Questions about IBM 1401 not answered by manuals

• The flow charts in the IBM 1401 Data Flow manual (G24-1477-0) don't show a check for a word mark in the I-3 and I-6 cycles (reading the fourth and seventh characters of an instruction). If a word mark is found in the fourth or seventh character of an instruction, does the processor ignore it?

• The flow charts also indicate that a B or V op code is processed without error if a word mark is discovered in the I-7 cycle (reading the eighth character of an instruction). When a BCE or BWZ instruction is processed, the A-register is assumed to contain the D-modifier. During the I-6 cycle, the memory contents are transferred from the B-register to the A-register. So a seven-character BCE or BWZ would use the units position of the B address as the D-modifier. Is this correct?

• It also looks like 1-character BWZ instructions work, with the D-modifier being the previous contents of the A register. The B address is whatever is in the B-address register. 2- or 4-character BWZ instructions probably don't do anything useful, as the I-1 to I-3 cycles put characters into both the A- and B- address registers. 5-character BWZ instructions probably don't work, as the I-4 cycle clears the B-address register to blanks, and then puts the character fetched from storage into the hundreds position of the B-address register. Is this correct? A 4-character BWZ wouldn't cause an error: It would branch to A if the character at A met the zone/word mark conditions specified by the low-order character of the A address.

• For instructions that are normally eight characters, if there's not a word mark after the eighth character, the data flow manual indicates that the processor scans for a word mark (in I-8 cycle). During the scan, it copies the B register to the A register. Instructions that interpret a D-modifier find it in the A register. This means that an instruction that is normally eight characters finds its D-modifier in the last character before the word mark, even if it's not the eighth character. Is this correct?

  What happens if you write a long instruction in place of one that is normally five characters, such as CCB or CU? Does the processor try to construct a valid B address, and check-stop if there isn't one? It doesn't do this for NOP, but is NOP special, or does the processor postpone checking that the B-address register contains a valid address until it needs to use it as an address?

• The question whether the processor scans for a word mark after the seventh character of a clear storage instruction is murky. In the first reprinting of the System Operation Reference Manual, A24-3067-1, it explicitly says a word mark isn't required after a clear storage and branch instruction. In the second reprinting, A-3067-2, this sentence has been removed. It's also absent from the fifth reprinting of the "Brownie Book," A24-1403-5. In the I-7 chart in the data flow manual, it only checks for set word mark. However, a correspondent has told me that the flow chart for the 1401 emulator that was microcoded for S/360/30 says it doesn't scan for a word mark starting at the eighth character if the op code is comma or slash. One possibility is that the behavior changed when the grand strategy for the 1401 logic was changed at around serial number 25,000. This would lead one to wonder why the S/360/30 microcode emulator used the old behavior. Is there one answer (and what is it), or are there two?

• It's clear that a group mark is put into core storage at the first character after the last character transmitted by a READ TAPE WITH WORD MARKS instruction, as the test for noise (i.e., really short) records is done by a chained BCE looking for a group mark. Studying the way the Autocoder assembler loads overlays has convinced me either that a READ TAPE WITH WORD MARKS instruction doesn't put a word mark with that group mark, or that it clears word marks as it goes, thereby not being stopped by sensing a group mark with a word mark in core storage. If the latter, the EXAMPLE paragraph on page 59 of the fifth printing of the Brownie Book (A24-1403-5) is incorrect. If a word mark is deposited with the group mark, and READ TAPE WITH WORD MARKS can be stopped by sensing a group mark with a word mark in core storage, subsequent transfers would be stopped at the length of the shortest previous record. The overlay reader in the Autocoder assembler doesn't clear a word mark at the end of a read operation, so one of the two above behaviors must hold. I suspect it's the latter.