https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118416
Bug ID: 118416
Summary: std::experimental::simd code detecting all zero is not
optimized to simple ptest on x86-64 avx
Product: gcc
Version: 14.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libstdc++
Assignee: unassigned at gcc dot gnu.org
Reporter: lee.imple at gmail dot com
Target Milestone: ---
The following code uses experimental c++ standard library simd, and wants to
detect several all-zero patterns that can be easily done with the vptest
instructions. All the code is available at https://godbolt.org/z/Kx68E1T6v .
```c++
#include <experimental/simd>
#include <cstdint>
namespace stdx = std::experimental;
template <class T, std::size_t N>
using simd_of = stdx::simd<T, stdx::simd_abi::deduce_t<T, N>>;
using data_t = simd_of<std::int32_t, 4>;
bool simple_ptest(data_t x) {
return all_of(x == 0);
}
bool ptest_and(data_t a, data_t b) {
return all_of((a & b) == 0);
}
bool ptest_andn(data_t a, data_t b) {
return all_of((a & ~b) == 0);
}
```
Equivalent assembly (hand-written):
```asm
simple_ptest:
vptest %xmm0, %xmm0
sete %al
ret
ptest_and:
vptest %xmm0, %xmm1
sete %al
ret
ptest_andn:
vptest %xmm0, %xmm1
setc %al
ret
```
But g++ generates the following code at `-O3 -march=x86-64-v3`, and clang++ and
even Intel icpx generates almost the same assembly.
```asm
simple_ptest(std::experimental::parallelism_v2::simd<int,
std::experimental::parallelism_v2::simd_abi::_VecBuiltin<16> >):
vpxor %xmm1, %xmm1, %xmm1
vpcmpeqd %xmm1, %xmm0, %xmm0
vpcmpeqd %xmm1, %xmm1, %xmm1
vptest %xmm1, %xmm0
setc %al
ret
ptest_and(std::experimental::parallelism_v2::simd<int,
std::experimental::parallelism_v2::simd_abi::_VecBuiltin<16> >,
std::experimental::parallelism_v2::simd<int,
std::experimental::parallelism_v2::simd_abi::_VecBuiltin<16> >):
vpand %xmm1, %xmm0, %xmm0
vpxor %xmm1, %xmm1, %xmm1
vpcmpeqd %xmm1, %xmm0, %xmm0
vpcmpeqd %xmm1, %xmm1, %xmm1
vptest %xmm1, %xmm0
setc %al
ret
ptest_andn(std::experimental::parallelism_v2::simd<int,
std::experimental::parallelism_v2::simd_abi::_VecBuiltin<16> >,
std::experimental::parallelism_v2::simd<int,
std::experimental::parallelism_v2::simd_abi::_VecBuiltin<16> >):
vpandn %xmm0, %xmm1, %xmm1
vpxor %xmm0, %xmm0, %xmm0
vpcmpeqd %xmm0, %xmm1, %xmm1
vpcmpeqd %xmm0, %xmm0, %xmm0
vptest %xmm0, %xmm1
setc %al
ret
```
I don't know whether this should be a missed optimization in g++ or a libstdc++
issue. Since these compilers generate the same output from the same library
code, I guess probably this should be a library issue.
Possibly related: PR58790 ?
