Skip to content

Added Paddle Patch File #8294

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
PrernaKumbhar20 wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Added Paddle Patch File #8294

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
120 changes: 120 additions & 0 deletions p/paddle/paddle_v3.0.0.patch
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,120 @@
diff --git a/paddle/phi/common/cpstring_impl.h b/paddle/phi/common/cpstring_impl.h
index 33780d1538..4b6cfd7889 100644
--- a/paddle/phi/common/cpstring_impl.h
+++ b/paddle/phi/common/cpstring_impl.h
@@ -203,7 +203,6 @@ HOSTDEVICE static inline void *PD_Memcpy(void *dst,
}
return dst;
}
-
HOSTDEVICE static inline void *PD_Malloc(size_t size) { return malloc(size); }

HOSTDEVICE static inline void *PD_Realloc(void *ptr,
diff --git a/paddle/phi/core/platform/denormal.cc b/paddle/phi/core/platform/denormal.cc
index 93c55e84be..f9c34c869a 100644
--- a/paddle/phi/core/platform/denormal.cc
+++ b/paddle/phi/core/platform/denormal.cc
@@ -30,7 +30,8 @@

#if !defined(GCC_WITHOUT_INTRINSICS) && !defined(PADDLE_WITH_ARM) && \
!defined(PADDLE_WITH_SW) && !defined(PADDLE_WITH_MIPS) && \
- !defined(_WIN32) && !defined(PADDLE_WITH_LOONGARCH)
+ !defined(_WIN32) && !defined(PADDLE_WITH_LOONGARCH) && \
+ !defined(__powerpc__) && !defined(__ppc__) && !defined(__PPC__)
#define DENORM_USE_INTRINSICS
#endif

diff --git a/paddle/phi/kernels/funcs/search_compute.h b/paddle/phi/kernels/funcs/search_compute.h
index b51f10cce9..ae25b45cfb 100644
--- a/paddle/phi/kernels/funcs/search_compute.h
+++ b/paddle/phi/kernels/funcs/search_compute.h
@@ -15,7 +15,8 @@
#pragma once

#if !defined(PADDLE_WITH_ARM) && !defined(PADDLE_WITH_SW) && \
- !defined(PADDLE_WITH_MIPS) && !defined(PADDLE_WITH_LOONGARCH)
+ !defined(PADDLE_WITH_MIPS) && !defined(PADDLE_WITH_LOONGARCH) && \
+ (defined(__x86_64__) || defined(__i386__) || defined(_M_X64) || defined(_M_IX86))
#include <immintrin.h>
#endif
#include <cfloat>
@@ -101,7 +102,8 @@ void call_gemm_batched(const Context& ctx,
}

#if !defined(PADDLE_WITH_ARM) && !defined(PADDLE_WITH_SW) && \
- !defined(PADDLE_WITH_MIPS) && !defined(PADDLE_WITH_LOONGARCH)
+ !defined(PADDLE_WITH_MIPS) && !defined(PADDLE_WITH_LOONGARCH) && \
+ (defined(__x86_64__) || defined(__i386__) || defined(_M_X64) || defined(_M_IX86))

#define __m256x __m256

@@ -144,7 +146,7 @@ inline void axpy(const T* x, T* y, size_t len, const T alpha) {
#elif defined(PADDLE_WITH_ARM) || defined(PADDLE_WITH_SW) || \
defined(PADDLE_WITH_MIPS) || defined(PADDLE_WITH_LOONGARCH)
PADDLE_THROW(common::errors::Unimplemented("axpy is not supported"));
-#else
+#elif defined(__x86_64__) || defined(__i386__) || defined(_M_X64) || defined(_M_IX86)
lll = len & ~SSE_CUT_LEN_MASK;
__m128x mm_alpha = _mm_load1_px(&alpha);
for (jjj = 0; jjj < lll; jjj += SSE_STEP_SIZE) {
@@ -174,7 +176,7 @@ inline void axpy_noadd(const T* x, T* y, size_t len, const T alpha) {
#elif defined(PADDLE_WITH_ARM) || defined(PADDLE_WITH_SW) || \
defined(PADDLE_WITH_MIPS) || defined(PADDLE_WITH_LOONGARCH)
PADDLE_THROW(common::errors::Unimplemented("axpy_noadd is not supported"));
-#else
+#elif defined(__x86_64__) || defined(__i386__) || defined(_M_X64) || defined(_M_IX86)
lll = len & ~SSE_CUT_LEN_MASK;
__m128x mm_alpha = _mm_load1_px(&alpha);
for (jjj = 0; jjj < lll; jjj += SSE_STEP_SIZE) {
diff --git a/paddle/phi/kernels/funcs/softmax_impl.h b/paddle/phi/kernels/funcs/softmax_impl.h
index 8f6b0fdd32..9b094fd237 100644
--- a/paddle/phi/kernels/funcs/softmax_impl.h
+++ b/paddle/phi/kernels/funcs/softmax_impl.h
@@ -237,14 +237,16 @@ class SoftmaxFunctor<DeviceContext, T, enable_if_CPU<DeviceContext>> {
const int axis_dim,
const phi::DenseTensor* X,
phi::DenseTensor* Y) {
+#if !defined(__powerpc__) && !defined(__ppc__) && !defined(__PPC__)
const auto& in_dims = X->dims();
constexpr int kBatchDim = 0;
constexpr int kClassDim = 1;
-
const int num_classes = in_dims[kClassDim];
const int batch_size = in_dims[kBatchDim];
const int num_remain = num_classes / axis_dim;

+ const int batch_size = in_dims[kBatchDim];
+ const int num_remain = num_classes / axis_dim;
if (num_remain == 1 &&
phi::backends::cpu::MayIUse(phi::backends::cpu::avx)) {
const T* in_data = X->data<T>();
@@ -267,7 +269,9 @@ class SoftmaxFunctor<DeviceContext, T, enable_if_CPU<DeviceContext>> {
in_data += num_classes;
out_data += num_classes;
}
- } else {
+ } else
+#endif
+ {
SoftmaxEigen<DeviceContext, T>()(context, axis_dim, X, Y);
}
}
@@ -394,6 +398,7 @@ class SoftmaxGradFunctor<DeviceContext, T, enable_if_CPU<DeviceContext>> {
const phi::DenseTensor* y,
const phi::DenseTensor* y_grad,
phi::DenseTensor* x_grad) {
+#if !defined(__powerpc__) && !defined(__ppc__) && !defined(__PPC__)
const auto& out_dims = y->dims();
constexpr int kBatchDim = 0;
constexpr int kClassDim = 1;
@@ -419,7 +424,9 @@ class SoftmaxGradFunctor<DeviceContext, T, enable_if_CPU<DeviceContext>> {
out_grad += num_classes;
in_grad += num_classes;
}
- } else {
+ } else
+#endif
+ {
SoftmaxGradEigen<DeviceContext, T>()(
context, axis_dim, y, y_grad, x_grad);
}
Loading