuiltinData{BuiltinType::kLdexp, CallParamType::kF32, "ldexp"},
+                    BuiltinData{BuiltinType::kLdexp, CallParamType::kF16, "ldexp"},
+                    BuiltinData{BuiltinType::kLength, CallParamType::kF32, "length"},
+                    BuiltinData{BuiltinType::kLength, CallParamType::kF16, "length"},
+                    BuiltinData{BuiltinType::kLog, CallParamType::kF32, "log"},
+                    BuiltinData{BuiltinType::kLog, CallParamType::kF16, "log"},
+                    BuiltinData{BuiltinType::kLog2, CallParamType::kF32, "log2"},
+                    BuiltinData{BuiltinType::kLog2, CallParamType::kF16, "log2"},
+                    BuiltinData{BuiltinType::kMax, CallParamType::kF32, "max"},
+                    BuiltinData{BuiltinType::kMax, CallParamType::kF16, "max"},
+                    BuiltinData{BuiltinType::kMin, CallParamType::kF32, "min"},
+                    BuiltinData{BuiltinType::kMin, CallParamType::kF16, "min"},
+                    BuiltinData{BuiltinType::kMix, CallParamType::kF32, "mix"},
+                    BuiltinData{BuiltinType::kMix, CallParamType::kF16, "mix"},
+                    BuiltinData{BuiltinType::kNormalize, CallParamType::kF32, "normalize"},
+                    BuiltinData{BuiltinType::kNormalize, CallParamType::kF16, "normalize"},
+                    BuiltinData{BuiltinType::kPow, CallParamType::kF32, "pow"},
+                    BuiltinData{BuiltinType::kPow, CallParamType::kF16, "pow"},
+                    BuiltinData{BuiltinType::kReflect, CallParamType::kF32, "reflect"},
+                    BuiltinData{BuiltinType::kReflect, CallParamType::kF16, "reflect"},
+                    BuiltinData{BuiltinType::kSign, CallParamType::kF32, "sign"},
+                    BuiltinData{BuiltinType::kSign, CallParamType::kF16, "sign"},
+                    BuiltinData{BuiltinType::kSin, CallParamType::kF32, "sin"},
+                    BuiltinData{BuiltinType::kSin, CallParamType::kF16, "sin"},
+                    BuiltinData{BuiltinType::kSinh, CallParamType::kF32, "sinh"},
+                    BuiltinData{BuiltinType::kSinh, CallParamType::kF16, "sinh"},
+                    BuiltinData{BuiltinType::kSmoothstep, CallParamType::kF32, "smoothstep"},
+                    BuiltinData{BuiltinType::kSmoothstep, CallParamType::kF16, "smoothstep"},
+                    BuiltinData{BuiltinType::kSqrt, CallParamType::kF32, "sqrt"},
+                    BuiltinData{BuiltinType::kSqrt, CallParamType::kF16, "sqrt"},
+                    BuiltinData{BuiltinType::kStep, CallParamType::kF32, "step"},
+                    BuiltinData{BuiltinType::kStep, CallParamType::kF16, "step"},
+                    BuiltinData{BuiltinType::kTan, CallParamType::kF32, "tan"},
+                    BuiltinData{BuiltinType::kTan, CallParamType::kF16, "tan"},
+                    BuiltinData{BuiltinType::kTanh, CallParamType::kF32, "tanh"},
+                    BuiltinData{BuiltinType::kTanh, CallParamType::kF16, "tanh"},
+                    BuiltinData{BuiltinType::kTrunc, CallParamType::kF32, "trunc"},
+                    BuiltinData{BuiltinType::kTrunc, CallParamType::kF16, "trunc"},
+                    /* Integer built-in */
+                    BuiltinData{BuiltinType::kAbs, CallParamType::kU32, "abs"},
+                    BuiltinData{BuiltinType::kClamp, CallParamType::kU32, "clamp"},
+                    BuiltinData{BuiltinType::kCountOneBits, CallParamType::kU32, "bitCount"},
+                    BuiltinData{BuiltinType::kMax, CallParamType::kU32, "max"},
+                    BuiltinData{BuiltinType::kMin, CallParamType::kU32, "min"},
+                    BuiltinData{BuiltinType::kReverseBits, CallParamType::kU32, "bitfieldReverse"},
+                    BuiltinData{BuiltinType::kRound, CallParamType::kU32, "round"},
+                    /* Matrix built-in */
+                    BuiltinData{BuiltinType::kDeterminant, CallParamType::kF32, "determinant"},
+                    BuiltinData{BuiltinType::kDeterminant, CallParamType::kF16, "determinant"},
+                    BuiltinData{BuiltinType::kTranspose, CallParamType::kF32, "transpose"},
+                    BuiltinData{BuiltinType::kTranspose, CallParamType::kF16, "transpose"},
+                    /* Vector built-in */
+                    BuiltinData{BuiltinType::kDot, CallParamType::kF32, "dot"},
+                    BuiltinData{BuiltinType::kDot, CallParamType::kF16, "dot"},
+                    /* Derivate built-in */
+                    BuiltinData{BuiltinType::kDpdx, CallParamType::kF32, "dFdx"},
+                    BuiltinData{BuiltinType::kDpdxCoarse, CallParamType::kF32, "dFdx"},
+                    BuiltinData{BuiltinType::kDpdxFine, CallParamType::kF32, "dFdx"},
+                    BuiltinData{BuiltinType::kDpdy, CallParamType::kF32, "dFdy"},
+                    BuiltinData{BuiltinType::kDpdyCoarse, CallParamType::kF32, "dFdy"},
+                    BuiltinData{BuiltinType::kDpdyFine, CallParamType::kF32, "dFdy"},
+                    BuiltinData{BuiltinType::kFwidth, CallParamType::kF32, "fwidth"},
+                    BuiltinData{BuiltinType::kFwidthCoarse, CallParamType::kF32, "fwidth"},
+                    BuiltinData{BuiltinType::kFwidthFine, CallParamType::kF32, "fwidth"}));
 
 TEST_F(GlslGeneratorImplTest_Builtin, Builtin_Call) {
     auto* call = Call("dot", "param1", "param2");
@@ -277,6 +374,41 @@ TEST_F(GlslGeneratorImplTest_Builtin, Select_Vector) {
     EXPECT_EQ(out.str(), "mix(ivec2(1, 2), ivec2(3, 4), bvec2(true, false))");
 }
 
+TEST_F(GlslGeneratorImplTest_Builtin, FMA_f32) {
+    auto* call = Call("fma", "a", "b", "c");
+
+    GlobalVar("a", ty.vec3<f32>(), ast::StorageClass::kPrivate);
+    GlobalVar("b", ty.vec3<f32>(), ast::StorageClass::kPrivate);
+    GlobalVar("c", ty.vec3<f32>(), ast::StorageClass::kPrivate);
+
+    WrapInFunction(CallStmt(call));
+
+    GeneratorImpl& gen = Build();
+
+    gen.increment_indent();
+    std::stringstream out;
+    ASSERT_TRUE(gen.EmitExpression(out, call)) << gen.error();
+    EXPECT_EQ(out.str(), "((a) * (b) + (c))");
+}
+
+TEST_F(GlslGeneratorImplTest_Builtin, FMA_f16) {
+    Enable(ast::Extension::kF16);
+
+    GlobalVar("a", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+    GlobalVar("b", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+    GlobalVar("c", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+
+    auto* call = Call("fma", "a", "b", "c");
+    WrapInFunction(CallStmt(call));
+
+    GeneratorImpl& gen = Build();
+
+    gen.increment_indent();
+    std::stringstream out;
+    ASSERT_TRUE(gen.EmitExpression(out, call)) << gen.error();
+    EXPECT_EQ(out.str(), "((a) * (b) + (c))");
+}
+
 TEST_F(GlslGeneratorImplTest_Builtin, Modf_Scalar) {
     auto* call = Call("modf", 1_f);
     WrapInFunction(CallStmt(call));
@@ -402,7 +534,7 @@ void main() {
 )"));
 }
 
-TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Scalar) {
+TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Scalar_f32) {
     auto* val = Var("val", ty.f32());
     auto* call = Call("degrees", val);
     WrapInFunction(val, call);
@@ -413,7 +545,7 @@ TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Scalar) {
     EXPECT_EQ(gen.result(), R"(#version 310 es
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -430,7 +562,7 @@ void main() {
 )");
 }
 
-TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Vector) {
+TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Vector_f32) {
     auto* val = Var("val", ty.vec3<f32>());
     auto* call = Call("degrees", val);
     WrapInFunction(val, call);
@@ -441,7 +573,7 @@ TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Vector) {
     EXPECT_EQ(gen.result(), R"(#version 310 es
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -458,7 +590,69 @@ void main() {
 )");
 }
 
-TEST_F(GlslGeneratorImplTest_Builtin, Radians_Scalar) {
+TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.f16());
+    auto* call = Call("degrees", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#version 310 es
+#extension GL_AMD_gpu_shader_half_float : require
+
+float16_t tint_degrees(float16_t param_0) {
+  return param_0 * 57.295779513082322865hf;
+}
+
+
+void test_function() {
+  float16_t val = 0.0hf;
+  float16_t tint_symbol = tint_degrees(val);
+}
+
+layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
+void main() {
+  test_function();
+  return;
+}
+)");
+}
+
+TEST_F(GlslGeneratorImplTest_Builtin, Degrees_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.vec3<f16>());
+    auto* call = Call("degrees", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#version 310 es
+#extension GL_AMD_gpu_shader_half_float : require
+
+f16vec3 tint_degrees(f16vec3 param_0) {
+  return param_0 * 57.295779513082322865hf;
+}
+
+
+void test_function() {
+  f16vec3 val = f16vec3(0.0hf, 0.0hf, 0.0hf);
+  f16vec3 tint_symbol = tint_degrees(val);
+}
+
+layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
+void main() {
+  test_function();
+  return;
+}
+)");
+}
+
+TEST_F(GlslGeneratorImplTest_Builtin, Radians_Scalar_f32) {
     auto* val = Var("val", ty.f32());
     auto* call = Call("radians", val);
     WrapInFunction(val, call);
@@ -469,7 +663,7 @@ TEST_F(GlslGeneratorImplTest_Builtin, Radians_Scalar) {
     EXPECT_EQ(gen.result(), R"(#version 310 es
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -486,7 +680,7 @@ void main() {
 )");
 }
 
-TEST_F(GlslGeneratorImplTest_Builtin, Radians_Vector) {
+TEST_F(GlslGeneratorImplTest_Builtin, Radians_Vector_f32) {
     auto* val = Var("val", ty.vec3<f32>());
     auto* call = Call("radians", val);
     WrapInFunction(val, call);
@@ -497,7 +691,7 @@ TEST_F(GlslGeneratorImplTest_Builtin, Radians_Vector) {
     EXPECT_EQ(gen.result(), R"(#version 310 es
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -514,6 +708,68 @@ void main() {
 )");
 }
 
+TEST_F(GlslGeneratorImplTest_Builtin, Radians_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.f16());
+    auto* call = Call("radians", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#version 310 es
+#extension GL_AMD_gpu_shader_half_float : require
+
+float16_t tint_radians(float16_t param_0) {
+  return param_0 * 0.017453292519943295474hf;
+}
+
+
+void test_function() {
+  float16_t val = 0.0hf;
+  float16_t tint_symbol = tint_radians(val);
+}
+
+layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
+void main() {
+  test_function();
+  return;
+}
+)");
+}
+
+TEST_F(GlslGeneratorImplTest_Builtin, Radians_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.vec3<f16>());
+    auto* call = Call("radians", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#version 310 es
+#extension GL_AMD_gpu_shader_half_float : require
+
+f16vec3 tint_radians(f16vec3 param_0) {
+  return param_0 * 0.017453292519943295474hf;
+}
+
+
+void test_function() {
+  f16vec3 val = f16vec3(0.0hf, 0.0hf, 0.0hf);
+  f16vec3 tint_symbol = tint_radians(val);
+}
+
+layout(local_size_x = 1, local_size_y = 1, local_size_z = 1) in;
+void main() {
+  test_function();
+  return;
+}
+)");
+}
+
 TEST_F(GlslGeneratorImplTest_Builtin, ExtractBits) {
     auto* v = Var("v", ty.vec3<u32>());
     auto* offset = Var("offset", ty.u32());
@@ -828,23 +1084,6 @@ void main() {
 )");
 }
 
-TEST_F(GlslGeneratorImplTest_Builtin, FMA) {
-    auto* call = Call("fma", "a", "b", "c");
-
-    GlobalVar("a", ty.vec3<f32>(), ast::StorageClass::kPrivate);
-    GlobalVar("b", ty.vec3<f32>(), ast::StorageClass::kPrivate);
-    GlobalVar("c", ty.vec3<f32>(), ast::StorageClass::kPrivate);
-
-    WrapInFunction(CallStmt(call));
-
-    GeneratorImpl& gen = Build();
-
-    gen.increment_indent();
-    std::stringstream out;
-    ASSERT_TRUE(gen.EmitExpression(out, call)) << gen.error();
-    EXPECT_EQ(out.str(), "((a) * (b) + (c))");
-}
-
 TEST_F(GlslGeneratorImplTest_Builtin, DotU32) {
     GlobalVar("v", ty.vec3<u32>(), ast::StorageClass::kPrivate);
     WrapInFunction(CallStmt(Call("dot", "v", "v")));

src/tint/writer/hlsl/generator_impl_builtin_test.cc

@@ -28,36 +28,40 @@ namespace {
 using BuiltinType = sem::BuiltinType;
 using HlslGeneratorImplTest_Builtin = TestHelper;
 
-enum class ParamType {
+enum class CallParamType {
     kF32,
     kU32,
     kBool,
+    kF16,
 };
 
 struct BuiltinData {
     BuiltinType builtin;
-    ParamType type;
+    CallParamType type;
     const char* hlsl_name;
 };
 inline std::ostream& operator<<(std::ostream& out, BuiltinData data) {
-    out << data.hlsl_name;
+    out << data.hlsl_name << "<";
     switch (data.type) {
-        case ParamType::kF32:
+        case CallParamType::kF32:
             out << "f32";
             break;
-        case ParamType::kU32:
+        case CallParamType::kU32:
             out << "u32";
             break;
-        case ParamType::kBool:
+        case CallParamType::kBool:
             out << "bool";
             break;
+        case CallParamType::kF16:
+            out << "f16";
+            break;
     }
     out << ">";
     return out;
 }
 
 const ast::CallExpression* GenerateCall(BuiltinType builtin,
-                                        ParamType type,
+                                        CallParamType type,
                                         ProgramBuilder* builder) {
     std::string name;
     std::ostringstream str(name);
@@ -95,29 +99,51 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
         case BuiltinType::kTanh:
         case BuiltinType::kTrunc:
         case BuiltinType::kSign:
-            return builder->Call(str.str(), "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2");
+            } else {
+                return builder->Call(str.str(), "f2");
+            }
         case BuiltinType::kLdexp:
-            return builder->Call(str.str(), "f2", "i2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "i2");
+            } else {
+                return builder->Call(str.str(), "f2", "i2");
+            }
         case BuiltinType::kAtan2:
         case BuiltinType::kDot:
         case BuiltinType::kDistance:
         case BuiltinType::kPow:
         case BuiltinType::kReflect:
         case BuiltinType::kStep:
-            return builder->Call(str.str(), "f2", "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2");
+            }
         case BuiltinType::kCross:
-            return builder->Call(str.str(), "f3", "f3");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h3", "h3");
+            } else {
+                return builder->Call(str.str(), "f3", "f3");
+            }
         case BuiltinType::kFma:
         case BuiltinType::kMix:
         case BuiltinType::kFaceForward:
         case BuiltinType::kSmoothstep:
-            return builder->Call(str.str(), "f2", "f2", "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "h2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2", "f2");
+            }
         case BuiltinType::kAll:
         case BuiltinType::kAny:
             return builder->Call(str.str(), "b2");
         case BuiltinType::kAbs:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2");
             } else {
                 return builder->Call(str.str(), "u2");
             }
@@ -126,32 +152,58 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
             return builder->Call(str.str(), "u2");
         case BuiltinType::kMax:
         case BuiltinType::kMin:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2", "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2");
             } else {
                 return builder->Call(str.str(), "u2", "u2");
             }
         case BuiltinType::kClamp:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2", "f2", "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "h2");
             } else {
                 return builder->Call(str.str(), "u2", "u2", "u2");
             }
         case BuiltinType::kSelect:
-            return builder->Call(str.str(), "f2", "f2", "b2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "b2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2", "b2");
+            }
         case BuiltinType::kDeterminant:
-            return builder->Call(str.str(), "m2x2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "hm2x2");
+            } else {
+                return builder->Call(str.str(), "m2x2");
+            }
         case BuiltinType::kTranspose:
-            return builder->Call(str.str(), "m3x2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "hm3x2");
+            } else {
+                return builder->Call(str.str(), "m3x2");
+            }
         default:
             break;
     }
     return nullptr;
 }
+
 using HlslBuiltinTest = TestParamHelper<BuiltinData>;
 TEST_P(HlslBuiltinTest, Emit) {
     auto param = GetParam();
 
+    if (param.type == CallParamType::kF16) {
+        Enable(ast::Extension::kF16);
+
+        GlobalVar("h2", ty.vec2<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("h3", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("hm2x2", ty.mat2x2<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("hm3x2", ty.mat3x2<f16>(), ast::StorageClass::kPrivate);
+    }
+
     GlobalVar("f2", ty.vec2<f32>(), ast::StorageClass::kPrivate);
     GlobalVar("f3", ty.vec3<f32>(), ast::StorageClass::kPrivate);
     GlobalVar("u2", ty.vec2<u32>(), ast::StorageClass::kPrivate);
@@ -179,64 +231,110 @@ TEST_P(HlslBuiltinTest, Emit) {
 INSTANTIATE_TEST_SUITE_P(
     HlslGeneratorImplTest_Builtin,
     HlslBuiltinTest,
-    testing::Values(BuiltinData{BuiltinType::kAbs, ParamType::kF32, "abs"},
-                    BuiltinData{BuiltinType::kAbs, ParamType::kU32, "abs"},
-                    BuiltinData{BuiltinType::kAcos, ParamType::kF32, "acos"},
-                    BuiltinData{BuiltinType::kAll, ParamType::kBool, "all"},
-                    BuiltinData{BuiltinType::kAny, ParamType::kBool, "any"},
-                    BuiltinData{BuiltinType::kAsin, ParamType::kF32, "asin"},
-                    BuiltinData{BuiltinType::kAtan, ParamType::kF32, "atan"},
-                    BuiltinData{BuiltinType::kAtan2, ParamType::kF32, "atan2"},
-                    BuiltinData{BuiltinType::kCeil, ParamType::kF32, "ceil"},
-                    BuiltinData{BuiltinType::kClamp, ParamType::kF32, "clamp"},
-                    BuiltinData{BuiltinType::kClamp, ParamType::kU32, "clamp"},
-                    BuiltinData{BuiltinType::kCos, ParamType::kF32, "cos"},
-                    BuiltinData{BuiltinType::kCosh, ParamType::kF32, "cosh"},
-                    BuiltinData{BuiltinType::kCountOneBits, ParamType::kU32, "countbits"},
-                    BuiltinData{BuiltinType::kCross, ParamType::kF32, "cross"},
-                    BuiltinData{BuiltinType::kDeterminant, ParamType::kF32, "determinant"},
-                    BuiltinData{BuiltinType::kDistance, ParamType::kF32, "distance"},
-                    BuiltinData{BuiltinType::kDot, ParamType::kF32, "dot"},
-                    BuiltinData{BuiltinType::kDpdx, ParamType::kF32, "ddx"},
-                    BuiltinData{BuiltinType::kDpdxCoarse, ParamType::kF32, "ddx_coarse"},
-                    BuiltinData{BuiltinType::kDpdxFine, ParamType::kF32, "ddx_fine"},
-                    BuiltinData{BuiltinType::kDpdy, ParamType::kF32, "ddy"},
-                    BuiltinData{BuiltinType::kDpdyCoarse, ParamType::kF32, "ddy_coarse"},
-                    BuiltinData{BuiltinType::kDpdyFine, ParamType::kF32, "ddy_fine"},
-                    BuiltinData{BuiltinType::kExp, ParamType::kF32, "exp"},
-                    BuiltinData{BuiltinType::kExp2, ParamType::kF32, "exp2"},
-                    BuiltinData{BuiltinType::kFaceForward, ParamType::kF32, "faceforward"},
-                    BuiltinData{BuiltinType::kFloor, ParamType::kF32, "floor"},
-                    BuiltinData{BuiltinType::kFma, ParamType::kF32, "mad"},
-                    BuiltinData{BuiltinType::kFract, ParamType::kF32, "frac"},
-                    BuiltinData{BuiltinType::kFwidth, ParamType::kF32, "fwidth"},
-                    BuiltinData{BuiltinType::kFwidthCoarse, ParamType::kF32, "fwidth"},
-                    BuiltinData{BuiltinType::kFwidthFine, ParamType::kF32, "fwidth"},
-                    BuiltinData{BuiltinType::kInverseSqrt, ParamType::kF32, "rsqrt"},
-                    BuiltinData{BuiltinType::kLdexp, ParamType::kF32, "ldexp"},
-                    BuiltinData{BuiltinType::kLength, ParamType::kF32, "length"},
-                    BuiltinData{BuiltinType::kLog, ParamType::kF32, "log"},
-                    BuiltinData{BuiltinType::kLog2, ParamType::kF32, "log2"},
-                    BuiltinData{BuiltinType::kMax, ParamType::kF32, "max"},
-                    BuiltinData{BuiltinType::kMax, ParamType::kU32, "max"},
-                    BuiltinData{BuiltinType::kMin, ParamType::kF32, "min"},
-                    BuiltinData{BuiltinType::kMin, ParamType::kU32, "min"},
-                    BuiltinData{BuiltinType::kMix, ParamType::kF32, "lerp"},
-                    BuiltinData{BuiltinType::kNormalize, ParamType::kF32, "normalize"},
-                    BuiltinData{BuiltinType::kPow, ParamType::kF32, "pow"},
-                    BuiltinData{BuiltinType::kReflect, ParamType::kF32, "reflect"},
-                    BuiltinData{BuiltinType::kReverseBits, ParamType::kU32, "reversebits"},
-                    BuiltinData{BuiltinType::kRound, ParamType::kU32, "round"},
-                    BuiltinData{BuiltinType::kSign, ParamType::kF32, "sign"},
-                    BuiltinData{BuiltinType::kSin, ParamType::kF32, "sin"},
-                    BuiltinData{BuiltinType::kSinh, ParamType::kF32, "sinh"},
-                    BuiltinData{BuiltinType::kSmoothstep, ParamType::kF32, "smoothstep"},
-                    BuiltinData{BuiltinType::kSqrt, ParamType::kF32, "sqrt"},
-                    BuiltinData{BuiltinType::kStep, ParamType::kF32, "step"},
-                    BuiltinData{BuiltinType::kTan, ParamType::kF32, "tan"},
-                    BuiltinData{BuiltinType::kTanh, ParamType::kF32, "tanh"},
-                    BuiltinData{BuiltinType::kTranspose, ParamType::kF32, "transpose"},
-                    BuiltinData{BuiltinType::kTrunc, ParamType::kF32, "trunc"}));
+    testing::Values(/* Logical built-in */
+                    BuiltinData{BuiltinType::kAll, CallParamType::kBool, "all"},
+                    BuiltinData{BuiltinType::kAny, CallParamType::kBool, "any"},
+                    /* Float built-in */
+                    BuiltinData{BuiltinType::kAbs, CallParamType::kF32, "abs"},
+                    BuiltinData{BuiltinType::kAbs, CallParamType::kF16, "abs"},
+                    BuiltinData{BuiltinType::kAcos, CallParamType::kF32, "acos"},
+                    BuiltinData{BuiltinType::kAcos, CallParamType::kF16, "acos"},
+                    BuiltinData{BuiltinType::kAsin, CallParamType::kF32, "asin"},
+                    BuiltinData{BuiltinType::kAsin, CallParamType::kF16, "asin"},
+                    BuiltinData{BuiltinType::kAtan, CallParamType::kF32, "atan"},
+                    BuiltinData{BuiltinType::kAtan, CallParamType::kF16, "atan"},
+                    BuiltinData{BuiltinType::kAtan2, CallParamType::kF32, "atan2"},
+                    BuiltinData{BuiltinType::kAtan2, CallParamType::kF16, "atan2"},
+                    BuiltinData{BuiltinType::kCeil, CallParamType::kF32, "ceil"},
+                    BuiltinData{BuiltinType::kCeil, CallParamType::kF16, "ceil"},
+                    BuiltinData{BuiltinType::kClamp, CallParamType::kF32, "clamp"},
+                    BuiltinData{BuiltinType::kClamp, CallParamType::kF16, "clamp"},
+                    BuiltinData{BuiltinType::kCos, CallParamType::kF32, "cos"},
+                    BuiltinData{BuiltinType::kCos, CallParamType::kF16, "cos"},
+                    BuiltinData{BuiltinType::kCosh, CallParamType::kF32, "cosh"},
+                    BuiltinData{BuiltinType::kCosh, CallParamType::kF16, "cosh"},
+                    BuiltinData{BuiltinType::kCross, CallParamType::kF32, "cross"},
+                    BuiltinData{BuiltinType::kCross, CallParamType::kF16, "cross"},
+                    BuiltinData{BuiltinType::kDistance, CallParamType::kF32, "distance"},
+                    BuiltinData{BuiltinType::kDistance, CallParamType::kF16, "distance"},
+                    BuiltinData{BuiltinType::kExp, CallParamType::kF32, "exp"},
+                    BuiltinData{BuiltinType::kExp, CallParamType::kF16, "exp"},
+                    BuiltinData{BuiltinType::kExp2, CallParamType::kF32, "exp2"},
+                    BuiltinData{BuiltinType::kExp2, CallParamType::kF16, "exp2"},
+                    BuiltinData{BuiltinType::kFaceForward, CallParamType::kF32, "faceforward"},
+                    BuiltinData{BuiltinType::kFaceForward, CallParamType::kF16, "faceforward"},
+                    BuiltinData{BuiltinType::kFloor, CallParamType::kF32, "floor"},
+                    BuiltinData{BuiltinType::kFloor, CallParamType::kF16, "floor"},
+                    BuiltinData{BuiltinType::kFma, CallParamType::kF32, "mad"},
+                    BuiltinData{BuiltinType::kFma, CallParamType::kF16, "mad"},
+                    BuiltinData{BuiltinType::kFract, CallParamType::kF32, "frac"},
+                    BuiltinData{BuiltinType::kFract, CallParamType::kF16, "frac"},
+                    BuiltinData{BuiltinType::kInverseSqrt, CallParamType::kF32, "rsqrt"},
+                    BuiltinData{BuiltinType::kInverseSqrt, CallParamType::kF16, "rsqrt"},
+                    BuiltinData{BuiltinType::kLdexp, CallParamType::kF32, "ldexp"},
+                    BuiltinData{BuiltinType::kLdexp, CallParamType::kF16, "ldexp"},
+                    BuiltinData{BuiltinType::kLength, CallParamType::kF32, "length"},
+                    BuiltinData{BuiltinType::kLength, CallParamType::kF16, "length"},
+                    BuiltinData{BuiltinType::kLog, CallParamType::kF32, "log"},
+                    BuiltinData{BuiltinType::kLog, CallParamType::kF16, "log"},
+                    BuiltinData{BuiltinType::kLog2, CallParamType::kF32, "log2"},
+                    BuiltinData{BuiltinType::kLog2, CallParamType::kF16, "log2"},
+                    BuiltinData{BuiltinType::kMax, CallParamType::kF32, "max"},
+                    BuiltinData{BuiltinType::kMax, CallParamType::kF16, "max"},
+                    BuiltinData{BuiltinType::kMin, CallParamType::kF32, "min"},
+                    BuiltinData{BuiltinType::kMin, CallParamType::kF16, "min"},
+                    BuiltinData{BuiltinType::kMix, CallParamType::kF32, "lerp"},
+                    BuiltinData{BuiltinType::kMix, CallParamType::kF16, "lerp"},
+                    BuiltinData{BuiltinType::kNormalize, CallParamType::kF32, "normalize"},
+                    BuiltinData{BuiltinType::kNormalize, CallParamType::kF16, "normalize"},
+                    BuiltinData{BuiltinType::kPow, CallParamType::kF32, "pow"},
+                    BuiltinData{BuiltinType::kPow, CallParamType::kF16, "pow"},
+                    BuiltinData{BuiltinType::kReflect, CallParamType::kF32, "reflect"},
+                    BuiltinData{BuiltinType::kReflect, CallParamType::kF16, "reflect"},
+                    BuiltinData{BuiltinType::kSign, CallParamType::kF32, "sign"},
+                    BuiltinData{BuiltinType::kSign, CallParamType::kF16, "sign"},
+                    BuiltinData{BuiltinType::kSin, CallParamType::kF32, "sin"},
+                    BuiltinData{BuiltinType::kSin, CallParamType::kF16, "sin"},
+                    BuiltinData{BuiltinType::kSinh, CallParamType::kF32, "sinh"},
+                    BuiltinData{BuiltinType::kSinh, CallParamType::kF16, "sinh"},
+                    BuiltinData{BuiltinType::kSmoothstep, CallParamType::kF32, "smoothstep"},
+                    BuiltinData{BuiltinType::kSmoothstep, CallParamType::kF16, "smoothstep"},
+                    BuiltinData{BuiltinType::kSqrt, CallParamType::kF32, "sqrt"},
+                    BuiltinData{BuiltinType::kSqrt, CallParamType::kF16, "sqrt"},
+                    BuiltinData{BuiltinType::kStep, CallParamType::kF32, "step"},
+                    BuiltinData{BuiltinType::kStep, CallParamType::kF16, "step"},
+                    BuiltinData{BuiltinType::kTan, CallParamType::kF32, "tan"},
+                    BuiltinData{BuiltinType::kTan, CallParamType::kF16, "tan"},
+                    BuiltinData{BuiltinType::kTanh, CallParamType::kF32, "tanh"},
+                    BuiltinData{BuiltinType::kTanh, CallParamType::kF16, "tanh"},
+                    BuiltinData{BuiltinType::kTrunc, CallParamType::kF32, "trunc"},
+                    BuiltinData{BuiltinType::kTrunc, CallParamType::kF16, "trunc"},
+                    /* Integer built-in */
+                    BuiltinData{BuiltinType::kAbs, CallParamType::kU32, "abs"},
+                    BuiltinData{BuiltinType::kClamp, CallParamType::kU32, "clamp"},
+                    BuiltinData{BuiltinType::kCountOneBits, CallParamType::kU32, "countbits"},
+                    BuiltinData{BuiltinType::kMax, CallParamType::kU32, "max"},
+                    BuiltinData{BuiltinType::kMin, CallParamType::kU32, "min"},
+                    BuiltinData{BuiltinType::kReverseBits, CallParamType::kU32, "reversebits"},
+                    BuiltinData{BuiltinType::kRound, CallParamType::kU32, "round"},
+                    /* Matrix built-in */
+                    BuiltinData{BuiltinType::kDeterminant, CallParamType::kF32, "determinant"},
+                    BuiltinData{BuiltinType::kDeterminant, CallParamType::kF16, "determinant"},
+                    BuiltinData{BuiltinType::kTranspose, CallParamType::kF32, "transpose"},
+                    BuiltinData{BuiltinType::kTranspose, CallParamType::kF16, "transpose"},
+                    /* Vector built-in */
+                    BuiltinData{BuiltinType::kDot, CallParamType::kF32, "dot"},
+                    BuiltinData{BuiltinType::kDot, CallParamType::kF16, "dot"},
+                    /* Derivate built-in */
+                    BuiltinData{BuiltinType::kDpdx, CallParamType::kF32, "ddx"},
+                    BuiltinData{BuiltinType::kDpdxCoarse, CallParamType::kF32, "ddx_coarse"},
+                    BuiltinData{BuiltinType::kDpdxFine, CallParamType::kF32, "ddx_fine"},
+                    BuiltinData{BuiltinType::kDpdy, CallParamType::kF32, "ddy"},
+                    BuiltinData{BuiltinType::kDpdyCoarse, CallParamType::kF32, "ddy_coarse"},
+                    BuiltinData{BuiltinType::kDpdyFine, CallParamType::kF32, "ddy_fine"},
+                    BuiltinData{BuiltinType::kFwidth, CallParamType::kF32, "fwidth"},
+                    BuiltinData{BuiltinType::kFwidthCoarse, CallParamType::kF32, "fwidth"},
+                    BuiltinData{BuiltinType::kFwidthFine, CallParamType::kF32, "fwidth"}));
 
 TEST_F(HlslGeneratorImplTest_Builtin, Builtin_Call) {
     auto* call = Call("dot", "param1", "param2");
@@ -380,7 +478,7 @@ void test_function() {
 )");
 }
 
-TEST_F(HlslGeneratorImplTest_Builtin, Degrees_Scalar) {
+TEST_F(HlslGeneratorImplTest_Builtin, Degrees_Scalar_f32) {
     auto* val = Var("val", ty.f32());
     auto* call = Call("degrees", val);
     WrapInFunction(val, call);
@@ -401,7 +499,7 @@ void test_function() {
 )");
 }
 
-TEST_F(HlslGeneratorImplTest_Builtin, Degrees_Vector) {
+TEST_F(HlslGeneratorImplTest_Builtin, Degrees_Vector_f32) {
     auto* val = Var("val", ty.vec3<f32>());
     auto* call = Call("degrees", val);
     WrapInFunction(val, call);
@@ -422,7 +520,53 @@ void test_function() {
 )");
 }
 
-TEST_F(HlslGeneratorImplTest_Builtin, Radians_Scalar) {
+TEST_F(HlslGeneratorImplTest_Builtin, Degrees_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.f16());
+    auto* call = Call("degrees", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(float16_t tint_degrees(float16_t param_0) {
+  return param_0 * 57.295779513082322865;
+}
+
+[numthreads(1, 1, 1)]
+void test_function() {
+  float16_t val = float16_t(0.0h);
+  const float16_t tint_symbol = tint_degrees(val);
+  return;
+}
+)");
+}
+
+TEST_F(HlslGeneratorImplTest_Builtin, Degrees_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.vec3<f16>());
+    auto* call = Call("degrees", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(vector<float16_t, 3> tint_degrees(vector<float16_t, 3> param_0) {
+  return param_0 * 57.295779513082322865;
+}
+
+[numthreads(1, 1, 1)]
+void test_function() {
+  vector<float16_t, 3> val = vector<float16_t, 3>(float16_t(0.0h), float16_t(0.0h), float16_t(0.0h));
+  const vector<float16_t, 3> tint_symbol = tint_degrees(val);
+  return;
+}
+)");
+}
+
+TEST_F(HlslGeneratorImplTest_Builtin, Radians_Scalar_f32) {
     auto* val = Var("val", ty.f32());
     auto* call = Call("radians", val);
     WrapInFunction(val, call);
@@ -443,7 +587,7 @@ void test_function() {
 )");
 }
 
-TEST_F(HlslGeneratorImplTest_Builtin, Radians_Vector) {
+TEST_F(HlslGeneratorImplTest_Builtin, Radians_Vector_f32) {
     auto* val = Var("val", ty.vec3<f32>());
     auto* call = Call("radians", val);
     WrapInFunction(val, call);
@@ -464,6 +608,52 @@ void test_function() {
 )");
 }
 
+TEST_F(HlslGeneratorImplTest_Builtin, Radians_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.f16());
+    auto* call = Call("radians", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(float16_t tint_radians(float16_t param_0) {
+  return param_0 * 0.017453292519943295474;
+}
+
+[numthreads(1, 1, 1)]
+void test_function() {
+  float16_t val = float16_t(0.0h);
+  const float16_t tint_symbol = tint_radians(val);
+  return;
+}
+)");
+}
+
+TEST_F(HlslGeneratorImplTest_Builtin, Radians_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.vec3<f16>());
+    auto* call = Call("radians", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(vector<float16_t, 3> tint_radians(vector<float16_t, 3> param_0) {
+  return param_0 * 0.017453292519943295474;
+}
+
+[numthreads(1, 1, 1)]
+void test_function() {
+  vector<float16_t, 3> val = vector<float16_t, 3>(float16_t(0.0h), float16_t(0.0h), float16_t(0.0h));
+  const vector<float16_t, 3> tint_symbol = tint_radians(val);
+  return;
+}
+)");
+}
+
 TEST_F(HlslGeneratorImplTest_Builtin, Pack4x8Snorm) {
     auto* call = Call("pack4x8snorm", "p1");
     GlobalVar("p1", ty.vec4<f32>(), ast::StorageClass::kPrivate);

src/tint/writer/msl/generator_impl_builtin_test.cc

@@ -25,36 +25,40 @@ using BuiltinType = sem::BuiltinType;
 
 using MslGeneratorImplTest = TestHelper;
 
-enum class ParamType {
+enum class CallParamType {
     kF32,
     kU32,
     kBool,
+    kF16,
 };
 
 struct BuiltinData {
     BuiltinType builtin;
-    ParamType type;
+    CallParamType type;
     const char* msl_name;
 };
 inline std::ostream& operator<<(std::ostream& out, BuiltinData data) {
     out << data.msl_name << "<";
     switch (data.type) {
-        case ParamType::kF32:
+        case CallParamType::kF32:
             out << "f32";
             break;
-        case ParamType::kU32:
+        case CallParamType::kU32:
             out << "u32";
             break;
-        case ParamType::kBool:
+        case CallParamType::kBool:
             out << "bool";
             break;
+        case CallParamType::kF16:
+            out << "f16";
+            break;
     }
     out << ">";
     return out;
 }
 
 const ast::CallExpression* GenerateCall(BuiltinType builtin,
-                                        ParamType type,
+                                        CallParamType type,
                                         ProgramBuilder* builder) {
     std::string name;
     std::ostringstream str(name);
@@ -92,31 +96,53 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
         case BuiltinType::kTanh:
         case BuiltinType::kTrunc:
         case BuiltinType::kSign:
-            return builder->Call(str.str(), "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2");
+            } else {
+                return builder->Call(str.str(), "f2");
+            }
         case BuiltinType::kLdexp:
-            return builder->Call(str.str(), "f2", "i2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "i2");
+            } else {
+                return builder->Call(str.str(), "f2", "i2");
+            }
         case BuiltinType::kAtan2:
         case BuiltinType::kDot:
         case BuiltinType::kDistance:
         case BuiltinType::kPow:
         case BuiltinType::kReflect:
         case BuiltinType::kStep:
-            return builder->Call(str.str(), "f2", "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2");
+            }
         case BuiltinType::kStorageBarrier:
             return builder->Call(str.str());
         case BuiltinType::kCross:
-            return builder->Call(str.str(), "f3", "f3");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h3", "h3");
+            } else {
+                return builder->Call(str.str(), "f3", "f3");
+            }
         case BuiltinType::kFma:
         case BuiltinType::kMix:
         case BuiltinType::kFaceForward:
         case BuiltinType::kSmoothstep:
-            return builder->Call(str.str(), "f2", "f2", "f2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "h2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2", "f2");
+            }
         case BuiltinType::kAll:
         case BuiltinType::kAny:
             return builder->Call(str.str(), "b2");
         case BuiltinType::kAbs:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2");
             } else {
                 return builder->Call(str.str(), "u2");
             }
@@ -131,21 +157,33 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
             return builder->Call(str.str(), "u2", "u2", "u1", "u1");
         case BuiltinType::kMax:
         case BuiltinType::kMin:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2", "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2");
             } else {
                 return builder->Call(str.str(), "u2", "u2");
             }
         case BuiltinType::kClamp:
-            if (type == ParamType::kF32) {
+            if (type == CallParamType::kF32) {
                 return builder->Call(str.str(), "f2", "f2", "f2");
+            } else if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "h2");
             } else {
                 return builder->Call(str.str(), "u2", "u2", "u2");
             }
         case BuiltinType::kSelect:
-            return builder->Call(str.str(), "f2", "f2", "b2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "h2", "h2", "b2");
+            } else {
+                return builder->Call(str.str(), "f2", "f2", "b2");
+            }
         case BuiltinType::kDeterminant:
-            return builder->Call(str.str(), "m2x2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "hm2x2");
+            } else {
+                return builder->Call(str.str(), "m2x2");
+            }
         case BuiltinType::kPack2x16snorm:
         case BuiltinType::kPack2x16unorm:
             return builder->Call(str.str(), "f2");
@@ -160,7 +198,11 @@ const ast::CallExpression* GenerateCall(BuiltinType builtin,
         case BuiltinType::kWorkgroupBarrier:
             return builder->Call(str.str());
         case BuiltinType::kTranspose:
-            return builder->Call(str.str(), "m3x2");
+            if (type == CallParamType::kF16) {
+                return builder->Call(str.str(), "hm3x2");
+            } else {
+                return builder->Call(str.str(), "m3x2");
+            }
         default:
             break;
     }
@@ -171,6 +213,15 @@ using MslBuiltinTest = TestParamHelper<BuiltinData>;
 TEST_P(MslBuiltinTest, Emit) {
     auto param = GetParam();
 
+    if (param.type == CallParamType::kF16) {
+        Enable(ast::Extension::kF16);
+
+        GlobalVar("h2", ty.vec2<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("h3", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("hm2x2", ty.mat2x2<f16>(), ast::StorageClass::kPrivate);
+        GlobalVar("hm3x2", ty.mat3x2<f16>(), ast::StorageClass::kPrivate);
+    }
+
     GlobalVar("f2", ty.vec2<f32>(), ast::StorageClass::kPrivate);
     GlobalVar("f3", ty.vec3<f32>(), ast::StorageClass::kPrivate);
     GlobalVar("f4", ty.vec4<f32>(), ast::StorageClass::kPrivate);
@@ -201,76 +252,122 @@ INSTANTIATE_TEST_SUITE_P(
     MslGeneratorImplTest,
     MslBuiltinTest,
     testing::Values(
-        BuiltinData{BuiltinType::kAbs, ParamType::kF32, "fabs"},
-        BuiltinData{BuiltinType::kAbs, ParamType::kU32, "abs"},
-        BuiltinData{BuiltinType::kAcos, ParamType::kF32, "acos"},
-        BuiltinData{BuiltinType::kAll, ParamType::kBool, "all"},
-        BuiltinData{BuiltinType::kAny, ParamType::kBool, "any"},
-        BuiltinData{BuiltinType::kAsin, ParamType::kF32, "asin"},
-        BuiltinData{BuiltinType::kAtan, ParamType::kF32, "atan"},
-        BuiltinData{BuiltinType::kAtan2, ParamType::kF32, "atan2"},
-        BuiltinData{BuiltinType::kCeil, ParamType::kF32, "ceil"},
-        BuiltinData{BuiltinType::kClamp, ParamType::kF32, "clamp"},
-        BuiltinData{BuiltinType::kClamp, ParamType::kU32, "clamp"},
-        BuiltinData{BuiltinType::kCos, ParamType::kF32, "cos"},
-        BuiltinData{BuiltinType::kCosh, ParamType::kF32, "cosh"},
-        BuiltinData{BuiltinType::kCountLeadingZeros, ParamType::kU32, "clz"},
-        BuiltinData{BuiltinType::kCountOneBits, ParamType::kU32, "popcount"},
-        BuiltinData{BuiltinType::kCountTrailingZeros, ParamType::kU32, "ctz"},
-        BuiltinData{BuiltinType::kCross, ParamType::kF32, "cross"},
-        BuiltinData{BuiltinType::kDeterminant, ParamType::kF32, "determinant"},
-        BuiltinData{BuiltinType::kDistance, ParamType::kF32, "distance"},
-        BuiltinData{BuiltinType::kDot, ParamType::kF32, "dot"},
-        BuiltinData{BuiltinType::kDpdx, ParamType::kF32, "dfdx"},
-        BuiltinData{BuiltinType::kDpdxCoarse, ParamType::kF32, "dfdx"},
-        BuiltinData{BuiltinType::kDpdxFine, ParamType::kF32, "dfdx"},
-        BuiltinData{BuiltinType::kDpdy, ParamType::kF32, "dfdy"},
-        BuiltinData{BuiltinType::kDpdyCoarse, ParamType::kF32, "dfdy"},
-        BuiltinData{BuiltinType::kDpdyFine, ParamType::kF32, "dfdy"},
-        BuiltinData{BuiltinType::kExp, ParamType::kF32, "exp"},
-        BuiltinData{BuiltinType::kExp2, ParamType::kF32, "exp2"},
-        BuiltinData{BuiltinType::kExtractBits, ParamType::kU32, "extract_bits"},
-        BuiltinData{BuiltinType::kFaceForward, ParamType::kF32, "faceforward"},
-        BuiltinData{BuiltinType::kFloor, ParamType::kF32, "floor"},
-        BuiltinData{BuiltinType::kFma, ParamType::kF32, "fma"},
-        BuiltinData{BuiltinType::kFract, ParamType::kF32, "fract"},
-        BuiltinData{BuiltinType::kFwidth, ParamType::kF32, "fwidth"},
-        BuiltinData{BuiltinType::kFwidthCoarse, ParamType::kF32, "fwidth"},
-        BuiltinData{BuiltinType::kFwidthFine, ParamType::kF32, "fwidth"},
-        BuiltinData{BuiltinType::kInsertBits, ParamType::kU32, "insert_bits"},
-        BuiltinData{BuiltinType::kInverseSqrt, ParamType::kF32, "rsqrt"},
-        BuiltinData{BuiltinType::kLdexp, ParamType::kF32, "ldexp"},
-        BuiltinData{BuiltinType::kLength, ParamType::kF32, "length"},
-        BuiltinData{BuiltinType::kLog, ParamType::kF32, "log"},
-        BuiltinData{BuiltinType::kLog2, ParamType::kF32, "log2"},
-        BuiltinData{BuiltinType::kMax, ParamType::kF32, "fmax"},
-        BuiltinData{BuiltinType::kMax, ParamType::kU32, "max"},
-        BuiltinData{BuiltinType::kMin, ParamType::kF32, "fmin"},
-        BuiltinData{BuiltinType::kMin, ParamType::kU32, "min"},
-        BuiltinData{BuiltinType::kNormalize, ParamType::kF32, "normalize"},
-        BuiltinData{BuiltinType::kPack4x8snorm, ParamType::kF32, "pack_float_to_snorm4x8"},
-        BuiltinData{BuiltinType::kPack4x8unorm, ParamType::kF32, "pack_float_to_unorm4x8"},
-        BuiltinData{BuiltinType::kPack2x16snorm, ParamType::kF32, "pack_float_to_snorm2x16"},
-        BuiltinData{BuiltinType::kPack2x16unorm, ParamType::kF32, "pack_float_to_unorm2x16"},
-        BuiltinData{BuiltinType::kPow, ParamType::kF32, "pow"},
-        BuiltinData{BuiltinType::kReflect, ParamType::kF32, "reflect"},
-        BuiltinData{BuiltinType::kReverseBits, ParamType::kU32, "reverse_bits"},
-        BuiltinData{BuiltinType::kRound, ParamType::kU32, "rint"},
-        BuiltinData{BuiltinType::kSelect, ParamType::kF32, "select"},
-        BuiltinData{BuiltinType::kSign, ParamType::kF32, "sign"},
-        BuiltinData{BuiltinType::kSin, ParamType::kF32, "sin"},
-        BuiltinData{BuiltinType::kSinh, ParamType::kF32, "sinh"},
-        BuiltinData{BuiltinType::kSmoothstep, ParamType::kF32, "smoothstep"},
-        BuiltinData{BuiltinType::kSqrt, ParamType::kF32, "sqrt"},
-        BuiltinData{BuiltinType::kStep, ParamType::kF32, "step"},
-        BuiltinData{BuiltinType::kTan, ParamType::kF32, "tan"},
-        BuiltinData{BuiltinType::kTanh, ParamType::kF32, "tanh"},
-        BuiltinData{BuiltinType::kTranspose, ParamType::kF32, "transpose"},
-        BuiltinData{BuiltinType::kTrunc, ParamType::kF32, "trunc"},
-        BuiltinData{BuiltinType::kUnpack4x8snorm, ParamType::kU32, "unpack_snorm4x8_to_float"},
-        BuiltinData{BuiltinType::kUnpack4x8unorm, ParamType::kU32, "unpack_unorm4x8_to_float"},
-        BuiltinData{BuiltinType::kUnpack2x16snorm, ParamType::kU32, "unpack_snorm2x16_to_float"},
-        BuiltinData{BuiltinType::kUnpack2x16unorm, ParamType::kU32, "unpack_unorm2x16_to_float"}));
+        /* Logical built-in */
+        BuiltinData{BuiltinType::kAll, CallParamType::kBool, "all"},
+        BuiltinData{BuiltinType::kAny, CallParamType::kBool, "any"},
+        BuiltinData{BuiltinType::kSelect, CallParamType::kF32, "select"},
+        /* Float built-in */
+        BuiltinData{BuiltinType::kAbs, CallParamType::kF32, "fabs"},
+        BuiltinData{BuiltinType::kAbs, CallParamType::kF16, "fabs"},
+        BuiltinData{BuiltinType::kAcos, CallParamType::kF32, "acos"},
+        BuiltinData{BuiltinType::kAcos, CallParamType::kF16, "acos"},
+        BuiltinData{BuiltinType::kAsin, CallParamType::kF32, "asin"},
+        BuiltinData{BuiltinType::kAsin, CallParamType::kF16, "asin"},
+        BuiltinData{BuiltinType::kAtan, CallParamType::kF32, "atan"},
+        BuiltinData{BuiltinType::kAtan, CallParamType::kF16, "atan"},
+        BuiltinData{BuiltinType::kAtan2, CallParamType::kF32, "atan2"},
+        BuiltinData{BuiltinType::kAtan2, CallParamType::kF16, "atan2"},
+        BuiltinData{BuiltinType::kCeil, CallParamType::kF32, "ceil"},
+        BuiltinData{BuiltinType::kCeil, CallParamType::kF16, "ceil"},
+        BuiltinData{BuiltinType::kClamp, CallParamType::kF32, "clamp"},
+        BuiltinData{BuiltinType::kClamp, CallParamType::kF16, "clamp"},
+        BuiltinData{BuiltinType::kCos, CallParamType::kF32, "cos"},
+        BuiltinData{BuiltinType::kCos, CallParamType::kF16, "cos"},
+        BuiltinData{BuiltinType::kCosh, CallParamType::kF32, "cosh"},
+        BuiltinData{BuiltinType::kCosh, CallParamType::kF16, "cosh"},
+        BuiltinData{BuiltinType::kCross, CallParamType::kF32, "cross"},
+        BuiltinData{BuiltinType::kCross, CallParamType::kF16, "cross"},
+        BuiltinData{BuiltinType::kDistance, CallParamType::kF32, "distance"},
+        BuiltinData{BuiltinType::kDistance, CallParamType::kF16, "distance"},
+        BuiltinData{BuiltinType::kExp, CallParamType::kF32, "exp"},
+        BuiltinData{BuiltinType::kExp, CallParamType::kF16, "exp"},
+        BuiltinData{BuiltinType::kExp2, CallParamType::kF32, "exp2"},
+        BuiltinData{BuiltinType::kExp2, CallParamType::kF16, "exp2"},
+        BuiltinData{BuiltinType::kFaceForward, CallParamType::kF32, "faceforward"},
+        BuiltinData{BuiltinType::kFaceForward, CallParamType::kF16, "faceforward"},
+        BuiltinData{BuiltinType::kFloor, CallParamType::kF32, "floor"},
+        BuiltinData{BuiltinType::kFloor, CallParamType::kF16, "floor"},
+        BuiltinData{BuiltinType::kFma, CallParamType::kF32, "fma"},
+        BuiltinData{BuiltinType::kFma, CallParamType::kF16, "fma"},
+        BuiltinData{BuiltinType::kFract, CallParamType::kF32, "fract"},
+        BuiltinData{BuiltinType::kFract, CallParamType::kF16, "fract"},
+        BuiltinData{BuiltinType::kInverseSqrt, CallParamType::kF32, "rsqrt"},
+        BuiltinData{BuiltinType::kInverseSqrt, CallParamType::kF16, "rsqrt"},
+        BuiltinData{BuiltinType::kLdexp, CallParamType::kF32, "ldexp"},
+        BuiltinData{BuiltinType::kLdexp, CallParamType::kF16, "ldexp"},
+        BuiltinData{BuiltinType::kLength, CallParamType::kF32, "length"},
+        BuiltinData{BuiltinType::kLength, CallParamType::kF16, "length"},
+        BuiltinData{BuiltinType::kLog, CallParamType::kF32, "log"},
+        BuiltinData{BuiltinType::kLog, CallParamType::kF16, "log"},
+        BuiltinData{BuiltinType::kLog2, CallParamType::kF32, "log2"},
+        BuiltinData{BuiltinType::kLog2, CallParamType::kF16, "log2"},
+        BuiltinData{BuiltinType::kMax, CallParamType::kF32, "fmax"},
+        BuiltinData{BuiltinType::kMax, CallParamType::kF16, "fmax"},
+        BuiltinData{BuiltinType::kMin, CallParamType::kF32, "fmin"},
+        BuiltinData{BuiltinType::kMin, CallParamType::kF16, "fmin"},
+        BuiltinData{BuiltinType::kNormalize, CallParamType::kF32, "normalize"},
+        BuiltinData{BuiltinType::kNormalize, CallParamType::kF16, "normalize"},
+        BuiltinData{BuiltinType::kPow, CallParamType::kF32, "pow"},
+        BuiltinData{BuiltinType::kPow, CallParamType::kF16, "pow"},
+        BuiltinData{BuiltinType::kReflect, CallParamType::kF32, "reflect"},
+        BuiltinData{BuiltinType::kReflect, CallParamType::kF16, "reflect"},
+        BuiltinData{BuiltinType::kSign, CallParamType::kF32, "sign"},
+        BuiltinData{BuiltinType::kSign, CallParamType::kF16, "sign"},
+        BuiltinData{BuiltinType::kSin, CallParamType::kF32, "sin"},
+        BuiltinData{BuiltinType::kSin, CallParamType::kF16, "sin"},
+        BuiltinData{BuiltinType::kSinh, CallParamType::kF32, "sinh"},
+        BuiltinData{BuiltinType::kSinh, CallParamType::kF16, "sinh"},
+        BuiltinData{BuiltinType::kSmoothstep, CallParamType::kF32, "smoothstep"},
+        BuiltinData{BuiltinType::kSmoothstep, CallParamType::kF16, "smoothstep"},
+        BuiltinData{BuiltinType::kSqrt, CallParamType::kF32, "sqrt"},
+        BuiltinData{BuiltinType::kSqrt, CallParamType::kF16, "sqrt"},
+        BuiltinData{BuiltinType::kStep, CallParamType::kF32, "step"},
+        BuiltinData{BuiltinType::kStep, CallParamType::kF16, "step"},
+        BuiltinData{BuiltinType::kTan, CallParamType::kF32, "tan"},
+        BuiltinData{BuiltinType::kTan, CallParamType::kF16, "tan"},
+        BuiltinData{BuiltinType::kTanh, CallParamType::kF32, "tanh"},
+        BuiltinData{BuiltinType::kTanh, CallParamType::kF16, "tanh"},
+        BuiltinData{BuiltinType::kTrunc, CallParamType::kF32, "trunc"},
+        BuiltinData{BuiltinType::kTrunc, CallParamType::kF16, "trunc"},
+        /* Integer built-in */
+        BuiltinData{BuiltinType::kAbs, CallParamType::kU32, "abs"},
+        BuiltinData{BuiltinType::kClamp, CallParamType::kU32, "clamp"},
+        BuiltinData{BuiltinType::kCountLeadingZeros, CallParamType::kU32, "clz"},
+        BuiltinData{BuiltinType::kCountOneBits, CallParamType::kU32, "popcount"},
+        BuiltinData{BuiltinType::kCountTrailingZeros, CallParamType::kU32, "ctz"},
+        BuiltinData{BuiltinType::kExtractBits, CallParamType::kU32, "extract_bits"},
+        BuiltinData{BuiltinType::kInsertBits, CallParamType::kU32, "insert_bits"},
+        BuiltinData{BuiltinType::kMax, CallParamType::kU32, "max"},
+        BuiltinData{BuiltinType::kMin, CallParamType::kU32, "min"},
+        BuiltinData{BuiltinType::kReverseBits, CallParamType::kU32, "reverse_bits"},
+        BuiltinData{BuiltinType::kRound, CallParamType::kU32, "rint"},
+        /* Matrix built-in */
+        BuiltinData{BuiltinType::kDeterminant, CallParamType::kF32, "determinant"},
+        BuiltinData{BuiltinType::kTranspose, CallParamType::kF32, "transpose"},
+        /* Vector built-in */
+        BuiltinData{BuiltinType::kDot, CallParamType::kF32, "dot"},
+        /* Derivate built-in */
+        BuiltinData{BuiltinType::kDpdx, CallParamType::kF32, "dfdx"},
+        BuiltinData{BuiltinType::kDpdxCoarse, CallParamType::kF32, "dfdx"},
+        BuiltinData{BuiltinType::kDpdxFine, CallParamType::kF32, "dfdx"},
+        BuiltinData{BuiltinType::kDpdy, CallParamType::kF32, "dfdy"},
+        BuiltinData{BuiltinType::kDpdyCoarse, CallParamType::kF32, "dfdy"},
+        BuiltinData{BuiltinType::kDpdyFine, CallParamType::kF32, "dfdy"},
+        BuiltinData{BuiltinType::kFwidth, CallParamType::kF32, "fwidth"},
+        BuiltinData{BuiltinType::kFwidthCoarse, CallParamType::kF32, "fwidth"},
+        BuiltinData{BuiltinType::kFwidthFine, CallParamType::kF32, "fwidth"},
+        /* Data packing builtin */
+        BuiltinData{BuiltinType::kPack4x8snorm, CallParamType::kF32, "pack_float_to_snorm4x8"},
+        BuiltinData{BuiltinType::kPack4x8unorm, CallParamType::kF32, "pack_float_to_unorm4x8"},
+        BuiltinData{BuiltinType::kPack2x16snorm, CallParamType::kF32, "pack_float_to_snorm2x16"},
+        BuiltinData{BuiltinType::kPack2x16unorm, CallParamType::kF32, "pack_float_to_unorm2x16"},
+        /* Data unpacking builtin */
+        BuiltinData{BuiltinType::kUnpack4x8snorm, CallParamType::kU32, "unpack_snorm4x8_to_float"},
+        BuiltinData{BuiltinType::kUnpack4x8unorm, CallParamType::kU32, "unpack_unorm4x8_to_float"},
+        BuiltinData{BuiltinType::kUnpack2x16snorm, CallParamType::kU32,
+                    "unpack_snorm2x16_to_float"},
+        BuiltinData{BuiltinType::kUnpack2x16unorm, CallParamType::kU32,
+                    "unpack_unorm2x16_to_float"}));
 
 TEST_F(MslGeneratorImplTest, Builtin_Call) {
     GlobalVar("param1", ty.vec2<f32>(), ast::StorageClass::kPrivate);
@@ -308,12 +405,12 @@ TEST_F(MslGeneratorImplTest, WorkgroupBarrier) {
     EXPECT_EQ(out.str(), "threadgroup_barrier(mem_flags::mem_threadgroup)");
 }
 
-TEST_F(MslGeneratorImplTest, Degrees_Scalar) {
+TEST_F(MslGeneratorImplTest, Degrees_Scalar_f32) {
     auto* val = Var("val", ty.f32());
     auto* call = Call("degrees", val);
     WrapInFunction(val, call);
 
-    GeneratorImpl& gen = Build();
+    GeneratorImpl& gen = SanitizeAndBuild();
 
     ASSERT_TRUE(gen.Generate()) << gen.error();
     EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
@@ -333,12 +430,12 @@ kernel void test_function() {
 )");
 }
 
-TEST_F(MslGeneratorImplTest, Degrees_Vector) {
+TEST_F(MslGeneratorImplTest, Degrees_Vector_f32) {
     auto* val = Var("val", ty.vec3<f32>());
     auto* call = Call("degrees", val);
     WrapInFunction(val, call);
 
-    GeneratorImpl& gen = Build();
+    GeneratorImpl& gen = SanitizeAndBuild();
 
     ASSERT_TRUE(gen.Generate()) << gen.error();
     EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
@@ -358,12 +455,66 @@ kernel void test_function() {
 )");
 }
 
-TEST_F(MslGeneratorImplTest, Radians_Scalar) {
+TEST_F(MslGeneratorImplTest, Degrees_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.f16());
+    auto* call = Call("degrees", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
+
+using namespace metal;
+
+half tint_degrees(half param_0) {
+  return param_0 * 57.295779513082322865;
+}
+
+kernel void test_function() {
+  half val = 0.0h;
+  half const tint_symbol = tint_degrees(val);
+  return;
+}
+
+)");
+}
+
+TEST_F(MslGeneratorImplTest, Degrees_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.vec3<f16>());
+    auto* call = Call("degrees", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
+
+using namespace metal;
+
+half3 tint_degrees(half3 param_0) {
+  return param_0 * 57.295779513082322865;
+}
+
+kernel void test_function() {
+  half3 val = 0.0h;
+  half3 const tint_symbol = tint_degrees(val);
+  return;
+}
+
+)");
+}
+
+TEST_F(MslGeneratorImplTest, Radians_Scalar_f32) {
     auto* val = Var("val", ty.f32());
     auto* call = Call("radians", val);
     WrapInFunction(val, call);
 
-    GeneratorImpl& gen = Build();
+    GeneratorImpl& gen = SanitizeAndBuild();
 
     ASSERT_TRUE(gen.Generate()) << gen.error();
     EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
@@ -383,12 +534,12 @@ kernel void test_function() {
 )");
 }
 
-TEST_F(MslGeneratorImplTest, Radians_Vector) {
+TEST_F(MslGeneratorImplTest, Radians_Vector_f32) {
     auto* val = Var("val", ty.vec3<f32>());
     auto* call = Call("radians", val);
     WrapInFunction(val, call);
 
-    GeneratorImpl& gen = Build();
+    GeneratorImpl& gen = SanitizeAndBuild();
 
     ASSERT_TRUE(gen.Generate()) << gen.error();
     EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
@@ -408,6 +559,60 @@ kernel void test_function() {
 )");
 }
 
+TEST_F(MslGeneratorImplTest, Radians_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.f16());
+    auto* call = Call("radians", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
+
+using namespace metal;
+
+half tint_radians(half param_0) {
+  return param_0 * 0.017453292519943295474;
+}
+
+kernel void test_function() {
+  half val = 0.0h;
+  half const tint_symbol = tint_radians(val);
+  return;
+}
+
+)");
+}
+
+TEST_F(MslGeneratorImplTest, Radians_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* val = Var("val", ty.vec3<f16>());
+    auto* call = Call("radians", val);
+    WrapInFunction(val, call);
+
+    GeneratorImpl& gen = SanitizeAndBuild();
+
+    ASSERT_TRUE(gen.Generate()) << gen.error();
+    EXPECT_EQ(gen.result(), R"(#include <metal_stdlib>
+
+using namespace metal;
+
+half3 tint_radians(half3 param_0) {
+  return param_0 * 0.017453292519943295474;
+}
+
+kernel void test_function() {
+  half3 val = 0.0h;
+  half3 const tint_symbol = tint_radians(val);
+  return;
+}
+
+)");
+}
+
 TEST_F(MslGeneratorImplTest, Pack2x16Float) {
     auto* call = Call("pack2x16float", "p1");
     GlobalVar("p1", ty.vec2<f32>(), ast::StorageClass::kPrivate);

src/tint/writer/spirv/builder_builtin_test.cc

@@ -131,6 +131,42 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_GLSLMethod_WithLoad_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* var = GlobalVar("ident", ty.f16(), ast::StorageClass::kPrivate);
+    auto* expr = Call("round", "ident");
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateGlobalVariable(var)) << b.error();
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect =
+        R"(%10 = OpExtInstImport "GLSL.std.450"
+OpName %1 "ident"
+OpName %7 "a_func"
+%3 = OpTypeFloat 16
+%2 = OpTypePointer Private %3
+%4 = OpConstantNull %3
+%1 = OpVariable %2 Private %4
+%6 = OpTypeVoid
+%5 = OpTypeFunction %6
+%7 = OpFunction %6 None %5
+%8 = OpLabel
+%11 = OpLoad %3 %1
+%9 = OpExtInst %3 %10 RoundEven %11
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 // Tests for Logical builtins
 namespace logical_builtin_tests {
 
@@ -485,6 +521,47 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_P(Builtin_Builder_SingleParam_Float_Test, Call_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto param = GetParam();
+    // Use a variable to prevent the function being evaluated as constant.
+    auto* scalar = Var("a", nullptr, Expr(1_h));
+    auto* expr = Call(param.name, scalar);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(scalar),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%11 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %7 "a"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%5 = OpTypeFloat 16
+%6 = OpConstant %5 0x1p+0
+%8 = OpTypePointer Function %5
+%9 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%7 = OpVariable %8 Function %9
+OpStore %7 %6
+%12 = OpLoad %5 %7
+%10 = OpExtInst %5 %11 )" +
+                  param.op +
+                  R"( %12
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_P(Builtin_Builder_SingleParam_Float_Test, Call_Vector_f32) {
     auto param = GetParam();
 
@@ -527,6 +604,50 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_P(Builtin_Builder_SingleParam_Float_Test, Call_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto param = GetParam();
+
+    // Use a variable to prevent the function being evaluated as constant.
+    auto* vec = Var("a", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call(param.name, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "a"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 )" +
+                  param.op +
+                  R"( %14
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 INSTANTIATE_TEST_SUITE_P(BuiltinBuilderTest,
                          Builtin_Builder_SingleParam_Float_Test,
                          testing::Values(BuiltinData{"abs", "FAbs"},
@@ -589,6 +710,43 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Length_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* scalar = Var("a", nullptr, Expr(1_h));
+    auto* expr = Call("length", scalar);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(scalar),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%11 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %7 "a"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%5 = OpTypeFloat 16
+%6 = OpConstant %5 0x1p+0
+%8 = OpTypePointer Function %5
+%9 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%7 = OpVariable %8 Function %9
+OpStore %7 %6
+%12 = OpLoad %5 %7
+%10 = OpExtInst %5 %11 Length %12
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Length_Vector_f32) {
     auto* vec = Var("a", nullptr, vec2<f32>(1_f, 1_f));
     auto* expr = Call("length", vec);
@@ -626,6 +784,45 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Length_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* vec = Var("a", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call("length", vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "a"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%12 = OpExtInst %6 %13 Length %14
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Normalize_f32) {
     auto* vec = Var("a", nullptr, vec2<f32>(1_f, 1_f));
     auto* expr = Call("normalize", vec);
@@ -663,6 +860,45 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Normalize_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* vec = Var("a", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call("normalize", vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "a"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 Normalize %14
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 using Builtin_Builder_DualParam_Float_Test = BuiltinBuilderTestWithParam<BuiltinData>;
 TEST_P(Builtin_Builder_DualParam_Float_Test, Call_Scalar_f32) {
     auto param = GetParam();
@@ -703,6 +939,47 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_P(Builtin_Builder_DualParam_Float_Test, Call_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto param = GetParam();
+    auto* scalar = Var("scalar", nullptr, Expr(1_h));
+    auto* expr = Call(param.name, scalar, scalar);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(scalar),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%11 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %7 "scalar"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%5 = OpTypeFloat 16
+%6 = OpConstant %5 0x1p+0
+%8 = OpTypePointer Function %5
+%9 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%7 = OpVariable %8 Function %9
+OpStore %7 %6
+%12 = OpLoad %5 %7
+%13 = OpLoad %5 %7
+%10 = OpExtInst %5 %11 )" +
+                  param.op +
+                  R"( %12 %13
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_P(Builtin_Builder_DualParam_Float_Test, Call_Vector_f32) {
     auto param = GetParam();
     auto* vec = Var("vec", nullptr, vec2<f32>(1_f, 1_f));
@@ -744,6 +1021,49 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_P(Builtin_Builder_DualParam_Float_Test, Call_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto param = GetParam();
+    auto* vec = Var("vec", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call(param.name, vec, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "vec"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%15 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 )" +
+                  param.op +
+                  R"( %14 %15
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 INSTANTIATE_TEST_SUITE_P(BuiltinBuilderTest,
                          Builtin_Builder_DualParam_Float_Test,
                          testing::Values(BuiltinData{"atan2", "Atan2"},
@@ -790,6 +1110,46 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Reflect_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* vec = Var("vec", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call("reflect", vec, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "vec"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%15 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 Reflect %14 %15
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Distance_Scalar_f32) {
     auto* scalar = Var("scalar", nullptr, Expr(1_f));
     auto* expr = Call("distance", scalar, scalar);
@@ -826,6 +1186,44 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Distance_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* scalar = Var("scalar", nullptr, Expr(1_h));
+    auto* expr = Call("distance", scalar, scalar);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(scalar),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%11 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %7 "scalar"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%5 = OpTypeFloat 16
+%6 = OpConstant %5 0x1p+0
+%8 = OpTypePointer Function %5
+%9 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%7 = OpVariable %8 Function %9
+OpStore %7 %6
+%12 = OpLoad %5 %7
+%13 = OpLoad %5 %7
+%10 = OpExtInst %5 %11 Distance %12 %13
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Distance_Vector_f32) {
     auto* vec = Var("vec", nullptr, vec2<f32>(1_f, 1_f));
     auto* expr = Call("distance", vec, vec);
@@ -864,6 +1262,46 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Distance_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* vec = Var("vec", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call("distance", vec, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "vec"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%15 = OpLoad %5 %9
+%12 = OpExtInst %6 %13 Distance %14 %15
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Cross_f32) {
     auto* vec = Var("vec", nullptr, vec3<f32>(1_f, 1_f, 1_f));
     auto* expr = Call("cross", vec, vec);
@@ -902,6 +1340,46 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Cross_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* vec = Var("vec", nullptr, vec3<f16>(1_h, 1_h, 1_h));
+    auto* expr = Call("cross", vec, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "vec"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 3
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%15 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 Cross %14 %15
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 using Builtin_Builder_ThreeParam_Float_Test = BuiltinBuilderTestWithParam<BuiltinData>;
 TEST_P(Builtin_Builder_ThreeParam_Float_Test, Call_Scalar_f32) {
     auto param = GetParam();
@@ -943,6 +1421,48 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_P(Builtin_Builder_ThreeParam_Float_Test, Call_Scalar_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto param = GetParam();
+    auto* scalar = Var("scalar", nullptr, Expr(1_h));
+    auto* expr = Call(param.name, scalar, scalar, scalar);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(scalar),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%11 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %7 "scalar"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%5 = OpTypeFloat 16
+%6 = OpConstant %5 0x1p+0
+%8 = OpTypePointer Function %5
+%9 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%7 = OpVariable %8 Function %9
+OpStore %7 %6
+%12 = OpLoad %5 %7
+%13 = OpLoad %5 %7
+%14 = OpLoad %5 %7
+%10 = OpExtInst %5 %11 )" +
+                  param.op +
+                  R"( %12 %13 %14
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_P(Builtin_Builder_ThreeParam_Float_Test, Call_Vector_f32) {
     auto param = GetParam();
     auto* vec = Var("vec", nullptr, vec2<f32>(1_f, 1_f));
@@ -985,6 +1505,50 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_P(Builtin_Builder_ThreeParam_Float_Test, Call_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto param = GetParam();
+    auto* vec = Var("vec", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call(param.name, vec, vec, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "vec"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%15 = OpLoad %5 %9
+%16 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 )" +
+                  param.op +
+                  R"( %14 %15 %16
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 INSTANTIATE_TEST_SUITE_P(BuiltinBuilderTest,
                          Builtin_Builder_ThreeParam_Float_Test,
                          testing::Values(BuiltinData{"clamp", "NClamp"},
@@ -1032,6 +1596,47 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_FaceForward_Vector_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* vec = Var("vec", nullptr, vec2<f16>(1_h, 1_h));
+    auto* expr = Call("faceForward", vec, vec, vec);
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Decl(vec),
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%13 = OpExtInstImport "GLSL.std.450"
+OpName %3 "a_func"
+OpName %9 "vec"
+%2 = OpTypeVoid
+%1 = OpTypeFunction %2
+%6 = OpTypeFloat 16
+%5 = OpTypeVector %6 2
+%7 = OpConstant %6 0x1p+0
+%8 = OpConstantComposite %5 %7 %7
+%10 = OpTypePointer Function %5
+%11 = OpConstantNull %5
+%3 = OpFunction %2 None %1
+%4 = OpLabel
+%9 = OpVariable %10 Function %11
+OpStore %9 %8
+%14 = OpLoad %5 %9
+%15 = OpLoad %5 %9
+%16 = OpLoad %5 %9
+%12 = OpExtInst %5 %13 FaceForward %14 %15 %16
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Modf) {
     auto* vec = Var("vec", nullptr, vec2<f32>(1_f, 2_f));
     auto* expr = Call("modf", vec);
@@ -2130,6 +2735,43 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Determinant_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* var = GlobalVar("var", ty.mat3x3<f16>(), ast::StorageClass::kPrivate);
+    auto* expr = Call("determinant", "var");
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateGlobalVariable(var)) << b.error();
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(%12 = OpExtInstImport "GLSL.std.450"
+OpName %1 "var"
+OpName %9 "a_func"
+%5 = OpTypeFloat 16
+%4 = OpTypeVector %5 3
+%3 = OpTypeMatrix %4 3
+%2 = OpTypePointer Private %3
+%6 = OpConstantNull %3
+%1 = OpVariable %2 Private %6
+%8 = OpTypeVoid
+%7 = OpTypeFunction %8
+%9 = OpFunction %8 None %7
+%10 = OpLabel
+%13 = OpLoad %3 %1
+%11 = OpExtInst %5 %12 Determinant %13
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 TEST_F(BuiltinBuilderTest, Call_Transpose_f32) {
     auto* var = GlobalVar("var", ty.mat2x3<f32>(), ast::StorageClass::kPrivate);
     auto* expr = Call("transpose", "var");
@@ -2166,6 +2808,44 @@ OpFunctionEnd
     EXPECT_EQ(got, expect);
 }
 
+TEST_F(BuiltinBuilderTest, Call_Transpose_f16) {
+    Enable(ast::Extension::kF16);
+
+    auto* var = GlobalVar("var", ty.mat2x3<f16>(), ast::StorageClass::kPrivate);
+    auto* expr = Call("transpose", "var");
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateGlobalVariable(var)) << b.error();
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    auto got = DumpBuilder(b);
+    auto expect = R"(OpName %1 "var"
+OpName %9 "a_func"
+%5 = OpTypeFloat 16
+%4 = OpTypeVector %5 3
+%3 = OpTypeMatrix %4 2
+%2 = OpTypePointer Private %3
+%6 = OpConstantNull %3
+%1 = OpVariable %2 Private %6
+%8 = OpTypeVoid
+%7 = OpTypeFunction %8
+%13 = OpTypeVector %5 2
+%12 = OpTypeMatrix %13 3
+%9 = OpFunction %8 None %7
+%10 = OpLabel
+%14 = OpLoad %3 %1
+%11 = OpTranspose %12 %14
+OpReturn
+OpFunctionEnd
+)";
+    EXPECT_EQ(got, expect);
+}
+
 }  // namespace matrix_builtin_tests
 
 // Tests for Numeric builtins with float and integer vector parameter, i.e. "dot"
@@ -2200,6 +2880,37 @@ OpReturn
 )");
 }
 
+TEST_F(BuiltinBuilderTest, Call_Dot_F16) {
+    Enable(ast::Extension::kF16);
+
+    auto* var = GlobalVar("v", ty.vec3<f16>(), ast::StorageClass::kPrivate);
+    auto* expr = Call("dot", "v", "v");
+    auto* func = Func("a_func", {}, ty.void_(),
+                      {
+                          Assign(Phony(), expr),
+                      });
+
+    spirv::Builder& b = Build();
+
+    ASSERT_TRUE(b.GenerateGlobalVariable(var)) << b.error();
+    ASSERT_TRUE(b.GenerateFunction(func)) << b.error();
+
+    EXPECT_EQ(DumpInstructions(b.types()), R"(%4 = OpTypeFloat 16
+%3 = OpTypeVector %4 3
+%2 = OpTypePointer Private %3
+%5 = OpConstantNull %3
+%1 = OpVariable %2 Private %5
+%7 = OpTypeVoid
+%6 = OpTypeFunction %7
+)");
+    EXPECT_EQ(DumpInstructions(b.functions()[0].instructions()),
+              R"(%11 = OpLoad %3 %1
+%12 = OpLoad %3 %1
+%10 = OpDot %4 %11 %12
+OpReturn
+)");
+}
+
 TEST_F(BuiltinBuilderTest, Call_Dot_U32) {
     auto* var = GlobalVar("v", ty.vec3<u32>(), ast::StorageClass::kPrivate);
     auto* expr = Call("dot", "v", "v");

test/tint/builtins/degrees.spvasm.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/gen.wgsl.tmpl

@@ -13,16 +13,20 @@ See:
 {{- range Sem.Builtins -}}
 {{-   range .Overloads  -}}
 {{-     range Permute . -}}
-{{- /*    Generate a ./literal/<function>/<permuataion-hash>.wgsl file using
-          the Permutation macro defined below                             */ -}}
-{{-       $file := printf "./literal/%v/%v.wgsl" .Intrinsic.Name .Hash -}}
-{{-       $content := Eval "Permutation" "Overload" . "Mode" "literal" -}}
-{{-       WriteFile $file $content -}}
-{{- /*    Generate a ./var/<function>/<permuataion-hash>.wgsl file using
-          the Permutation macro defined below                             */ -}}
-{{-       $file := printf "./var/%v/%v.wgsl" .Intrinsic.Name .Hash -}}
-{{-       $content := Eval "Permutation" "Overload" . "Mode" "var" -}}
-{{-       WriteFile $file $content -}}
+{{- /*    TODO(crbug.com/tint/1502): Remove the bodge below and emit "enable f16;" after
+          getting ready for F16 end-to-end tests.                                        */ -}}
+{{-       if not (OverloadUsesF16 .Overload) -}}
+{{- /*      Generate a ./literal/<function>/<permuataion-hash>.wgsl file using
+            the Permutation macro defined below                             */ -}}
+{{-         $file := printf "./literal/%v/%v.wgsl" .Intrinsic.Name .Hash -}}
+{{-         $content := Eval "Permutation" "Overload" . "Mode" "literal" -}}
+{{-         WriteFile $file $content -}}
+{{- /*      Generate a ./var/<function>/<permuataion-hash>.wgsl file using
+            the Permutation macro defined below                             */ -}}
+{{-         $file := printf "./var/%v/%v.wgsl" .Intrinsic.Name .Hash -}}
+{{-         $content := Eval "Permutation" "Overload" . "Mode" "var" -}}
+{{-         WriteFile $file $content -}}
+{{-       end  }}
 {{-     end  }}
 {{-   end  }}
 {{- end  }}
@@ -319,4 +323,4 @@ enable chromium_experimental_dp4a;
 {{-    end -}}
 >
 {{-   end -}}
-{{- end -}}
+{{- end -}}

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl

@@ -1,47 +0,0 @@
-// Copyright 2021 The Tint Authors.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-////////////////////////////////////////////////////////////////////////////////
-// File generated by tools/src/cmd/gen
-// using the template:
-//   test/tint/builtins/gen/gen.wgsl.tmpl
-//
-// Do not modify this file directly
-////////////////////////////////////////////////////////////////////////////////
-
-struct SB_RO {
-  arg_0: array<f16>,
-};
-@group(0) @binding(1) var<storage, read> sb_ro : SB_RO;
-
-// fn arrayLength(ptr<storage, array<f16>, read>) -> u32
-fn arrayLength_8421b9() {
-  var res: u32 = arrayLength(&sb_ro.arg_0);
-}
-
-@vertex
-fn vertex_main() -> @builtin(position) vec4<f32> {
-  arrayLength_8421b9();
-  return vec4<f32>();
-}
-
-@fragment
-fn fragment_main() {
-  arrayLength_8421b9();
-}
-
-@compute @workgroup_size(1)
-fn compute_main() {
-  arrayLength_8421b9();
-}

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl.expected.dxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl.expected.fxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl.expected.glsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl.expected.msl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl.expected.spvasm

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/8421b9.wgsl.expected.wgsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl

@@ -1,47 +0,0 @@
-// Copyright 2021 The Tint Authors.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-////////////////////////////////////////////////////////////////////////////////
-// File generated by tools/src/cmd/gen
-// using the template:
-//   test/tint/builtins/gen/gen.wgsl.tmpl
-//
-// Do not modify this file directly
-////////////////////////////////////////////////////////////////////////////////
-
-struct SB_RW {
-  arg_0: array<f16>,
-};
-@group(0) @binding(0) var<storage, read_write> sb_rw : SB_RW;
-
-// fn arrayLength(ptr<storage, array<f16>, read_write>) -> u32
-fn arrayLength_cbd6b5() {
-  var res: u32 = arrayLength(&sb_rw.arg_0);
-}
-
-@vertex
-fn vertex_main() -> @builtin(position) vec4<f32> {
-  arrayLength_cbd6b5();
-  return vec4<f32>();
-}
-
-@fragment
-fn fragment_main() {
-  arrayLength_cbd6b5();
-}
-
-@compute @workgroup_size(1)
-fn compute_main() {
-  arrayLength_cbd6b5();
-}

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl.expected.dxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl.expected.fxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl.expected.glsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl.expected.msl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl.expected.spvasm

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/arrayLength/cbd6b5.wgsl.expected.wgsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/literal/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/literal/degrees/0d170c.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/literal/degrees/1ad5df.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec2 tint_degrees(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 vec2 tint_degrees(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 vec2 tint_degrees(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/literal/degrees/2af623.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/literal/degrees/51f705.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/literal/radians/09b7fc.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec4 tint_radians(vec4 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 vec4 tint_radians(vec4 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 vec4 tint_radians(vec4 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/literal/radians/61687a.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec2 tint_radians(vec2 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 vec2 tint_radians(vec2 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 vec2 tint_radians(vec2 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/literal/radians/6b0ff2.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/literal/radians/f96258.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -26,7 +26,7 @@ void main() {
 precision mediump float;
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -45,7 +45,7 @@ void main() {
 #version 310 es
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl

@@ -1,47 +0,0 @@
-// Copyright 2021 The Tint Authors.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-////////////////////////////////////////////////////////////////////////////////
-// File generated by tools/src/cmd/gen
-// using the template:
-//   test/tint/builtins/gen/gen.wgsl.tmpl
-//
-// Do not modify this file directly
-////////////////////////////////////////////////////////////////////////////////
-
-struct SB_RO {
-  arg_0: array<f16>,
-};
-@group(0) @binding(1) var<storage, read> sb_ro : SB_RO;
-
-// fn arrayLength(ptr<storage, array<f16>, read>) -> u32
-fn arrayLength_8421b9() {
-  var res: u32 = arrayLength(&sb_ro.arg_0);
-}
-
-@vertex
-fn vertex_main() -> @builtin(position) vec4<f32> {
-  arrayLength_8421b9();
-  return vec4<f32>();
-}
-
-@fragment
-fn fragment_main() {
-  arrayLength_8421b9();
-}
-
-@compute @workgroup_size(1)
-fn compute_main() {
-  arrayLength_8421b9();
-}

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl.expected.dxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl.expected.fxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl.expected.glsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl.expected.msl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl.expected.spvasm

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/8421b9.wgsl.expected.wgsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/8421b9.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl

@@ -1,47 +0,0 @@
-// Copyright 2021 The Tint Authors.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-////////////////////////////////////////////////////////////////////////////////
-// File generated by tools/src/cmd/gen
-// using the template:
-//   test/tint/builtins/gen/gen.wgsl.tmpl
-//
-// Do not modify this file directly
-////////////////////////////////////////////////////////////////////////////////
-
-struct SB_RW {
-  arg_0: array<f16>,
-};
-@group(0) @binding(0) var<storage, read_write> sb_rw : SB_RW;
-
-// fn arrayLength(ptr<storage, array<f16>, read_write>) -> u32
-fn arrayLength_cbd6b5() {
-  var res: u32 = arrayLength(&sb_rw.arg_0);
-}
-
-@vertex
-fn vertex_main() -> @builtin(position) vec4<f32> {
-  arrayLength_cbd6b5();
-  return vec4<f32>();
-}
-
-@fragment
-fn fragment_main() {
-  arrayLength_cbd6b5();
-}
-
-@compute @workgroup_size(1)
-fn compute_main() {
-  arrayLength_cbd6b5();
-}

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl.expected.dxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl.expected.fxc.hlsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl.expected.glsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl.expected.msl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl.expected.spvasm

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/arrayLength/cbd6b5.wgsl.expected.wgsl

@@ -1,6 +0,0 @@
-SKIP: FAILED
-
-builtins/gen/var/arrayLength/cbd6b5.wgsl:26:16 error: f16 used without 'f16' extension enabled
-  arg_0: array<f16>,
-               ^^^
-

test/tint/builtins/gen/var/degrees/0d170c.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/var/degrees/1ad5df.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec2 tint_degrees(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 vec2 tint_degrees(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 vec2 tint_degrees(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/var/degrees/2af623.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 vec3 tint_degrees(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/var/degrees/51f705.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 float tint_degrees(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

test/tint/builtins/gen/var/radians/09b7fc.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec4 tint_radians(vec4 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 vec4 tint_radians(vec4 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 vec4 tint_radians(vec4 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/var/radians/61687a.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec2 tint_radians(vec2 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 vec2 tint_radians(vec2 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 vec2 tint_radians(vec2 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/var/radians/6b0ff2.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/gen/var/radians/f96258.wgsl.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -27,7 +27,7 @@ void main() {
 precision mediump float;
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 
@@ -47,7 +47,7 @@ void main() {
 #version 310 es
 
 vec3 tint_radians(vec3 param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/radians.spvasm.expected.glsl

@@ -1,7 +1,7 @@
 #version 310 es
 
 float tint_radians(float param_0) {
-  return param_0 * 0.017453292519943295474;
+  return param_0 * 0.017453292519943295474f;
 }
 
 

test/tint/builtins/repeated_use.wgsl.expected.glsl

@@ -1,19 +1,19 @@
 #version 310 es
 
 vec4 tint_degrees(vec4 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 vec3 tint_degrees_1(vec3 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 vec2 tint_degrees_2(vec2 param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 float tint_degrees_3(float param_0) {
-  return param_0 * 57.295779513082322865;
+  return param_0 * 57.295779513082322865f;
 }
 
 

tools/src/tint/intrinsic/gen/gen.go

@@ -448,6 +448,8 @@ func SplitDisplayName(displayName string) []string {
 // If the type is not a composite type, then the fully qualified name is returned
 func ElementType(fqn sem.FullyQualifiedName) sem.FullyQualifiedName {
 	switch fqn.Target.GetName() {
+	case "vec2", "vec3", "vec4":
+		return fqn.TemplateArguments[0].(sem.FullyQualifiedName)
 	case "vec":
 		return fqn.TemplateArguments[1].(sem.FullyQualifiedName)
 	case "mat":
@@ -462,12 +464,16 @@ func ElementType(fqn sem.FullyQualifiedName) sem.FullyQualifiedName {
 // fully qualified name.
 func DeepestElementType(fqn sem.FullyQualifiedName) sem.FullyQualifiedName {
 	switch fqn.Target.GetName() {
+	case "vec2", "vec3", "vec4":
+		return fqn.TemplateArguments[0].(sem.FullyQualifiedName)
 	case "vec":
 		return fqn.TemplateArguments[1].(sem.FullyQualifiedName)
 	case "mat":
 		return DeepestElementType(fqn.TemplateArguments[2].(sem.FullyQualifiedName))
 	case "array":
 		return DeepestElementType(fqn.TemplateArguments[0].(sem.FullyQualifiedName))
+	case "ptr":
+		return DeepestElementType(fqn.TemplateArguments[1].(sem.FullyQualifiedName))
 	}
 	return fqn
 }
@@ -489,7 +495,7 @@ func IsDeclarable(fqn sem.FullyQualifiedName) bool {
 }
 
 // OverloadUsesF16 returns true if the overload uses the f16 type anywhere in the signature.
-func OverloadUsesF16(overload *sem.Overload, typename string) bool {
+func OverloadUsesF16(overload *sem.Overload) bool {
 	for _, param := range overload.Parameters {
 		if DeepestElementType(param.Type).Target.GetName() == "f16" {
 			return true