Bottom acceleration structure reconstruction method and device, graphics processor, equipment and medium

Abstract

The application discloses a bottom-layer acceleration structure reconstruction method and device, a graphics processor, equipment and a medium. The method comprises the following steps: screening a reconstructable bottom-layer acceleration structure based on structure feature information; determining a reconstruction cost of the reconstructable bottom-layer acceleration structure; performing reconstruction scoring according to the reconstruction cost and the structure feature information, and determining a structure reconstruction score; determining a target bottom-layer acceleration structure according to the structure reconstruction score and a preset score threshold; and splitting and reconstructing the target bottom-layer acceleration structure to obtain a plurality of sub bottom-layer acceleration structures. The technical scheme provided by the application splits and reconstructs the target bottom-layer acceleration structure to generate a plurality of sub bottom-layer acceleration structures with more compact spatial distribution, significantly reduces the spatial overlap degree between different instance bounding boxes, reduces the number of redundant node access and intersection test times during the traversal of the top-layer acceleration structure in the ray tracing process, improves the traversal efficiency of the ray tracing, and reduces the waste of computing power resources.

Description

Technical Field

[0001] This application belongs to the field of graphics processor technology, and particularly relates to methods, apparatuses, graphics processors, devices and media for reconstructing underlying acceleration structures. Background Technology

[0002] In the field of graphics processing unit (GPU) technology, ray tracing achieves highly realistic rendering by simulating the interaction between light rays and scene objects. Traversal efficiency is highly dependent on the quality of the acceleration structure construction. Acceleration structures are divided into bottom-level acceleration structures (BLAS) that describe the spatial distribution of geometric primitives, and top-level acceleration structures (TLAS) that organize the position and pose of each instance in the scene.

[0003] In existing ray tracing methods, the traversal cost and intersection test count of the top-level acceleration structure are affected by the spatial relationship between instances. When there are large overlapping areas between the bounding boxes of instances from different bottom-level acceleration structures, the ray traversal process needs to visit a large number of redundant nodes, causing the number of tests to increase exponentially. Most existing solutions address this issue by optimizing the layout of the top-level acceleration structure. However, if the coverage area of ​​the bottom-level acceleration structure is too large, resulting in extensive overlap, the layout optimization effect is limited, and it is difficult to eliminate the computational waste and impact on ray traversal efficiency caused by the extensive overlap of the bottom-level acceleration structures.

[0004] Therefore, how to effectively improve the traversal efficiency of ray tracing is an important problem that urgently needs to be solved. Summary of the Invention

[0005] This application provides a method, apparatus, graphics processor, device, and medium for accelerating underlying structure reconstruction, which can effectively improve the traversal efficiency of ray tracing.

[0006] A first aspect of this application provides a method for reconstructing a low-level accelerated structure, applied to ray tracing, the method comprising: Obtain structural feature information corresponding to multiple candidate underlying acceleration structures; Based on structural feature information, at least one reconfigurable underlying acceleration structure is selected from multiple candidate underlying acceleration structures. For each reconfigurable underlying acceleration structure, determine the reconfiguration cost of the reconfigurable underlying acceleration structure, and based on the reconfiguration cost and structural feature information, score the reconfigurable underlying acceleration structure to determine the corresponding structural reconfiguration score. Based on the structural reconstruction score and the preset score threshold, the target underlying acceleration structure is determined from at least one reconfigurable underlying acceleration structure. The target underlying acceleration structure is split and reconstructed to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

[0007] In some embodiments, for each candidate underlying acceleration structure, the structural feature information includes the depth of the candidate underlying acceleration structure tree, the number of instantiations, and the dynamic update frequency. Based on structural feature information, at least one reconfigurable underlying acceleration structure is selected from multiple candidate underlying acceleration structures, including: For each candidate underlying acceleration structure, if the depth of the candidate underlying acceleration structure's structure tree is greater than a preset depth threshold, the number of instantiations is less than a preset number of times threshold, and the dynamic update frequency is zero, the candidate underlying acceleration structure is determined to be a reconfigurable underlying acceleration structure.

[0008] In some embodiments, for each candidate bottom-level acceleration structure, the structural feature information also includes a geometric metric of the candidate bottom-level acceleration structure, which is the surface area of ​​the structure or the bounding box volume. Based on the reconstruction cost and structural feature information, a reconstruction score is calculated for the reconfigurable underlying acceleration structure to determine the corresponding structural reconstruction score, including: For each reconfigurable bottom-level acceleration structure, the structural overlap rate of the reconfigurable bottom-level acceleration structure is determined based on the geometric metric of the reconfigurable bottom-level acceleration structure and the geometric metric of the adjacent bottom-level acceleration structures in the top-level acceleration structure. Calculate the structural reconstruction score based on the reconstruction cost and structural overlap rate.

[0009] In some embodiments, the method further includes: For each candidate underlying acceleration structure, if the depth of the candidate underlying acceleration structure's tree structure is greater than a preset depth threshold and the dynamic update frequency is zero, but the number of instantiations is greater than or equal to a preset number threshold, the candidate underlying acceleration structure is determined to be a reconfigurable underlying acceleration structure whose score needs to be corrected, and the difference between the number of instantiations and the preset number threshold is determined. If the depth of the candidate underlying acceleration structure is greater than a preset depth threshold and the number of instantiations is less than a preset number threshold, but the dynamic update frequency is not zero, the candidate underlying acceleration structure is determined to be a reconfigurable underlying acceleration structure whose threshold needs to be corrected.

[0010] In some embodiments, a target underlying acceleration structure is determined from at least one reconfigurable underlying acceleration structure based on a structural reconstruction score and a preset score threshold, including: For each reconfigurable underlying acceleration structure whose score needs to be corrected, the score correction coefficient of the reconfigurable underlying acceleration structure is determined based on the difference in the number of iterations. Based on the score correction coefficient and the structural reconstruction score, the corrected reconstruction score of the reconfigurable underlying accelerated structure is determined, and the corrected reconstruction score is less than the structural reconstruction score. If the reconstructed score is greater than the preset score threshold, the reconstructable underlying acceleration structure is determined as the target underlying acceleration structure.

[0011] In some embodiments, a target underlying acceleration structure is determined from at least one reconfigurable underlying acceleration structure based on a structural reconstruction score and a preset score threshold, including: For each reconfigurable underlying acceleration structure to be corrected at a threshold, the dynamic structure scoring threshold corresponding to the dynamic underlying acceleration structure is used as the preset scoring threshold. If the structural reconstruction score is greater than the dynamic structural score threshold, the reconfigurable underlying acceleration structure is determined as the target underlying acceleration structure.

[0012] In some embodiments, determining the reconfiguration cost of the reconfigurable underlying acceleration structure includes: Determine the reconstruction time for splitting and reconstructing the reconstructable underlying acceleration structure, as well as the additional storage usage of multiple sub-underlying acceleration structures compared to the reconstructable underlying acceleration structure, and the synchronization time for multi-threaded synchronization of multiple sub-underlying acceleration structures. Based on preset cost weights, the reconstruction time, additional storage usage, and synchronization time are weighted and summed to obtain the reconstruction cost.

[0013] In some embodiments, the target underlying acceleration structure is decomposed and reconstructed to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure, including: Based on the preset reconstruction hierarchy, at least one candidate reconstruction layer is determined in the target underlying acceleration structure; For each candidate reconstruction layer, the set of nodes constituting the candidate reconstruction layer and the layer volume of the candidate reconstruction layer are determined based on the structural feature information. The target reconstruction layer is determined based on the layer volume corresponding to at least one candidate reconstruction layer; The node set corresponding to the target reconstruction layer is split and reconstructed by node subtree splitting to obtain multiple sub-bottom-level acceleration structures corresponding to the target reconstruction layer.

[0014] In some embodiments, determining the target reconstruction layer based on the layer volume corresponding to at least one candidate reconstruction layer includes: For each candidate reconstruction layer, the volume decay rate of the candidate reconstruction layer is calculated based on the layer volume corresponding to the candidate reconstruction layer and the layer volume corresponding to the upper structural layer of the candidate reconstruction layer in the target bottom acceleration structure. The candidate reconstruction layer whose volume decay rate meets the preset decay constraint is selected as the target reconstruction layer.

[0015] In some embodiments, the method further includes: Each sub-bottom-level acceleration structure is used as a candidate bottom-level acceleration structure. The process returns to obtain the structural feature information corresponding to multiple candidate bottom-level acceleration structures until the overlap attenuation rate after a single split and reconstruction is lower than the preset attenuation threshold, or the number of corresponding layers of the sub-bottom-level acceleration structure after a single split and reconstruction reaches the preset reconstruction level upper limit in the target bottom-level acceleration structure, thus obtaining multiple sub-bottom-level acceleration structures corresponding to the target bottom-level acceleration structure.

[0016] A second aspect of this application provides a low-level accelerated structure reconstruction apparatus, applied to ray tracing, the apparatus comprising: The information acquisition module is used to acquire structural feature information corresponding to multiple candidate underlying acceleration structures; The first screening module is used to select at least one reconfigurable underlying acceleration structure from multiple candidate underlying acceleration structures based on structural feature information. The structure scoring module is used to determine the reconstruction cost of each reconfigurable underlying acceleration structure, and to score the reconfigurable underlying acceleration structure based on the reconstruction cost and structural feature information, thereby determining the corresponding structure reconstruction score of the reconfigurable underlying acceleration structure. The second screening module is used to determine the target underlying acceleration structure from at least one reconfigurable underlying acceleration structure based on the structural reconstruction score and a preset score threshold. The structure reconstruction module is used to split and reconstruct the target underlying acceleration structure to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

[0017] A third aspect of this application provides a graphics processor that includes the underlying acceleration structure reconstruction apparatus described in the second aspect.

[0018] A fourth aspect of this application provides an electronic device, the electronic device comprising: at least one graphics processor; and a memory communicatively connected to the at least one graphics processor; wherein the memory stores a computer program executable by the at least one graphics processor, the computer program being executed by the at least one graphics processor to enable the at least one graphics processor to execute the underlying accelerated structure reconstruction method of the first aspect.

[0019] A fifth aspect of this application provides a computer-readable storage medium storing computer instructions that cause a graphics processor to implement the underlying acceleration structure reconstruction method of the first aspect when executed.

[0020] The underlying acceleration structure reconstruction method provided in this application includes: selecting at least one reconfigurable underlying acceleration structure based on the structural feature information corresponding to each candidate underlying acceleration structure; determining the reconstruction cost of each reconfigurable underlying acceleration structure, and scoring the reconfigurable underlying acceleration structure based on the reconstruction cost and structural feature information to determine a structural reconstruction score; then, determining a target underlying acceleration structure based on the structural reconstruction score and a preset score threshold; and finally, splitting and reconstructing the target underlying acceleration structure to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

[0021] The technical solution provided in this application uses structural feature information of candidate underlying acceleration structures for screening, which can accurately identify reconfigurable underlying acceleration structures and avoid the waste of computing power caused by blind reconstruction. Furthermore, this application can score the reconstruction benefit based on reconstruction cost and feature information, and determine the target underlying acceleration structure according to a preset scoring threshold, ensuring that only underlying acceleration structures whose reconstruction benefit outweighs the cost are reconstructed. This effectively prevents excessive computing power consumption during the reconstruction process and enhances the beneficial impact of the reconstruction process on ray tracing. This application generates multiple sub-underlying acceleration structures with a more compact spatial distribution by splitting and reconstructing the target underlying acceleration structure, significantly reducing the spatial overlap between different instance bounding boxes, reducing the number of redundant node visits and intersection tests during the traversal of the top-level acceleration structure in ray tracing, improving the traversal efficiency of ray tracing, and reducing the waste of computing resources. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating a method for accelerating the reconstruction of a low-level structure, provided in one embodiment of this application; Figure 2 A schematic diagram of a low-level accelerated structure reconstruction device provided in another embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in yet another embodiment of this application. Detailed Implementation

[0024] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0025] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0026] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0027] In the field of graphics processing unit (GPU) technology, ray tracing achieves highly realistic rendering by simulating the interaction between light rays and scene objects. Traversal efficiency is highly dependent on the quality of the acceleration structure construction. The acceleration structure is divided into a low-level acceleration structure that describes the spatial distribution of geometric primitives and a high-level acceleration structure that organizes the position and pose of each instance in the scene. A single low-level acceleration structure can be instantiated into multiple instances with different position and rotation settings.

[0028] In existing ray tracing methods, the traversal cost and intersection test count of the top-level acceleration structure are affected by the spatial relationship between instances. When there are many overlapping regions between the bounding boxes of instances in different bottom-level acceleration structures, the ray traversal process needs to visit a large number of redundant nodes, causing the number of tests to increase exponentially. Most existing solutions address this issue by optimizing the layout of the top-level acceleration structure, such as hierarchical rearrangement, adjusting instance order, and local reconstruction.

[0029] However, if the coverage of the underlying acceleration structure is too large, or the internal node space is discrete, the bounding box volume will be too large. In the top-level acceleration structure, it is very easy to overlap with the corresponding instances of other underlying acceleration structures. In this scenario, the optimization effect of layout optimization methods is limited, and it is difficult to eliminate the invalid traversal caused by the overlap of underlying acceleration structures, which increases the consumption of useless computing power and reduces the efficiency of ray traversal.

[0030] Based on the aforementioned technical problems, embodiments of this application provide a method, apparatus, graphics processor, device, and medium for reconstructing underlying acceleration structures. The method includes: selecting at least one reconstructable underlying acceleration structure based on structural feature information corresponding to each candidate underlying acceleration structure; determining the reconstruction cost of each reconstructable underlying acceleration structure, and scoring the reconstructable underlying acceleration structure based on the reconstruction cost and structural feature information to determine a structural reconstruction score; then, determining a target underlying acceleration structure based on the structural reconstruction score and a preset score threshold; and finally, splitting and reconstructing the target underlying acceleration structure to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

[0031] The technical solution provided in this application is based on the structural feature information of candidate underlying acceleration structures and is used for screening. This allows for accurate identification of reconfigurable underlying acceleration structures, avoiding the waste of computing power caused by blind reconstruction. Furthermore, this application can score the reconstruction benefit based on the reconstruction cost and feature information, and determine the target underlying acceleration structure according to a preset scoring threshold. This ensures that only underlying acceleration structures whose reconstruction benefit outweighs the cost are reconstructed, effectively preventing excessive computing power consumption during the reconstruction process and enhancing the beneficial impact of the reconstruction process on ray tracing. By splitting and reconstructing the target underlying acceleration structure, this application generates multiple sub-underlying acceleration structures with a more compact spatial distribution. This significantly reduces the spatial overlap between bounding boxes of different instances, reduces the number of redundant node visits and intersection tests during the traversal of the top-level acceleration structure in ray tracing, improves the traversal efficiency of ray tracing, and reduces the waste of computing resources.

[0032] The execution entity used in the embodiments of this application can specifically be a graphics processing unit (GPU), or a terminal device that can control the GPU, such as a desktop computer, laptop computer, or server. In addition, the execution entity in the embodiments of this application can also be a software entity, such as a client or software program installed in the GPU or terminal device. The specific type of execution entity corresponding to the underlying acceleration structure reconstruction method, apparatus, GPU, device, and medium provided in the embodiments of this application is not strictly limited here; it can be flexibly selected and set according to the application scenario and actual needs.

[0033] It should be noted that the embodiments provided in this application do not limit the specific application scenarios of the underlying acceleration structure reconstruction method, apparatus, graphics processor, device, and medium provided above. The technical solutions provided in the embodiments of this application can be flexibly applied to various practical application scenarios of ray tracing technology according to actual needs. For example, in the fields of film or animation production, video games, industrial design and architectural visualization, the technical solutions provided in the embodiments of this application can effectively reduce the overlap between instances in the top-level acceleration structure through underlying acceleration structure reconstruction.

[0034] Figure 1 This is a flowchart illustrating a method for accelerating the reconstruction of a low-level structure, as provided in one embodiment of this application.

[0035] like Figure 1 As shown, the underlying accelerated structure reconstruction method provided in this application embodiment includes steps S101 to S105.

[0036] S101: Obtain structural feature information corresponding to multiple candidate underlying acceleration structures.

[0037] S102: Based on structural feature information, select at least one reconfigurable underlying acceleration structure from multiple candidate underlying acceleration structures.

[0038] As shown in steps S101 and S102, in the embodiments provided in this application, structural feature information corresponding to each candidate bottom-level acceleration structure in the top-level acceleration structure to be constructed can be obtained. Furthermore, based on the structural feature information, it can be determined whether each candidate bottom-level acceleration structure can be subsequently split and reconstructed, thus identifying at least one reconstructable bottom-level acceleration structure.

[0039] The structural feature information corresponding to each underlying acceleration structure can include multi-dimensional feature data such as space occupation, structural complexity, reuse in the top-level acceleration structure, and whether it dynamically changes. In the embodiments provided in this application, the structural feature information may include the tree structure depth, instantiation count, and dynamic update frequency of the corresponding candidate underlying acceleration structure. In other embodiments, the structural feature information may also include the geometric metric of the candidate underlying acceleration structure. In some embodiments, the geometric metric can be determined according to the actual structural dimension of the underlying acceleration structure. If the underlying acceleration structure is a two-dimensional structure, the geometric metric can be the surface area; if the underlying acceleration structure is a three-dimensional structure, the geometric metric can be the bounding box volume. The geometric metric can be used in subsequent steps to determine the structural reconstruction score for the reconfigurable underlying acceleration structure.

[0040] The structure tree depth represents the number of levels within the underlying acceleration structure's internal structure tree, from the root node to the leaf node. The instantiation count represents the number of instance pointers corresponding to the underlying acceleration structure within the top-level acceleration structure. The dynamic update frequency represents the frequency with which the objects corresponding to the underlying acceleration structure undergo actions such as position changes, rotations, and scaling; a value of zero indicates that the objects corresponding to the underlying acceleration structure are static objects, while a value greater than zero indicates that the objects corresponding to the underlying acceleration structure are dynamic objects.

[0041] Regarding the specific screening process for reconfigurable underlying acceleration structures, in the embodiments provided in this application, multiple feature constraints that reconfigurable underlying acceleration structures must satisfy can be set according to the application scenario and actual needs. Based on these multiple feature constraints, it is determined whether a candidate underlying acceleration structure can be used as a reconfigurable underlying acceleration structure.

[0042] Specifically, in some embodiments, the reconfigurable underlying acceleration structure may be subject to feature constraints such as whether the depth of the candidate underlying acceleration structure's tree structure is greater than a preset depth threshold, whether the number of instantiations is less than a preset number threshold, and whether the dynamic update frequency is zero.

[0043] In some embodiments, a candidate underlying acceleration structure can be used as a reconfigurable underlying acceleration structure only when it satisfies all feature constraints. That is, when the depth of the candidate underlying acceleration structure's tree structure is greater than a preset depth threshold, the number of instantiations is less than a preset number threshold, and the dynamic update frequency is zero, the candidate underlying acceleration structure can be used as a reconfigurable underlying acceleration structure for further subsequent structure reconstruction scoring and may be reconfigured as a target underlying acceleration structure.

[0044] In some embodiments, the threshold limitation on the depth of the tree structure is because a shallow tree structure results in lower ray tracing gains from reconstructing candidate bottom-level accelerated structures and increases the number of instances in the top-level accelerated structures. The preset depth threshold can be flexibly set, for example, 2 layers, 3 layers, etc.

[0045] The threshold for the number of instantiations is set because if the candidate underlying acceleration structure has too many instances in the top-level acceleration structure, splitting it into multiple sub-underlying acceleration structures after reconstruction would require a corresponding increase in the number of instances in the top-level acceleration structure, significantly increasing the construction cost of the top-level acceleration structure. The preset threshold can be flexibly set, such as 20 times, 15 times, etc.

[0046] The limitation on dynamic update frequency is because if the object corresponding to the candidate underlying acceleration structure is a dynamic object (i.e., the dynamic update frequency is greater than zero), the candidate underlying acceleration structure needs to be reconstructed every frame. If the candidate underlying acceleration structure of a dynamic object is reconstructed into multiple sub-underlying acceleration structures that are also dynamic objects, the graphics processor needs to reconstruct multiple sub-underlying acceleration structures simultaneously every frame, which significantly increases the load and computing power consumption.

[0047] In some embodiments, the reconfigurable underlying acceleration structure can be determined based on the bounding box volume relationship between the underlying acceleration structures in the top-level acceleration structure. Specifically, the bounding box volume of each candidate underlying acceleration structure can be compared with that of the largest underlying acceleration structure in the top-level acceleration structure. While satisfying the above-mentioned multiple feature constraints, if the bounding box volume of the candidate underlying acceleration structure accounts for a significant proportion of the volume of the largest underlying acceleration structure, such as exceeding 50% or 60%, this candidate underlying acceleration structure can be regarded as a reconfigurable underlying acceleration structure. Other feature constraints can also be used in other embodiments, and can be flexibly set according to actual needs and application scenarios.

[0048] In the above embodiments, a triple constraint is set on the candidate underlying acceleration structures, including the depth of the tree structure, the number of instantiations, and the dynamic update frequency. This accurately filters out reconfigurable underlying acceleration structures that are complex, have a low number of instantiations, and are static. This embodiment effectively reduces the computational waste and memory waste caused by subsequent invalid reconstructions, ensuring that the subsequent reconstruction process only applies to underlying acceleration structures with clear optimization potential, maximizing the reconstruction benefits, and significantly improving the overall execution efficiency and resource utilization rationality of the ray tracing process.

[0049] In other embodiments, to further reduce the overlap area and volume of instances in the top-level acceleration structure and improve the traversal efficiency of ray tracing, candidate bottom-level acceleration structures corresponding to objects with a high number of instantiations or dynamic objects can also be used as special reconfigurable bottom-level acceleration structures.

[0050] Specifically, in the embodiments provided in this application, candidate underlying acceleration structures with a structure tree depth greater than a preset depth threshold, a dynamic update frequency of zero, but an instantiation count higher than a preset threshold can be designated as reconfigurable underlying acceleration structures with scores to be corrected. "Score to be corrected" means that when determining the structure reconstruction score for such reconfigurable underlying acceleration structures, the score can be lowered to a certain extent based on the instantiation count to reduce the reconstruction priority of such structures, while not completely eliminating the reconstruction opportunity, thereby minimizing instance overlap and improving the traversal efficiency of ray tracing.

[0051] Meanwhile, for candidate underlying acceleration structures whose tree structure depth is greater than a preset depth threshold and whose instantiation count is less than a preset count threshold, but whose dynamic update frequency is not zero, they can be considered as reconfigurable underlying acceleration structures with thresholds to be corrected. "Thresholds to be corrected" means that when determining the target underlying acceleration structure based on the structure reconstruction score for such reconfigurable underlying acceleration structures, a higher scoring threshold can be selected to determine whether such structures can be used as target underlying acceleration structures. This allows for a more conservative reconstruction process for such structures while not completely eliminating their reconstruction opportunities, minimizing instance overlap, and improving ray traversal efficiency.

[0052] In the above embodiments, targeted refactoring decisions are implemented for underlying acceleration structures with high instantiation counts or dynamic objects. Lowering the score for structures with high instantiation counts effectively prevents the expansion of refactoring overhead caused by large-scale instance updates in the top-level acceleration structure. A higher score threshold is used for dynamic objects to avoid reduced refactoring benefits due to frequent updates. This embodiment effectively expands the identification range of refactorable underlying acceleration structures and achieves precise control over potential refactoring benefit risks, maintaining a stable balance between refactoring coverage and execution security.

[0053] S103: For each reconfigurable underlying acceleration structure, determine the reconfiguration cost of the reconfigurable underlying acceleration structure, and based on the reconfiguration cost and structural feature information, perform a reconfiguration score on the reconfigurable underlying acceleration structure to determine the corresponding structural reconfiguration score.

[0054] As shown in step S103, in the embodiments provided in this application, the reconstruction cost can represent the estimated additional overhead and storage overhead required to perform split reconstruction on the reconfigurable underlying acceleration structure. The structure reconstruction score can represent the difference between the expected benefit of performing split reconstruction on the reconfigurable underlying acceleration structure and the reconstruction cost; the higher the score, the more worthwhile it is to perform reconfiguration and splitting on the reconfigurable underlying acceleration structure.

[0055] In the embodiments provided in this application, the reconstruction cost of each reconfigurable underlying acceleration structure can be comprehensively determined from multiple dimensions. Specifically, for each reconfigurable underlying acceleration structure, the reconstruction time required to split and reconstruct the reconfigurable underlying acceleration structure can be estimated, as can the additional storage occupation of the multiple sub-underlying acceleration structures obtained after splitting and reconstructing the reconfigurable underlying acceleration structure compared to the unreconstructed reconfigurable underlying acceleration structure, and the synchronization time consumed when performing multi-threaded synchronization of the multiple sub-underlying acceleration structures after the reconfigurable underlying acceleration structure is reconstructed. Thus, estimated overhead data in three dimensions—reconstruction time, additional storage occupation, and synchronization time—are obtained.

[0056] Furthermore, based on preset cost weights, the reconstruction time, additional storage usage, and synchronization time corresponding to the reconfigurable underlying acceleration structure can be weighted and summed to obtain the reconstruction cost corresponding to the reconfigurable underlying acceleration structure.

[0057] The above embodiments, by quantifying and weighting the reconstruction time, additional storage usage, and multi-threaded synchronization time overhead, achieve an accurate prediction of the total cost required to decompose and reconstruct the reconstructable underlying accelerated structure. This ensures that the structure reconstruction score determined in this way can truly reflect the relationship between reconstruction benefits and costs. This embodiment effectively avoids invalid reconstruction or resource overruns that may be caused by distorted reconstruction cost estimation, ensuring that the reconstruction process is always executed within a controllable memory and latency budget. This effectively guarantees the stability and efficiency of ray tracing, as well as the reliability of resource scheduling during the reconstruction process.

[0058] Regarding the specific calculation process of the above-mentioned structural reconstruction score, in the embodiments provided in this application, for each reconfigurable bottom-level acceleration structure, the structural overlap rate of the reconfigurable bottom-level acceleration structure can be calculated based on the geometric measure (structural surface area or bounding box volume) in the structural feature information of the reconfigurable bottom-level acceleration structure mentioned in the above embodiments, and the geometric measure of the adjacent bottom-level acceleration structures of the reconfigurable bottom-level acceleration structure in the top-level acceleration structure.

[0059] The structural overlap rate can specifically be the proportion of the overlapping area or volume in the reconfigurable bottom-level acceleration structure. If this reconfigurable bottom-level acceleration structure is used as the target bottom-level acceleration structure for splitting and reconstruction, the structural overlap rate can be significantly reduced. Therefore, the magnitude of the structural overlap rate can be used to characterize the traversal efficiency gain brought by the reconstruction process to the top-level acceleration structure. In some embodiments, the structural overlap rate can also be calculated directly based on the geometric metrics of each candidate bottom-level acceleration structure when obtaining the structural feature information.

[0060] Then, a structural reconstruction score for the reconfigurable underlying accelerated structure can be calculated based on the reconstruction cost and the structural overlap rate. In some embodiments, the structural reconstruction score can be obtained by subtracting the reconstruction cost, which represents the loss in the reconstruction process, from the structural overlap rate, which represents the traversal efficiency gain.

[0061] In some embodiments, the calculation process of the structural reconstruction score can refer to the following formula (1): (1) in, Indicates the structural reconstruction score. Indicates the structural overlap rate. , as well as These represent the reconstruction time, additional storage usage, and synchronization time, respectively. , as well as These are the reconstruction times respectively. Additional storage usage and synchronization time The corresponding preset cost weight. This refers to the estimated reconstruction cost.

[0062] In the above embodiments, the structural overlap rate is calculated based on geometric metrics, and a precise score is given in conjunction with the reconstruction cost. This accurately quantifies the degree of traversal redundancy caused by structural overlap in the reconfigurable lower-level acceleration structure, and the scoring results accurately reflect the expected actual benefits of splitting and reconstructing. This embodiment ensures that reconstruction resources are always prioritized for lower-level acceleration structures with severe overlap and controllable reconstruction costs, significantly improving the traversal efficiency of the top-level acceleration structure during ray tracing.

[0063] S104: Determine the target underlying acceleration structure from at least one reconfigurable underlying acceleration structure based on the structural reconstruction score and the preset score threshold.

[0064] As shown in step S104, in the embodiments provided in this application, for each reconfigurable underlying acceleration structure, it can be determined whether the reconfigurable underlying acceleration structure can be used as a target underlying acceleration structure based on the relationship between its corresponding structural reconstruction score and a preset score threshold, thereby performing subsequent splitting and reconstruction processing. In some embodiments, reconfigurable underlying acceleration structures with structural reconstruction scores greater than the preset score threshold can be used as the aforementioned target underlying acceleration structures.

[0065] In addition, in some embodiments, for the reconfigurable underlying acceleration structure with a score to be corrected as mentioned in the above embodiments, before determining whether such a structure can be used as a target underlying acceleration structure, the structure reconstruction score can be corrected based on the difference between the number of instantiations exceeding the threshold and the preset threshold, thus determining the corrected reconstruction score. Further, based on the corrected reconstruction score and the preset score threshold, it is determined whether the corresponding reconfigurable underlying acceleration structure can be used as the target underlying acceleration structure.

[0066] Specifically, in the embodiments provided in this application, the scoring correction coefficient corresponding to the reconfigurable underlying acceleration structure whose scoring needs to be corrected can be determined based on the aforementioned difference in the number of times. The scoring correction coefficient can represent the degree of attenuation of reconstruction benefits caused by large-scale instance updates after reconstruction due to a higher number of instantiations. It is inversely proportional to the difference in the number of times; the larger the difference in the number of times, the smaller the scoring correction coefficient. The value can be between 0 and 1.

[0067] Then, the score correction factor can be multiplied by the structural refactoring score to calculate the corrected refactoring score for this type of refactorable underlying accelerated structure. The corrected refactoring score is the original structural refactoring score minus the refactoring benefit loss caused by the high number of instantiations, so the corrected refactoring score will be less than the structural refactoring score.

[0068] For reconfigurable underlying acceleration structures after score correction, the determination of whether they can serve as target underlying acceleration structures can be based on the relationship between the corrected reconstruction score and a preset score threshold. That is, when the corrected reconstruction score is greater than the preset score threshold, it can be determined that this reconfigurable underlying acceleration structure can serve as the target underlying acceleration structure.

[0069] The above embodiments can implement flexible penalties for underlying acceleration structures that exceed the threshold in the number of instantiations but perform well in other features. By adjusting the reconstruction score, the actual reconstruction risk can be objectively reflected. This embodiment effectively avoids missing out on underlying acceleration structures with high reconstruction value due to hard exclusion, while preventing highly reusable structures from occupying reconstruction resources and generating effective benefits due to inflated reconstruction scores. It ensures that reconstruction resources are prioritized for underlying acceleration structures with controllable instance impact and stable reconstruction benefits.

[0070] In other embodiments, for the reconfigurable underlying acceleration structure with threshold to be corrected provided in the above embodiments, the objects corresponding to such structures are dynamic objects. Considering that dynamic objects need to be reconstructed every frame, and the reconstruction will further increase the load on the graphics processor, a more conservative scoring threshold can be set for the underlying acceleration structure of dynamic objects to fully consider the increased load that dynamic objects may bring after reconstruction.

[0071] Specifically, in the embodiments provided in this application, when determining whether a reconfigurable underlying acceleration structure with a threshold to be corrected can be used as a target underlying acceleration structure, the preset dynamic structure scoring threshold for dynamic underlying acceleration structures (i.e., the underlying acceleration structure corresponding to dynamic objects) can be used as the scoring threshold for the reconfigurable underlying acceleration structure with a threshold to be corrected.

[0072] Then, when the structural reconstruction score is greater than the dynamic structural score threshold, the corresponding reconfigurable underlying acceleration structure can be used as the target underlying acceleration structure for subsequent splitting and reconstruction.

[0073] In the above embodiments, a higher dynamic structure scoring threshold is set independently for the underlying acceleration structure of dynamic objects. This enables stricter reconstruction eligibility judgment for dynamically changing but otherwise well-performing underlying acceleration structures. This effectively reduces the decay and fluctuation of reconstruction benefits caused by frequent updates of dynamic objects, while preserving reconstruction opportunities for dynamic objects and improving the adaptability and robustness of reconstructing dynamic objects.

[0074] In other embodiments, besides determining the target bottom-level acceleration structure based on the structural reconstruction score and score threshold, the overlapping area (or volume) between each candidate bottom-level acceleration structure and its adjacent bottom-level acceleration structure, or the number of overlapping triangles (leaf nodes of the bottom-level acceleration structure), can be determined directly based on structural feature information to see if it exceeds a preset threshold. If it does, the candidate bottom-level acceleration structure is then split and reconstructed as the target bottom-level acceleration structure. The specific determination process for the target bottom-level acceleration structure can be flexibly selected according to actual needs and application scenarios.

[0075] S105: Decompose and reconstruct the target underlying acceleration structure to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

[0076] As shown in step S105, in the embodiments provided in this application, at least one candidate reconstruction layer can be determined from the target underlying structure, and based on the layer volume, the target reconstruction layer with the greatest reconstruction benefit is determined from the candidate reconstruction layers. The target reconstruction layer is then split into node subtrees, dividing the target underlying structure into multiple sub-underlying acceleration structures.

[0077] Specifically, in some embodiments, at least one candidate reconstruction layer can be determined from the target underlying acceleration structure based on a preset reconstruction level. The preset reconstruction level can represent a subset of levels in the underlying acceleration structure that are suitable for splitting and reconstruction, such as layers 3 to 5, or layers 4 to 6, in the direction from the root node to the leaf node. It should be noted that the leaf nodes of the target underlying acceleration structure cannot be considered candidate reconstruction layers because they are already at the lowest level and cannot be split and reconstructed.

[0078] Then, for each candidate structure layer, based on the structural feature information corresponding to the target bottom acceleration structure, the nodes contained in the candidate structure layer can be determined, the node set can be obtained, and the layer volume of the candidate structure layer can be determined.

[0079] In some embodiments, the structural feature information may also include the distribution and relationships of nodes in the underlying acceleration structure, as well as the bounding box volume of each node. All nodes in a single candidate structure layer constitute a node set, and the bounding box volume of the node set can be used as the layer volume of that candidate structure layer.

[0080] Furthermore, based on the layer volume corresponding to each candidate reconstruction layer, the candidate reconstruction layer with the optimal reconstruction benefit can be determined as the target reconstruction layer. In some embodiments, the target reconstruction layer can be determined based on the degree of volume reduction between upper and lower layers.

[0081] Then, for the target reconstruction layer, the subtrees with nodes in the node set as root nodes are split into sub-bottom-level acceleration structures, thereby splitting the target bottom-level acceleration structure into multiple sub-bottom-level acceleration structures, the number of which is the same as the number of nodes in the target reconstruction layer.

[0082] In the above embodiments, candidate reconstruction layers are determined based on preset reconstruction levels, and target reconstruction layers are selected by combining layer volume. This allows for precise positioning of the layer with the greatest reconstruction benefit in the target bottom-level acceleration structure. This embodiment enables the reconstructed sub-bottom-level acceleration structures to have more compact bounding boxes and lower spatial redundancy. Under controllable storage and computational overhead, it minimizes instance overlap and redundant traversal accesses in the top-level acceleration structure, significantly improving the intersection testing efficiency of ray tracing and overall rendering performance.

[0083] Regarding the specific process of determining the target reconstruction layer, in the embodiments provided in this application, the candidate reconstruction layer with the most significant volume decay can be selected as the target reconstruction layer based on the volume decay between adjacent layers.

[0084] Specifically, in some embodiments, for each candidate reconstruction layer, the volume decay rate from the upper structural layer to the candidate reconstruction layer is calculated based on the layer volume corresponding to the candidate reconstruction layer and the layer volume of the upper structural layer corresponding to the candidate reconstruction layer in the target bottom acceleration structure.

[0085] Then, the candidate reconstruction layer with the largest volume decay rate can be determined from each candidate reconstruction layer as the target reconstruction layer mentioned above, so as to minimize the overlap rate between instances in the top-level acceleration structure, while maximizing the efficiency improvement of the subsequent split reconstruction process on the ray tracing process.

[0086] In the above embodiments, based on the volume decay rate between different layers, the optimal reconstruction layer in the target bottom-level acceleration structure where the volume of the upper and lower layers changes significantly is accurately located, ensuring that the splitting and reconstruction operation occurs at the layer with the best benefit. This embodiment effectively avoids the situation where the volume of the sub-bottom-level acceleration structure is still too large due to selecting too shallow a layer, or the sub-bottom-level acceleration structure is too fragmented or the number of instances surges due to selecting too deep a layer. This allows the split sub-bottom-level acceleration structures to achieve an optimal balance between spatial compactness and quantity, maximizing traversal efficiency benefits within a limited reconstruction budget.

[0087] In some embodiments, to further improve the traversal efficiency of the ray tracing process, multiple rounds of differential reconstruction can be performed on the target underlying acceleration structure to maximize the reconstruction benefits. Specifically, in the embodiments provided in this application, after reconstructing the target underlying acceleration structure into multiple corresponding sub-underlying acceleration structures through the above-mentioned multiple embodiments, each sub-underlying acceleration structure can be used as a candidate underlying acceleration structure in step S101, and the contents of steps S101 to S105 can be executed.

[0088] Furthermore, the sub-sub-bottom acceleration structure corresponding to the sub-sub-bottom acceleration structure can be reconstructed, and the sub-sub-bottom acceleration structure reconstruction method provided in this application can be executed again for the sub-sub-bottom acceleration structure, repeating multiple rounds to maximize the reconstruction benefits and significantly improve the traversal efficiency of ray tracing.

[0089] The termination conditions for multiple rounds of reconstruction can be flexibly set according to actual needs and application scenarios. In the embodiments provided in this application, the overlap rate between multiple sub-bottom-level acceleration structures and adjacent bottom-level acceleration structures after each reconstruction can be statistically analyzed, and the ratio of this overlap rate to the overlap rate of the target bottom-level acceleration structure before reconstruction can be calculated as the overlap attenuation rate. When the overlap attenuation rate is lower than a preset attenuation threshold, it indicates that the benefit after this round of reconstruction is low, and further reconstruction is unnecessary, so the reconstruction can be terminated.

[0090] Alternatively, in some embodiments, when the sub-bottom-level acceleration structure after a single split and reconstruction reaches a preset upper limit for the corresponding layer in the initial target bottom-level acceleration structure, the reconstruction can be terminated. The preset upper limit for the reconstruction layer can be flexibly set, such as the second-to-last layer or the third-to-last layer of the tree structure depth. Alternatively, in other embodiments, when the reconstruction cost exceeds a preset cost threshold, subsequent reconstruction processing can be stopped in step S102.

[0091] In the above embodiments, the sub-bottom-level acceleration structure is re-selected as a candidate structure and the reconstruction process is executed iteratively. This enables multi-round progressive splitting and reconstruction of complex and large-volume bottom-level acceleration structures. Through the iterative reconstruction process of this embodiment, the structural overlap problem in the top-level acceleration structure is continuously resolved in the layer-by-layer refinement. Compared with a single split, it can adapt to deeper and more complex structural distributions, significantly improving the traversal efficiency of the top-level acceleration structure, and achieving better performance gains with controllable reconstruction overhead.

[0092] The multiple sub-bottom-level acceleration structures obtained from the reconstruction and splitting can be used together with other unreconstructed bottom-level acceleration structures for the reconstruction of the top-level acceleration structure. Since the target bottom-level acceleration structure is reconstructed into multiple sub-bottom-level acceleration structures, the bounding box volume and structural overlap rate are significantly reduced. The reconstructed top-level acceleration structure exhibits more accurate and effective intersection points between rays and instances during intersection testing, effectively reducing invalid intersection nodes, significantly improving traversal efficiency, enhancing the execution efficiency of the ray tracing process, and saving computational resources. In some embodiments, bottom-level acceleration structures that are not used as reconstructable bottom-level acceleration structures or target bottom-level acceleration structures, or whose volume decay rates are low across all candidate reconstruction layers, can directly participate in the reconstruction of the top-level acceleration structure based on the original complete structure.

[0093] The above describes the specific implementation of the underlying acceleration structure reconstruction method provided in this application. The technical solution provided in this application is based on the structural feature information of candidate underlying acceleration structures and is used for screening. This can accurately identify reconstructable underlying acceleration structures and avoid the waste of computing power caused by blind reconstruction. Furthermore, this application can score the reconstruction benefit based on the reconstruction cost and feature information, and determine the target underlying acceleration structure according to a preset scoring threshold. This ensures that only underlying acceleration structures whose reconstruction benefit outweighs the cost are reconstructed, effectively preventing excessive computing power consumption during the reconstruction process and enhancing the beneficial impact of the reconstruction process on ray tracing. By splitting and reconstructing the target underlying acceleration structure, this application generates multiple sub-underlying acceleration structures with a more compact spatial distribution. This significantly reduces the spatial overlap between different instance bounding boxes, reduces the number of redundant node visits and intersection tests during the traversal of the top-level acceleration structure in the ray tracing process, improves the traversal efficiency of ray tracing, and reduces the waste of computing power resources.

[0094] Based on the same technical concept as the aforementioned graphics processor, this application also provides a low-level acceleration structure reconstruction device, which is applied to ray tracing. For details, please refer to... Figure 2 The structure diagram is shown below.

[0095] Figure 2 This is a schematic diagram of a low-level accelerated structure reconstruction device provided in another embodiment of this application.

[0096] like Figure 2 As shown, the underlying accelerated structure reconstruction device 200 provided in this application embodiment includes: Information acquisition module 201 is used to acquire structural feature information corresponding to multiple candidate underlying acceleration structures; The first screening module 202 is used to screen at least one reconfigurable underlying acceleration structure from multiple candidate underlying acceleration structures based on structural feature information. The structure scoring module 203 is used to determine the reconstruction cost of each reconfigurable underlying acceleration structure, and to score the reconfigurable underlying acceleration structure based on the reconstruction cost and structural feature information, thereby determining the corresponding structure reconstruction score of the reconfigurable underlying acceleration structure. The second screening module 204 is used to determine the target underlying acceleration structure from at least one reconfigurable underlying acceleration structure based on the structural reconstruction score and a preset score threshold. The structure reconstruction module 205 is used to split and reconstruct the target underlying acceleration structure to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

[0097] In some embodiments, for each candidate underlying acceleration structure, the structural feature information includes the depth of the candidate underlying acceleration structure tree, the number of instantiations, and the dynamic update frequency. The first filtering module 202 mentioned above is used for: For each candidate underlying acceleration structure, if the depth of the candidate underlying acceleration structure's structure tree is greater than a preset depth threshold, the number of instantiations is less than a preset number of times threshold, and the dynamic update frequency is zero, the candidate underlying acceleration structure is determined to be a reconfigurable underlying acceleration structure.

[0098] In some embodiments, for each candidate bottom-level acceleration structure, the structural feature information also includes a geometric metric of the candidate bottom-level acceleration structure, which is the surface area of ​​the structure or the bounding box volume. The above-mentioned structural scoring module 203 is used for: For each reconfigurable bottom-level acceleration structure, the structural overlap rate of the reconfigurable bottom-level acceleration structure is determined based on the geometric metric of the reconfigurable bottom-level acceleration structure and the geometric metric of the adjacent bottom-level acceleration structures in the top-level acceleration structure. Calculate the structural reconstruction score based on the reconstruction cost and structural overlap rate.

[0099] In some embodiments, the first screening module 202 described above is used for: For each candidate underlying acceleration structure, if the depth of the candidate underlying acceleration structure's tree structure is greater than a preset depth threshold and the dynamic update frequency is zero, but the number of instantiations is greater than or equal to a preset number threshold, the candidate underlying acceleration structure is determined to be a reconfigurable underlying acceleration structure whose score needs to be corrected, and the difference between the number of instantiations and the preset number threshold is determined. If the depth of the candidate underlying acceleration structure is greater than a preset depth threshold and the number of instantiations is less than a preset number threshold, but the dynamic update frequency is not zero, the candidate underlying acceleration structure is determined to be a reconfigurable underlying acceleration structure whose threshold needs to be corrected.

[0100] In some embodiments, the second screening module 204 described above is used for: For each reconfigurable underlying acceleration structure whose score needs to be corrected, the score correction coefficient of the reconfigurable underlying acceleration structure is determined based on the difference in the number of iterations. Based on the score correction coefficient and the structural reconstruction score, the corrected reconstruction score of the reconfigurable underlying accelerated structure is determined, and the corrected reconstruction score is less than the structural reconstruction score. If the reconstructed score is greater than the preset score threshold, the reconstructable underlying acceleration structure is determined as the target underlying acceleration structure.

[0101] In some embodiments, the second screening module 204 described above is used for: For each reconfigurable underlying acceleration structure to be corrected at a threshold, the dynamic structure scoring threshold corresponding to the dynamic underlying acceleration structure is used as the preset scoring threshold. If the structural reconstruction score is greater than the dynamic structural score threshold, the reconfigurable underlying acceleration structure is determined as the target underlying acceleration structure.

[0102] In some embodiments, the above-described structure scoring module 203 is used for: Determine the reconstruction time for splitting and reconstructing the reconstructable underlying acceleration structure, as well as the additional storage usage of multiple sub-underlying acceleration structures compared to the reconstructable underlying acceleration structure, and the synchronization time for multi-threaded synchronization of multiple sub-underlying acceleration structures. Based on preset cost weights, the reconstruction time, additional storage usage, and synchronization time are weighted and summed to obtain the reconstruction cost.

[0103] In some embodiments, the above-described structural reconstruction module 205 is used for: Based on the preset reconstruction hierarchy, at least one candidate reconstruction layer is determined in the target underlying acceleration structure; For each candidate reconstruction layer, the set of nodes constituting the candidate reconstruction layer and the layer volume of the candidate reconstruction layer are determined based on the structural feature information. The target reconstruction layer is determined based on the layer volume corresponding to at least one candidate reconstruction layer; The node set corresponding to the target reconstruction layer is split and reconstructed by node subtree splitting to obtain multiple sub-bottom-level acceleration structures corresponding to the target reconstruction layer.

[0104] In some embodiments, the above-described structural reconstruction module 205 is used for: For each candidate reconstruction layer, the volume decay rate of the candidate reconstruction layer is calculated based on the layer volume corresponding to the candidate reconstruction layer and the layer volume corresponding to the upper structural layer of the candidate reconstruction layer in the target bottom acceleration structure. The candidate reconstruction layer whose volume decay rate meets the preset decay constraint is selected as the target reconstruction layer.

[0105] In some embodiments, the above-described structural reconstruction module 205 is used for: Each sub-bottom-level acceleration structure is used as a candidate bottom-level acceleration structure. The process returns to obtain the structural feature information corresponding to multiple candidate bottom-level acceleration structures until the overlap attenuation rate after a single split and reconstruction is lower than the preset attenuation threshold, or the number of corresponding layers of the sub-bottom-level acceleration structure after a single split and reconstruction reaches the preset reconstruction level upper limit in the target bottom-level acceleration structure, thus obtaining multiple sub-bottom-level acceleration structures corresponding to the target bottom-level acceleration structure.

[0106] Based on the same technical concept, this application embodiment also provides a graphics processor, which includes the aforementioned underlying acceleration structure reconstruction device 200.

[0107] Figure 3This is a schematic diagram of the structure of an electronic device provided in yet another embodiment of this application.

[0108] Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present application described and / or claimed herein.

[0109] like Figure 3 As shown, the electronic device 300 includes at least one graphics processor 301 and a memory, such as a read-only memory 302 (ROM) or a random access memory 303 (RAM), communicatively connected to the graphics processor 301. The memory stores computer programs executable by the graphics processor 301. The graphics processor 301 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 302 or loaded from the storage unit 308 into the random access memory 303. The random access memory 303 can also store various programs and data required for the operation of the electronic device 300. The graphics processor 301, the read-only memory 302, and the random access memory 303 are interconnected via a bus 304. An input / output interface 305 (I / O) is also connected to the bus 304.

[0110] Multiple components in electronic device 300 are connected to input / output interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of monitors, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0111] The graphics processor 301 can be any dedicated graphics processing component with processing and computing capabilities. The graphics processor 301 executes the various methods and processes described above, such as the key-value cache read / write method mentioned above.

[0112] In some embodiments, the key-value cache read / write method described above can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 300 via read-only memory 302 and / or communication unit 309. When the computer program is loaded into random access memory 303 and executed by graphics processor 301, one or more steps of the key-value cache read / write method described above can be performed. Alternatively, in other embodiments, graphics processor 301 can be configured to perform the key-value cache read / write method described above by any other suitable means (e.g., by means of firmware).

[0113] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: implementation in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable graphics processor, which may be an application-specific programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0114] Computer programs used to implement the methods of this application may be written in any combination of one or more programming languages. These computer programs may be provided to the graphics processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the graphics processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0115] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory (EPROM), optical fiber, compact disc read-only memory (CD ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0116] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0117] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include: Local Area Networks (LANs), Wide Area Networks (WANs), blockchain networks, and the Internet.

[0118] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product within the cloud computing service system. It addresses the shortcomings of traditional physical hosts and Virtual Private Server (VPS) services, such as high management difficulty and weak business scalability.

[0119] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.

[0120] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for accelerating the reconstruction of a low-level structure, characterized in that, The method is applied to ray tracing, and the method includes: Obtain structural feature information corresponding to multiple candidate underlying acceleration structures; Based on the structural feature information, at least one reconfigurable underlying acceleration structure is selected from the plurality of candidate underlying acceleration structures. For each of the reconfigurable underlying acceleration structures, the reconfiguration cost of the reconfigurable underlying acceleration structure is determined, and a reconfiguration score is given to the reconfigurable underlying acceleration structure based on the reconfiguration cost and the structural feature information, thereby determining the structural reconfiguration score corresponding to the reconfigurable underlying acceleration structure. Based on the structural reconstruction score and the preset score threshold, a target underlying acceleration structure is determined from the at least one reconfigurable underlying acceleration structure; The target underlying acceleration structure is split and reconstructed to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

2. The method according to claim 1, characterized in that, For each candidate underlying acceleration structure, the structural feature information includes the tree structure depth, instantiation count, and dynamic update frequency of the candidate underlying acceleration structure; Based on the structural feature information, at least one reconfigurable underlying acceleration structure is selected from the plurality of candidate underlying acceleration structures, including: For each candidate underlying acceleration structure, if the depth of the candidate underlying acceleration structure is greater than a preset depth threshold, the number of instantiations is less than a preset number threshold, and the dynamic update frequency is zero, the candidate underlying acceleration structure is determined to be the reconfigurable underlying acceleration structure.

3. The method according to claim 2, characterized in that, For each of the candidate bottom-level acceleration structures, the structural feature information also includes the geometric metric of the candidate bottom-level acceleration structure, which is the surface area of ​​the structure or the bounding box volume; Based on the reconstruction cost and the structural feature information, a reconstruction score is calculated for the reconfigurable underlying acceleration structure to determine the corresponding structural reconstruction score, including: For each of the reconfigurable bottom-level acceleration structures, the structural overlap rate of the reconfigurable bottom-level acceleration structure is determined based on the geometric metric of the reconfigurable bottom-level acceleration structure and the geometric metric of the adjacent bottom-level acceleration structures in the top-level acceleration structure. The structural reconstruction score is calculated based on the reconstruction cost and the structural overlap rate.

4. The method according to claim 2, characterized in that, The method further includes: For each candidate underlying acceleration structure, if the depth of the tree structure of the candidate underlying acceleration structure is greater than the preset depth threshold, the dynamic update frequency is zero, but the number of instantiations is greater than or equal to the preset number threshold, the candidate underlying acceleration structure is determined to be the reconfigurable underlying acceleration structure whose score needs to be corrected, and the difference between the number of instantiations and the preset number threshold is determined. If the depth of the tree structure of the candidate underlying acceleration structure is greater than the preset depth threshold, and the number of instantiations is less than the preset number of times threshold, but the dynamic update frequency is not zero, then the candidate underlying acceleration structure is determined to be the reconfigurable underlying acceleration structure with threshold correction.

5. The method according to claim 4, characterized in that, Based on the structural reconstruction score and a preset scoring threshold, a target underlying acceleration structure is determined from the at least one reconfigurable underlying acceleration structure, including: For each reconfigurable underlying acceleration structure whose score needs to be corrected, a score correction coefficient for the reconfigurable underlying acceleration structure is determined based on the difference in the number of times. Based on the scoring correction coefficient and the structural reconstruction score, the corrected reconstruction score of the reconfigurable underlying acceleration structure is determined, wherein the corrected reconstruction score is less than the structural reconstruction score; If the corrected reconstructed score is greater than the preset score threshold, the reconstructable underlying acceleration structure is determined to be the target underlying acceleration structure.

6. The method according to claim 4, characterized in that, Based on the structural reconstruction score and a preset scoring threshold, a target underlying acceleration structure is determined from the at least one reconfigurable underlying acceleration structure, including: For each reconfigurable underlying acceleration structure to be corrected at a threshold, the dynamic structure scoring threshold corresponding to the dynamic underlying acceleration structure is used as the preset scoring threshold. If the structural reconstruction score is greater than the dynamic structural score threshold, the reconfigurable underlying acceleration structure is determined to be the target underlying acceleration structure.

7. The method according to claim 1, characterized in that, Determining the reconfigurable underlying acceleration structure's reconfiguration cost includes: Determine the reconstruction time for splitting and reconstructing the reconfigurable underlying acceleration structure, determine the additional storage usage of the multiple sub-underlying acceleration structures compared to the reconfigurable underlying acceleration structure, and determine the synchronization time for multi-threaded synchronization of the multiple sub-underlying acceleration structures. The reconstruction cost is obtained by weighting and summing the reconstruction time, the additional storage usage, and the synchronization time according to the preset cost weights.

8. The method according to claim 1, characterized in that, The target underlying acceleration structure is decomposed and reconstructed to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure, including: Based on the preset reconstruction level, at least one candidate reconstruction layer in the target bottom-level acceleration structure is determined; For each candidate reconstruction layer, the set of nodes constituting the candidate reconstruction layer and the layer volume of the candidate reconstruction layer are determined based on the structural feature information. The target reconstruction layer is determined based on the layer volume corresponding to the at least one candidate reconstruction layer; The node set corresponding to the target reconstruction layer is split and reconstructed by node subtree splitting to obtain multiple sub-bottom layer acceleration structures corresponding to the target reconstruction layer.

9. The method according to claim 8, characterized in that, Determining the target reconstruction layer based on the layer volume corresponding to the at least one candidate reconstruction layer includes: For each candidate reconstruction layer, the volume decay rate of the candidate reconstruction layer is calculated based on the layer volume corresponding to the candidate reconstruction layer and the layer volume corresponding to the upper structural layer of the candidate reconstruction layer in the target bottom acceleration structure. The candidate reconstruction layer whose volume decay rate meets the preset decay constraint is selected as the target reconstruction layer.

10. The method according to claim 8, characterized in that, The method further includes: Each of the sub-bottom-level acceleration structures is used as a candidate bottom-level acceleration structure. The process of obtaining the structural feature information corresponding to multiple candidate bottom-level acceleration structures is repeated until the overlap attenuation rate after a single split and reconstruction is lower than a preset attenuation threshold, or the number of corresponding layers of the sub-bottom-level acceleration structure after a single split and reconstruction reaches the preset reconstruction level upper limit in the target bottom-level acceleration structure, thereby obtaining multiple sub-bottom-level acceleration structures corresponding to the target bottom-level acceleration structure.

11. A device for reconfiguring a bottom-level accelerated structure, characterized in that, The device is used for ray tracing, and the device includes: The information acquisition module is used to acquire structural feature information corresponding to multiple candidate underlying acceleration structures; The first screening module is used to select at least one reconfigurable underlying acceleration structure from the plurality of candidate underlying acceleration structures based on the structural feature information. The structure scoring module is used to determine the reconstruction cost of each reconfigurable underlying acceleration structure, and to score the reconfigurable underlying acceleration structure based on the reconstruction cost and the structure feature information, thereby determining the structure reconstruction score corresponding to the reconfigurable underlying acceleration structure. The second filtering module is used to determine the target underlying acceleration structure from the at least one reconfigurable underlying acceleration structure based on the structure reconstruction score and the preset score threshold. The structure reconstruction module is used to split and reconstruct the target underlying acceleration structure to obtain multiple sub-underlying acceleration structures corresponding to the target underlying acceleration structure.

12. A graphics processor, characterized in that, Includes the underlying accelerated structure reconstruction device as described in claim 11.

13. An electronic device, characterized in that, The electronic device includes: at least one graphics processor; and a memory communicatively connected to the at least one graphics processor; wherein the memory stores a computer program executable by the at least one graphics processor, the computer program being executed by the at least one graphics processor to enable the at least one graphics processor to perform the underlying accelerated structure reconstruction method according to any one of claims 1-10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the graphics processor to implement the underlying acceleration structure reconstruction method according to any one of claims 1-10 when executed.

Images