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

Add image GPU loading in preprocimage #44

Merged
merged 13 commits into from
Aug 16, 2022
Merged

Add image GPU loading in preprocimage #44

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
489870f
Add TI kernel load
Aug 10, 2022
6491b62
Fix typo
Aug 10, 2022
937a9cb
Refact image processing in imfilter
ClaroHenrique Aug 13, 2022
9d30911
Add view_kernel function
ClaroHenrique Aug 13, 2022
4b0999f
Convert Array to CudaArray when using imfilter_gpu
ClaroHenrique Aug 13, 2022
891ce0f
Remove Float32 convertion
ClaroHenrique Aug 13, 2022
4142b24
RemovRefact imagepreproc
ClaroHenrique Aug 13, 2022
b3bf61c
Update src/imfilter.jl
juliohm Aug 14, 2022
cfdbcac
Update src/utils.jl
juliohm Aug 14, 2022
6cd4836
Update src/utils.jl
juliohm Aug 14, 2022
f5731e9
Update src/utils.jl
juliohm Aug 14, 2022
69c71dd
Refact array kernel convertion
Aug 15, 2022
3367837
Fix GPU tests
ClaroHenrique Aug 15, 2022
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
2 changes: 1 addition & 1 deletion src/imfilter.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ function imfilter_gpu(img, krn)
padkrn = padarray(krn, Fill(zero(T), ntuple(i->0, N), padsize))

# perform ifft(fft(img) .* conj.(fft(krn)))
fftimg = img |> CuArray |> CUFFT.fft
fftimg = img |> CUFFT.fft
fftkrn = padkrn |> CuArray |> CUFFT.fft
juliohm marked this conversation as resolved.
Show resolved Hide resolved
result = (fftimg .* conj.(fftkrn)) |> CUFFT.ifft

Expand Down
2 changes: 1 addition & 1 deletion src/iqsim.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -226,7 +226,7 @@ function iqsim(trainimg::AbstractArray{T,N}, tilesize::Dims{N},
rtile = CartesianIndex(start):CartesianIndex(finish)

# selected training image dataevent
TIdev = view(TI, rtile)
TIdev = view_kernel(TI, rtile)

# boundary cut mask
cutmask .= false
Expand Down
20 changes: 19 additions & 1 deletion src/utils.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,18 @@ function activation(hard, tile)
buff
end

array_cpu(array) = array

array_gpu(array) = CuArray(array)

const array_kernel = CUDA.functional() ? array_gpu : array_cpu

view_cpu(array, I) = view(array, I)

view_gpu(array, I) = Array(array[I])

const view_kernel = CUDA.functional() ? view_gpu : view_cpu

function imagepreproc(trainimg, soft, geoconfig)
padsize = geoconfig.padsize

Expand All @@ -73,7 +85,13 @@ function imagepreproc(trainimg, soft, geoconfig)
AUX, AUXTI
end

TI, SOFT
# load array to GPU, when available
TI_kernel = array_kernel(TI)
SOFT_kernel = map(SOFT) do (AUX, AUXTI)
AUX, array_kernel(AUXTI)
juliohm marked this conversation as resolved.
Show resolved Hide resolved
end

TI_kernel, SOFT_kernel
end

function finddisabled(trainimg, geoconfig)
Expand Down
4 changes: 2 additions & 2 deletions test/lowapi.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,7 @@ if CUDA.functional()
krn = rand(30, 10)

result_cpu = ImageQuilting.imfilter_cpu(img, krn)
result_gpu = ImageQuilting.imfilter_gpu(img, krn)
result_gpu = ImageQuilting.imfilter_gpu(CuArray(img), krn)
@test size(result_cpu) == size(result_gpu)
@test isapprox(result_cpu[:], result_gpu[:], atol=tolerance)

Expand All @@ -160,7 +160,7 @@ if CUDA.functional()
krn = rand(10, 20, 30)

result_cpu = ImageQuilting.imfilter_cpu(img, krn)
result_gpu = ImageQuilting.imfilter_gpu(img, krn)
result_gpu = ImageQuilting.imfilter_gpu(CuArray(img), krn)
@test size(result_cpu) == size(result_gpu)
@test isapprox(result_cpu[:], result_gpu[:], atol=tolerance)
end
Expand Down