Shmuel wrote: >Are you running releases of CP and CMS that support vector instructions?
Yes. Same as the CP and CMS I'm running z/OS on at IBM Dallas (CP 7.2, CMS 30). >How is the CMS virtual machine defined? ESA, same as our z/OS guest in Dallas. >Where is the HLASM documentation for the VL instruction? Does M_3 default to 0? Good question; I don't have HLASM doc handy, but note that: -> 0000E072 ???? E710C0A00006 CC 1 *** 0000E072 PROG 0007 -> 0139BCF0 DATA doesn't even translate the instruction, so that makes it pretty clear that it's not even recognizing the instruction. CP is pretty smart about such things-to the point of being irritating, since it COULD say "Hey, this is a VL instruction, but I'm not simulating it" (which I'm sensitive to from SSCH in 370 mode, back in the day, or SIO now, I guess), rather than pretending it doesn't know what it is. At least, ISTR that being the behavior; someone could convince me I'm wrong pretty easily. Anyway, it's sure not translating THIS one. In fact: vgfmg x x -> 0000E072 ???? E710C0A00006 CC 1 *** 0000E072 PROG 0007 -> 0139BCF0 DATA d ie072.20 R0000E072 ******* E710 LARL C0A00006E720 LARL C0900006E730 E6 R0000E080 IIHF C0980006E712 LPER 3000 LPER 30B4 R0000E08A ******* 0000 ******* 0000 ******* 0000 R0000E090 ******* E740 Now THIS is interesting, on Dallas machine (not zPDT): vmfhlasm vgfmg dmsvm VMFASM2760I VMFHLASM processing started DMSUPD181E No update files were found DMSOPN002E File HCPOM1 MACLIB * not found DMSOPN002E File HCPOM2 MACLIB * not found VMFASM1907I Assembling VGFMG 00E06E 0000 0000 41 VL 1,WORK1,4 Set 64-bit output area VGF00310 ASMA029E Incorrect register specification - WORK1 ASMA173S Delimiter error, expected blank - ,4 ASMA435I Record 31 in $VGFMG ASSEMBLE A1 on volume: ETP191 00E072 0000 0000 42 VL 2,PARM1,4 Load 64-bit value #1 VGF00320 ASMA029E Incorrect register specification - PARM1 ASMA173S Delimiter error, expected blank - ,4 ASMA435I Record 32 in $VGFMG ASSEMBLE A1 on volume: ETP191 00E076 0000 0000 43 VL 3,PARM2 Load 64-bit value #2 VGF00330 ASMA029E Incorrect register specification - PARM2 ASMA435I Record 33 in $VGFMG ASSEMBLE A1 on volume: ETP191 44 VGFMG 1,2,3 VECTOR GALOIS FIELD MULTIPLY SUM VGF00340 ASMA057E Undefined operation code - VGFMG Same source code. Another wild hare-I copied the load module I'd created on the zPDT to z/OS and ran it (under z/VM, with PER PROG again): vgfmg x x -> 0000E072 ???? E710C0A00006 *** 0000E072 PROG 0001 -> 0139B410 OPERATION That's different-PROG 1 instead of PROG 7. Then I did a SET MACHINE Z and IPL ZCMS and I'm back to a PROG 7 again, DXC FE. Same on the zPDT, but I suspect that the fact that I get a PROG 7 instead of a PROG 1 is a buglet. Meanwhile, in Z mode I can do D XG0: CRG 0 = 00000000000010E0 (z14) CRG 0 = 00000000000010E2 (zPDT) So it appears that, indeed, the hardware supports not the vector stuff (bits 45/46 are both zero).. I didn't think that was optional on a z14! Maybe an option in HMC, though why is beyond me. My conclusions: 1. zPDT and real hardware behave differently here 2. I do NOT have a machine that does the vector instructions 3. The documentation on this is atrocious Unless someone has another suggestion, I'm going to give this up at this point, other than pointing out the inconsistency to IBM. I've learned a few things-thanks for all the help! ...phsiii ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
