gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[IR] Preliminary support for BF16 (#129)

Browse Source 
This PR adds a BF16 data-type, along with FP32 <-> BF16 conversion instructions in the LLVM codegen. Other kinds of ops on bfloat16 are not yet supported.
This commit is contained in:
Philippe Tillet
2021-06-25 10:19:29 -04:00
committed by Philippe Tillet
parent 9b4e2cae2d
commit 8cea583109
18 changed files with 173 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
lib/driver/module.cc
View File

@@ -283,8 +283,6 @@ std::string cu_module::compile_llvm_module(llvm::Module* module, driver::device*
void cu_module::init_from_ptx(const std::string& ptx, driver::cu_device* device) {
// JIT compile source-code
// std::cout << ptx << std::endl;
try{
std::string ptxas = tools::getenv("TRITON_PTXAS");
@@ -324,7 +322,7 @@ void cu_module::init_from_ptx(const std::string& ptx, driver::cu_device* device)
}
catch(exception::cuda::invalid_ptx const &){
//#ifdef TRITON_LOG_PTX_ERROR
// std::cout << ptx << std::endl;
std::cout << ptx << std::endl;
std::cerr << "It appears that Triton produced invalid PTX code:" << std::endl;
// exit(1);
//#endif