Number: 482

Date:  3-Apr-84 17':26':40

Submitter: Sannella.PA

Source: JONL.PA

Subject: Want FIXP <-> FLOATP coercions done in uCode

Lisp Version: 

Description: '
Date': 18 MAR 84 17':19 PST'
From': JONL.PA'
Subject': Re': Floating arith optimizatio'
To':   kaplan, lispsupport'
cc':   JONL'
'
In response to the message sent  18-Mar-84  9':31':15 PST from kaplan.pa'
'
These factors, 2 to 5, of course apply only to the low-level functions '
mentioned, but in some cases these do add up to a surprising overall '
performance degradation;  it all just goes to show critical even one "leaky '
hole in the dike" can be.  Larry has already requested an AR on the obvious '
compiler optimization, but that''s still not the root of the matter.  Two '
main points spring to mind':'
'
1) ucode, where possible, should do the coercions.  The compiler simply'
   can''t optimize out cases like (FPLUS X Y) where X is fixp and Y floatp.'
...'


Workaround: 

Test Case: 

Edit-By: 

Edit-Date: 

Attn: Charn, purc, jonl

Assigned To: 

In/By: 

Disposition: 

System: Language Support

Subsystem: Microcode

Machine: 

Disk: 

Microcode Version: 

Memory Size: 

File Server: 

Server Software Version: 

Difficulty: 

Frequency: Everytime

Impact: Moderate

Priority: Unlikely

Status: Wish

Problem Type: Performance

Source Files: