Skip to content
This repository was archived by the owner on Feb 25, 2025. It is now read-only.

[Impeller] Fix 1-d grid computation for compute #42516

Merged
auto-submit merged 11 commits into from
Jun 3, 2023
Merged

[Impeller] Fix 1-d grid computation for compute #42516

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
cdf0fde
++
Jun 2, 2023
eed6d9d
Merge branch 'master' of github.com:flutter/engine into fix_grid_comp…
Jun 2, 2023
ab061d6
fix vulkan computation
Jun 2, 2023
8b3fed1
Update compute_pass_vk.cc
Jun 2, 2023
8966c2a
provide vulkan compute shaders with specialization constant for x wg …
Jun 2, 2023
93c1827
++
Jun 2, 2023
0bb2e9f
++
Jun 2, 2023
3cb5394
Update compute_pass_vk.cc
Jun 2, 2023
d8ed07a
licenses
Jun 2, 2023
7844258
Merge branch 'fix_grid_computation' of github.com:jonahwilliams/engin…
Jun 2, 2023
d02e1bb
maliocdiff
Jun 2, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions impeller/renderer/backend/metal/compute_pass_mtl.mm
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -258,8 +258,10 @@ static bool Bind(ComputePassBindingsCache& pass,

// Special case for linear processing.
if (height == 1) {
int64_t threadGroups =
std::max(width / maxTotalThreadsPerThreadgroup, 1LL);
int64_t threadGroups = std::max(

@jonahwilliams jonahwilliams Jun 2, 2023

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Forgot these were ints so we were capped at 1.

@dnfield dnfield Jun 2, 2023

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you add a test for this?

@jonahwilliams jonahwilliams Jun 2, 2023

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes. Doing so required me to update the vulkan compute shaders to provide a specialization constant which contains the actual wg size, as these need to agree with the value in compute_pass.vk for anything to work correctly.

static_cast<int64_t>(
std::ceil(width * 1.0 / maxTotalThreadsPerThreadgroup * 1.0)),
1LL);
[encoder dispatchThreadgroups:MTLSizeMake(threadGroups, 1, 1)
threadsPerThreadgroup:MTLSizeMake(maxTotalThreadsPerThreadgroup,
1, 1)];
Expand Down
21 changes: 14 additions & 7 deletions impeller/renderer/backend/vulkan/compute_pass_vk.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -252,14 +252,21 @@ bool ComputePassVK::OnEncodeCommands(const Context& context,
int64_t width = grid_size.width;
int64_t height = grid_size.height;

while (width > max_wg_size[0]) {
width = std::max(static_cast<int64_t>(1), width / 2);
// Special case for linear processing.
if (height == 1) {
int64_t threadGroups = std::max(
static_cast<int64_t>(std::ceil(width * 1.0 / max_wg_size[0] * 1.0)),
1LL);
cmd_buffer.dispatch(threadGroups, 1, 1);
} else {
while (width > max_wg_size[0]) {
width = std::max(static_cast<int64_t>(1), width / 2);
}
while (height > max_wg_size[1]) {
height = std::max(static_cast<int64_t>(1), height / 2);
}
cmd_buffer.dispatch(width, height, 1);
}
while (height > max_wg_size[1]) {
height = std::max(static_cast<int64_t>(1), height / 2);
}

cmd_buffer.dispatch(width, height, 1);
}
}

Expand Down