JOhn, all >The issue is that we have three formats (binary, decimal, and hexadecimal) >and three lengths (4, 8, and 16). >So we need to determine what characters should be used for the various >formats and lengths.
Now that I better understand the question, there is a way to do just that defined in the IBM High Level Assembler Reference Manual plus PTF updates for DFP: http://publibz.boulder.ibm.com/epubs/pdf/asmr1010.pdf http://www-1.ibm.com/support/docview.wss? rs=12&context=SSENW6&dc=DB550&uid=isg1PK18170&loc=en_US&cs=utf- 8&lang=en Explicit definitions for the 9 floating point literal types and 3 fixed point types are shown below used in a conteXt where the explicit type and length is indeed required. These sorts of tests are used in some of the regression tests for z390 found in rt\test\TESTFP1.MLC, rt\test\TESTDFP1.MLC, and rt\test\TESTINS1.MLC: 1. CLC =EH'1E2',=EH'10E+1' HFP SHORT 32 BIT 2. CLC =DH'100',=DH'1000E-1' HFP LONG 64 BIT 3. CLC =LH'0.1E3',=LH'0.01E+4' HFP EXTENDED 128 BIT 4. CLC =EB'1E2',=EB'10E+1' BFP SHORT 32 BIT 5. CLC =DB'100',=DB'1000E-1' BFP LONG 64 BIT 6. CLC =LB'0.1E3',=LB'0.01E+4' BFP EXTENDED 128 BIT 7. CLC =ED'1E2',=ED'10E+1' FP SHORT 32 BIT 8. CLC =DD'100',=DD'1000E-1' BDFP LONG 64 BIT 9. CLC =LD'0.1E3',=LD'0.01E+4' BFP EXTENDED 128 BIT 10 CLC =F'1E9',=F'10E8' FIXED SHORT 32 BIT 12.CLC =FD'1E15',=FD'10E14' FIXED LONG 64 BIT 13.CLC =FL16'1E30',=FL16'10E29' FIXED EXTENDED 128 BIT Can you guess which of the above logical compares fail? The DFP compares fail because DFP does not normalize all equivalent values to the same format. If the DFP equivalent values are loaded into FPR registers and compared using a DFP arithmetic compare such as CEDTR, then they do compare equal. Also the FL16 128 bit integer format is an extension in z390 for use in testing instructions like DLGR which divides 128 bit dividend by 64 bit dvisor giving 64 bit quotent and 64 bit remainder. Absent an international standard for explicit definition of BFP and DFP literals, I would use the IBM HLASM definitions as a defacto standard which also includes HFP which is an IBM unique format. For fixed point I'd propose switch ing to EF, DF, and LF type prefix for consisency (The current HLASM use of F and FD does not allow addition of FL due to conflict with length submodifier L). Don Higgins [EMAIL PROTECTED] www.z390.org ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
