From 507d2bb9b93a7ac51dfb32633a8eb86e75d80ea1 Mon Sep 17 00:00:00 2001
From: Sergii Dymchenko <sedymche@microsoft.com>
Date: Thu, 9 Apr 2020 18:09:46 -0700
Subject: [PATCH] Get onnxruntime/core/providers/cuda/math/matmul_integer.cc
 from ort_training.

---
 .../providers/cuda/math/matmul_integer.cc     | 23 -------------------
 1 file changed, 23 deletions(-)

diff --git a/onnxruntime/core/providers/cuda/math/matmul_integer.cc b/onnxruntime/core/providers/cuda/math/matmul_integer.cc
index 61164a1ddb..e4547da039 100644
--- a/onnxruntime/core/providers/cuda/math/matmul_integer.cc
+++ b/onnxruntime/core/providers/cuda/math/matmul_integer.cc
@@ -6,7 +6,6 @@
 #include "core/providers/cpu/math/matmul_helper.h"
 #include "core/providers/cuda/shared_inc/fpgeneric.h"
 #include "core/providers/cuda/cuda_allocator.h"
-#include "core/providers/cuda/igemm.h"
 #include "core/providers/common.h"
 
 namespace onnxruntime {
@@ -107,28 +106,6 @@ Status MatMulInteger<int8_t, int8_t>::ComputeInternal(OpKernelContext* ctx) cons
     beta = 1;
   }
 
-#if CUDA_VERSION >= 10010
-  if (DeviceProp::GetDeviceProps().major >= 7 && DeviceProp::GetDeviceProps().minor >= 5) {
-    for (size_t batch = 0; batch < helper.OutputOffsets().size(); batch++) {
-      LtIgemmTensor(
-          static_cast<int>(helper.M()),
-          static_cast<int>(helper.N()),
-          static_cast<int>(helper.K()),
-          alpha,
-          beta,
-          a_ptr + helper.LeftOffsets()[batch],
-          static_cast<int>(helper.K()),
-          b_ptr + helper.RightOffsets()[batch],
-          static_cast<int>(helper.N()),
-          output_ptr + helper.OutputOffsets()[batch],
-          static_cast<int>(helper.N()),
-          this,
-          Base::CublasLtHandle());
-    }
-    return Status::OK();
-  }
-#endif
-
   // pad A and B to make their leading dimension be multiples of 32
   // because cublasGemmEx requires:
   // 1. leading dimension is multiples of 4