after all callees are inlined into the current graph.
NOTE: There's also a major bug fix for the BU pass in DataStructure.cpp,
which ensures that resolvable indirect calls are not moved out to the
globals graph, so that they are eventually inlined (if possible).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7189 91177308-0d34-0410-b5e6-96231b3b80d8
after all callers are inlined into the current graph.
(2) Optimize the way a graph is inlined into its callees in the TD phase:
(a) Use DSGraph::cloneReachableSubgraph to clone only a subgraph at
each call site, for faster inlining.
(b) Clone separately for the same callee at different call sites,
since only the reachable subgraph is being cloned, not the entire
caller graph.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7188 91177308-0d34-0410-b5e6-96231b3b80d8
and (2) faster inlining by cloning only reachable nodes. In particular:
(1) Added DSGraph::cloneReachableSubgraph and DSGraph::cloneReachableNodes
to clone the subgraph reachable from a set of root nodes, into the
current graph, merging the global nodes into thos in the current graph.
The TD pass now uses this for faster inlining, and so does the
next function.
(2) Added DSGraph::updateFromGlobalGraph() to rematerialize nodes from the
globals graph into the current graph in both BU and TD passes.
(3) `I' flags are removed from all nodes in the globals graph, because they
are difficult to maintain correctly and are not needed anyway.
(4) Aux. function calls are only removed to the globals graph if they
will never be resovled. (This is what fixed gap.) The immediate
reason is that if we took these out of a function (and moved them to
the globals graph) we would need to rematerialize these nodes into the
function graph for every function in the BU pass. The longer term
problem is that we would need to find a way to remove them from the
globals graph iff they have been resolved on all paths through the
call graph.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7187 91177308-0d34-0410-b5e6-96231b3b80d8
isUnresolvableFunc() (I thought I needed this externally.
I don't, but it's still nicer this way.)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7186 91177308-0d34-0410-b5e6-96231b3b80d8
remembered in valuesStoredInFunction, but never traced at function return,
and that's too late to be finding the error anyway).
Stores trace both the value and the address being stored to,
but after some experience I think only values should be traced.
The pointer hash table just fills up far too quickly if every
store address were traced.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7169 91177308-0d34-0410-b5e6-96231b3b80d8
(1) Cannot use ANDN(ot), ORN, and XORN for boolean ops, only bitwise ops.
(2) Conditional move instructions must distinguish signed and unsigned
condition codes, e.g., MOVLE vs. MOVLEU.
(3) Conditional-move-on-register was using the cond-move-on-cc opcodes,
which produces a valid-looking instruction with bogus registers!
(4) Here's a really cute one: dividing-by-2^k for negative numbers needs to
add 2^k-1 before shifting, not add 1 after shifting. Sadly, these
are the same when k=0 so our poor test case worked fine.
(5) Casting between signed and unsigned values was not correct:
completely reimplemented.
(6) Zero-extension on unsigned values was bogus: I was only doing the
SRL and not the SLLX before it. Don't know WHAT I was thinking!
(7) And the most important class of changes: Sign-extensions on signed values.
Signed values are not sign-extended after ordinary operations,
so they must be sign-extended before the following cases:
-- passing to an external or unknown function
-- returning from a function
-- using as operand 2 of DIV or REM
-- using as either operand of condition-code setting operation
(currently only SUBCC), with smaller than 32-bit operands
Also, a couple of improvements:
(1) Fold cast-to-bool into Not(bool). Need to do this for And, Or, XOR also.
(2) Convert SetCC-Const into a conditional-move-on-register (case 41)
if the constant is 0. This was only being done for branch-on-SetCC-Const
when the branch is folded with the SetCC-Const.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7159 91177308-0d34-0410-b5e6-96231b3b80d8
(1) An int CC live range must be spilled if there are any interferences,
even if no other "neighbour" in the interf. graph has been allocated
that reg. yet. This is actually true of any class with only one reg!
(2) SparcIntCCRegClass::colorIGNode sets the color even if the LR must
be spilled so that the machine-independent spill code doesn't have to
make the machine-dependent decision of which CC name to use based on
operand type: %xcc or %icc. (These are two halves of the same
register.)
(3) LR->isMarkedForSpill() is no longer the same as LR->hasColor().
These should never have been the same, and this is necessary now for #2.
(4) All RDCCR and WRCCR instructions are directly generated with the
phony number for %ccr so that EmitAssembly/EmitBinary doesn't have to
deal with this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7152 91177308-0d34-0410-b5e6-96231b3b80d8
(1) An int CC live range must be spilled if there are any interferences,
even if no other "neighbour" in the interf. graph has been allocated
that reg. yet. This is actually true of any class with only one reg!
(2) SparcIntCCRegClass::colorIGNode sets the color even if the LR must
be spilled so that the machine-independent spill code doesn't have to
make the machine-dependent decision of which CC name to use based on
operand type: %xcc or %icc. (These are two halves of the same register.)
(3) LR->isMarkedForSpill() is no longer the same as LR->hasColor().
These should never have been the same, and this is necessary now for #2.
(4) All RDCCR and WRCCR instructions are directly generated with the
phony number for %ccr so that EmitAssembly/EmitBinary doesn't have to
deal with this.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7151 91177308-0d34-0410-b5e6-96231b3b80d8
Unfortunately, reallocation also means that the pointer numbering will
change, so increase table size to try to avoid it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7130 91177308-0d34-0410-b5e6-96231b3b80d8
taking out the rule for compiling the test driver.
We need the native libraries for libinstr because they are directly
linked into the native CBE or LLC code. That is the only practical
way to debug them!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7129 91177308-0d34-0410-b5e6-96231b3b80d8
integer overflow):
We need to use %icc and not %xcc for comparisons on 32-bit or smaller
integer values.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7111 91177308-0d34-0410-b5e6-96231b3b80d8
area to avoid using up precious stack space within the 4095 offset limit
from %fp. Such objects that would themselves live at a large offset
were being put there already so this is a simple change.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7066 91177308-0d34-0410-b5e6-96231b3b80d8
deletes it, but we were merrily trying to fix the operands of that
instruction anyway! Instead, fix the replacement instruction.
(2) An Improvement: Check for and extract global values in all operands,
not just in known pointer operands. For example, they can occur in
call arguments, and probably other unforeseeable places as well.
This also eliminates the special-case handling of Load and Store.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7053 91177308-0d34-0410-b5e6-96231b3b80d8
not just an Instruction*, at least in one unfortunate case:
the first operand to the va_arg instruction.
Modify ValueToDefVecMap to map from Value*, not Instruction*.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7052 91177308-0d34-0410-b5e6-96231b3b80d8
(1) select: Ok to convert a pointer to a float or double.
(2) regalloc: Some MachineInstr* for caller-saving code before a call
were being inserted before and after the call!
(3) Don't insert the caller-saving instructions in the
MachineCodeForInstruction for the Call instruction.
*All* instructions generated by register allocation need to be
recorded in those maps, but it needs to be done uniformly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@7051 91177308-0d34-0410-b5e6-96231b3b80d8
-- added new targets %.trace-out-llc and %.trace-out-cbe
-- either TRACE=yes or TRACEM=yes is still needed and chooses how to trace
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6585 91177308-0d34-0410-b5e6-96231b3b80d8
TmpInstruction constructors because every TmpInstruction object has
to be registered with a MachineCodeForInstruction to prevent leaks.
This simplifies the user's code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6469 91177308-0d34-0410-b5e6-96231b3b80d8
preallocated. While reg-to-reg dependences were already handled, this
change required new code for adding edges to/from call instructions.
This was part of the extensive changes to the way code generation occurs
for function call arguments and return values.
See log for CodeGen/PhyRegAlloc.cpp.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6467 91177308-0d34-0410-b5e6-96231b3b80d8
globals in some other places may not have been pulled out either;
globals in phi operands were being put just before the phi instead of
in the predecessor basic blocks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6466 91177308-0d34-0410-b5e6-96231b3b80d8
call arguments and return values:
Now all copy operations before and after a call are generated during
selection instead of during register allocation.
The values are copied to virtual registers (or to the stack), but
in the former case these operands are marked with the correct physical
registers according to the calling convention.
Although this complicates scheduling and does not work well with
live range analysis, it simplifies the machine-dependent part of
register allocation.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6465 91177308-0d34-0410-b5e6-96231b3b80d8
-- setCC of FP type used by a Phi: have to save in reg.
-- setNE of FP type used by a branch: cannot use result directly in branch!
-- setCC used outside the same basic block: have to save in reg. for now
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6382 91177308-0d34-0410-b5e6-96231b3b80d8
Fixed spilling of %fcc[0-3] which are part of %fsr.
(2) Moved some machine-independent reg-class code to class TargetRegInfo
from SparcReg{Class,}Info.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6343 91177308-0d34-0410-b5e6-96231b3b80d8
and related functions and flags. Fixed several bugs where only
"isDef" was being checked, not "isDefAndUse".
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6342 91177308-0d34-0410-b5e6-96231b3b80d8
Fixed spilling of %fcc[0-3] which are part of %fsr.
(2) Moved some machine-independent reg-class code to class TargetRegInfo
from SparcReg{Class,}Info.
(3) Renamed MachienOperand::opIsDef to MachineOperand::opIsDefOnly()
and related functions and flags. Fixed several bugs where only
"isDef" was being checked, not "isDefAndUse".
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6341 91177308-0d34-0410-b5e6-96231b3b80d8
Fixed spilling of %fcc[0-3] which are part of %fsr.
Moved some machine-independent reg-class code to class TargetRegInfo
from SparcReg{Class,}Info.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@6339 91177308-0d34-0410-b5e6-96231b3b80d8
virtual register computed by an Instruction& X and replaces it with
a slot in the stack frame, allocated via alloca.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4964 91177308-0d34-0410-b5e6-96231b3b80d8
virtual register computed by an Instruction& X and replaces it with
a slot in the stack frame, allocated via alloca.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4963 91177308-0d34-0410-b5e6-96231b3b80d8
using the Cilk multi-threaded runtime system to execute parallel code.
The current version inserts too many sync() operations in the program
because it does not attempt to optimize their placement.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4962 91177308-0d34-0410-b5e6-96231b3b80d8
Cilk multi-threaded runtime system to execute parallel code. This file
exposes some routines needed for code generation for that pass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4961 91177308-0d34-0410-b5e6-96231b3b80d8
computes a dependence graph for data dependences on memory locations
using interprocedural Mod/Ref information.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4957 91177308-0d34-0410-b5e6-96231b3b80d8
(1) Make entries for Constant values in the ScalarMap.
(2) Set MOD bit for the node pointed to by the
argument of a free instruction.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4948 91177308-0d34-0410-b5e6-96231b3b80d8
fact that the incoming nodes may be merged away at intermediate
steps. Use an extra level of indirection via DSNodeHandles
to track the nodes being merged. All this now happens in a
static helper function MergeNodes().
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4947 91177308-0d34-0410-b5e6-96231b3b80d8
been missed if node *this got merged away due to recursive merging!
Also, links were not moved correctly if a node is collapsed.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4933 91177308-0d34-0410-b5e6-96231b3b80d8
(1) Applied patch from Casey to implement iterator::operator= correctly:
it should use a pointer, not a reference.
(2) Added operators == and !=, and method all().
(3) Important bug fix: excess bits need to be ignored in operations
like ==, count(), and all(). We do this by ensuring excess bits
in the last bitset are always 0.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4837 91177308-0d34-0410-b5e6-96231b3b80d8
before inlining their graphs into a function. To support this,
added flags to CloneFlags to strip/keep Mod/Ref bits.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4836 91177308-0d34-0410-b5e6-96231b3b80d8
We run removeDeadNodes() on the TD graph up front before using it.
(2) Major enhancement to printing of results: now we list the actual objects
that are mod/ref instead of just printing the bit vectors.
Also an important bug fix in TDDataStructures pass (no change here):
clear Mod/Ref bits of callers before inlining into a function.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4833 91177308-0d34-0410-b5e6-96231b3b80d8
For now, this just computes the #indirect call sites and
the avg. #callees per indirect call site (actually it prints
both totals and the average is their ratio).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4705 91177308-0d34-0410-b5e6-96231b3b80d8
because this is usually such a long list that it makes the output
useless anyway.
Disable checking exit status until we can find something more precise.
The exit status is too coarse-grain to be useful.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4468 91177308-0d34-0410-b5e6-96231b3b80d8
Vikram S. Adve