Skip to content

Port equal from THC to ATen (CUDA) #36483

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

Closed
Baranowski wants to merge 1 commit into from
+18 −49
Closed

Port equal from THC to ATen (CUDA) #36483

Show file tree
Hide file tree
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
13 changes: 0 additions & 13 deletions aten/src/ATen/Declarations.cwrap
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,19 +146,6 @@
- THIndexTensor* index
- real value
]]
[[
name: _th_equal
cname: equal
backends:
- CUDA
cuda_bool: True
variants:
- function
return: bool
arguments:
- THTensor* self
- THTensor* other
]]
[[
name: _th_mode
variants: function
Expand Down
17 changes: 17 additions & 0 deletions aten/src/ATen/native/cuda/ReduceLogicKernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,4 +24,21 @@ void or_kernel_cuda(TensorIterator& iter) {
REGISTER_DISPATCH(and_stub, &and_kernel_cuda);
REGISTER_DISPATCH(or_stub, &or_kernel_cuda);

bool cuda_equal(const Tensor& self, const Tensor &src) {
if (!at::namedinference::are_names_equal(
self.unsafeGetTensorImpl(), src.unsafeGetTensorImpl())) {
return false;
}
at::NoNamesGuard guard;
TORCH_CHECK(self.device() == src.device(), "Cannot compare two tensors on "
"different devices. Got: ", self.device(), " and ", src.device());
if (self.sizes() != src.sizes()) {
return false;
}
if (self.numel() == 0) {
return true;
}
return at::native::eq(self, src).all().item().to<bool>();
}

}} // namespace at::native
2 changes: 1 addition & 1 deletion aten/src/ATen/native/native_functions.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5014,7 +5014,7 @@
variants: method, function
dispatch:
CPU: cpu_equal
CUDA: legacy::cuda::_th_equal
CUDA: cuda_equal
QuantizedCPU: quantized_equal_cpu

- func: pow.Tensor_Tensor_out(Tensor self, Tensor exponent, *, Tensor(a!) out) -> Tensor(a!)
Expand Down
35 changes: 0 additions & 35 deletions aten/src/THC/generic/THCTensorMathPairwise.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,41 +4,6 @@

#include <ATen/NamedTensorUtils.h>

static int THCTensor_(equalImpl)(THCState *state, THCTensor *self_, THCTensor *src_)
{
THCAssertSameGPU(THCTensor_(checkGPU)(state, 2, self_, src_));
if (!THCTensor_(isSameSizeAs(state, self_, src_))) {
return 0;
}

// This is not as efficient as TH, but the basic idea: create a buffer that stores
// 1 if the two tensors are equal at a position, otherwise 0. If the minimum value
// in this buffer is 1, the two tensors are equal, otherwise they are not

// Both tensors are empty
if(THTensor_(nElement)(self_) == 0) return true;

THCudaByteTensor *buf = at::empty_like(THTensor_wrap(self_), at::kByte).unsafeReleaseTensorImpl();

if (!THC_pointwiseApply3<uint8_t, scalar_t, scalar_t>(state, buf, self_, src_, TensorEQOp<scalar_t, unsigned char>())) {
THArgCheck(false, 2, CUTORCH_DIM_WARNING);
}

unsigned char min = THTensor_wrap(buf).min().item<unsigned char>();

THCudaByteTensor_free(state, buf);

return min != 0;
}

int THCTensor_(equal)(THCState *state, THCTensor *self_, THCTensor *src_) {
if (!at::namedinference::are_names_equal(self_, src_)) {
return 0;
}
at::NoNamesGuard guard;
return THCTensor_(equalImpl)(state, self_, src_);
}

#if !defined(THC_REAL_IS_BOOL)

void THCTensor_(mul)(THCState *state, THCTensor *self_, THCTensor *src_, scalar_t value)
Expand Down