On Thu, 20 Aug 2015 19:34:37 +0100, Bill Spitzak wrote:
Could this be whether some "bad" instruction ends up next to or split
by a cache line boundary? That would produce a random-looking plot,
though it really is a plot of the location of the bad instructions in
the measured function.
If this
On Thu, Aug 20, 2015 at 6:58 AM, Pekka Paalanen wrote:
> A thing that explains a great deal of these anomalies, but not all of it,
> has
> something to do with function addresses. There are hypotheses that it might
> have to do with the branch predictor and its cache. We made a test
> targeting
From: Pekka Paalanen
Enable the fast path added in the previous patch by moving the lookup
table entries to their proper locations.
Lowlevel-blt-bench benchmark statistics with 30 iterations, showing the
effect of adding this one patch on top of
"pixman-fast-path: Add over_n_ fast path (disa
From: Ben Avison
This new fast path is initially disabled by putting the entries in the
lookup table after the sentinel. The compiler cannot tell the new code
is not used, so it cannot eliminate the code. Also the lookup table size
will include the new fast path. When the follow-up patch then ena
From: Pekka Paalanen
Enable the fast path added in the previous patch by moving the lookup
table entries to their proper locations.
Lowlevel-blt-bench benchmark statistics with 30 iterations, showing the
effect of adding this one patch on top of
"pixman-fast-path: Add over_n_ fast path (disa
From: Ben Avison
This is a C fast path, useful for reference or for platforms that don't
have their own fast path for this operation.
This new fast path is initially disabled by putting the entries in the
lookup table after the sentinel. The compiler cannot tell the new code
is not used, so it c
From: Pekka Paalanen
Hi,
I and Ben have been fighting with benchmarking on the Raspberry Pi 1 for a long
time. The problem is unexpected performance differences. Insignificant changes
to the code can cause significant changes to lowlevel-blt-bench performance
results. The change can be a speed u
---
I've been doing a fair bit of work on scaled plots again recently, and I was
finding that scaling-test omitted to mention lots of useful information
(notably whether a mask was used, and what the mask parameers were).
test/scaling-test.c | 63 +---