Jens Johansen
7f8e6f8bb6
Be more verbose in loops etc in kernel
Benchmarks have shown that old-style for-loops are faster than for-in
for instance.
Golem says
ia32
whole stack (i.e. including utf8 and change to VariableIndexer)
AstToBinaryP90 (Intel Core i5) 6.617% (0.4 noise)
AstToBinary (Intel Xeon) 5.158% (0.4 noise)
AstToBinaryP90 (Intel Xeon) 5.104% (0.4 noise)
AstToBinaryP50 (Intel Xeon) 5.148% (0.4 noise)
AstToBinary (Intel Core i5) 6.579% (0.4 noise)
AstToBinaryP50 (Intel Core i5) 6.741% (0.4 noise)
by itself (i.e. compared to a build with utf8 and change to VariableIndexer)
AstToBinaryP90 (Intel Core i5) 0.7724% (0.0 noise)
AstToBinaryP50 (Intel Core i5) 0.7848% (0.0 noise)
AstToBinary (Intel Core i5) 0.8559% (0.0 noise)
AstToBinaryP90 (Intel Xeon) 0.7730% (0.1 noise)
AstToBinaryP50 (Intel Xeon) 0.8195% (0.1 noise)
AstToBinary (Intel Xeon) 0.9319% (0.1 noise)
x64
whole stack (i.e. including utf8 and change to VariableIndexer)
AstToBinaryP90 (Intel Core i5) 5.555% (0.2 noise)
AstToBinaryP50 (Intel Xeon) 6.352% (0.4 noise)
AstToBinaryP50 (Intel Core i5) 5.395% (0.5 noise)
AstToBinary (Intel Xeon) 6.928% (0.5 noise)
AstToBinaryP90 (Intel Xeon) 7.180% (0.5 noise)
AstToBinary (Intel Core i5) 5.824% (0.5 noise)
by itself (i.e. compared to a build with utf8 and change to VariableIndexer)
AstToBinaryP90 (Intel Core i5) 0.1039% (0.0 noise)
AstToBinaryP50 (Intel Core i5) 0.6154% (0.0 noise)
AstToBinary (Intel Core i5) 0.7640% (0.1 noise)
AstToBinaryP50 (Intel Xeon) 1.268% (0.1 noise)
AstToBinary (Intel Xeon) 1.457% (0.1 noise)
AstToBinaryP90 (Intel Xeon) 1.512% (0.1 noise)
Change-Id: If091a3a01f2fe6574712946a01644413033739b6
Reviewed-on: https://dart-review.googlesource.com/c/85341
Commit-Queue: Jens Johansen <jensj@google.com>
Reviewed-by: Peter von der Ahé <ahe@google.com>
2018-11-30 13:15:47 +00:00
..
2018-11-30 13:15:47 +00:00
2018-11-30 13:15:47 +00:00
2018-11-30 13:15:47 +00:00
2018-10-25 11:40:37 +00:00