Viktor 062cb210ea gpu: optimize opCache to avoid expensive map lookups ...

Benchmarking showed that the double map access calls
were a bottleneck. Rework the cache to avoid half of them.

The simplest, naive approach would have been to store a
pointer to a struct with a keep field in the map, allowing cheap
update and frame() operation. Benchmarking showed that the
increased GC pressure of that approach decreased performance
however.

Signed-off-by: Viktor <viktor.ogeman@gmail.com>

2020-06-21 11:20:36 +02:00

..

backend

gpu: saturate instead of overflowing depth buffer

2020-06-08 16:28:56 +02:00

gl

gpu: saturate instead of overflowing depth buffer

2020-06-08 16:28:56 +02:00

shaders

gpu: reintroduce reuse of offset-only stenciling

2020-06-21 11:20:36 +02:00

caches.go

gpu: optimize opCache to avoid expensive map lookups

2020-06-21 11:20:36 +02:00

clip.go

op/clip, gpu: split complex curves in package gpu instead

2020-06-21 11:17:27 +02:00

gen.go

internal/cmd/convertshaders: move shader converter to separate package

2020-02-27 20:34:22 +01:00

gpu.go

gpu: exploit pathCache in collectOps

2020-06-21 11:20:36 +02:00

pack.go

gpu: expose the rendering implementation

2020-02-07 21:21:38 +01:00

path.go

gpu: reintroduce reuse of offset-only stenciling

2020-06-21 11:20:36 +02:00

shaders.go

gpu: reintroduce reuse of offset-only stenciling

2020-06-21 11:20:36 +02:00

timer.go

gpu/backend: move backend interface types to a separate package

2020-02-27 20:34:22 +01:00