1. Be careful with cse "cheap" expressions. e.g. constant materialization. Only cse them when the common expression is local or in a direct predecessor. We don't want cse of cheap instruction causing other expressions to be spilled.
2. Watch out for the case where the expression doesn't itself uses a virtual register. e.g. lea of frame object. If the common expression itself is used by copies (common for passing addresses to function calls), don't perform the cse. Since these expressions do not use a register, it creates a live range but doesn't close any, we want to be very careful with increasing register pressure.
Note these are heuristics so machine cse doesn't make register allocator unhappy. Once we have proper live range splitting and re-materialization support in place, these should be evaluated again.
Now machine cse is almost always a win on llvm nightly tests on x86 and x86_64.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98121 91177308-0d34-0410-b5e6-96231b3b80d8
need to be MCized, but the last debug info thing are LEB and
cygwin specific (which the MC api doesn't support yet) and
one specific form of EmitReference which I'll tackle next.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98116 91177308-0d34-0410-b5e6-96231b3b80d8
method. With this, comments should end up on the same lines as the .byte
directives (for example) and we now get no output with:
$ llc CodeGen/X86/2009-02-12-DebugInfoVLA.ll -o - -filetype=null -asm-verbose
woot.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98105 91177308-0d34-0410-b5e6-96231b3b80d8
significant debug info testcases are now all going through
MCStreamer, though they print a lot of extraneous newlines to "O".
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98101 91177308-0d34-0410-b5e6-96231b3b80d8
The Caller cost info would be reset everytime a callee was inlined. If the
caller has lots of calls and there is some mutual recursion going on, the
caller cost info could be calculated many times.
This patch reduces inliner runtime from 240s to 0.5s for a function with 20000
small function calls.
This is a more conservative version of r98089 that doesn't break the clang
test CodeGenCXX/temp-order.cpp. That test relies on rather extreme inlining
for constant folding.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98099 91177308-0d34-0410-b5e6-96231b3b80d8
The Caller cost info would be reset everytime a callee was inlined. If the
caller has lots of calls and there is some mutual recursion going on, the
caller cost info could be calculated many times.
This patch reduces inliner runtime from 240s to 0.5s for a function with 20000
small function calls.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98089 91177308-0d34-0410-b5e6-96231b3b80d8
is preparatory to having PEI's scavenged frame index value reuse logic
properly distinguish types of frame values (e.g., whether the value is
stack-pointer relative or frame-pointer relative).
No functionality change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98086 91177308-0d34-0410-b5e6-96231b3b80d8
Make it so. (This patch is in LowerCall_Darwin, which seems
to be used by SVR4 code as well; since that doesn't belong here,
I haven't worried about this case.)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98077 91177308-0d34-0410-b5e6-96231b3b80d8
register is involved for thumb1. Work around this for the moment by only
re-using SP-relative offsets. This is temporary 'til the code can distinguish
multiple base registers.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98071 91177308-0d34-0410-b5e6-96231b3b80d8
Place the LSDA into the TEXT section for ARM platforms. This involves making the
encoding indirect, pcrel, and sdata4 instead of an absolute pointer. The
references to the type infos are then non-lazy pointers. Revision 98019 changed
the encoding of non-lazy pointers to add the symbol to the non-lazy pointer
definition if it's a local symbol (otherwise, it's external and set to '0' so
that the loader can adjust it to the real value). This paved the way for this
change to work on ARM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98068 91177308-0d34-0410-b5e6-96231b3b80d8
parallel builds: the gold plugin fails to link because the lto
library is in the middle of being written out by the linker.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@98054 91177308-0d34-0410-b5e6-96231b3b80d8
