mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-11-18 09:58:18 +08:00
CUDA 13.2 samples update (#432)
- Added Python samples for CUDA Python 1.0 release - Renamed top-level `Samples` directory to `cpp` to accommodate Python samples.
This commit is contained in:
27
cpp/7_libNVVM/ptxgen/README.md
Normal file
27
cpp/7_libNVVM/ptxgen/README.md
Normal file
@@ -0,0 +1,27 @@
|
||||
Introduction
|
||||
============
|
||||
|
||||
ptxgen is a simple IR compiler that generates PTX code in stdout with the
|
||||
input NVVM IR files. It generates warnings, errors, and other messages in
|
||||
stderr. When multiple input files are given, it links them into a single
|
||||
module before the compilation, and generates a single PTX module.
|
||||
|
||||
ptxgen always links the libDevice library with the input NVVM IR program.
|
||||
|
||||
Before compiling the input IR, ptxgen will verify the IR for conformance
|
||||
to the NVVM IR specification.
|
||||
|
||||
Usage
|
||||
-----
|
||||
|
||||
The command-line options, except for the program name and the input file
|
||||
names, are directly passed to nvvmCompileProgram without modification.
|
||||
Each input NVVM IR file can be either in the bitcode representation or
|
||||
in the text representation. Input file names and command-line options can be
|
||||
interleaved.
|
||||
|
||||
For example,
|
||||
|
||||
$ ptxgen a.ll -arch=compute_75 b.bc
|
||||
|
||||
links a.ll and b.bc, and generates PTX code for the compute_75 architecture.
|
||||
Reference in New Issue
Block a user