diff options
author | Matthias Braun <matze@braunis.de> | 2018-08-29 23:12:42 +0000 |
---|---|---|
committer | Matthias Braun <matze@braunis.de> | 2018-08-29 23:12:42 +0000 |
commit | b7b5860657ed4e31ab0e5472c4998eadbc084f05 (patch) | |
tree | 8e2b7d7e88d711ff67f0415745eb4405dae1c865 | |
parent | d0ab67f7d02023482c5dc06fcc14ea8488929684 (diff) | |
download | bcm5719-llvm-b7b5860657ed4e31ab0e5472c4998eadbc084f05.tar.gz bcm5719-llvm-b7b5860657ed4e31ab0e5472c4998eadbc084f05.zip |
Reverse subregister saved loops in register usage info collector; NFC
On AMDGPU we have 70 register classes, so iterating over all 70
each time and exiting is costly on the CPU, this flips the loop
around so that it loops over the 70 register classes first,
and exits without doing the inner loop if needed.
On my test just starting radv this takes
RegUsageInfoCollector::runOnMachineFunction
from 6.0% of total time to 2.7% of total time,
and reduces the startup from 2.24s to 2.19s
Patch by David Airlie!
Differential Revision: https://reviews.llvm.org/D48582
llvm-svn: 340993
-rw-r--r-- | llvm/lib/CodeGen/RegUsageInfoCollector.cpp | 45 |
1 files changed, 22 insertions, 23 deletions
diff --git a/llvm/lib/CodeGen/RegUsageInfoCollector.cpp b/llvm/lib/CodeGen/RegUsageInfoCollector.cpp index f1c442ac38a..9db2af9f962 100644 --- a/llvm/lib/CodeGen/RegUsageInfoCollector.cpp +++ b/llvm/lib/CodeGen/RegUsageInfoCollector.cpp @@ -166,28 +166,27 @@ computeCalleeSavedRegs(BitVector &SavedRegs, MachineFunction &MF) { } // Insert any register fully saved via subregisters. - for (unsigned PReg = 1, PRegE = TRI.getNumRegs(); PReg < PRegE; ++PReg) { - if (SavedRegs.test(PReg)) - continue; - - // Check if PReg is fully covered by its subregs. - bool CoveredBySubRegs = false; - for (const TargetRegisterClass *RC : TRI.regclasses()) - if (RC->CoveredBySubRegs && RC->contains(PReg)) { - CoveredBySubRegs = true; - break; - } - if (!CoveredBySubRegs) - continue; - - // Add PReg to SavedRegs if all subregs are saved. - bool AllSubRegsSaved = true; - for (MCSubRegIterator SR(PReg, &TRI, false); SR.isValid(); ++SR) - if (!SavedRegs.test(*SR)) { - AllSubRegsSaved = false; - break; - } - if (AllSubRegsSaved) - SavedRegs.set(PReg); + for (const TargetRegisterClass *RC : TRI.regclasses()) { + if (!RC->CoveredBySubRegs) + continue; + + for (unsigned PReg = 1, PRegE = TRI.getNumRegs(); PReg < PRegE; ++PReg) { + if (SavedRegs.test(PReg)) + continue; + + // Check if PReg is fully covered by its subregs. + if (!RC->contains(PReg)) + continue; + + // Add PReg to SavedRegs if all subregs are saved. + bool AllSubRegsSaved = true; + for (MCSubRegIterator SR(PReg, &TRI, false); SR.isValid(); ++SR) + if (!SavedRegs.test(*SR)) { + AllSubRegsSaved = false; + break; + } + if (AllSubRegsSaved) + SavedRegs.set(PReg); + } } } |