https://sourceware.org/bugzilla/show_bug.cgi?id=31685
Bug ID: 31685
Summary: Issues when casting _Float16 to __bf16
Product: binutils
Version: 2.42
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: gas
Assignee: unassigned at sourceware dot org
Reporter: mat.mcroci at gmail dot com
Target Milestone: ---
When compiling the following code (assuming it's in a file called test.cpp)
#include <stdlib.h>
#include <cstdint>
#include <array>
#include <iostream>
#define SIZE 8
typedef _Float16 T;
//typedef volatile float T;
void fp16tobf16(_Float16 * f) {
__bf16 * b = reinterpret_cast<__bf16*>(f);
for(int i=0; i<SIZE; i++){
T temp = f[i];
b[i] = (__bf16) temp;
}
}
with:
g++ -c -S -march=sapphirerapids -O2 -std=c++23 -o test.s test.cpp
using gcc 13.2.0 compiled with binutils 2.42 using spack on a Sapphire Rapids
machine, the assembler does not do any casting and simply returns. You can also
check this behaviour on godbolt.
When using -O1 or -O0, the assembler generates the command __extendhfbf2 which
I assume correctly does the casting. However, when compiling the following very
similar code (assuming it's in a file called test2.cpp)
#include <stdlib.h>
#include <cstdint>
#include <array>
#include <iostream>
#define SIZE 8
typedef _Float16 T;
//typedef volatile float T;
void fp16tobf16(_Float16 * f) {
__bf16 * b = reinterpret_cast<__bf16*>(f);
for(int i=0; i<SIZE; i++){
T temp = f[i];
b[i] = (__bf16) temp;
}
}
int main(){
std::array<_Float16, SIZE> a{};
std::fill(a.begin(), a.end(), (_Float16) 1.7653432432424324);
fp16tobf16(a.data());
__bf16 * b = reinterpret_cast<__bf16*>(a.data());
std::cout << "\n";
for(int i=0; i<SIZE; i++)
std::cout << ((double) b[i]) << "\n";
return 0;
}
with:
g++ -march=sapphirerapids -O0 -std=c++23 -o test2.run test2.cpp
The binutils ld triggers the error:
binutils-2.42-xfrnc4ulsgaca6jysot5cb7reykux4t6/bin/ld: /tmp/ccJJ5ihK.o: in
function `fp16tobf16(_Float16*)':
main_fp16_to_bf16.cpp:15:(.text+0x39): undefined reference to `__extendhfbf2'
--
You are receiving this mail because:
You are on the CC list for the bug.
