This blog delves into the notable enhancements brought by ABO 2.2 and ABO 2.1 and compares its performance gains against different COBOL versions on IBM z15 and z16. Explore how these advancements translate into significant CPU time reductions across various benchmarks, marking a notable stride in COBOL application efficiency.

ABO 2.2 performance improvement data on IBM z16 with ARCH=14

ABO 2.2 ships with advanced optimization technology. It has the capability to generate code that can take full advantage of the IBM z16 processor - especially the vector packed-decimal enhancement facility 2 feature — to provide a significant performance boost for business-critical COBOL applications.

The new vector packed-decimal enhancement facility 2 adds performance improvements for COBOL programs that contain one or more of the following types of statements:

• Exponentiation operations on packed or zoned decimal data items where the exponent is declared with one or more fractional digits
• Arithmetic statements involving mixed decimal and floating-point data items
• Statements using numeric-edited data items

Over a varied range of benchmarks, ABO 2.2 shows a significant reduction in CPU time on z16 with ARCH=14 than Enterprise COBOL:

• Compared to Enterprise COBOL 4.2 OPT with ARCH14, ABO 2.2 reduces CPU time by 56% on average for a mix of compute and I/O intensive benchmarks.
• Compared to Enterprise COBOL 5.2 OPT with ARCH14, ABO 2.2 reduces CPU time by 28% on average for a mix of compute and I/O intensive benchmarks.
• Compared to Enterprise COBOL 6.1 OPT with ARCH14, ABO 2.2 reduces CPU time by 7% on average for a mix of compute and I/O intensive benchmarks.
• Compared to ABO 2.1 with ARCH=13, ABO 2.2 reduces CPU time by 5% on average for a mix of compute and I/O-intensive benchmarks

ABO 2.1 performance improvement data on IBM z15 with ARCH=13

ABO 2.1 can generate code that takes full advantage of the new z15 processor — especially the Vector Packed Decimal Enhancement Facility feature — to provide a significant performance boost for business-critical COBOL applications.

z15 features the Vector Packed Decimal Enhancement Facility to accelerate packed and zoned decimal computations. On z14 we introduced Vector Packed Decimal Facility where vector registers and new instructions were introduced to replace memory-to-memory instructions, significantly speeding up packed decimal arithmetic. 

On z15, the Vector Packed Decimal Enhancement Facility further enhanced the performance of packed decimal arithmetic. This facility uses wide vector registers to perform decimal computations much more efficiently. It was designed from the ground up to accelerate performance and provide equivalent functionality for the most important decimal data types found in COBOL.

• On z15, ABO 2.1 with ARCH=13 optimization reduces the CPU usage of compute-intensive applications by an average of 57% compared with the same code compiled with COBOL 4.2 with OPT. ABO 2.1 has significantly closed the performance gap with COBOL 6.3.
• ABO 2.1 with ARCH=13 reduces the CPU usage of compute-intensive applications by an average of 10% compared with the same code optimized with ABO 1.3 with ARCH=12.

For applications that use inter-language calls (e.g. COBOL calling Java), applications encountering decimal overflow conditions would see up to 22% CPU reduction on z15.

Related resource:

Unlock the full potential of ABO with the ABO Essentials Badge V2 today! Dive into a series of customized courses meticulously crafted to arm you with the indispensable knowledge and skills needed to supercharge your utilization of IBM ABO. Don't miss out—get started now!