CUTLASS 2.1 (#83)

CUTLASS 2.1 contributes:
- BLAS-style host-side API added to CUTLASS Library
- Planar Complex GEMM kernels targeting Volta and Turing Tensor Cores
- Minor enhancements and bug fixes
This commit is contained in:
Andrew Kerr
2020-04-07 13:51:25 -07:00
committed by GitHub
parent 7c0cd26d13
commit 96dab34ad9
196 changed files with 20653 additions and 1995 deletions

View File

@ -27,10 +27,11 @@
*/
#include <string>
#include <iostream>
#include <stdexcept>
#include <sstream>
#include "cutlass/library/util.h"
#include "problem_space.h"
/////////////////////////////////////////////////////////////////////////////////////////////////
@ -849,17 +850,16 @@ bool arg_as_OpcodeClassID(
}
/////////////////////////////////////////////////////////////////////////////////////////////////
/// Lexically casts an argument to a given type stored in a byte array. Returns true if not null.
bool arg_as_scalar(
std::vector<uint8_t> &bytes,
library::NumericTypeID numeric_type,
KernelArgument::Value const *value_ptr) {
if (value_ptr->not_null) {
if (value_ptr->argument->description->type == ArgumentTypeID::kInteger) {
int64_t int_value = static_cast<IntegerArgument::IntegerValue const *>(value_ptr)->value;
// TODO - convert int64_t => destination type
}
else if (value_ptr->argument->description->type == ArgumentTypeID::kScalar) {