"... In PAGE 7: ... TIME (seconds) DYNAMIC INSTRUCTION INST. TIME (seconds) Jmark N/A 14 17 JLex 15 89 89 EspressoGrinder N/A 233 174 CUP 24 55 88 BIT 15 32 31 Table2 shows the execution time of the instrumented programs for each tool. The increase in execution time of the instrumented programs ranged from 23% to 150%.... ..."

"... In PAGE 8: ... TIME (seconds) DYNAMIC INSTRUCTION INST. TIME (seconds) Jmark N/A 14 17 JLex 15 89 89 EspressoGrinder N/A 233 174 CUP 24 55 88 BIT 15 32 31 Table2 shows the execution time of the instrumented programs for each tool. The increase in execution time of the instrumented programs ranged from 23% to 150%.... ..."

"... In PAGE 40: ... BIT has more classes and has larger code size, and this explains why it took more time to instrument BIT than the other two applications. Table2 shows the execution time of the instrumented programs for each tool. Table 2.... ..."

"... In PAGE 40: ... BIT has more classes and has larger code size, and this explains why it took more time to instrument BIT than the other two applications. Table2 shows the execution time of the instrumented programs for each tool. Table 2.... ..."

"... In PAGE 7: ... Jmark consists of 19 class files and benchmarks 11 different areas of Java performance, JLex consists of 23 class files, EspressoGrinder is composed of 105 class files, CUP consists of 40 class files, and BIT consists of 43 class files. Table1 summarizes the time taken for each tool to build the instrumented programs. As shown in Table 1, the time taken to instrument each of the five applications was under four minutes for both of the customized tools.... ..."

"... In PAGE 40: ... Jmark consists of 19 class files and benchmarks 11 different areas, JLex consists of 20 class files, and BIT consists of 43 class files. Table1 summarizes the time taken for each tool to build the instrumented programs. Table 1.... In PAGE 40: ...As shown in Table1 , the time taken to instrument each of the three applications was under three minutes for both of the customized tools. The average time taken to instrument a single class file was about four seconds.... ..."

"... In PAGE 40: ... Jmark consists of 19 class files and benchmarks 11 different areas, JLex consists of 20 class files, and BIT consists of 43 class files. Table1 summarizes the time taken for each tool to build the instrumented programs. Table 1.... In PAGE 40: ...As shown in Table1 , the time taken to instrument each of the three applications was under three minutes for both of the customized tools. The average time taken to instrument a single class file was about four seconds.... ..."

"... In PAGE 8: ... Jmark consists of 19 class files and benchmarks 11 different areas of Java performance, JLex consists of 23 class files, EspressoGrinder is composed of 105 class files, CUP consists of 40 class files, and BIT consists of 43 class files. Table1 summarizes the time taken for each tool to build the instrumented programs. As shown in Table 1, the time taken to instrument each of the five applications was under four minutes for both of the customized tools.... ..."

"... In PAGE 8: ...8% in BIT. BIT instrumentation caused an increase in the program size as well, and the overhead ranged from 14% to 37% as shown in Table3 . This increase in size is explained by the addition of entries in the constant pool table, which includes static arguments to the analysis routines, the names of class and analysis methods, and method descriptors, and by the addition of actual bytecodes to invoke the analysis routines.... ..."

"... In PAGE 9: ...8% in BIT. BIT instrumentation caused an increase in the program size as well, and the overhead ranged from 14% to 37% as shown in Table3 . This increase in size is explained by the addition of entries in the constant pool table, which includes static arguments to the analysis routines, the names of class and analysis methods, and method descriptors, and by the addition of actual bytecodes to invoke the analysis routines.... ..."