[HLSL] Add reversebits library function

[bob80905]

Summary

This change exposes the reversebits library function for HLSL, excluding floating point types.
The reversebits function is supported for all scalar, vector, and matrix types.

The full documentation of the HLSL reversebits function is available here:
https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/reversebits

[bob80905]

• change builtin name to reversebits

[python3kgae]

clang/test/CodeGenHLSL/builtins/bitreverse.hlsl

32 ↗

(On Diff #546576)

Not NO_HALF, it is integer :).

And __HLSL_ENABLE_16_BIT only defined once.
Maybe just one RUN line is enough.

[bob80905]

• rename file, make merge 2 Runs into 1 with enable16bit defined

[python3kgae]

Don't need -fnative-half-type in the RUN line.

[bob80905]

• remove fnative half type from run line

clang/lib/Headers/hlsl/hlsl_intrinsics.h

 min(double, double);
 __attribute__((clang_builtin_alias(__builtin_elementwise_min)))
 double2 min(double2, double2);
 __attribute__((clang_builtin_alias(__builtin_elementwise_min)))
 double3 min(double3, double3);
 __attribute__((clang_builtin_alias(__builtin_elementwise_min)))
 double4 min(double4, double4);
 
+// reversebits builtins
+#ifdef __HLSL_ENABLE_16_BIT
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int16_t reversebits(int16_t);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int16_t2 reversebits(int16_t2);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int16_t3 reversebits(int16_t3);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int16_t4 reversebits(int16_t4);
+
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint16_t reversebits(uint16_t);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint16_t2 reversebits(uint16_t2);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint16_t3 reversebits(uint16_t3);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint16_t4 reversebits(uint16_t4);
+#endif
+
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse))) int
+reversebits(int);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int2 reversebits(int2);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int3 reversebits(int3);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int4 reversebits(int4);
+
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint reversebits(uint);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint2 reversebits(uint2);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint3 reversebits(uint3);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint4 reversebits(uint4);
+
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int64_t reversebits(int64_t);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int64_t2 reversebits(int64_t2);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int64_t3 reversebits(int64_t3);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+int64_t4 reversebits(int64_t4);
+
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint64_t reversebits(uint64_t);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint64_t2 reversebits(uint64_t2);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint64_t3 reversebits(uint64_t3);
+__attribute__((clang_builtin_alias(__builtin_elementwise_bitreverse)))
+uint64_t4 reversebits(uint64_t4);
 } // namespace hlsl
 #endif //_HLSL_HLSL_INTRINSICS_H_

clang/test/CodeGenHLSL/builtins/reversebits.hlsl

+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
+// RUN:   dxil-pc-shadermodel6.3-library %s -fnative-half-type \
+// RUN:   -D__HLSL_ENABLE_16_BIT -emit-llvm -disable-llvm-passes -O3 -o - | FileCheck %s
+
+#ifdef __HLSL_ENABLE_16_BIT
+// CHECK: define noundef i16 @
+// CHECK: call i16 @llvm.bitreverse.i16(
+int16_t test_bitreverse_short(int16_t p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <2 x i16> @
+// CHECK: call <2 x i16> @llvm.bitreverse.v2i16(
+int16_t2 test_bitreverse_short2(int16_t2 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <3 x i16> @
+// CHECK: call <3 x i16> @llvm.bitreverse.v3i16
+int16_t3 test_bitreverse_short3(int16_t3 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <4 x i16> @
+// CHECK: call <4 x i16> @llvm.bitreverse.v4i16
+int16_t4 test_bitreverse_short4(int16_t4 p0)
+{
+  return reversebits(p0);
+}
+
+// CHECK: define noundef i16 @
+// CHECK: call i16 @llvm.bitreverse.i16(
+uint16_t test_bitreverse_ushort(uint16_t p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <2 x i16> @
+// CHECK: call <2 x i16> @llvm.bitreverse.v2i16
+uint16_t2 test_bitreverse_ushort2(uint16_t2 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <3 x i16> @
+// CHECK: call <3 x i16> @llvm.bitreverse.v3i16
+uint16_t3 test_bitreverse_ushort3(uint16_t3 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <4 x i16> @
+// CHECK: call <4 x i16> @llvm.bitreverse.v4i16
+uint16_t4 test_bitreverse_ushort4(uint16_t4 p0)
+{
+  return reversebits(p0);
+}
+#endif
+
+// CHECK: define noundef i32 @
+// CHECK: call i32 @llvm.bitreverse.i32(
+int test_bitreverse_int(int p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <2 x i32> @
+// CHECK: call <2 x i32> @llvm.bitreverse.v2i32
+int2 test_bitreverse_int2(int2 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <3 x i32> @
+// CHECK: call <3 x i32> @llvm.bitreverse.v3i32
+int3 test_bitreverse_int3(int3 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <4 x i32> @
+// CHECK: call <4 x i32> @llvm.bitreverse.v4i32
+int4 test_bitreverse_int4(int4 p0)
+{
+  return reversebits(p0);
+}
+
+// CHECK: define noundef i32 @
+// CHECK: call i32 @llvm.bitreverse.i32(
+int test_bitreverse_uint(uint p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <2 x i32> @
+// CHECK: call <2 x i32> @llvm.bitreverse.v2i32
+uint2 test_bitreverse_uint2(uint2 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <3 x i32> @
+// CHECK: call <3 x i32> @llvm.bitreverse.v3i32
+uint3 test_bitreverse_uint3(uint3 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <4 x i32> @
+// CHECK: call <4 x i32> @llvm.bitreverse.v4i32
+uint4 test_bitreverse_uint4(uint4 p0)
+{
+  return reversebits(p0);
+}
+
+// CHECK: define noundef i64 @
+// CHECK: call i64 @llvm.bitreverse.i64(
+int64_t test_bitreverse_long(int64_t p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <2 x i64> @
+// CHECK: call <2 x i64> @llvm.bitreverse.v2i64
+int64_t2 test_bitreverse_long2(int64_t2 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <3 x i64> @
+// CHECK: call <3 x i64> @llvm.bitreverse.v3i64
+int64_t3 test_bitreverse_long3(int64_t3 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <4 x i64> @
+// CHECK: call <4 x i64> @llvm.bitreverse.v4i64
+int64_t4 test_bitreverse_long4(int64_t4 p0)
+{
+  return reversebits(p0);
+}
+
+// CHECK: define noundef i64 @
+// CHECK: call i64 @llvm.bitreverse.i64(
+uint64_t test_bitreverse_long(uint64_t p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <2 x i64> @
+// CHECK: call <2 x i64> @llvm.bitreverse.v2i64
+uint64_t2 test_bitreverse_long2(uint64_t2 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <3 x i64> @
+// CHECK: call <3 x i64> @llvm.bitreverse.v3i64
+uint64_t3 test_bitreverse_long3(uint64_t3 p0)
+{
+  return reversebits(p0);
+}
+// CHECK: define noundef <4 x i64> @
+// CHECK: call <4 x i64> @llvm.bitreverse.v4i64
+uint64_t4 test_bitreverse_long4(uint64_t4 p0)
+{
+  return reversebits(p0);
+}