Fixed mod function specialization #281 Fixed bitscan detection

glm/detail/func_common.inl

@@ -162,6 +162,16 @@ namespace detail
 			return (x >> Shift) | y;
 		}
 	};
+
+	template <typename T, precision P, template <class, precision> class vecType, typename genType, bool isFloat = true>
+	struct compute_mod
+	{
+		GLM_FUNC_QUALIFIER static vecType<T, P> call(vecType<T, P> const & a, genType const & b)
+		{
+			GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'mod' only accept floating-point inputs. Include <glm/gtc/integer.hpp> for integer inputs.");
+			return a - b * floor(a / b);
+		}
+	};
 }//namespace detail
 
 	// abs
@@ -334,15 +344,13 @@ namespace detail
 	template <typename T, precision P, template <typename, precision> class vecType>
 	GLM_FUNC_QUALIFIER vecType<T, P> mod(vecType<T, P> const & x, T y)
 	{
-		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'mod' only accept floating-point inputs");
-		return x - y * floor(x / y);
+		return detail::compute_mod<T, P, vecType, T, std::numeric_limits<T>::is_iec559>::call(x, y);
 	}
 
 	template <typename T, precision P, template <typename, precision> class vecType>
 	GLM_FUNC_QUALIFIER vecType<T, P> mod(vecType<T, P> const & x, vecType<T, P> const & y)
 	{
-		GLM_STATIC_ASSERT(std::numeric_limits<T>::is_iec559, "'mod' only accept floating-point inputs");
-		return x - y * floor(x / y);
+		return detail::compute_mod<T, P, vecType, vecType<T, P>, std::numeric_limits<T>::is_iec559>::call(x, y);
 	}
 
 	// modf

glm/detail/func_integer.inl

@@ -104,7 +104,7 @@ namespace detail
 		}
 	};
 
-#	if(GLM_ARCH != GLM_ARCH_PURE) && ((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & (GLM_COMPILER_LLVM | GLM_COMPILER_INTEL)) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+#	if GLM_HAS_BITSCAN_WINDOWS
 		template <typename genIUType>
 		struct compute_findLSB<genIUType, 32>
 		{
@@ -126,7 +126,7 @@ namespace detail
 				return IsNotNull ? int(Result) : -1;
 			}
 		};
-#	endif//GLM_ARCH != GLM_ARCH_PURE
+#	endif//GLM_HAS_BITSCAN_WINDOWS
 
 	template <typename T, glm::precision P, template <class, glm::precision> class vecType, bool EXEC = true>
 	struct compute_findMSB_step_vec
@@ -162,7 +162,7 @@ namespace detail
 		}
 	};
 
-#	if(GLM_ARCH != GLM_ARCH_PURE) && ((GLM_COMPILER & GLM_COMPILER_VC) || ((GLM_COMPILER & (GLM_COMPILER_LLVM | GLM_COMPILER_INTEL)) && (GLM_PLATFORM & GLM_PLATFORM_WINDOWS)))
+#	if GLM_HAS_BITSCAN_WINDOWS
 		template <typename genIUType>
 		GLM_FUNC_QUALIFIER int compute_findMSB_32(genIUType Value)
 		{
@@ -196,7 +196,7 @@ namespace detail
 				return detail::functor1<int, T, P, vecType>::call(compute_findMSB_64, x);
 			}
 		};
-#	endif//GLM_ARCH != GLM_ARCH_PURE
+#	endif//GLM_HAS_BITSCAN_WINDOWS
 }//namespace detail
 
 	// uaddCarry

glm/detail/setup.hpp

@@ -525,14 +525,19 @@
 	(GLM_LANG & GLM_LANG_CXX11_FLAG) || \
 	((GLM_LANG & GLM_LANG_CXX0X_FLAG) && (GLM_COMPILER & GLM_COMPILER_GCC) && (GLM_COMPILER >= GLM_COMPILER_GCC49)))
 
-#define GLM_HAS_TRIVIAL_QUERIES (\
+#define GLM_HAS_TRIVIAL_QUERIES ( \
 	((GLM_LANG & GLM_LANG_CXX11_FLAG) && !(GLM_COMPILER & GLM_COMPILER_GCC)) || \
 	((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC2013)))
 
-#define GLM_HAS_MAKE_SIGNED (\
+#define GLM_HAS_MAKE_SIGNED ( \
 	(GLM_LANG & GLM_LANG_CXX11_FLAG) || \
 	((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC2013)))
 
+#define GLM_HAS_BITSCAN_WINDOWS ( \
+	(GLM_ARCH != GLM_ARCH_PURE) && \
+	(GLM_PLATFORM & GLM_PLATFORM_WINDOWS) && \
+	(GLM_COMPILER & (GLM_COMPILER_VC | GLM_COMPILER_LLVM | GLM_COMPILER_INTEL)))
+
 // OpenMP
 #ifdef _OPENMP 
 #	if GLM_COMPILER & GLM_COMPILER_GCC

glm/gtc/integer.hpp

@@ -45,6 +45,7 @@
 // Dependencies
 #include "../detail/setup.hpp"
 #include "../detail/precision.hpp"
+#include "../detail/func_common.hpp"
 #include "../detail/func_integer.hpp"
 #include "../detail/func_exponential.hpp"
 #include <limits>
@@ -58,11 +59,46 @@ namespace glm
 	/// @addtogroup gtc_integer
 	/// @{
 
-	/// Returns the log2 of x. Can be reliably using to compute mipmap count from the texture size.
-	/// From GLM_GTC_integer extension.
+	/// Returns the log2 of x for integer values. Can be reliably using to compute mipmap count from the texture size.
+	/// @see gtc_integer
 	template <typename genIUType>
 	GLM_FUNC_DECL genIUType log2(genIUType x);
 
+	/// Modulus. Returns x % y
+	/// for each component in x using the floating point value y.
+	///
+	/// @tparam genIUType Integer-point scalar or vector types.
+	///
+	/// @see gtc_integer
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/mod.xml">GLSL mod man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template <typename genIUType>
+	GLM_FUNC_DECL genIUType mod(genIUType x, genIUType y);
+
+	/// Modulus. Returns x % y
+	/// for each component in x using the floating point value y.
+	///
+	/// @tparam T Integer scalar types.
+	/// @tparam vecType vector types.
+	///
+	/// @see gtc_integer
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/mod.xml">GLSL mod man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template <typename T, precision P, template <typename, precision> class vecType>
+	GLM_FUNC_DECL vecType<T, P> mod(vecType<T, P> const & x, T y);
+
+	/// Modulus. Returns x % y
+	/// for each component in x using the floating point value y.
+	///
+	/// @tparam T Integer scalar types.
+	/// @tparam vecType vector types.
+	///
+	/// @see gtc_integer
+	/// @see <a href="http://www.opengl.org/sdk/docs/manglsl/xhtml/mod.xml">GLSL mod man page</a>
+	/// @see <a href="http://www.opengl.org/registry/doc/GLSLangSpec.4.20.8.pdf">GLSL 4.20.8 specification, section 8.3 Common Functions</a>
+	template <typename T, precision P, template <typename, precision> class vecType>
+	GLM_FUNC_DECL vecType<T, P> mod(vecType<T, P> const & x, vecType<T, P> const & y);
+
 	/// @}
 } //namespace glm
 

glm/gtc/integer.inl

@@ -44,24 +44,31 @@ namespace detail
 		}
 	};
 
-#	if(GLM_ARCH != GLM_ARCH_PURE) && (GLM_COMPILER & (GLM_COMPILER_VC | GLM_COMPILER_APPLE_CLANG | GLM_COMPILER_LLVM))
-
-	template <precision P>
-	struct compute_log2<int, P, tvec4, false>
-	{
-		GLM_FUNC_QUALIFIER static tvec4<int, P> call(tvec4<int, P> const & vec)
+#	if GLM_HAS_BITSCAN_WINDOWS
+		template <precision P>
+		struct compute_log2<int, P, tvec4, false>
 		{
-			tvec4<int, P> Result(glm::uninitialize);
+			GLM_FUNC_QUALIFIER static tvec4<int, P> call(tvec4<int, P> const & vec)
+			{
+				tvec4<int, P> Result(glm::uninitialize);
+
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.x), vec.x);
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.y), vec.y);
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.z), vec.z);
+				_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.w), vec.w);
 
-			_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.x), vec.x);
-			_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.y), vec.y);
-			_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.z), vec.z);
-			_BitScanReverse(reinterpret_cast<unsigned long*>(&Result.w), vec.w);
+				return Result;
+			}
+		};
+#	endif//GLM_HAS_BITSCAN_WINDOWS
 
-			return Result;
+	template <typename T, precision P, template <class, precision> class vecType, typename genType>
+	struct compute_mod<T, P, vecType, genType, false>
+	{
+		GLM_FUNC_QUALIFIER static vecType<T, P> call(vecType<T, P> const & a, genType const & b)
+		{
+			return a % b;
 		}
 	};
-
-#	endif//GLM_ARCH != GLM_ARCH_PURE
 }//namespace detail
 }//namespace glm

test/core/core_func_common.cpp

@@ -162,6 +162,40 @@ namespace modf_
 	}
 }//namespace modf
 
+namespace mod_
+{
+	int test()
+	{
+		int Error(0);
+
+		{
+			float A(3.0);
+			float B(2.0f);
+			float C = glm::mod(A, B);
+
+			Error += glm::abs(C - 1.0f) < 0.00001f ? 0 : 1;
+		}
+
+		{
+			glm::vec4 A(3.0);
+			float B(2.0f);
+			glm::vec4 C = glm::mod(A, B);
+
+			Error += glm::all(glm::epsilonEqual(C, glm::vec4(1.0f), 0.00001f)) ? 0 : 1;
+		}
+
+		{
+			glm::vec4 A(3.0);
+			glm::vec4 B(2.0f);
+			glm::vec4 C = glm::mod(A, B);
+
+			Error += glm::all(glm::epsilonEqual(C, glm::vec4(1.0f), 0.00001f)) ? 0 : 1;
+		}
+
+		return Error;
+	}
+}//namespace mod_
+
 namespace floatBitsToInt
 {
 	int test()
@@ -1109,6 +1143,7 @@ int main()
 
 	Error += sign::test();
 	Error += floor_::test();
+	Error += mod_::test();
 	Error += modf_::test();
 	Error += floatBitsToInt::test();
 	Error += floatBitsToUint::test();

test/core/core_func_integer.cpp

@@ -578,6 +578,7 @@ namespace findMSB
 		genType		Return;
 	};
 
+#	if GLM_HAS_BITSCAN_WINDOWS
 	template <typename genIUType>
 	GLM_FUNC_QUALIFIER int findMSB_intrinsic(genIUType Value)
 	{
@@ -590,6 +591,20 @@ namespace findMSB
 		_BitScanReverse(&Result, Value);
 		return int(Result);
 	}
+#	endif//GLM_HAS_BITSCAN_WINDOWS
+
+#	if GLM_ARCH & GLM_ARCH_AVX
+	template <typename genIUType>
+	GLM_FUNC_QUALIFIER int findMSB_avx(genIUType Value)
+	{
+		GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'findMSB' only accept integer values");
+
+		if(Value == 0)
+			return -1;
+
+		return int(_tzcnt_u32(Value));
+	}
+#	endif
 
 	template <typename genIUType>
 	GLM_FUNC_QUALIFIER int findMSB_095(genIUType Value)
@@ -698,7 +713,7 @@ namespace findMSB
 		};
 
 		int Error(0);
-		std::size_t const Count(1000000);
+		std::size_t const Count(10000000);
 
 		std::clock_t Timestamps0 = std::clock();
 
@@ -738,12 +753,14 @@ namespace findMSB
 
 		std::clock_t Timestamps4 = std::clock();
 
-		for(std::size_t k = 0; k < Count; ++k)
-		for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
-		{
-			int Result = findMSB_intrinsic(Data[i].Value);
-			Error += Data[i].Return == Result ? 0 : 1;
-		}
+#		if GLM_HAS_BITSCAN_WINDOWS
+			for(std::size_t k = 0; k < Count; ++k)
+			for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
+			{
+				int Result = findMSB_intrinsic(Data[i].Value);
+				Error += Data[i].Return == Result ? 0 : 1;
+			}
+#		endif//GLM_HAS_BITSCAN_WINDOWS
 
 		std::clock_t Timestamps5 = std::clock();
 
@@ -756,13 +773,31 @@ namespace findMSB
 
 		std::clock_t Timestamps6 = std::clock();
 
+#		if GLM_ARCH & GLM_ARCH_AVX
+			for(std::size_t k = 0; k < Count; ++k)
+			for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
+			{
+				int Result = findMSB_avx(Data[i].Value);
+				Error += Data[i].Return == Result ? 0 : 1;
+			}
+#		endif
+
+		std::clock_t Timestamps7 = std::clock();
+
 		std::printf("glm::findMSB: %d clocks\n", static_cast<unsigned int>(Timestamps1 - Timestamps0));
 		std::printf("findMSB - nlz1: %d clocks\n", static_cast<unsigned int>(Timestamps2 - Timestamps1));
 		std::printf("findMSB - nlz2: %d clocks\n", static_cast<unsigned int>(Timestamps3 - Timestamps2));
 		std::printf("findMSB - 0.9.5: %d clocks\n", static_cast<unsigned int>(Timestamps4 - Timestamps3));
-		std::printf("findMSB - intrinsics: %d clocks\n", static_cast<unsigned int>(Timestamps5 - Timestamps4));
+
+#		if GLM_HAS_BITSCAN_WINDOWS
+			std::printf("findMSB - intrinsics: %d clocks\n", static_cast<unsigned int>(Timestamps5 - Timestamps4));
+#		endif//GLM_HAS_BITSCAN_WINDOWS
 		std::printf("findMSB - pop: %d clocks\n", static_cast<unsigned int>(Timestamps6 - Timestamps5));
 
+#		if GLM_ARCH & GLM_ARCH_AVX
+			std::printf("findMSB - avx tzcnt: %d clocks\n", static_cast<unsigned int>(Timestamps7 - Timestamps6));
+#		endif
+
 		return Error;
 	}
 
@@ -888,6 +923,8 @@ namespace findMSB
 		for(std::size_t i = 0; i < sizeof(Data) / sizeof(type<int>); ++i)
 		{
 			int Result0 = findMSB_intrinsic(Data[i].Value);
+			//unsigned int A = _lzcnt_u32(Data[i].Value);
+			//unsigned int B = _tzcnt_u32(Data[i].Value);
 			Error += Data[i].Return == Result0 ? 0 : 1;
 		}
 
@@ -1527,6 +1564,8 @@ int main()
 	Error += ::bitfieldInsert::test();
 	Error += ::bitfieldExtract::test();
 
+	Error += ::findMSB::perf();
+
 #	ifdef GLM_TEST_ENABLE_PERF
 		Error += ::bitCount::perf();
 		Error += ::bitfieldReverse::perf();

test/gtc/gtc_integer.cpp

@@ -30,6 +30,7 @@
 ///////////////////////////////////////////////////////////////////////////////////
 
 #define GLM_FORCE_INLINE
+#include <glm/gtc/epsilon.hpp>
 #include <glm/gtc/integer.hpp>
 #include <glm/gtc/type_precision.hpp>
 #include <glm/gtc/vec1.hpp>
@@ -102,7 +103,7 @@ namespace log2_
 			printf("glm::log2<ivec4>: %d clocks\n", End - Begin);
 		}
 
-#		if(GLM_ARCH != GLM_ARCH_PURE) && (GLM_COMPILER & (GLM_COMPILER_VC | GLM_COMPILER_APPLE_CLANG | GLM_COMPILER_LLVM))
+#		if GLM_HAS_BITSCAN_WINDOWS
 		{
 			std::vector<glm::ivec4> Result;
 			Result.resize(Count);
@@ -163,7 +164,7 @@ namespace log2_
 
 			printf("glm::log2<ivec4> reinterpret: %d clocks\n", End - Begin);
 		}
-#		endif//GLM_ARCH != GLM_ARCH_PURE
+#		endif//GLM_HAS_BITSCAN_WINDOWS
 
 		{
 			std::vector<float> Result;
@@ -197,12 +198,70 @@ namespace log2_
 	}
 }//namespace log2_
 
+namespace mod_
+{
+	int test()
+	{
+		int Error(0);
+
+		{
+			float A(3.0);
+			float B(2.0f);
+			float C = glm::mod(A, B);
+
+			Error += glm::abs(C - 1.0f) < 0.00001f ? 0 : 1;
+		}
+
+		{
+			glm::vec4 A(3.0);
+			float B(2.0f);
+			glm::vec4 C = glm::mod(A, B);
+
+			Error += glm::all(glm::epsilonEqual(C, glm::vec4(1.0f), 0.00001f)) ? 0 : 1;
+		}
+
+		{
+			glm::vec4 A(3.0);
+			glm::vec4 B(2.0f);
+			glm::vec4 C = glm::mod(A, B);
+
+			Error += glm::all(glm::epsilonEqual(C, glm::vec4(1.0f), 0.00001f)) ? 0 : 1;
+		}
+
+		{
+			int A(3);
+			int B(2);
+			int C = glm::mod(A, B);
+
+			Error += C == 1 ? 0 : 1;
+		}
+
+		{
+			glm::ivec4 A(3);
+			int B(2);
+			glm::ivec4 C = glm::mod(A, B);
+
+			Error += glm::all(glm::equal(C, glm::ivec4(1))) ? 0 : 1;
+		}
+
+		{
+			glm::ivec4 A(3);
+			glm::ivec4 B(2);
+			glm::ivec4 C = glm::mod(A, B);
+
+			Error += glm::all(glm::equal(C, glm::ivec4(1))) ? 0 : 1;
+		}
+
+		return Error;
+	}
+}//namespace mod_
 
 int main()
 {
 	int Error(0);
 
 	Error += ::log2_::test();
+	Error += ::mod_::test();
 
 #	ifdef GLM_TEST_ENABLE_PERF
 		Error += ::log2_::perf();