This patch is an incremental step towards supporting a flat symbol table.
It de-overloads the intrinsic functions by providing type-specific intrinsics
and arranging for automatically upgrading from the old overloaded name to
the new non-overloaded name. Specifically:
llvm.isunordered -> llvm.isunordered.f32, llvm.isunordered.f64
llvm.sqrt -> llvm.sqrt.f32, llvm.sqrt.f64
llvm.ctpop -> llvm.ctpop.i8, llvm.ctpop.i16, llvm.ctpop.i32, llvm.ctpop.i64
llvm.ctlz -> llvm.ctlz.i8, llvm.ctlz.i16, llvm.ctlz.i32, llvm.ctlz.i64
llvm.cttz -> llvm.cttz.i8, llvm.cttz.i16, llvm.cttz.i32, llvm.cttz.i64
New code should not use the overloaded intrinsic names. Warnings will be
emitted if they are used.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25366 91177308-0d34-0410-b5e6-96231b3b80d8
This file makes the helper functions for auto-upgrade of llvm assembly and
bytecode more accessible. This is part of de-overloading of intrinsic
functions to support the flat symbol table (no type planes).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25365 91177308-0d34-0410-b5e6-96231b3b80d8
"Cut up the runPasses method into smaller pieces. The small private
helpers should be easier to deal with when code shuffling arising
from creating the new specialised batchers, not to mention, they're
much easier to understand.
I inlined them, in case function call overhead would be noticeable --
doubtful."
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25213 91177308-0d34-0410-b5e6-96231b3b80d8
First step in refactoring the SymbolTable is to split it into two classes,
one for a symbol table of types and one for a symbol table of Values.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25175 91177308-0d34-0410-b5e6-96231b3b80d8
This fixes a "gccass" regression. The -debug-pass=Structure option now
prints all the appropriate output
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25151 91177308-0d34-0410-b5e6-96231b3b80d8
"added the asserts and casts, fixed the comments and started the break
down of the larger methods. A few more patches and the breakdown
should be complete."
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25142 91177308-0d34-0410-b5e6-96231b3b80d8
Fold:
seteq ({ short }* cast (int 1 to { short }*), { short }* null)
setlt ({ short }* cast (int 1 to { short }*), { short }* cast (int 2 to { short }*))
to false/true. These last two commonly occur in the output of compilers that
tag integers, like cozmic's scheme compiler.
Tested by Regression/Assembler/ConstantExprFold.llx
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25112 91177308-0d34-0410-b5e6-96231b3b80d8
This sanitises the world, blows away the specialisations and adds
traits per passmanager type -- seemed most natural.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25085 91177308-0d34-0410-b5e6-96231b3b80d8
"All this should do is create what will eventually be the specialised
passmanagers. Currently, the templates are inheriting them, once the
template specialisations' methods have been absorbed, patches
submitted method by method. I'll nuke the specialisations and have
the new objects inherit directly from passmanagert, and sanitise the
world of all references to templates.
"
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25053 91177308-0d34-0410-b5e6-96231b3b80d8
Implement the suggested check to ensure that out-of-range float constants
don't get accepted by LLVM accidentally. Adjust the supporting test cases
as well.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@24963 91177308-0d34-0410-b5e6-96231b3b80d8
packed types with an element count of 1, although more generic support is
coming. This allows LLVM to turn the following code:
void %foo(<1 x float> * %a) {
entry:
%tmp1 = load <1 x float> * %a;
%tmp2 = add <1 x float> %tmp1, %tmp1
store <1 x float> %tmp2, <1 x float> *%a
ret void
}
Into:
_foo:
lfs f0, 0(r3)
fadds f0, f0, f0
stfs f0, 0(r3)
blr
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@24416 91177308-0d34-0410-b5e6-96231b3b80d8
type when it gets refined. This allows us to hash on this crucial value.
* Fix several issues in TypeMap::RefineAbstractType that prevent it from
handling hash values that change correctly.
* Define hashTypeStructure to not always return 0. :)
This last part (which depends on the first two) speeds up gccld time on eon
from 3.78s to 2.75s with a release build (a 28% speedup!). This resolves
PR474.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@24372 91177308-0d34-0410-b5e6-96231b3b80d8
former of which takes much less work than the later. This speeds up linking
eon from 3.749 to 3.637s with a release build (about 3%).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@24338 91177308-0d34-0410-b5e6-96231b3b80d8
