475849 - Parens around for-condition lead to inefficient bytecode

Status: RESOLVED DUPLICATE of bug 461269

(Core :: JavaScript Engine, defect)

Description

[Jesse Ruderman]

This is a simple-infinite-loop example, but similar things happen in other cases.

js> dis(function() { for (; (1); ) { } })    
flags: LAMBDA INTERPRETED
main:
00000:  nop
00001:  goto 4 (3)
00004:  true
00005:  ifne 4 (-1)
00008:  stop

js> dis(function() { for (; 1; ) { } })   
flags: LAMBDA INTERPRETED
main:
00000:  nop
00001:  goto 1 (0)
00004:  stop

js> dis(function() { for (; (true); ) { } })
flags: LAMBDA INTERPRETED
main:
00000:  nop
00001:  goto 1 (0)
00004:  stop

Comment 1

[Jason Orendorff [:jorendorff]]

In current tracemonkey tip:

js> dis(function() { for (; (1); ) { } })
flags: LAMBDA NULL_CLOSURE
main:
00000:  nop
00001:  loop
00002:  goto 1 (-1)
00005:  stop

Source notes:
  0:     0 [   0] for      cond 1 update 1 tail 1
js> dis(function() { for (; 1; ) { } })
flags: LAMBDA NULL_CLOSURE
main:
00000:  nop
00001:  loop
00002:  goto 1 (-1)
00005:  stop

Source notes:
  0:     0 [   0] for      cond 1 update 1 tail 1
js> dis(function() { for (; (true); ) { } })
flags: LAMBDA NULL_CLOSURE
main:
00000:  nop
00001:  loop
00002:  goto 1 (-1)
00005:  stop

Source notes:
  0:     0 [   0] for      cond 1 update 1 tail 1

(The "loop" instruction is a no-op that we emit for the benefit of the JIT, particularly when blacklisting a loop so that the JIT doesn't waste any more time trying to compile it.)