From 760a31c8481922cbc12fa6e0b2d13f67c175aaa9 Mon Sep 17 00:00:00 2001
From: Tianlei Wu <tlwu@microsoft.com>
Date: Tue, 16 Jul 2024 15:58:11 -0700
Subject: [PATCH] Exclude blkq4_fp16_gemm_sm80_test in cuda 12.5 build (#21373)

There is build errors when build with CUDA 12.5 and
`--cmake_extra_defines onnxruntime_BUILD_UNIT_TESTS=ON
onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS=ON`.

Temporally exclude blkq4_fp16_gemm_sm80_test to unblock cuda 12.5 build.
---
 .../test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80.h    | 3 ++-
 .../providers/cuda/test_cases/blkq4_fp16_gemm_sm80_test.cc   | 3 ++-
 .../providers/cuda/test_cases/blkq4_fp16_gemm_sm80_testcu.cu | 5 +++++
 3 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80.h b/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80.h
index fa1c739c04e3a..f96c8ce9ce729 100644
--- a/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80.h
+++ b/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80.h
@@ -13,7 +13,7 @@
  */
 
 #pragma once
-
+#if defined(CUDA_VERSION) && CUDA_VERSION <= 12030
 #include "test/cuda_host/blkq4_fp16_quant_sm80.h"
 
 #include <random>
@@ -197,3 +197,4 @@ void run_blkq4_small_gemm(int m, int n, int k);
 }  // namespace test
 }  // namespace cuda
 }  // namespace onnxruntime
+#endif
diff --git a/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_test.cc b/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_test.cc
index b95e093e41eab..3fcb9045ee7e6 100644
--- a/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_test.cc
+++ b/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_test.cc
@@ -10,7 +10,7 @@
  *   This part requires gtest header files, which do not play
  *   well with CUTLASS headers.
  */
-
+#if defined(CUDA_VERSION) && CUDA_VERSION <= 12030
 #include "blkq4_fp16_gemm_sm80.h"
 
 #include "gtest/gtest.h"
@@ -341,3 +341,4 @@ TEST(BlkQ4_GEMM, Sm80SmallTileKernelTest) {
 
 }  // namespace test
 }  // namespace onnxruntime
+#endif
diff --git a/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_testcu.cu b/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_testcu.cu
index f5600ca9885a3..8b27c3d8c3aed 100644
--- a/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_testcu.cu
+++ b/onnxruntime/test/providers/cuda/test_cases/blkq4_fp16_gemm_sm80_testcu.cu
@@ -11,6 +11,9 @@
  *   well with gtest headers.
  */
 
+// This test has build error with cuda 12.5
+#if defined(CUDA_VERSION) && CUDA_VERSION <= 12030
+
 #include "blkq4_fp16_gemm_sm80.h"
 
 #include <random>
@@ -532,3 +535,5 @@ template void run_blkq4_small_gemm<128, false, false>(int m, int n, int k);
 }  // namespace test
 }  // namespace cuda
 }  // namespace onnxruntime
+
+#endif