From fac4e96940d9f2ac8dde8fc864b4c76cbdfd0e2d Mon Sep 17 00:00:00 2001 From: Hashem Hashemi <159079214+amd-hhashemi@users.noreply.github.com> Date: Thu, 12 Feb 2026 12:26:36 -0800 Subject: [PATCH] small adjustment to wvSplitKrc (#34410) Signed-off-by: Hashem Hashemi --- csrc/rocm/skinny_gemms.cu | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/csrc/rocm/skinny_gemms.cu b/csrc/rocm/skinny_gemms.cu index ecd94cacc..976874e6f 100644 --- a/csrc/rocm/skinny_gemms.cu +++ b/csrc/rocm/skinny_gemms.cu @@ -1568,8 +1568,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS) { #endif unsigned int kOff = k + (thrd * A_CHUNK); - unsigned int kOffcp = - k_str + kOff; // min__(K - A_CHUNK, k_str + kOff); + unsigned int kOffcp = min__(K - A_CHUNK, k_str + kOff); for (unsigned int n = 0; n < N; n += CHUNKK * sprdN) { __builtin_amdgcn_global_load_lds( (int*)(&A[min__(