From aa3abd4bc689d6412979c7f55f9cd132848c9c6a Mon Sep 17 00:00:00 2001
From: Anerudhan Gopal <agopal@nvidia.com>
Date: Tue, 18 Jun 2024 01:46:41 +0000
Subject: [PATCH] Release notes for cudnn-frontend 1.5.1: (#84)

[Bug fix] Fixed an issue, where cudnn-frontend (1.5.0) when built with
cudnn version 9.1.1 and below, runs into issues when run with 9.2.0 and
above.
---
 CMakeLists.txt                 | 2 +-
 include/cudnn_frontend.h       | 2 +-
 include/cudnn_frontend/plans.h | 4 ++--
 include/cudnn_frontend_shim.h  | 5 +++++
 python/cudnn/__init__.py       | 2 +-
 5 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index f52185f1..28eb9617 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.17)
 
-project(cudnn_frontend VERSION 1.5.0)
+project(cudnn_frontend VERSION 1.5.1)
 
 option(CUDNN_FRONTEND_SKIP_JSON_LIB "Defines whether FE should not include nlohmann/json.hpp." OFF)
 option(CUDNN_FRONTEND_BUILD_SAMPLES "Defines if samples are built or not." ON)
diff --git a/include/cudnn_frontend.h b/include/cudnn_frontend.h
index 114c8a98..7bca4256 100644
--- a/include/cudnn_frontend.h
+++ b/include/cudnn_frontend.h
@@ -125,7 +125,7 @@
 
 #define CUDNN_FRONTEND_MAJOR_VERSION 1
 #define CUDNN_FRONTEND_MINOR_VERSION 5
-#define CUDNN_FRONTEND_PATCH_VERSION 0
+#define CUDNN_FRONTEND_PATCH_VERSION 1
 #define CUDNN_FRONTEND_VERSION \
     ((CUDNN_FRONTEND_MAJOR_VERSION * 10000) + (CUDNN_FRONTEND_MINOR_VERSION * 100) + CUDNN_FRONTEND_PATCH_VERSION)
 
diff --git a/include/cudnn_frontend/plans.h b/include/cudnn_frontend/plans.h
index de37ca65..370b7424 100644
--- a/include/cudnn_frontend/plans.h
+++ b/include/cudnn_frontend/plans.h
@@ -225,7 +225,7 @@ class Execution_plan_list {
         }
 
         // workspace check for 9.2+ is already done at engine config level
-        if (detail::get_backend_version() < 90200) {
+        if (detail::get_backend_version() < 90200 || detail::get_compiled_version() < 90200) {
             if (execution_plans[index]->getWorkspaceSize() > max_workspace_allowed) {
                 barred_indices[index] = true;
                 return {error_code_t::GRAPH_EXECUTION_PLAN_CREATION_FAILED,
@@ -427,7 +427,7 @@ class Execution_plan_list {
                         std::to_string(index)};
         }
 
-        if (detail::get_backend_version() >= 90200) {
+        if (detail::get_backend_version() >= 90200 && detail::get_compiled_version() >= 90200) {
             // Ignore kernels that require larger than tolerable shared memory.
             int32_t shared_memory_size = INT32_MAX;
             auto status                = detail::get_shared_memory_size(engine_configs[index], shared_memory_size);
diff --git a/include/cudnn_frontend_shim.h b/include/cudnn_frontend_shim.h
index de6dbf71..be185569 100644
--- a/include/cudnn_frontend_shim.h
+++ b/include/cudnn_frontend_shim.h
@@ -190,6 +190,11 @@ get_backend_version(void) {
 #endif
 }
 
+inline constexpr size_t
+get_compiled_version(void) {
+    return CUDNN_VERSION;
+}
+
 inline std::string
 convert_version_to_str(size_t const version) {
     // The multiplier for major version pre-v9 and post-v9 are different.
diff --git a/python/cudnn/__init__.py b/python/cudnn/__init__.py
index cb7fdf6c..ad29d101 100644
--- a/python/cudnn/__init__.py
+++ b/python/cudnn/__init__.py
@@ -25,7 +25,7 @@
 
 from .datatypes import _library_type, _is_torch_tensor
 
-__version__ = "1.5.0"
+__version__ = "1.5.1"
 
 
 def _tensor(