From 3e3a26e0da526de9f577c007a61d5ee7b5343e69 Mon Sep 17 00:00:00 2001 From: Cedric Nugteren Date: Fri, 20 Apr 2018 21:50:36 +0200 Subject: Fixes for the CUDA API --- src/clblast_cuda.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/clblast_cuda.cpp b/src/clblast_cuda.cpp index 0ba57056..8927014b 100644 --- a/src/clblast_cuda.cpp +++ b/src/clblast_cuda.cpp @@ -2593,7 +2593,8 @@ StatusCode GemmTempBufferSize(const Layout layout, const Transpose a_transpose, else { temp_buffer_size = Xgemm::GetTempSize(layout, a_transpose, b_transpose, m, n, k, a_offset, a_ld, b_offset, b_ld, c_offset, c_ld, - db["MWG"], db["NWG"], db["KWG"]); + db["MWG"], db["NWG"], db["KWG"] * db["KREG"], + db["GEMMK"]); } temp_buffer_size *= sizeof(T); // translate from num-elements to bytes return StatusCode::kSuccess; -- cgit v1.2.3