Search for: All records

Programming to achieve high performance for NVIDIA GPUs using CUDA has been known to be challenging. A GPU has hundreds or thousands of cores that a program must exhibit sufficient parallelism to achieve maximum GPU utilization. A system with GPU accelerators has a heterogeneous and deep memory system that programmers must effectively and correctly use to fully take advantage of the GPU's parallelism capability. In this paper, we present CUDAMicroBench, a collection of fourteen microbenchmarks that demonstrate performance challenges in CUDA programming and techniques to optimize the CUDA programs to address these challenges. It also includes examples and techniques for using advanced CUDA features such as data shuffling between threads, dynamic parallelism, etc that can help users optimize the CUDA program for performance. The microbenchmark can be used for evaluating the performance of GPU architectures, the memory systems of GPU itself and of the whole system architectures, and for evaluating the effectiveness of compiler and performance tools for performance analysis. It can be used to help users understand the complexity of heterogeneous GPU-accelerator systems through examples and guide users for performance optimization. It is released as BSD-licensed open-source from https://github.com/passlab/CUDAMicroBench.git. 

YiiP is a secondary transporter that couples Zn2+ transport to the proton motive force. Structural studies of YiiP from prokaryotes and Znt8 from humans have revealed three different Zn2+ sites and a conserved homodimeric architecture. These structures define the inward-facing and outward-facing states that characterize the archetypal alternating access mechanism of transport. To study the effects of Zn2+ binding on the conformational transition, we use cryo-EM together with molecular dynamics simulation to compare structures of YiiP from Shewanella oneidensis in the presence and absence of Zn2+. To enable single-particle cryo-EM, we used a phage-display library to develop a Fab antibody fragment with high affinity for YiiP, thus producing a YiiP/Fab complex. To perform MD simulations, we developed a nonbonded dummy model for Zn2+ and validated its performance with known Zn2+-binding proteins. Using these tools, we find that, in the presence of Zn2+, YiiP adopts an inward-facing conformation consistent with that previously seen in tubular crystals. After removal of Zn2+ with high-affinity chelators, YiiP exhibits enhanced flexibility and adopts a novel conformation that appears to be intermediate between inward-facing and outward-facing states. This conformation involves closure of a hydrophobic gate that has been postulated to control access to the primary transport site. Comparison of several independent cryo-EM maps suggests that the transition from the inward-facing state is controlled by occupancy of a secondary Zn2+ site at the cytoplasmic membrane interface. This work enhances our understanding of individual Zn2+ binding sites and their role in the conformational dynamics that govern the transport cycle.

 

Bor1p is a secondary transporter in yeast that is responsible for boron transport. Bor1p belongs to the SLC4 family which controls bicarbonate exchange and pH regulation in animals as well as borate uptake in plants. The SLC4 family is more distantly related to members of the Amino acid‐Polyamine‐organoCation (APC) superfamily, which includes well studied transporters such as LeuT, Mhp1, AdiC, vSGLT, UraA, SLC26Dg. Their mechanism generally involves relative movements of two domains: a core domain that binds substrate and a gate domain that in many cases mediates dimerization. To shed light on conformational changes governing transport by the SLC4 family, we grew helical membrane crystals of Bor1p fromSaccharomyces mikataeand determined a structure at ∼6 Å resolution using cryo‐electron microscopy. To evaluate the conformation of Bor1p in these crystals, a homology model was built based on the related anion exchanger from red blood cells (AE1). This homology model was fitted to the cryo‐EM density map using the Molecular Dynamics (MD) Flexible Fitting method and then relaxed by all‐atom MD simulation in explicit solvent and membrane. Mapping of water accessibility indicates that the resulting structure represents an inward‐facing conformation. Comparisons of the resulting Bor1p model with the X‐ray structure of AE1 in an outward‐facing conformation, together with MD simulations of inward‐facing and outward‐facing Bor1p models, suggest rigid body movements of the core domain relative to the gate domain. These movements are consistent with the rocking‐bundle transport mechanism described for other members of the APC superfamily.