replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[SE] Clean up device and host memory slices 

Summary:
* Add LLVM_ATTRIBUTE_UNUSED_RESULT used to slicing methods in order to
  emphasize that the slicing is not done in place.
* Change device memory slice function name from `drop_front` to `slice`
  in order to match the naming convention of `llvm::ArrayRef` and host
  memory slice.
* Change the parameter names of host memory slice functions to
  `DropCount` and `TakeCount` to match device memory slice declarations.

Reviewers: jlebar

Subscribers: jprice, parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D24464

llvm-svn: 281239
This commit is contained in:
Jason Henline 2016-09-12 17:20:43 +00:00
parent 8d4be3aacf
commit c16fb8748d
4 changed files with 35 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
parallel-libs/streamexecutor/include/streamexecutor/DeviceMemory.h
View File

@ -77,7 +77,8 @@ public:
size_t getByteCount() const { return ElementCount * sizeof(ElemT); }
/// Creates a slice of the memory with the first DropCount elements removed.
GlobalDeviceMemorySlice<ElemT> drop_front(size_t DropCount) const {
LLVM_ATTRIBUTE_UNUSED_RESULT
GlobalDeviceMemorySlice<ElemT> slice(size_t DropCount) const {
assert(DropCount <= ElementCount &&
"dropping more than the size of a slice");
return GlobalDeviceMemorySlice<ElemT>(BaseMemory, ElementOffset + DropCount,
@ -85,6 +86,7 @@ public:
}
/// Creates a slice of the memory with the last DropCount elements removed.
LLVM_ATTRIBUTE_UNUSED_RESULT
GlobalDeviceMemorySlice<ElemT> drop_back(size_t DropCount) const {
assert(DropCount <= ElementCount &&
"dropping more than the size of a slice");
@ -94,6 +96,7 @@ public:
/// Creates a slice of the memory that chops off the first DropCount elements
/// and keeps the next TakeCount elements.
LLVM_ATTRIBUTE_UNUSED_RESULT
GlobalDeviceMemorySlice<ElemT> slice(size_t DropCount,
size_t TakeCount) const {
assert(DropCount + TakeCount <= ElementCount &&

28
parallel-libs/streamexecutor/include/streamexecutor/HostMemory.h
View File

@ -47,19 +47,26 @@ public:
ElemT *getPointer() const { return MutableArrayRef.data(); }
size_t getElementCount() const { return MutableArrayRef.size(); }
/// Chops off the first N elements of the slice.
MutableRegisteredHostMemorySlice slice(size_t N) const {
return MutableRegisteredHostMemorySlice(MutableArrayRef.slice(N));
/// Chops off the first DropCount elements of the slice.
LLVM_ATTRIBUTE_UNUSED_RESULT
MutableRegisteredHostMemorySlice slice(size_t DropCount) const {
return MutableRegisteredHostMemorySlice(MutableArrayRef.slice(DropCount));
}
/// Chops off the first N elements of the slice and keeps the next M elements.
MutableRegisteredHostMemorySlice slice(size_t N, size_t M) const {
return MutableRegisteredHostMemorySlice(MutableArrayRef.slice(N, M));
/// Chops off the first DropCount elements of the slice and keeps the next
/// TakeCount elements.
LLVM_ATTRIBUTE_UNUSED_RESULT
MutableRegisteredHostMemorySlice slice(size_t DropCount,
size_t TakeCount) const {
return MutableRegisteredHostMemorySlice(
MutableArrayRef.slice(DropCount, TakeCount));
}
/// Chops off the last N elements of the slice.
MutableRegisteredHostMemorySlice drop_back(size_t N) const {
return MutableRegisteredHostMemorySlice(MutableArrayRef.drop_back(N));
/// Chops off the last DropCount elements of the slice.
LLVM_ATTRIBUTE_UNUSED_RESULT
MutableRegisteredHostMemorySlice drop_back(size_t DropCount) const {
return MutableRegisteredHostMemorySlice(
MutableArrayRef.drop_back(DropCount));
}
private:
@ -91,16 +98,19 @@ public:
size_t getElementCount() const { return ArrayRef.size(); }
/// Chops off the first N elements of the slice.
LLVM_ATTRIBUTE_UNUSED_RESULT
RegisteredHostMemorySlice slice(size_t N) const {
return RegisteredHostMemorySlice(ArrayRef.slice(N));
}
/// Chops off the first N elements of the slice and keeps the next M elements.
LLVM_ATTRIBUTE_UNUSED_RESULT
RegisteredHostMemorySlice slice(size_t N, size_t M) const {
return RegisteredHostMemorySlice(ArrayRef.slice(N, M));
}
/// Chops off the last N elements of the slice.
LLVM_ATTRIBUTE_UNUSED_RESULT
RegisteredHostMemorySlice drop_back(size_t N) const {
return RegisteredHostMemorySlice(ArrayRef.drop_back(N));
}

16
parallel-libs/streamexecutor/unittests/CoreTests/DeviceTest.cpp
View File

@ -141,7 +141,7 @@ TEST_F(DeviceTest, SyncCopyD2HToPointer) {
TEST_F(DeviceTest, SyncCopyD2HSliceToMutableArrayRefByCount) {
EXPECT_NO_ERROR(Device.synchronousCopyD2H(
DeviceA5.asSlice().drop_front(1), MutableArrayRef<int>(Host5 + 1, 4), 4));
DeviceA5.asSlice().slice(1), MutableArrayRef<int>(Host5 + 1, 4), 4));
for (int I = 1; I < 5; ++I) {
EXPECT_EQ(HostA5[I], Host5[I]);
}
@ -177,8 +177,8 @@ TEST_F(DeviceTest, SyncCopyD2HSliceToMutableArrayRef) {
}
TEST_F(DeviceTest, SyncCopyD2HSliceToPointer) {
EXPECT_NO_ERROR(Device.synchronousCopyD2H(DeviceA5.asSlice().drop_front(1),
Host5 + 1, 4));
EXPECT_NO_ERROR(
Device.synchronousCopyD2H(DeviceA5.asSlice().slice(1), Host5 + 1, 4));
for (int I = 1; I < 5; ++I) {
EXPECT_EQ(HostA5[I], Host5[I]);
}
@ -227,8 +227,8 @@ TEST_F(DeviceTest, SyncCopyH2DToPointer) {
}
TEST_F(DeviceTest, SyncCopyH2DSliceToArrayRefByCount) {
EXPECT_NO_ERROR(Device.synchronousCopyH2D(
ArrayRef<int>(Host5 + 1, 4), DeviceA5.asSlice().drop_front(1), 4));
EXPECT_NO_ERROR(Device.synchronousCopyH2D(ArrayRef<int>(Host5 + 1, 4),
DeviceA5.asSlice().slice(1), 4));
for (int I = 1; I < 5; ++I) {
EXPECT_EQ(getDeviceValue(DeviceA5, I), Host5[I]);
}
@ -305,7 +305,7 @@ TEST_F(DeviceTest, SyncCopyD2D) {
TEST_F(DeviceTest, SyncCopySliceD2DByCount) {
EXPECT_NO_ERROR(
Device.synchronousCopyD2D(DeviceA5.asSlice().drop_front(1), DeviceB5, 4));
Device.synchronousCopyD2D(DeviceA5.asSlice().slice(1), DeviceB5, 4));
for (int I = 0; I < 4; ++I) {
EXPECT_EQ(getDeviceValue(DeviceA5, I + 1), getDeviceValue(DeviceB5, I));
}
@ -331,7 +331,7 @@ TEST_F(DeviceTest, SyncCopySliceD2D) {
}
EXPECT_ERROR(
Device.synchronousCopyD2D(DeviceA7.asSlice().drop_front(1), DeviceB5));
Device.synchronousCopyD2D(DeviceA7.asSlice().slice(1), DeviceB5));
EXPECT_ERROR(
Device.synchronousCopyD2D(DeviceA5.asSlice().drop_back(1), DeviceB7));
@ -339,7 +339,7 @@ TEST_F(DeviceTest, SyncCopySliceD2D) {
TEST_F(DeviceTest, SyncCopyD2DSliceByCount) {
EXPECT_NO_ERROR(
Device.synchronousCopyD2D(DeviceA5, DeviceB7.asSlice().drop_front(2), 5));
Device.synchronousCopyD2D(DeviceA5, DeviceB7.asSlice().slice(2), 5));
for (int I = 0; I < 5; ++I) {
EXPECT_EQ(getDeviceValue(DeviceA5, I), getDeviceValue(DeviceB7, I + 2));
}

8
parallel-libs/streamexecutor/unittests/CoreTests/StreamTest.cpp
View File

@ -114,7 +114,7 @@ TEST_F(StreamTest, CopyD2HToRegistered) {
}
TEST_F(StreamTest, CopyD2HSliceToRegiseredSliceByCount) {
Stream.thenCopyD2H(DeviceA5.asSlice().drop_front(1),
Stream.thenCopyD2H(DeviceA5.asSlice().slice(1),
RegisteredHost5.asSlice().slice(1, 4), 4);
EXPECT_TRUE(Stream.isOK());
for (int I = 1; I < 5; ++I) {
@ -174,7 +174,7 @@ TEST_F(StreamTest, CopyH2DFromRegistered) {
TEST_F(StreamTest, CopyH2DFromRegisteredSliceToSlice) {
Stream.thenCopyH2D(RegisteredHost5.asSlice().slice(1, 4),
DeviceA5.asSlice().drop_front(1), 4);
DeviceA5.asSlice().slice(1), 4);
EXPECT_TRUE(Stream.isOK());
for (int I = 1; I < 5; ++I) {
EXPECT_EQ(getDeviceValue(DeviceA5, I), Host5[I]);
@ -232,7 +232,7 @@ TEST_F(StreamTest, CopyD2D) {
}
TEST_F(StreamTest, CopySliceD2DByCount) {
Stream.thenCopyD2D(DeviceA5.asSlice().drop_front(1), DeviceB5, 4);
Stream.thenCopyD2D(DeviceA5.asSlice().slice(1), DeviceB5, 4);
EXPECT_TRUE(Stream.isOK());
for (int I = 0; I < 4; ++I) {
EXPECT_EQ(getDeviceValue(DeviceA5, I + 1), getDeviceValue(DeviceB5, I));
@ -260,7 +260,7 @@ TEST_F(StreamTest, CopySliceD2D) {
}
TEST_F(StreamTest, CopyD2DSliceByCount) {
Stream.thenCopyD2D(DeviceA5, DeviceB7.asSlice().drop_front(2), 5);
Stream.thenCopyD2D(DeviceA5, DeviceB7.asSlice().slice(2), 5);
EXPECT_TRUE(Stream.isOK());
for (int I = 0; I < 5; ++I) {
EXPECT_EQ(getDeviceValue(DeviceA5, I), getDeviceValue(DeviceB7, I + 2));