From e86e04f6d6cf1bcdaf6a7f9cac1930dde2c8b6ab Mon Sep 17 00:00:00 2001
From: LordMathis <matus@namesny.com>
Date: Thu, 11 Apr 2024 19:49:09 +0200
Subject: [PATCH] Add clearing kernel

---
 include/kernels/matmul.cuh |  5 +++++
 src/kernels/matmul.cu      | 11 +++++++++++
 2 files changed, 16 insertions(+)

diff --git a/include/kernels/matmul.cuh b/include/kernels/matmul.cuh
index 75100ba..de70f1f 100644
--- a/include/kernels/matmul.cuh
+++ b/include/kernels/matmul.cuh
@@ -62,6 +62,11 @@ __global__ void vec_scalar_sub(
     const unsigned int w
 );
 
+__global__ void clear(
+    float* __restrict__ d_vector,
+    const unsigned int w
+);
+
 }  // namespace CUDANet::Kernels
 
 #endif  // CUDANET_MATMUL_H
\ No newline at end of file
diff --git a/src/kernels/matmul.cu b/src/kernels/matmul.cu
index 190b8b4..80e3c00 100644
--- a/src/kernels/matmul.cu
+++ b/src/kernels/matmul.cu
@@ -83,4 +83,15 @@ __global__ void Kernels::vec_scalar_sub(
         return;
     }
     d_output[tid] = d_vector[tid] - d_scalar[0];
+}
+
+__global__ void Kernels::clear(
+    float* __restrict__ d_vector,
+    const unsigned int w
+) {
+    int tid = blockDim.x * blockIdx.x + threadIdx.x;
+    if (tid >= w) {
+        return;
+    }
+    d_vector[tid] = 0.0f;
 }
\ No newline at end of file