PLATFORM SPECIFIC OPTIMIZATIONS IN STATIC COMPILERS

• US 20140380289A1
• Filed: 06/21/2013
• Published: 12/25/2014
• Est. Priority Date: 06/21/2013
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: KALOGEROPULOS, SPIROS, TIRUMALAI, PARTHA P.
• Granted Patent: US 9,032,379 B2
• Time in Patent Office: Days
• US Class Current: 717/152
• CPC Class Codes: G06F 8/443 Optimisation
• CPC Class Codes: G06F 8/4441 Reducing the execution time required by the program code
• CPC Class Codes: G06F 8/447 Target code generation

Abstract

Embodiments include systems and methods for generating an application code binary that exploits new platform-specific capabilities, while maintaining backward compatibility with other older platforms. For example, application code is profiled to determine which code regions are main contributors to the runtime execution of the application. For each hot code region, a determination is made as to whether multiple versions of the hot code region should be produced for different target platform models. Each hot code region can be analyzed to determine if benefits can be achieved by exploiting platform-specific capabilities corresponding to each of N platform models, which can result in between one and N versions of that particular hot code region. Navigation instructions are generated as part of the application code binary to permit a target machine to select appropriate versions of the hot code sections at load time, according to the target machine'"'"'s capabilities.

Claims

1. A system for static compilation of application code, the system comprising:

a data store that operates to store application code comprising a plurality of code regions and to store a plurality of machine models that each models capabilities of a corresponding target platform;

a computer-implemented code profiler that operates to identify hot code regions from the plurality of code regions of the application code, such that each hot code region is a primary contributor to runtime execution of the application code; and

a computer-implemented code versioner that operates, for each hot code region, to;

compile a first version of the hot code region according to a first of the machine models, the first version having a first estimated execution time;

compile a second version of the hot code region according to a second of the machine models, the second version having a second estimated execution time; and

compile the application code to include both the first and the second versions of the hot code region when a difference between the first and second estimated execution times is greater than a predetermined threshold, the application code being compiled in such a way that permits selection by a target machine at load time between the first and second versions of the hot code region.

2. The system of claim 1, wherein the computer-implemented code versioner further operates, for each hot code region, to compile the application code to include only the first version of the hot code region when the difference between the first and second estimated execution times is not greater than the predetermined threshold.

3. The system of claim 1, wherein the first of the machine models represents a lowest capability target platform, wherein the predetermined threshold is a first predetermined threshold, and the computer-implemented code versioner further operates, for each hot code region, to:

compile a third version of the hot code region according to a third of the machine models, the third of the machine models representing a highest capability target platform, the third version having a third estimated execution time; and

determine whether a difference between the first and third estimated execution times is greater than a second predetermined threshold,

wherein the computer-implemented code versioner operates to compile the second version only when the difference between the first and third estimated execution times is greater than the second predetermined threshold.

4. The system of claim 3, wherein the plurality of machine models is ordered from 1 to N, and the computer-implemented code versioner further operates, when the difference between the first and third estimated execution times is greater than the second predetermined threshold, to:

set a purported best execution time (Tbest) to the first execution time; and

for i=1 to N;

compile an ith version of the hot code region according to an ith of the machine models, the ith version having an ith estimated execution time;

compile the application code to further include the ith version of the hot code region when a difference between the Tbest and ith estimated execution times is greater than the first predetermined threshold; and

set Tbest to the ith execution time when the difference between the Tbest and ith estimated execution times is greater than the first predetermined threshold.

5. The system of claim 4, wherein the plurality of machine models is ordered from 1 to N according to the respective capabilities of the target platforms represented by the machine models.

6. The system of claim 1, wherein the computer-implemented code profiler operates to identify the hot code regions from the plurality of code regions of the application code by dynamic profiling of the application code.

7. The system of claim 1, wherein the computer-implemented code profiler operates to identify the hot code regions from the plurality of code regions of the application code by static profiling of the application code.

8. The system of claim 1, wherein the computer-implemented code profiler operates to identify the hot code regions from the plurality of code regions of the application code by directive-based analysis of the application code.

9. The system of claim 1, wherein the computer-implemented code versioner operates to compile the application code in such a way that permits selection by the target machine at load time between the first and second versions of the hot code region by:

generating navigation functionality in the application code that permits the target machine to load one of the versions of each hot code region according to capabilities of the target machine.

10. The system of claim 9, wherein generating the navigation functionality in the application code comprises:

outlining each version into a corresponding function in such a way that permits linking of one of the corresponding functions by the target machine at load time.

11. The system of claim 1, wherein the plurality of machine models comprises a generic machine model representing a least common denominator of platform capabilities relative to the other machine models.

12. The system of claim 1, wherein each of the plurality of code regions comprises at least one of a basic code block, a loop, a set of nested loops, or a function.

13. A computer-implemented method for static compilation of application code, the method comprising:

profiling application code to identify hot code regions from a plurality of code regions of the application code, such that each hot code region is a primary contributor to runtime execution of the application code; and

for each identified hot code region;

compiling a first version of the hot code region according to a first of a plurality of machine models that each models capabilities of a corresponding target platform, the first version having a first estimated execution time;

compiling a second version of the hot code region according to a second of the machine models, the second version having a second estimated execution time; and

compiling the application code to include both the first and the second versions of the hot code region when a difference between the first and second estimated execution times is greater than a predetermined threshold, the application code being compiled in such a way that permits selection by a target machine at load time between the first and second versions of the hot code region.

14. The method of claim 13, further comprising, for each hot code region, compiling the application code to include only the first version of the hot code region when the difference between the first and second estimated execution times is not greater than the predetermined threshold.

15. The method of claim 13, wherein the first of the machine models represents a lowest capability target platform, wherein the predetermined threshold is a first predetermined threshold, and further comprising, for each hot code region:

compiling a third version of the hot code region according to a third of the machine models, the third of the machine models representing a highest capability target platform, the third version having a third estimated execution time; and

determining whether a difference between the first and third estimated execution times is greater than a second predetermined threshold,

wherein the compiling the second version step is performed only when the difference between the first and third estimated execution times is greater than the second predetermined threshold.

16. The method of claim 15, wherein the plurality of machine models is ordered from 1 to N, and further comprising, when the difference between the first and third estimated execution times is greater than the second predetermined threshold:

setting a purported best execution time (Tbest) to the first execution time; and

for i=1 to N;

compiling an ith version of the hot code region according to an ith of the machine models, the ith version having an ith estimated execution time;

compiling the application code to further include the ith version of the hot code region when a difference between the Tbest and ith estimated execution times is greater than the first predetermined threshold; and

setting Tbest to the ith execution time when the difference between the Tbest and ith estimated execution times is greater than the first predetermined threshold.

17. The method of claim 16, wherein the plurality of machine models is ordered from 1 to N according to the respective capabilities of the target platforms represented by the machine models.

18. The method of claim 15, wherein the first and second predetermined thresholds are substantially equivalent.

19. The method of claim 13, wherein compiling the application code in such a way that permits selection by a target machine at load time between the first and second versions of the hot code region comprises outlining each version into a corresponding function in such a way that permits linking of one of the corresponding functions by the target machine at load time.

20. The method of claim 13, wherein profiling the application code to identify the hot code regions from the plurality of code regions comprises performing at least one of dynamic profiling of the application code, static profiling of the application code, or directive-based analysis of the application code.