The conditions for which Clang emits the `unsafe-fp-math` function
attribute has been modified as part of
`84a9ec2ff1ee97fd7e8ed988f5e7b197aab84a7`.
In the backend code generators `"unsafe-fp-math"="true"` enable floating
point contraction for the whole function.
The intent of the change in `84a9ec2ff1ee97fd7e8ed988f5e7b197aab84a7`
was to prevent backend code generators performing contractions when that
is not expected.
However the change is inaccurate and incomplete because it allows
`unsafe-fp-math` to be set also when only in-statement contraction is
allowed.
Consider the following example
```
float foo(float a, float b, float c) {
float tmp = a * b;
return tmp + c;
}
```
and compile it with the command line
```
clang -fno-math-errno -funsafe-math-optimizations -ffp-contract=on \
-O2 -mavx512f -S -o -
```
The resulting assembly has a `vfmadd213ss` instruction which corresponds
to a fused multiply-add. From the user perspective there shouldn't be
any contraction because the multiplication and the addition are not in
the same statement.
The optimized IR is:
```
define float @test(float noundef %a, float noundef %b, float noundef %c) #0 {
%mul = fmul reassoc nsz arcp afn float %b, %a
%add = fadd reassoc nsz arcp afn float %mul, %c
ret float %add
}
attributes #0 = {
[...]
"no-signed-zeros-fp-math"="true"
"no-trapping-math"="true"
[...]
"unsafe-fp-math"="true"
}
```
The `"unsafe-fp-math"="true"` function attribute allows the backend code
generator to perform `(fadd (fmul a, b), c) -> (fmadd a, b, c)`.
In the current IR representation there is no way to determine the
statement boundaries from the original source code.
Because of this for in-statement only contraction the generated IR
doesn't have instructions with the `contract` fast-math flag and
`llvm.fmuladd` is being used to represent contractions opportunities
that occur within a single statement.
Therefore `"unsafe-fp-math"="true"` can only be emitted when contraction
across statements is allowed.
Moreover the change in `84a9ec2ff1ee97fd7e8ed988f5e7b197aab84a7` doesn't
take into account that the floating point math function attributes can
be refined during IR code generation of a function to handle the cases
where the floating point math options are modified within a compound
statement via pragmas (see `CGFPOptionsRAII`).
For consistency `unsafe-fp-math` needs to be disabled if the contraction
mode for any scope/operation is not `fast`.
Similarly for consistency reason the initialization of `UnsafeFPMath` of
in `TargetOptions` for the backend code generation should take into
account the contraction mode as well.
Reviewed By: zahiraam
Differential Revision: https://reviews.llvm.org/D136786
IRgen optimization opportunities.
//===---------------------------------------------------------------------===//
The common pattern of
--
short x; // or char, etc
(x == 10)
--
generates an zext/sext of x which can easily be avoided.
//===---------------------------------------------------------------------===//
Bitfields accesses can be shifted to simplify masking and sign
extension. For example, if the bitfield width is 8 and it is
appropriately aligned then is is a lot shorter to just load the char
directly.
//===---------------------------------------------------------------------===//
It may be worth avoiding creation of alloca's for formal arguments
for the common situation where the argument is never written to or has
its address taken. The idea would be to begin generating code by using
the argument directly and if its address is taken or it is stored to
then generate the alloca and patch up the existing code.
In theory, the same optimization could be a win for block local
variables as long as the declaration dominates all statements in the
block.
NOTE: The main case we care about this for is for -O0 -g compile time
performance, and in that scenario we will need to emit the alloca
anyway currently to emit proper debug info. So this is blocked by
being able to emit debug information which refers to an LLVM
temporary, not an alloca.
//===---------------------------------------------------------------------===//
We should try and avoid generating basic blocks which only contain
jumps. At -O0, this penalizes us all the way from IRgen (malloc &
instruction overhead), all the way down through code generation and
assembly time.
On 176.gcc:expr.ll, it looks like over 12% of basic blocks are just
direct branches!
//===---------------------------------------------------------------------===//
b7d7c448df