[DRIVER] Added options for developers to cache PTX file so that ti can

be manually modified
This commit is contained in:
Philippe Tillet
2021-02-09 00:09:10 -05:00
parent 53fd9631ef
commit c847cc6320
3 changed files with 32 additions and 5 deletions

View File

@@ -184,7 +184,7 @@ float triton_dot(drv::context* context, drv::stream* stream,
rt::add_arg(oss, *dlocks->cu());
// function
rt::function function(src::dot, opt, device);
// std::cout << function.get_kernels()[0].second->get_asm(rt::ASM_LLIR) << std::endl;
// std::cout << function.get_kernels()[0].second->get_asm(rt::ASM_NV_PTX) << std::endl;
// grid
auto ceil = [](size_t x, size_t y) { return (x + y - 1) / y; };
auto grid = [ceil, M, N](const rt::options_t& x) {