Richard E. Sweet
[igor]<sweet>79performance.bravo

1.Compiler Development -- I am continuing to clean up the code generators and to produce better object code. The lowest level data structures of the code generators have been around since Mesa 1, and hence many new language and architecture features are clumsy to represent. For Mesa 6, I have redesigned the lexeme representation and stack modelling. In addition to fitting the current language and architecture better, these new facilities make more extensive use of the Mesa type system. At present, about four thousand lines of new code have been written. I have kept the debugging userproc CGenDebug current with the new data structures to facilitate debugging this mass of new code.

2.Teak -- I performed a number of tasks either directly or indirectly related to the Teak effort.

I produced special features in the Mesa 5.0 compiler to facilitate bootstraping to the D0.

I interfaced the MarkIID character printer to the D0, including writing several "tools" that allowed us to decipher the documentation that came with the machines. I wrote Pilot Test Package routines, and a Pilot Functional Specification chapter on the printer. The lowest level tool also allowed us to diagnose a hardware problem with the El Segundo MarkIID printers.

I helped CIOS to demonstrate the feasibility of the 1750 printer (and hence the Beach) as the character printer for the Star product.

I went to El Segundo as a consultant early in the Teak system test period, where my low level knowledge of Mesa proved particularly useful.

3.Compiler Maintenance & Support -- I finished the Mesa 5.0 compiler implementation, including such new features as indexing constants in the code segment, safer constructors, code for expanded inline bodies, cross jumping, and D* object format. I helped CSL to produce several compilers with special features to enable Cedar development. I helped CSL to produce a Pascal P-code to Mesa byte code translator. To facilitate compiler debugging, I wrote a userproc, CGenDebug, to display internal data structures such as symbol tables, trees, and lexemes. This consists of approximately three thousand lines of Mesa code. Subsets of this userproc will be useful to others for debugging the compiler front end, the source formatter, the debugger, etc.

4.Mesa Class -- I helped SDSupport give a beginning Mesa class. My help included last minute rewriting of laboratory hand-outs, simplifying and solving exercises, and lecturing in Palo Alto and El Segundo.

5.Technical Review -- Richard Johnsson and I gave a portion of the Technical Review describing Thistle. I wrote the original talk outline, produced all of the slides, and gave the demonstration portion of the talk. We also repeated the talk for a Computer Forum.

6.Manual Editing -- With help from the entire Mesa group, especially Ed Satterthwaite and Bruce Malasky, I coordinated the editing of the Mesa 5.0 manual for the release. This activity was complicated by the large press run of two thousand, the release of new fonts (which necessitated massive re-editing), and unexpected slowness on the part of the print shop (which caused the manual to arrive several weeks later than expected).