Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(291)

Issue 1656543002: NEON for table lookups? (Closed)

Created:
4 years, 10 months ago by msarett
Modified:
4 years, 10 months ago
Reviewers:
mtklein
CC:
reviews_skia.org
Base URL:
https://skia.googlesource.com/skia.git@index
Target Ref:
refs/heads/master
Project:
skia
Visibility:
Public.

Description

NEON for table lookups? BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1656543002 CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Patch Set 1 #

Patch Set 2 : Use VTBL and OR #

Total comments: 1

Patch Set 3 : Performance test #

Patch Set 4 : Inline assembly #

Unified diffs Side-by-side diffs Delta from patch set Stats (+494 lines, -14 lines) Patch
A bench/IndexBench.cpp View 1 2 3 1 chunk +411 lines, -0 lines 0 comments Download
M bench/nanobench.cpp View 3 chunks +7 lines, -7 lines 0 comments Download
M dm/DM.cpp View 3 chunks +5 lines, -5 lines 0 comments Download
M src/codec/SkSwizzler.cpp View 1 2 chunks +69 lines, -0 lines 0 comments Download
M src/opts/SkBlitRow_opts_arm.cpp View 2 chunks +2 lines, -2 lines 0 comments Download

Depends on Patchset:

Messages

Total messages: 14 (3 generated)
msarett
Thought you might find this interesting (though it's probably a dead end). I didn't have ...
4 years, 10 months ago (2016-01-30 20:43:18 UTC) #3
msarett
Second patch is better, but still too slow. Out of curiosity, I think I'll write ...
4 years, 10 months ago (2016-01-30 21:03:37 UTC) #4
msarett
No hope for this. Way slower than the scalar code.
4 years, 10 months ago (2016-02-01 15:01:18 UTC) #5
mtklein
On 2016/02/01 15:01:18, msarett wrote: > No hope for this. Way slower than the scalar ...
4 years, 10 months ago (2016-02-01 15:08:50 UTC) #7
msarett
Any way to look at the disassembly on ARM?
4 years, 10 months ago (2016-02-01 15:19:59 UTC) #8
mtklein
On 2016/02/01 15:19:59, msarett wrote: > Any way to look at the disassembly on ARM? ...
4 years, 10 months ago (2016-02-01 15:32:15 UTC) #9
mtklein
On 2016/02/01 15:32:15, mtklein wrote: > On 2016/02/01 15:19:59, msarett wrote: > > Any way ...
4 years, 10 months ago (2016-02-01 15:44:00 UTC) #10
mtklein
https://codereview.chromium.org/1656543002/diff/20001/src/codec/SkSwizzler.cpp File src/codec/SkSwizzler.cpp (right): https://codereview.chromium.org/1656543002/diff/20001/src/codec/SkSwizzler.cpp#newcode247 src/codec/SkSwizzler.cpp:247: t0 = vld4q_u8((const uint8_t*) (table + 0)); // rgba ...
4 years, 10 months ago (2016-02-01 16:30:37 UTC) #11
msarett
On 2016/02/01 16:30:37, mtklein wrote: > https://codereview.chromium.org/1656543002/diff/20001/src/codec/SkSwizzler.cpp > File src/codec/SkSwizzler.cpp (right): > > https://codereview.chromium.org/1656543002/diff/20001/src/codec/SkSwizzler.cpp#newcode247 > ...
4 years, 10 months ago (2016-02-01 16:42:11 UTC) #12
msarett
The latest Patch Set uses inline assembly and works just well enough to benchmark. There ...
4 years, 10 months ago (2016-02-02 00:03:46 UTC) #13
mtklein
4 years, 10 months ago (2016-02-02 00:37:13 UTC) #14
Message was sent while issue was closed.
On 2016/02/02 00:03:46, msarett wrote:
> The latest Patch Set uses inline assembly and works just well enough to
> benchmark.
> 
> There is no performance gain.  I think it's time for me to stop wasting time
on
> this :).

SGTM.  Just think how happily we'd land a CL that turned a giant mess of inline
assembly into
   SkPMColor c = table[index];

Best to focus on the positive aspects of this:
  - learning there's no more vswp
  - learning inline asm (and that it's miserable)

Powered by Google App Engine
This is Rietveld 408576698