triton-windows 3.3.1.post19__cp312-cp312-win_amd64.whl → 3.4.0.post20__cp312-cp312-win_amd64.whl

triton/backends/amd/include/hip/amd_detail/amd_hip_bfloat16.h

@@ -1,293 +0,0 @@
-/**
- * MIT License
- *
- * Copyright (c) 2019 - 2022 Advanced Micro Devices, Inc. All rights reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to deal
- * in the Software without restriction, including without limitation the rights
- * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
- * copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-
-/*!\file
- * \brief hip_bfloat16.h provides struct for hip_bfloat16 typedef
- */
-
-#ifndef _HIP_INCLUDE_HIP_AMD_DETAIL_HIP_BFLOAT16_H_
-#define _HIP_INCLUDE_HIP_AMD_DETAIL_HIP_BFLOAT16_H_
-
-#include "host_defines.h"
-#if defined(__HIPCC_RTC__)
-    #define __HOST_DEVICE__ __device__
-#else
-    #define __HOST_DEVICE__ __host__ __device__
-#endif
-
-#if __cplusplus < 201103L || !defined(__HIPCC__)
-
-// If this is a C compiler, C++ compiler below C++11, or a host-only compiler, we only
-// include a minimal definition of hip_bfloat16
-
-#include <stdint.h>
-/*! \brief Struct to represent a 16 bit brain floating point number. */
-typedef struct
-{
-    uint16_t data;
-} hip_bfloat16;
-
-#else // __cplusplus < 201103L || !defined(__HIPCC__)
-
-#include <hip/hip_runtime.h>
-
-#pragma clang diagnostic push
-#pragma clang diagnostic ignored "-Wshadow"
-struct hip_bfloat16
-{
-    __hip_uint16_t data;
-
-    enum truncate_t
-    {
-        truncate
-    };
-
-    __HOST_DEVICE__ hip_bfloat16() = default;
-
-    // round upper 16 bits of IEEE float to convert to bfloat16
-    explicit __HOST_DEVICE__ hip_bfloat16(float f)
-        : data(float_to_bfloat16(f))
-    {
-    }
-
-    explicit __HOST_DEVICE__ hip_bfloat16(float f, truncate_t)
-        : data(truncate_float_to_bfloat16(f))
-    {
-    }
-
-    // zero extend lower 16 bits of bfloat16 to convert to IEEE float
-    __HOST_DEVICE__ operator float() const
-    {
-        union
-        {
-            uint32_t int32;
-            float    fp32;
-        } u = {uint32_t(data) << 16};
-        return u.fp32;
-    }
-
-    __HOST_DEVICE__ hip_bfloat16 &operator=(const float& f)
-    {
-       data = float_to_bfloat16(f);
-       return *this;
-    }
-
-    static  __HOST_DEVICE__ hip_bfloat16 round_to_bfloat16(float f)
-    {
-        hip_bfloat16 output;
-        output.data = float_to_bfloat16(f);
-        return output;
-    }
-
-    static  __HOST_DEVICE__ hip_bfloat16 round_to_bfloat16(float f, truncate_t)
-    {
-        hip_bfloat16 output;
-        output.data = truncate_float_to_bfloat16(f);
-        return output;
-    }
-
-private:
-    static __HOST_DEVICE__ __hip_uint16_t float_to_bfloat16(float f)
-    {
-        union
-        {
-            float    fp32;
-            uint32_t int32;
-        } u = {f};
-        if(~u.int32 & 0x7f800000)
-        {
-            // When the exponent bits are not all 1s, then the value is zero, normal,
-            // or subnormal. We round the bfloat16 mantissa up by adding 0x7FFF, plus
-            // 1 if the least significant bit of the bfloat16 mantissa is 1 (odd).
-            // This causes the bfloat16's mantissa to be incremented by 1 if the 16
-            // least significant bits of the float mantissa are greater than 0x8000,
-            // or if they are equal to 0x8000 and the least significant bit of the
-            // bfloat16 mantissa is 1 (odd). This causes it to be rounded to even when
-            // the lower 16 bits are exactly 0x8000. If the bfloat16 mantissa already
-            // has the value 0x7f, then incrementing it causes it to become 0x00 and
-            // the exponent is incremented by one, which is the next higher FP value
-            // to the unrounded bfloat16 value. When the bfloat16 value is subnormal
-            // with an exponent of 0x00 and a mantissa of 0x7F, it may be rounded up
-            // to a normal value with an exponent of 0x01 and a mantissa of 0x00.
-            // When the bfloat16 value has an exponent of 0xFE and a mantissa of 0x7F,
-            // incrementing it causes it to become an exponent of 0xFF and a mantissa
-            // of 0x00, which is Inf, the next higher value to the unrounded value.
-            u.int32 += 0x7fff + ((u.int32 >> 16) & 1); // Round to nearest, round to even
-        }
-        else if(u.int32 & 0xffff)
-        {
-            // When all of the exponent bits are 1, the value is Inf or NaN.
-            // Inf is indicated by a zero mantissa. NaN is indicated by any nonzero
-            // mantissa bit. Quiet NaN is indicated by the most significant mantissa
-            // bit being 1. Signaling NaN is indicated by the most significant
-            // mantissa bit being 0 but some other bit(s) being 1. If any of the
-            // lower 16 bits of the mantissa are 1, we set the least significant bit
-            // of the bfloat16 mantissa, in order to preserve signaling NaN in case
-            // the bloat16's mantissa bits are all 0.
-            u.int32 |= 0x10000; // Preserve signaling NaN
-        }
-        return __hip_uint16_t(u.int32 >> 16);
-    }
-
-    // Truncate instead of rounding, preserving SNaN
-    static __HOST_DEVICE__ __hip_uint16_t truncate_float_to_bfloat16(float f)
-    {
-        union
-        {
-            float    fp32;
-            uint32_t int32;
-        } u = {f};
-        return __hip_uint16_t(u.int32 >> 16) | (!(~u.int32 & 0x7f800000) && (u.int32 & 0xffff));
-    }
-};
-#pragma clang diagnostic pop
-
-typedef struct
-{
-    __hip_uint16_t data;
-} hip_bfloat16_public;
-
-static_assert(__hip_internal::is_standard_layout<hip_bfloat16>{},
-              "hip_bfloat16 is not a standard layout type, and thus is "
-              "incompatible with C.");
-
-static_assert(__hip_internal::is_trivial<hip_bfloat16>{},
-              "hip_bfloat16 is not a trivial type, and thus is "
-              "incompatible with C.");
-#if !defined(__HIPCC_RTC__)
-static_assert(sizeof(hip_bfloat16) == sizeof(hip_bfloat16_public)
-                  && offsetof(hip_bfloat16, data) == offsetof(hip_bfloat16_public, data),
-              "internal hip_bfloat16 does not match public hip_bfloat16");
-
-inline std::ostream& operator<<(std::ostream& os, const hip_bfloat16& bf16)
-{
-  return os << float(bf16);
-}
-#endif
-
-inline __HOST_DEVICE__ hip_bfloat16 operator+(hip_bfloat16 a)
-{
-    return a;
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator-(hip_bfloat16 a)
-{
-    a.data ^= 0x8000;
-    return a;
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator+(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return hip_bfloat16(float(a) + float(b));
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator-(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return hip_bfloat16(float(a) - float(b));
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator*(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return hip_bfloat16(float(a) * float(b));
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator/(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return hip_bfloat16(float(a) / float(b));
-}
-inline __HOST_DEVICE__ bool operator<(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return float(a) < float(b);
-}
-inline __HOST_DEVICE__ bool operator==(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return float(a) == float(b);
-}
-inline __HOST_DEVICE__ bool operator>(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return b < a;
-}
-inline __HOST_DEVICE__ bool operator<=(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return !(a > b);
-}
-inline __HOST_DEVICE__ bool operator!=(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return !(a == b);
-}
-inline __HOST_DEVICE__ bool operator>=(hip_bfloat16 a, hip_bfloat16 b)
-{
-    return !(a < b);
-}
-inline __HOST_DEVICE__ hip_bfloat16& operator+=(hip_bfloat16& a, hip_bfloat16 b)
-{
-    return a = a + b;
-}
-inline __HOST_DEVICE__ hip_bfloat16& operator-=(hip_bfloat16& a, hip_bfloat16 b)
-{
-    return a = a - b;
-}
-inline __HOST_DEVICE__ hip_bfloat16& operator*=(hip_bfloat16& a, hip_bfloat16 b)
-{
-    return a = a * b;
-}
-inline __HOST_DEVICE__ hip_bfloat16& operator/=(hip_bfloat16& a, hip_bfloat16 b)
-{
-    return a = a / b;
-}
-inline __HOST_DEVICE__ hip_bfloat16& operator++(hip_bfloat16& a)
-{
-    return a += hip_bfloat16(1.0f);
-}
-inline __HOST_DEVICE__ hip_bfloat16& operator--(hip_bfloat16& a)
-{
-    return a -= hip_bfloat16(1.0f);
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator++(hip_bfloat16& a, int)
-{
-    hip_bfloat16 orig = a;
-    ++a;
-    return orig;
-}
-inline __HOST_DEVICE__ hip_bfloat16 operator--(hip_bfloat16& a, int)
-{
-    hip_bfloat16 orig = a;
-    --a;
-    return orig;
-}
-
-namespace std
-{
-    constexpr __HOST_DEVICE__ bool isinf(hip_bfloat16 a)
-    {
-        return !(~a.data & 0x7f80) && !(a.data & 0x7f);
-    }
-    constexpr __HOST_DEVICE__ bool isnan(hip_bfloat16 a)
-    {
-        return !(~a.data & 0x7f80) && +(a.data & 0x7f);
-    }
-    constexpr __HOST_DEVICE__ bool iszero(hip_bfloat16 a)
-    {
-        return !(a.data & 0x7fff);
-    }
-}
-
-#endif // __cplusplus < 201103L || !defined(__HIPCC__)
-
-#endif // _HIP_BFLOAT16_H_

triton/backends/amd/include/hip/amd_detail/amd_hip_common.h

@@ -1,32 +0,0 @@
-/*
-Copyright (c) 2019 - 2021 Advanced Micro Devices, Inc. All rights reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a copy of
-this software and associated documentation files (the "Software"), to deal in
-the Software without restriction, including without limitation the rights to
-use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
-of the Software, and to permit persons to whom the Software is furnished to do
-so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
-*/
-
-#ifndef HIP_INCLUDE_HIP_AMD_DETAIL_HIP_COMMON_H
-#define HIP_INCLUDE_HIP_AMD_DETAIL_HIP_COMMON_H
-
-#if defined(__clang__) && defined(__HIP__)
-#define __HIP_CLANG_ONLY__ 1
-#else
-#define __HIP_CLANG_ONLY__ 0
-#endif
-
-#endif // HIP_INCLUDE_HIP_AMD_DETAIL_HIP_COMMON_H

triton/backends/amd/include/hip/amd_detail/amd_hip_complex.h

@@ -1,174 +0,0 @@
-/*
-Copyright (c) 2015 - 2023 Advanced Micro Devices, Inc. All rights reserved.
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in
-all copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-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
-
-#if !defined(__HIPCC_RTC__)
-#include "hip/amd_detail/amd_hip_vector_types.h"
-#endif
-
-#if defined(__HIPCC_RTC__)
-#define __HOST_DEVICE__ __device__
-#else
-#define __HOST_DEVICE__ __host__ __device__
-// TODO: Clang has a bug which allows device functions to call std functions
-// when std functions are introduced into default namespace by using statement.
-// math.h may be included after this bug is fixed.
-#if __cplusplus
-#include <cmath>
-#else
-#include "math.h"
-#endif
-#endif // !defined(__HIPCC_RTC__)
-
-typedef float2 hipFloatComplex;
-
-__HOST_DEVICE__ static inline float hipCrealf(hipFloatComplex z) { return z.x; }
-
-__HOST_DEVICE__ static inline float hipCimagf(hipFloatComplex z) { return z.y; }
-
-__HOST_DEVICE__ static inline hipFloatComplex make_hipFloatComplex(float a, float b) {
-    hipFloatComplex z;
-    z.x = a;
-    z.y = b;
-    return z;
-}
-
-__HOST_DEVICE__ static inline hipFloatComplex hipConjf(hipFloatComplex z) {
-    hipFloatComplex ret;
-    ret.x = z.x;
-    ret.y = -z.y;
-    return ret;
-}
-
-__HOST_DEVICE__ static inline float hipCsqabsf(hipFloatComplex z) {
-    return z.x * z.x + z.y * z.y;
-}
-
-__HOST_DEVICE__ static inline hipFloatComplex hipCaddf(hipFloatComplex p, hipFloatComplex q) {
-    return make_hipFloatComplex(p.x + q.x, p.y + q.y);
-}
-
-__HOST_DEVICE__ static inline hipFloatComplex hipCsubf(hipFloatComplex p, hipFloatComplex q) {
-    return make_hipFloatComplex(p.x - q.x, p.y - q.y);
-}
-
-__HOST_DEVICE__ static inline hipFloatComplex hipCmulf(hipFloatComplex p, hipFloatComplex q) {
-    return make_hipFloatComplex(p.x * q.x - p.y * q.y, p.y * q.x + p.x * q.y);
-}
-
-__HOST_DEVICE__ static inline hipFloatComplex hipCdivf(hipFloatComplex p, hipFloatComplex q) {
-    float sqabs = hipCsqabsf(q);
-    hipFloatComplex ret;
-    ret.x = (p.x * q.x + p.y * q.y) / sqabs;
-    ret.y = (p.y * q.x - p.x * q.y) / sqabs;
-    return ret;
-}
-
-__HOST_DEVICE__ static inline float hipCabsf(hipFloatComplex z) { return sqrtf(hipCsqabsf(z)); }
-
-
-typedef double2 hipDoubleComplex;
-
-__HOST_DEVICE__ static inline double hipCreal(hipDoubleComplex z) { return z.x; }
-
-__HOST_DEVICE__ static inline double hipCimag(hipDoubleComplex z) { return z.y; }
-
-__HOST_DEVICE__ static inline hipDoubleComplex make_hipDoubleComplex(double a, double b) {
-    hipDoubleComplex z;
-    z.x = a;
-    z.y = b;
-    return z;
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipConj(hipDoubleComplex z) {
-    hipDoubleComplex ret;
-    ret.x = z.x;
-    ret.y = -z.y;
-    return ret;
-}
-
-__HOST_DEVICE__ static inline double hipCsqabs(hipDoubleComplex z) {
-    return z.x * z.x + z.y * z.y;
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipCadd(hipDoubleComplex p, hipDoubleComplex q) {
-    return make_hipDoubleComplex(p.x + q.x, p.y + q.y);
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipCsub(hipDoubleComplex p, hipDoubleComplex q) {
-    return make_hipDoubleComplex(p.x - q.x, p.y - q.y);
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipCmul(hipDoubleComplex p, hipDoubleComplex q) {
-    return make_hipDoubleComplex(p.x * q.x - p.y * q.y, p.y * q.x + p.x * q.y);
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipCdiv(hipDoubleComplex p, hipDoubleComplex q) {
-    double sqabs = hipCsqabs(q);
-    hipDoubleComplex ret;
-    ret.x = (p.x * q.x + p.y * q.y) / sqabs;
-    ret.y = (p.y * q.x - p.x * q.y) / sqabs;
-    return ret;
-}
-
-__HOST_DEVICE__ static inline double hipCabs(hipDoubleComplex z) { return sqrt(hipCsqabs(z)); }
-
-typedef hipFloatComplex hipComplex;
-
-__HOST_DEVICE__ static inline hipComplex make_hipComplex(float x, float y) {
-    return make_hipFloatComplex(x, y);
-}
-
-__HOST_DEVICE__ static inline hipFloatComplex hipComplexDoubleToFloat(hipDoubleComplex z) {
-    return make_hipFloatComplex((float)z.x, (float)z.y);
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipComplexFloatToDouble(hipFloatComplex z) {
-    return make_hipDoubleComplex((double)z.x, (double)z.y);
-}
-
-__HOST_DEVICE__ static inline hipComplex hipCfmaf(hipComplex p, hipComplex q, hipComplex r) {
-    float real = (p.x * q.x) + r.x;
-    float imag = (q.x * p.y) + r.y;
-
-    real = -(p.y * q.y) + real;
-    imag = (p.x * q.y) + imag;
-
-    return make_hipComplex(real, imag);
-}
-
-__HOST_DEVICE__ static inline hipDoubleComplex hipCfma(hipDoubleComplex p, hipDoubleComplex q,
-                                                           hipDoubleComplex r) {
-    double real = (p.x * q.x) + r.x;
-    double imag = (q.x * p.y) + r.y;
-
-    real = -(p.y * q.y) + real;
-    imag = (p.x * q.y) + imag;
-
-    return make_hipDoubleComplex(real, imag);
-}
-
-#endif //HIP_INCLUDE_HIP_AMD_DETAIL_HIP_COMPLEX_H