Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

SIMD-based blitting using ARM Neon and Intel SSE4 #5428

Open
nfeske opened this issue Jan 21, 2025 · 5 comments
Open

SIMD-based blitting using ARM Neon and Intel SSE4 #5428

nfeske opened this issue Jan 21, 2025 · 5 comments
Labels
feature fixed

Comments

@nfeske
Copy link
Member

nfeske commented Jan 21, 2025 •
edited
Loading

At the moment, we use SIMD instructions for 2D-pixel copying only on the x86 platform. I'd like to foster the use of SIMD on ARMv8 as well, and also leverage more modern SIMD variants (SSE4) on x86_64.
@nfeske nfeske added the feature label Jan 21, 2025
@nfeske nfeske mentioned this issue Jan 21, 2025
Closed
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
blit: SIMD-based back2front copy
3bf26a6 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
nitpicker: alloc padded capture buffer
e19bda2 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
c5ef6ef 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
fb_sdl: support rotate and flip
7a28ed0 
Issue genodelabs#5428
@nfeske
Copy link
Member Author

nfeske commented Jan 21, 2025

Implemented for ARM Neon and SSE on my simd branch https://github.com/nfeske/genode/commits/simd/.

nfeske added a commit to nfeske/genode-allwinner that referenced this issue Jan 21, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
a19ed76 
Issue genodelabs/genode#5428
nfeske added a commit to nfeske/genode-rpi that referenced this issue Jan 21, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
9effb1a 
Issue genodelabs/genode#5428
nfeske added a commit to nfeske/genode-imx that referenced this issue Jan 21, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
01bf582 
Issue genodelabs/genode#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
blit: SIMD-based back2front copy
59649b0 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
nitpicker: alloc padded capture buffer
459f0d6 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
17d23e9 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 21, 2025
@nfeske
fb_sdl: support rotate and flip
191517e 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
blit: SIMD-based back2front copy
dcaf515 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
nitpicker: alloc padded capture buffer
996317d 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
53df286 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
fb_sdl: support rotate and flip
32eebd2 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
fixup "blit: SIMD-based back2front copy" (Neon cache locality and pre…
9f592c7 
…fetching)

Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode that referenced this issue Jan 23, 2025
@nfeske @cnuke
blit: SIMD-based back2front copy
a708866 
Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode that referenced this issue Jan 23, 2025
@nfeske @cnuke
nitpicker: alloc padded capture buffer
d58a4cb 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode that referenced this issue Jan 23, 2025
@nfeske @cnuke
Capture::Connection::Screen: use Blit::back2front
2305b50 
Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode that referenced this issue Jan 23, 2025
@nfeske @cnuke
fb_sdl: support rotate and flip
e5dfdc2 
Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode that referenced this issue Jan 23, 2025
@nfeske @cnuke
fixup "blit: SIMD-based back2front copy" (Neon cache locality and pre…
482f51a 
…fetching)

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
fixup "blit: SIMD-based back2front copy" (Neon cache locality and pre…
61a712a 
…fetching)

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
fixup "blit: SIMD-based back2front copy" (slow: read sequentially)
4c9ad57 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
fixup "blit: SIMD-based back2front copy" (neon: read sequentially)
3f608b7 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
nitpicker: alloc padded capture buffer
72f01be 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
6ebd217 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 23, 2025
@nfeske
fb_sdl: support rotate and flip
bd460b7 
Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode that referenced this issue Jan 28, 2025
@nfeske @cnuke
blit: Neon specialization of Blend::xrgb_a
a42a488 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 28, 2025
@nfeske
blit: SIMD-based back2front copy
a7efdb4 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 28, 2025
@nfeske
nitpicker: alloc padded capture buffer
19128b7 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 28, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
8207409 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 28, 2025
@nfeske
fb_sdl: support rotate and flip
03071d5 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 28, 2025
@nfeske
blit: Blit::blend_xrgb_a
3bb478f 
This commit adds support for SIMD-based alpha blending, which speeds up
the alpha-compositing of the nitpicker GUI server by circa 300% on ARM
Neon and x86_64 using SSE4.1

Issue genodelabs#5428
cnuke pushed a commit to cnuke/genode-allwinner that referenced this issue Jan 29, 2025
@nfeske @cnuke
Capture::Connection::Screen: use Blit::back2front
360dc8c 
Issue genodelabs/genode#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 30, 2025
@nfeske
blit: SIMD-based back2front copy
2222786 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 30, 2025
@nfeske
nitpicker: alloc padded capture buffer
5a08839 
This is a precondition for using Blit::back2front at the driver side.

Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 30, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
4cffa4f 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 30, 2025
@nfeske
fb_sdl: support rotate and flip
465b83f 
Issue genodelabs#5428
nfeske added a commit to nfeske/genode that referenced this issue Jan 30, 2025
@nfeske
blit: Blit::blend_xrgb_a
afde65b 
This commit adds support for SIMD-based alpha blending, which speeds up
the alpha-compositing of the nitpicker GUI server by circa 300% on ARM
Neon and x86_64 using SSE4.1

Issue genodelabs#5428
@nfeske nfeske mentioned this issue Jan 30, 2025
Open
nfeske added a commit that referenced this issue Jan 30, 2025
@nfeske
blit: SIMD-based back2front copy
0a756e4 
Issue #5428
nfeske added a commit that referenced this issue Jan 30, 2025
@nfeske
nitpicker: alloc padded capture buffer
ef3cc09 
This is a precondition for using Blit::back2front at the driver side.

Issue #5428
nfeske added a commit that referenced this issue Jan 30, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
d0d0b69 
Issue #5428
nfeske added a commit that referenced this issue Jan 30, 2025
@nfeske
fb_sdl: support rotate and flip
1eca184 
Issue #5428
nfeske added a commit that referenced this issue Jan 30, 2025
@nfeske
blit: Blit::blend_xrgb_a
c9863e5 
This commit adds support for SIMD-based alpha blending, which speeds up
the alpha-compositing of the nitpicker GUI server by circa 300% on ARM
Neon and x86_64 using SSE4.1

Issue #5428
@nfeske
Copy link
Member Author

nfeske commented Jan 30, 2025

The commit series for ARM Neon and SSE4 for x86_64 went to staging today. I'll defer the AVX topic to a later point.

@nfeske nfeske changed the title SIMD-based blitting using ARM Neon and Intel AVX SIMD-based blitting using ARM Neon and Intel SSE4 Jan 30, 2025
@nfeske nfeske added the fixed label Jan 30, 2025
nfeske added a commit to genodelabs/genode-allwinner that referenced this issue Jan 31, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
2901ecf 
Issue genodelabs/genode#5428
nfeske added a commit to genodelabs/genode-imx that referenced this issue Jan 31, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
bb62f37 
Issue genodelabs/genode#5428
nfeske added a commit to genodelabs/genode-rpi that referenced this issue Jan 31, 2025
@nfeske
Capture::Connection::Screen: use Blit::back2front
d7f4bb4 
Issue genodelabs/genode#5428
@nfeske
Copy link
Member Author

nfeske commented Jan 31, 2025

Yesterday, I missed to push the adaptations in the rpi, allwinner, and imx repository. I updated those just now.

chelmuth added a commit that referenced this issue Jan 31, 2025
@chelmuth
fixup "blit: SIMD-based back2front copy" (test-blit used_apis)
1cf2b9f 
Issue #5428
@chelmuth
Copy link
Member

Besides driver/framebuffer/virtio and server/vmm also test/fb_bench is broken on arm_v8a.

nfeske added a commit that referenced this issue Feb 4, 2025
@nfeske
framebuffer/virtio: resolve uint64_t ambiguity
38c5ad2 
This ambiguity is caused by the inclusion of arm_neon.h by the blit
library header.

Issue #5428
nfeske added a commit that referenced this issue Feb 4, 2025
@nfeske
ARM vmm: resolve uint64_t ambiguity
013a9e9 
Issue #5428
nfeske added a commit that referenced this issue Feb 5, 2025
@nfeske
test/fb_bench: resolve uint64_t ambiguity
3fc1fff 
Issue #5428
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
feature fixed
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@chelmuth @nfeske