Jiafeng Zhu, Rongcai Zhao, Lin Han, Yunlong Hao, Lili Bai
{"title":"SIMD扩展导出分支的矢量化方法","authors":"Jiafeng Zhu, Rongcai Zhao, Lin Han, Yunlong Hao, Lili Bai","doi":"10.1109/ICIS.2011.49","DOIUrl":null,"url":null,"abstract":"The main popular vectorization methods for the SIMD extension dig the parallelism of the programs relying on the compiler's data dependence analysis. But the data dependence analysis can not deal with the non-structured control flow statements. Therefore, the up-to-date compilers are extremely limited to vectorize these statements. Here is a vectorization method of the export branch for the SIMD extension, which can automatically and effectively vectorize the export branch within the vector length. And the results of performance test show that this method can both fully ensure the semantic correctness of the control flow and exploit the parallelism of the data flow.","PeriodicalId":256762,"journal":{"name":"2011 10th IEEE/ACIS International Conference on Computer and Information Science","volume":"59 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2011-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"A Vectorization Method of Export Branch for SIMD Extension\",\"authors\":\"Jiafeng Zhu, Rongcai Zhao, Lin Han, Yunlong Hao, Lili Bai\",\"doi\":\"10.1109/ICIS.2011.49\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The main popular vectorization methods for the SIMD extension dig the parallelism of the programs relying on the compiler's data dependence analysis. But the data dependence analysis can not deal with the non-structured control flow statements. Therefore, the up-to-date compilers are extremely limited to vectorize these statements. Here is a vectorization method of the export branch for the SIMD extension, which can automatically and effectively vectorize the export branch within the vector length. And the results of performance test show that this method can both fully ensure the semantic correctness of the control flow and exploit the parallelism of the data flow.\",\"PeriodicalId\":256762,\"journal\":{\"name\":\"2011 10th IEEE/ACIS International Conference on Computer and Information Science\",\"volume\":\"59 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2011-05-16\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2011 10th IEEE/ACIS International Conference on Computer and Information Science\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICIS.2011.49\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2011 10th IEEE/ACIS International Conference on Computer and Information Science","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICIS.2011.49","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
A Vectorization Method of Export Branch for SIMD Extension
The main popular vectorization methods for the SIMD extension dig the parallelism of the programs relying on the compiler's data dependence analysis. But the data dependence analysis can not deal with the non-structured control flow statements. Therefore, the up-to-date compilers are extremely limited to vectorize these statements. Here is a vectorization method of the export branch for the SIMD extension, which can automatically and effectively vectorize the export branch within the vector length. And the results of performance test show that this method can both fully ensure the semantic correctness of the control flow and exploit the parallelism of the data flow.