rename spirv subgroup funcs, minor config changes

Author: keptsecret

include/nbl/builtin/hlsl/spirv_intrinsics/subgroup_arithmetic.hlsl

@@ -55,28 +55,28 @@ T groupBitwiseXor(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T
 template<typename T>
 [[vk::ext_capability( spv::CapabilityGroupNonUniformArithmetic )]]
 [[vk::ext_instruction( spv::OpGroupNonUniformSMin )]]
-enable_if_t<!is_matrix_v<T> && is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupBitwiseMin(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
+enable_if_t<!is_matrix_v<T> && is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupSMin(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
 template<typename T>
 [[vk::ext_capability( spv::CapabilityGroupNonUniformArithmetic )]]
 [[vk::ext_instruction( spv::OpGroupNonUniformUMin )]]
-enable_if_t<!is_matrix_v<T> && !is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupBitwiseMin(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
+enable_if_t<!is_matrix_v<T> && !is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupUMin(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
 template<typename T>
 [[vk::ext_capability( spv::CapabilityGroupNonUniformArithmetic )]]
 [[vk::ext_instruction( spv::OpGroupNonUniformFMin )]]
-enable_if_t<!is_matrix_v<T> && is_floating_point_v<typename vector_traits<T>::scalar_type>, T> groupBitwiseMin(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
+enable_if_t<!is_matrix_v<T> && is_floating_point_v<typename vector_traits<T>::scalar_type>, T> groupFMin(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
 
 template<typename T>
 [[vk::ext_capability( spv::CapabilityGroupNonUniformArithmetic )]]
 [[vk::ext_instruction( spv::OpGroupNonUniformSMax )]]
-enable_if_t<!is_matrix_v<T> && is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupBitwiseMax(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
+enable_if_t<!is_matrix_v<T> && is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupSMax(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
 template<typename T>
 [[vk::ext_capability( spv::CapabilityGroupNonUniformArithmetic )]]
 [[vk::ext_instruction( spv::OpGroupNonUniformUMax )]]
-enable_if_t<!is_matrix_v<T> && !is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupBitwiseMax(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
+enable_if_t<!is_matrix_v<T> && !is_signed_v<T> && is_integral_v<typename vector_traits<T>::scalar_type>, T> groupUMax(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
 template<typename T>
 [[vk::ext_capability( spv::CapabilityGroupNonUniformArithmetic )]]
 [[vk::ext_instruction( spv::OpGroupNonUniformFMax )]]
-enable_if_t<!is_matrix_v<T> && is_floating_point_v<typename vector_traits<T>::scalar_type>, T> groupBitwiseMax(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
+enable_if_t<!is_matrix_v<T> && is_floating_point_v<typename vector_traits<T>::scalar_type>, T> groupFMax(uint32_t groupScope, [[vk::ext_literal]] uint32_t operation, T value);
 
 }
 }

include/nbl/builtin/hlsl/subgroup2/arithmetic_portability.hlsl

@@ -19,16 +19,18 @@ namespace hlsl
 namespace subgroup2
 {
 
-template<typename Config, class BinOp, int32_t _ItemsPerInvocation=1, class device_capabilities=void NBL_PRIMARY_REQUIRES(is_configuration_v<Config>)
+template<typename Config, class BinOp, int32_t _ItemsPerInvocation=1, class device_capabilities=void NBL_PRIMARY_REQUIRES(is_configuration_v<Config> && is_scalar_v<typename BinOp::type_t>)
 struct ArithmeticParams
 {
     using config_t = Config;
     using binop_t = BinOp;
-    using scalar_t = typename BinOp::type_t;    // BinOp should be with scalar type
+    using scalar_t = typename BinOp::type_t;
     using type_t = vector<scalar_t, _ItemsPerInvocation>;
+    using device_traits = device_capabilities_traits<device_capabilities>;
 
     NBL_CONSTEXPR_STATIC_INLINE int32_t ItemsPerInvocation = _ItemsPerInvocation;
     NBL_CONSTEXPR_STATIC_INLINE bool UseNativeIntrinsics = device_capabilities_traits<device_capabilities>::shaderSubgroupArithmetic /*&& /*some heuristic for when its faster*/;
+    // TODO add a IHV enum to device_capabilities_traits to check !is_nvidia
 };
 
 template<typename Params>