[PYTHON][TESTS][DOC] Various improvement of the API and code quality:

* Simplified `triton.kernel` API to achieve lower latency:
  > .data_ptr() must now be passed as kernel argument. No more implicit
conversion from torch.tensor
  > compilation options are now constant attributes, i.e., opt.d('VAR')
becomes opt.VAR
  > torch.device must now be passed explicitly to triton.kernel (no
longer inferred from torch.tensor arguments)
* C++ tests moved to `python/tests/`
* C++ tutorial created in `tutorials/`
* Python tutorial created in python/tutorials/
* Version changed to 1.0alpha
* No longer copying C++ headers into the Python package
* added python/triton/ops/ package for pre-written Triton ops
This commit is contained in:
Philippe Tillet
2021-01-29 17:27:16 -05:00
parent a5a477c36b
commit 269ebc12e5
63 changed files with 2255 additions and 3883 deletions

View File

@@ -76,7 +76,7 @@ void host_stream::synchronize() {
hst_->args.clear();
}
void host_stream::enqueue(driver::kernel* kernel, std::array<size_t, 3> grid, std::array<size_t, 3> block, void **args, size_t args_size) {
void host_stream::enqueue(driver::kernel* kernel, std::array<size_t, 3> grid, std::array<size_t, 3> block, void* args, size_t args_size) {
auto hst = kernel->module()->hst();
hst_->futures->reserve(hst_->futures->size() + grid[0]*grid[1]*grid[2]);
char* params = new char[args_size];
@@ -113,7 +113,7 @@ void cu_stream::synchronize() {
dispatch::cuStreamSynchronize(*cu_);
}
void cu_stream::enqueue(driver::kernel* kernel, std::array<size_t, 3> grid, std::array<size_t, 3> block, void** args, size_t args_size) {
void cu_stream::enqueue(driver::kernel* kernel, std::array<size_t, 3> grid, std::array<size_t, 3> block, void* args, size_t args_size) {
void *config[] = {
CU_LAUNCH_PARAM_BUFFER_POINTER, args,
CU_LAUNCH_PARAM_BUFFER_SIZE, &args_size,