S0CF: Floating-point divide exception

What it means

The mainframe has three floating-point formats. Hexadecimal floating point (HFP) is the original IBM format. Binary floating point (BFP) follows the IEEE 754 standard, and there is also a decimal floating-point format. S0CF belongs to HFP. When an HFP divide instruction finds that the divisor's fraction is zero, the processor raises a floating-point divide exception, program interruption code 15 (hex F), and z/OS ends the task with system completion code 0CF and reason code F.

This exception can't be switched off with the program mask. A zero HFP divisor always interrupts the program.

Programs that use IEEE binary floating point don't get S0CF for a zero divisor. IEEE arithmetic either returns infinity and carries on, or, if the program enabled that trap, reports the error as a data exception (S0C7) with a data exception code that names the IEEE condition.

COBOL COMP-1 and COMP-2 fields use HFP by default. So do many older PL/I, C and assembler programs.

Common causes

• A floating-point divisor is zero, for example an average computed over no records, or a ratio whose denominator was never set.
• A COMP-1 or COMP-2 field was never initialized and holds low-values. Low-values are a valid HFP zero, so the divide fails. Spaces aren't zero in HFP, so they don't cause this.
• A very small intermediate result underflowed to zero with the exponent underflow mask off, and that zero was later used as a divisor.
• A COBOL DIVIDE or COMPUTE on floating-point fields has no ON SIZE ERROR clause.

How to find the problem

1. Find CEE3215S (Language Environment) or the IEA995I symptom dump in the job log. Note the program name and the offset.
2. Match the offset to the compiler listing to find the division.
3. Check the divisor in the dump. An HFP value has a zero fraction when every byte after the first is zero.
4. Trace where the divisor got its value, including any earlier calculation that could have underflowed to zero.

How to fix it

• Test the divisor before dividing, and decide what the result should be when it is zero.
• Add ON SIZE ERROR to COBOL DIVIDE and COMPUTE statements that use floating-point fields.
• Initialize floating-point fields to a known value before use.
• If underflow is producing the zero, rescale the calculation, or let the program detect underflow so it can react before dividing.

What it looks like

IEA995I SYMPTOM DUMP OUTPUT
  SYSTEM COMPLETION CODE=0CF  REASON CODE=0000000F
IEF450I ACCTLIB1 STEP020 - ABEND=S0CF U0000 REASON=0000000F

Technical reviewer: Albert Joseph, WND Software.

Last reviewed 2026-10-18 by the WND Software team. Written in our own words. For the full reference, check IBM’s z/OS documentation for your release.