VHDL-200x Meeting In Attendance: Tej Singh Dave Bishop Chuck Swart John Ries Jim Lewis Eric Marschner Patent policy reviewed. Question/Answer on PSL addition John questioned about PSL declarations (endpoints, properties, and sequences) being required to be placed in in VHDL declarative part. Currently, vunits allow declarations and directives to be mixed, so do meta comment implementations. Eric responded that he considered this and didn't see requiring PSL declaration is the declarative region this as too restrictive. Issue here is upward compatibility of people moving from meta comments PSL. Eric's belief is that there are trade-offs, some good and some bad, but overall there is an improvement for the user. Chuck: Adding of section 12.8.1, default clock declarations, where is default clock declaration syntax? Defined in PSL LRM. There is an issue that the default clock should be at most one per declarative part not region. Tej: Limitation on shared variables why? Answer: Shared variable are effected by process interactions and these interactions are not specified. Eric is worried there may be race conditions. Eric suspects there are issues here but he is not sure. No problem with process variables because you can't get at them in properties. Tej: In case of parameterized sequences what types are allowed? Should VHDL allow PSL type BOOLEAN which includes std_logic and such? Eric feels for portability that this merging of boolean/bit/std_logic should be kept. Also it keeps down the verboseness of the language Tej: VHPI question some of the attributes on PSL items make sense only on directives and not declarations. Eric thinks that the reason was to allow 1.0 PSL merging of directives and declarations. This may no longer be the case with 1.1. Eric will the forward question on to Francoise. Eric: If you have issues, please send them to Eric. FT5 Dave Bishop, has two differences from proposal. to_string(preferred declaration), to_hstring, to_ostring are defined for all predefined types. to_dstring isn't defined for some types don't have a signed/unsigned value. Package is not implemented. Discussion on FT on to_string with radix. Discussion of defaulting to unsigned, but allow for choice of signed format/radix. Key is that the to_string in 1164 would have an additional argument for signed vs unsigned to allow for overload selection. Dave still thinks there are some issues here but will investigate. Eric pointed out the creation of to_XXX. Should we look at overloading type names as functions. There are issues here with overloading and making current type conversions occur. IEEE package copyright issue. Chuck proposed that the bodies be labeled as sample implementations only and place the bodies in the public domain. Much discussion conclusion is we are not going to resolve issue within this group. FT8 NL and CRLF are defined as character substitution syntax. SREAD don't have strlen parameter, use the 'length on the line output. Intent was to only have NL or CRLF. CRLF will be dropped. NL will be deferred constant. Don't have versions with outs of integer and real. Moved the functions to packages that have representation defined. Discussion on the interpretation of text characters to integer. Jim, Eric, and John hand issues on defining hread to an integer in the numeric_unsigned packages. One issue is range for reading integer values. Is the string "FF" a signed object or not? Do we need to define the formats and then do the read/write operations for those formats? For integers, issues are number of characters read, signed or unsigned. Should we have a hread/hwrite on integer at all? Should we add another parameter signed/unsigned on hread or hwrite. David will modify, will review at later date. Eric: On reading and writing, do we skip blank spaces (yes), number of character read and valid values written down in English? No, package source is the only definition. Charlie: how do we read 9 character bit vector? David: skip all spaces then read the next 9 character or until next white space. Current implementation will consider FFX an error because X is an illegal character. Eric believes we should read until space or a character that is not valid for this format. 1993 LRM page 204, section 14.3 already defines a lot of formating for different packages. Consensus is that we need to define formats that read takes. Defer hread/hwrite for integer and dread/dwrite std_logic. Charlie - remove restriction of multiple of 4 for hex and 3 for octal. Dave will update package for this.