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

better fix for the freezing issue #153 #290

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

better fix for the freezing issue #153 #290

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
27 changes: 0 additions & 27 deletions src/JPEGView/JPEGImage.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,10 +14,6 @@
#include <math.h>
#include <assert.h>

// Hacky workaround. Look at comment block in CJPEGImage::Resample()
// undefine this flag to investigate which optimization might cause that particular failure (TODO)
#define AVX_SSE_FREEZE_FALLBACK

///////////////////////////////////////////////////////////////////////////////////
// Static helpers
///////////////////////////////////////////////////////////////////////////////////
Expand Down Expand Up @@ -573,29 +569,6 @@ void* CJPEGImage::Resample(CSize fullTargetSize, CSize clippingSize, CPoint targ
EProcessingFlags eProcFlags, double dSharpen, double dRotation, EResizeType eResizeType) {

Helpers::CPUType cpu = CSettingsProvider::This().AlgorithmImplementation();
// NOTE: Hacky workaround... there is probably a very obscure bug in the AVX2 implementation
// which causes WaitForSingleObject to wait indefinitely on SampleUp_HQ_SIMD()
// when window dimensions are > 3224 pixels wide.
//
// This obscure bug ONLY manifests itself on RELEASE builds, and not DEBUG builds, meaning
// it happens only when optimization is turned on. I haven't had the chance to trial and error exactly
// which optimization flag causes the freeze, but that's a future TODO
//
// A lot of trial and error tests has a freeze happening at 3226 pixels wide, but not at 3224 pixels. (client rect pixels)
// (I was not able to test how this all behaves if it's 3224 pixels high, so, when displays get that cool, then we might have to revisit this hack)
// The AVX2 code is very advanced, and I don't have the expertise to debug it -sylikc
//
// The known workarounds based on GitHub issues is to either set CPUType=SSE or HighQualityResampling=false
//
// So, here, we detect and fallback to SSE when the conditions are met. To be safe, I set the limit at 3200 pixels
#ifdef AVX_SSE_FREEZE_FALLBACK
if (cpu == Helpers::CPU_AVX2 && clippingSize.cx > 3200) {
// only override the usage for SSE for these specific conditions
// AVX2 is supposed to be ~2.4x faster than SSE
cpu = Helpers::CPU_SSE;
}
#endif

EFilterType filter = CSettingsProvider::This().DownsamplingFilter();

if (fullTargetSize.cx > 65535 || fullTargetSize.cy > 65535) return NULL;
Expand Down
3 changes: 2 additions & 1 deletion src/JPEGView/ProcessingThreadPool.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,7 @@ bool CProcessingThreadPool::Process(CProcessingRequest* pRequest) {
if (m_nNumThreads == 0) {
CProcessingThread::DoProcess(pRequest, 0, nTargetCY);
} else {
if (nTargetCX * nTargetCY < 100000 || nTargetCY <= 12) {
if (nTargetCX * nTargetCY < 100000 || nTargetCY < 2 * pRequest->StripPadding) {
CProcessingThread::DoProcess(pRequest, 0, nTargetCY);
} else {
// Important: All slices must have a height dividable by 'StripPadding', except the last one
Expand All @@ -87,6 +87,7 @@ bool CProcessingThreadPool::Process(CProcessingRequest* pRequest) {
while ((nSliceCY = ~(pRequest->StripPadding - 1) & (nTargetCY / nNumThreadsUsed)) < pRequest->StripPadding) {
nNumThreadsUsed--;
}
assert(nNumThreadsUsed > 1);
int nLastCY = nTargetCY - (nNumThreadsUsed - 1)*nSliceCY;
volatile LONG nRequestThreadCounter = nNumThreadsUsed - 1;
int nCurrCY = 0;
Expand Down