S.A. Mahlke et al., "Characterizing the Impact of Predicated Execution on Branch Prediction," Proc. 27th Int'l Symp. Microarchitecture, ACM Press, 1994, pp. 217--227.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to remain as predicated branches in the region and incorporate the frequent path of execution as part of the region. Our region formation methods will target the removal of only unpredictable branches rather than unbiased branches. 2. 2 Interaction BetweenPredication andBranch Mahlke et al. [11] investigated the interaction of predicated hyperblock region formation and branch prediction using two branch prediction architectures: a BTB with a 2 bit counter, and a BTB with profile based direction prediction. Over a subset of SPEC92 benchmarks and UNIX utilities, they showed a reduction in ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual Intl. Symp. on Microarchitecture, pages 217--227, December 1994.

....will be shown in Chapter 8.1.3. With specialized BTB support for indirect jumps [9] even better results should be obtained. Some machines provide other special architectural support for speculative execution of instructions dependent on branches, such as boosting [27] and predicted execution [24, 19]. The relative cost of an indirect jump versus the set of branches it replaces will be affected by such support. The compiler writer must use appropriate cost estimates based on the architectural support available for branches and indirect jumps on the target machine. An optimizer could also later ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th International Symposium on Microarchitecture, pages 217--227, December 1994.

....and the else part in the inner loop body into one basic block. The result of the if test is then used as a guard or predicate. Recently, much research is being done into the use of guarding and predicates in general to increase the number of instruction that can be executed simultaneously [24,11,23]. The transformations shown should increase the potential for exploiting these techniques. It should be noted that the SPARC processor that we considered, includes a fairly simple branch prediction scheme (2 bit prediction scheme) and that 2 level branch history based prediction schemes [21,29] ....

S.A. Mahlke, R.E. Hank, R.A. Bringmann, J.C. Gyllenhaal, D.M. Gallagher, and W.W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual International Symposium on Microarchitecture, pages 217--227. ACM SIGMICRO and IEEE Computer Society TC-MICRO, December 1994.

....history lengths for 2 level predictors performed better for different classes of branches. Although the design space for branch predictors grew tremendously, an understanding of how and why these predictors worked did not. More recently, however, a number of papers, such as [23] 12] 6] [16], and [4] have provided insights into mechanisms for creating more accurate branch predictors. Our goal is to continue this process towards a deeper understanding about the nature of branch behavior and branch predictor performance. In this paper, we introduce a new metric for branch behavior, ....

S. Mahlke, R. Hank, R. Bringmann, J. Gyllenhaal, D. Gallagher, and W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Intl. Symposium on Microarchitecture, pages 217--227, December 1994.

....with the prior branch instruction and increases the possibility of higher functional unit utilization between branch instructions. Previous studies have shown that selectively applying predicated execution on such illpredicted branches can improve the overall branch prediction in the hardware [11, 4]. This paper proposes a new selective if conversion algorithm using profiling on a dynamically scheduled processor supporting speculative execution with guarded execution support. The branch predictor used is the McFarling s combined (gshare bimodal) branch predictor [15] Profiling is done ....

....improve dramatically. Mahlke et al. 5] studied the effects of predicated execution on a VLIW (or statically scheduled) machine. Static estimates or profiling data derived from simple dynamic branch predictor are used to determine hard to predict branches which are then selectively if converted [4]. Their experiments were done on a VLIW machine which doesn t support OOO execution, dynamic renaming and other features prevalant in modern superscalar processors. Tyson [11] studied the effects of predicated execution by predicating all short forward branches which were hard to predict. He, then ....

S.A. Mahlke, R.E. Hank, R.A. Bringmann, J.C. Gyllenhall, D.M. Gallagher and W.W. Hwu, "Characterizing the Impact of Predicated Execution on Branch Prediction," Proceedings of the 27th International Symposium and Workshop on Microarchitecture (MICRO-27), Nov 30 - Dec 2, 1994

....to double the performance, we should both double the fetch rate and decrease four fold the number of mispredicted branches. The conclusion on the tuning of the instruction fetch engine is that when the threshold fetch is reached, we should rather try to decrease the number of control flow breaks [10, 7, 8] or improve the execution core [9, 15, 17, 6] Finally, we would like to conclude on the methodology in computer architecture research. Most microarchitecture studies use simulators modeling a complete processor as accurately as possible, that is, modeling all mechanisms present in a real ....

Scott A. Mahlke, Richard E. Hank, Roger A. Bringmann, John C. Gyllenhaal, David M. Gallagher, and Wen-mei W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual International Symposium on Microarchitecture, 1994.

.... call for every polymorphic call [6] Indirect branches commonly appear in applications that use dynamically linked libraries since they implement DLL calls [7] 3 Moreover, the relative performance impact of indirect branches will be further increased as techniques such as predicated execution [8], branch elimination [9] and coalescing of conditional branches into indirect jumps [10] become increasingly popular. By reducing the number of conditional branches (sometimes at the expense of introducing extra indirect jumps [10] these techniques will generate a higher percentage of indirect ....

S.A. Mahlke and R.E. Hank et.al. Characterizing the Impact of Predicated Execution on Branch Prediction. In Proceedings of the International Symposium on Microarchitecture, pages 217--227, November 1994. 32

.... the branch prediction rate further (e.g. guarding, dual path prediction, and recently, agree prediction) Guarding has been shown to reduce the number of branch mispredictions between 20 to 75 for the SPEC92 integer applications relative to a prediction scheme similar to our bimodal scheme [26]. However, this increase occurred with an corresponding increase in the total number of instructions executed dynamically. A recent study on selective dual path execution achieved 10 improvements in execution time [27] However, clustering of branch mispredictions, and exponential growth in the ....

Scott Mahlke et al. Characterizing the Impact of Predicated Execution on Branch Prediction. In Proceedings of the 27th International Symposium on Microarchitecture, 1994. 25

....the preferred mechanism for implementing polymorphic calls in most C and Java compilers [7] Also, indirect function calls are often used to implement DLL calls. Moreover, the relative performance impact of indirect branches will be further increased as techniques such as predicated execution [8] and branch elimination [9] become increasingly popular. By reducing the number of conditional branches, these techniques will generate a higher frequency of indirect branches (relative to conditional branches) and thus, our ability to correctly predict indirect branches become increasingly ....

S.A. Mahlke and R.E. Hank et.al. Characterizing the Impact of Predicated Execution on Branch Prediction. In Proceedings of the International Symposium on Microarchitecture, pages 217--227, November 1994. 14

....execution to predicated instructions while also allowing for selective inclusion of basic blocks based on frequency and size. Mahlke then studied the impact of predicated execution on branch prediction and discovered that 56 of dynamic branch mispredictions are eliminated with predication support [18]. Our work may be viewed as an extension of the work done by Mahlke into the dynamic domain. In addition, several other researchers have delved into if conversion. One example is the work of Klauser et al. 15] They describe a method for dynamically introducing predication to architectures that ....

Mahlke, S., R. Hank, R. Bringmann, J. Gyllenhaal, D. Gallagher, W. Hwu. "Characterizing the Impact of Predicated Execution on Branch Prediction." Proc. of the 27th Annual International Symposium on Microarchitecture, 1994, pp. 217 - 227.

.... is based on program profiling of the original basic blocks which includes information such as execution frequency, basic block size, operation latencies, and other characteristics [22] A typical code section to include in a hyperblock is one that contains a hard to predict (unbiased) branch [21], as shown in Figure 1. After If conversion, the Control Flow Graph (CFG) in Figure 1(a) which is comprised of four basic blocks, results in the predicated hyperblock shown in Figure 1(b) All operations in the hyperblock are now guarded, either by a predicate register set to the constant value ....

....code will have on the memory system. Dynamic code expansion indicates an increase in the working set size to be supported by the instruction cache. 19 7 Related Work Predicated execution presents challenges and prospects that researchers have addressed in a variety of ways. Mahlke et al. [21] showed that predicated execution can be used to remove an average of 27 of the executed branches and 56 of the branch mispredictions. Tyson also found similar results and correlated the relationship between predication and branch prediction [29] In an effort to relieve some of the ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual Intl. Symp. on Microarchitecture, pages 217--227, December 1994.

....the processor state. 7 The removal of branches yields performance bene ts in the execution of the ILP code, the most notable of which is the elimination of branch misprediction penalties. In particular, the removal of frequently mispredicted branches yields large performance gains [13] [14], 15] Predicated execution also provides an ecient mechanism by which a compiler can explicitly present the overlapping execution of multiple control paths to the hardware. In this manner, processor performance is increased by the compiler s ability to nd ILP across distant multiple program ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu, \Characterizing the impact of predicated execution on branch prediction, " in Proceedings of the 27th International Symposium on Microarchitecture, December 1994, pp. 217-227.

....instruction set, 8 GUARD TRUE and GUARD FALSE instructions will be able to predicate about 20 instructions while the GUARD BOTH instruction will only be able to guard 12. But, they show that the GUARD BOTH introduces less overhead than the separate form of the instruction. A few studies [5][6] 7] on the effects of predicated execution on branch prediction have been performed. Mahlke et al. [5] examine the effects of hyperblock techniques on the number of branches and mispredictions on three branch prediction schemes: a 1K entry branch target buffer (BTB) with 2 bit saturating ....

....instructions while the GUARD BOTH instruction will only be able to guard 12. But, they show that the GUARD BOTH introduces less overhead than the separate form of the instruction. A few studies [5] 6] 7] on the effects of predicated execution on branch prediction have been performed. Mahlke et al. [5] examine the effects of hyperblock techniques on the number of branches and mispredictions on three branch prediction schemes: a 1K entry branch target buffer (BTB) with 2 bit saturating counters, a 1K BTB with profile based direction prediction, and a 1K entry branch target cache. Not only do ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Cyllenhaal, D. M. Gallagher, and W. W. Hwu. "Characterizing the Impact of Predicated Execution on Branch Prediction." In Proceedings of the 27th International Symposium on Microarchitecture. pp. 217-227. November 1994.

....are the preferred mechanism for implemeting polymorphic calls in most C and Java compilers) Also, indirect function calls are often used to implement DLL calls. Moreover, the relative performance impact of indirect branches will be further increased as techniques such as predicated execution [6] and branch elimination [7] become increasingly popular. By reducing the number of conditional branches, these techniques will generate a higher frequency of indirect branches (relative to conditional branches) and thus, our ability to correctly predict indirect branches become increasingly ....

S.A. Mahlke and R.E. Hank et.al. Characterizing the Impact of Predicated Execution on Branch Prediction. In Proceedings of the International Symposium on Microarchitecture, pages 217--227, November 1994.

....of increasing the branch prediction accuracy. By increasing the distance between mispredicted branches, these schemes require a higher fetch bandwidth (measured in useful instructions per cycle) Techniques to increase the distance between mispredicted branches include predication (both static [11] and dynamic [8] and selective eager execution [9] Techniques to Alleviate Data Dependencies Some recently proposed techniques have been effective in reducing the negative performance impact of data dependencies. By cutting data dependencies, these techniques increase the amount of parallelism ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. mei W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual International Symposium on Microarchitecture, 1994.

....blocks to be included in the hyperblock is based on program profiling which includes information such as execution frequency, basic block size, operation latencies, and other characteristics. A typical code section to include in a hyperblock is one that contains a hard to predict (unbiased) branch [18], as shown in Figure 1. After predication, the Control Flow Graph (CFG) in Figure 1(b) which is comprised of five basic blocks, results in the predicated hyperblock shown in Figure 1(c) All operations in the hyperblock are now guarded, either by a predicate register set to the constant value of ....

....utilization still remains well within the reported IA 64 register file sizes (128 general purpose, 128 floating point, 8 branch, and 64 predicate) 3] 6 Related Work Predicated execution presents challenges and prospects that researchers have addressed in a variety of ways. Mahlke et al. [18] showed that predicated execution can be used to remove an average of 27 of the executed branches and 56 of the branch mispredictions. Tyson also found similar results and correlated the relationship between predication and branch prediction [26] In an effort to relieve some of the difficulties ....

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual Intl. Symp. on Microarchitecture, pages 217--227, December 1994.

No context found.

S.A. Mahlke et al., "Characterizing the Impact of Predicated Execution on Branch Prediction," Proc. 27th Int'l Symp. Microarchitecture, ACM Press, 1994, pp. 217--227.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th ACM/IEEE International Symposium on Microarchitecture, pages 217--227, 1994.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Hyllenhaal, D. M. Hallagher, and W. mei W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th International Symposium on Microarchitecture, pages 217--227, 1994.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual Intl. Symp. on Microarchitecture, pages 217--227, December 1994.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proceedings of the 27th Annual Intl. Symp. on Microarchitecture, pages 217--227, December 1994.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu, \Characterizing the Impact of Predicated Execution on Branch Prediction," Proc. 27th International Symposium on Microarchitecture, pp. 217-227, 1994.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the impact of predicated execution on branch prediction. In Proc. of the 27th Annual Intl. Symp. on Microarchitecture, pages 217--227, December 1994.

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the Impact of Predicated Execution on Branch Prediction. Proceedings of the 27th International Symposium on Microarchitecture, December 1994, pp. 217-227

No context found.

S. A. Mahlke, R. E. Hank, R. A. Bringmann, J. C. Gyllenhaal, D. M. Gallagher, and W. W. Hwu. Characterizing the Impact of Predicated Execution on Branch Prediction. Proceedings of the 27th International Symposium on Microarchitecture, December 1994, pp. 217-227

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC