From eae030c9d1d607d1c14e4ade99cb5caea6403efd Mon Sep 17 00:00:00 2001 From: aamir Date: Sat, 3 Nov 2018 17:03:58 -0700 Subject: merged with memory subsytem. Regression is passing but tensorcore kernel is stuck in deadlock --- src/cuda-sim/ptx_ir.h | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src/cuda-sim/ptx_ir.h') diff --git a/src/cuda-sim/ptx_ir.h b/src/cuda-sim/ptx_ir.h index 0767379..4c36cfc 100644 --- a/src/cuda-sim/ptx_ir.h +++ b/src/cuda-sim/ptx_ir.h @@ -1333,6 +1333,7 @@ public: const struct gpgpu_ptx_sim_info* get_kernel_info () const { + assert (m_kernel_info.maxthreads == maxnt_id); return &m_kernel_info; } @@ -1340,6 +1341,8 @@ public: m_kernel_info = info; m_kernel_info.ptx_version = 10*get_ptx_version().ver(); m_kernel_info.sm_target = get_ptx_version().target(); + // THIS DEPENDS ON ptxas being called after the PTX is parsed. + m_kernel_info.maxthreads = maxnt_id; } symbol_table *get_symtab() { @@ -1363,7 +1366,11 @@ public: } bool is_entry_point() const { return m_entry_point; } + void set_maxnt_id(unsigned maxthreads) { maxnt_id = maxthreads;} + unsigned get_maxnt_id() { return maxnt_id;} + private: + unsigned maxnt_id; unsigned m_uid; unsigned m_local_mem_framesize; bool m_entry_point; -- cgit v1.3