AI/rocm-systems
2
0
انشعاب 0
کد مسائل تقاضاهای واکشی Actions Packages پروژه‌ها انتشارها دانشنامه فعالیت

SWDEV-426270 - Remove duplicated operators

فهرست منبع 
Remove duplicated operators of hipComplexFloat and
hipComplexDouble.
If users need complex number multiplication and division,
they should call
hipCmulf() and hipCdivf() for hipComplexFloat,
hipCmul() and hipCdiv() for hipComplexComplex

SWDEV-428198 - Add missing operators

Add missing operators of vectors in host

Change-Id: Ie58d1642d579e7119997db49a9fd6a6641b666fd


[ROCm/clr commit: d4799b2a3f]
This commit is contained in:
taosang2
2023-10-17 13:39:07 -04:00
کامیت شده توسط Tao Sang
والد 16c18d5498
کامیت 6ea1a64c56
3فایلهای تغییر یافته به همراه55 افزوده شده و 148 حذف شده
دانلود پرونده وصله دانلود فایل تغییرات diff Expand all files Collapse all files
projects/clr/CMakeLists.txt
+1 -1
مشاهده پرونده
@@ -18,7 +18,7 @@
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
# THE SOFTWARE.
cmake_minimum_required(VERSION 3.16.8)
cmake_minimum_required(VERSION 3.16.3)
project(clr)
##########
projects/clr/hipamd/include/hip/amd_detail/amd_hip_complex.h
+2 -144
مشاهده پرونده
@@ -20,6 +20,8 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
*/
/* The header defines complex numbers and related functions*/
#ifndef HIP_INCLUDE_HIP_AMD_DETAIL_HIP_COMPLEX_H
#define HIP_INCLUDE_HIP_AMD_DETAIL_HIP_COMPLEX_H
@@ -53,100 +55,6 @@ THE SOFTWARE.
#endif
#endif // !defined(__HIPCC_RTC__)
#if __cplusplus
#define COMPLEX_NEG_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type operator-(const type& op) { \
type ret; \
ret.x = -op.x; \
ret.y = -op.y; \
return ret; \
}
#define COMPLEX_EQ_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline bool operator==(const type& lhs, const type& rhs) { \
return lhs.x == rhs.x && lhs.y == rhs.y; \
}
#define COMPLEX_NE_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline bool operator!=(const type& lhs, const type& rhs) { \
return !(lhs == rhs); \
}
#define COMPLEX_ADD_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type operator+(const type& lhs, const type& rhs) { \
type ret; \
ret.x = lhs.x + rhs.x; \
ret.y = lhs.y + rhs.y; \
return ret; \
}
#define COMPLEX_SUB_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type operator-(const type& lhs, const type& rhs) { \
type ret; \
ret.x = lhs.x - rhs.x; \
ret.y = lhs.y - rhs.y; \
return ret; \
}
#define COMPLEX_MUL_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type operator*(const type& lhs, const type& rhs) { \
type ret; \
ret.x = lhs.x * rhs.x - lhs.y * rhs.y; \
ret.y = lhs.x * rhs.y + lhs.y * rhs.x; \
return ret; \
}
#define COMPLEX_DIV_OP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type operator/(const type& lhs, const type& rhs) { \
type ret; \
ret.x = (lhs.x * rhs.x + lhs.y * rhs.y); \
ret.y = (rhs.x * lhs.y - lhs.x * rhs.y); \
ret.x = ret.x / (rhs.x * rhs.x + rhs.y * rhs.y); \
ret.y = ret.y / (rhs.x * rhs.x + rhs.y * rhs.y); \
return ret; \
}
#define COMPLEX_ADD_PREOP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type& operator+=(type& lhs, const type& rhs) { \
lhs.x += rhs.x; \
lhs.y += rhs.y; \
return lhs; \
}
#define COMPLEX_SUB_PREOP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type& operator-=(type& lhs, const type& rhs) { \
lhs.x -= rhs.x; \
lhs.y -= rhs.y; \
return lhs; \
}
#define COMPLEX_MUL_PREOP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type& operator*=(type& lhs, const type& rhs) { \
type temp{lhs}; \
lhs.x = rhs.x * temp.x - rhs.y * temp.y; \
lhs.y = rhs.y * temp.x + rhs.x * temp.y; \
return lhs; \
}
#define COMPLEX_DIV_PREOP_OVERLOAD(type) \
__HOST_DEVICE__ static inline type& operator/=(type& lhs, const type& rhs) { \
type temp; \
temp.x = (lhs.x*rhs.x + lhs.y * rhs.y) / (rhs.x*rhs.x + rhs.y*rhs.y); \
temp.y = (lhs.y * rhs.x - lhs.x * rhs.y) / (rhs.x*rhs.x + rhs.y*rhs.y); \
lhs = temp; \
return lhs; \
}
#define COMPLEX_SCALAR_PRODUCT(type, type1) \
__HOST_DEVICE__ static inline type operator*(const type& lhs, type1 rhs) { \
type ret; \
ret.x = lhs.x * rhs; \
ret.y = lhs.y * rhs; \
return ret; \
}
#endif
typedef float2 hipFloatComplex;
__HOST_DEVICE__ static inline float hipCrealf(hipFloatComplex z) { return z.x; }
@@ -240,56 +148,6 @@ __HOST_DEVICE__ static inline hipDoubleComplex hipCdiv(hipDoubleComplex p, hipDo
__HOST_DEVICE__ static inline double hipCabs(hipDoubleComplex z) { return sqrt(hipCsqabs(z)); }
#if __cplusplus
COMPLEX_NEG_OP_OVERLOAD(hipFloatComplex)
COMPLEX_EQ_OP_OVERLOAD(hipFloatComplex)
COMPLEX_NE_OP_OVERLOAD(hipFloatComplex)
COMPLEX_ADD_OP_OVERLOAD(hipFloatComplex)
COMPLEX_SUB_OP_OVERLOAD(hipFloatComplex)
COMPLEX_MUL_OP_OVERLOAD(hipFloatComplex)
COMPLEX_DIV_OP_OVERLOAD(hipFloatComplex)
COMPLEX_ADD_PREOP_OVERLOAD(hipFloatComplex)
COMPLEX_SUB_PREOP_OVERLOAD(hipFloatComplex)
COMPLEX_MUL_PREOP_OVERLOAD(hipFloatComplex)
COMPLEX_DIV_PREOP_OVERLOAD(hipFloatComplex)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, unsigned short)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, signed short)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, unsigned int)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, signed int)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, float)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, unsigned long)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, signed long)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, double)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, signed long long)
COMPLEX_SCALAR_PRODUCT(hipFloatComplex, unsigned long long)
COMPLEX_NEG_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_EQ_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_NE_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_ADD_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_SUB_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_MUL_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_DIV_OP_OVERLOAD(hipDoubleComplex)
COMPLEX_ADD_PREOP_OVERLOAD(hipDoubleComplex)
COMPLEX_SUB_PREOP_OVERLOAD(hipDoubleComplex)
COMPLEX_MUL_PREOP_OVERLOAD(hipDoubleComplex)
COMPLEX_DIV_PREOP_OVERLOAD(hipDoubleComplex)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, unsigned short)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, signed short)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, unsigned int)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, signed int)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, float)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, unsigned long)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, signed long)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, double)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, signed long long)
COMPLEX_SCALAR_PRODUCT(hipDoubleComplex, unsigned long long)
#endif
typedef hipFloatComplex hipComplex;
__HOST_DEVICE__ static inline hipComplex make_hipComplex(float x, float y) {
projects/clr/hipamd/include/hip/amd_detail/amd_hip_vector_types.h
+52 -3
مشاهده پرونده
@@ -38,6 +38,7 @@ THE SOFTWARE.
#if defined(__has_attribute)
#if __has_attribute(ext_vector_type)
#define __HIP_USE_NATIVE_VECTOR__ 1
#define __NATIVE_VECTOR__(n, T) T __attribute__((ext_vector_type(n)))
#else
#define __NATIVE_VECTOR__(n, T) T[n]
@@ -427,7 +428,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator+=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data += x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] += x.data[i];
#endif
return *this;
}
template<
@@ -443,7 +448,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator-=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data -= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] -= x.data[i];
#endif
return *this;
}
template<
@@ -459,7 +468,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator*=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data *= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] *= x.data[i];
#endif
return *this;
}
@@ -488,7 +501,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator/=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data /= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] /= x.data[i];
#endif
return *this;
}
template<
@@ -508,7 +525,11 @@ THE SOFTWARE.
HIP_vector_type operator-() const noexcept
{
auto tmp(*this);
#if __HIP_USE_NATIVE_VECTOR__
tmp.data = -tmp.data;
#else
for (auto i = 0u; i != rank; ++i) tmp.data[i] = -tmp.data[i];
#endif
return tmp;
}
@@ -519,7 +540,11 @@ THE SOFTWARE.
HIP_vector_type operator~() const noexcept
{
HIP_vector_type r{*this};
#if __HIP_USE_NATIVE_VECTOR__
r.data = ~r.data;
#else
for (auto i = 0u; i != rank; ++i) r.data[i] = ~r.data[i];
#endif
return r;
}
@@ -529,7 +554,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator%=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data %= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] %= x.data[i];
#endif
return *this;
}
@@ -539,7 +568,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator^=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data ^= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] ^= x.data[i];
#endif
return *this;
}
@@ -549,7 +582,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator|=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data |= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] |= x.data[i];
#endif
return *this;
}
@@ -559,7 +596,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator&=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data &= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] &= x.data[i];
#endif
return *this;
}
@@ -569,7 +610,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator>>=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data >>= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] >>= x.data[i];
#endif
return *this;
}
@@ -579,7 +624,11 @@ THE SOFTWARE.
__HOST_DEVICE__
HIP_vector_type& operator<<=(const HIP_vector_type& x) noexcept
{
#if __HIP_USE_NATIVE_VECTOR__
data <<= x.data;
#else
for (auto i = 0u; i != rank; ++i) data[i] <<= x.data[i];
#endif
return *this;
}
};
@@ -682,10 +731,10 @@ THE SOFTWARE.
__HOST_DEVICE__
inline
constexpr
bool _hip_any_zero(const V& x, int n) noexcept
bool _hip_compare(const V& x, const V& y, int n) noexcept
{
return
(n == -1) ? true : ((x[n] == 0) ? false : _hip_any_zero(x, n - 1));
(n == -1) ? true : ((x[n] != y[n]) ? false : _hip_compare(x, y, n - 1));
}
template<typename T, unsigned int n>
@@ -695,7 +744,7 @@ THE SOFTWARE.
bool operator==(
const HIP_vector_type<T, n>& x, const HIP_vector_type<T, n>& y) noexcept
{
return _hip_any_zero(x.data == y.data, n - 1);
return _hip_compare(x.data, y.data, n - 1);
}
template<typename T, unsigned int n, typename U>
__HOST_DEVICE__