Rounding error in subnormal division in compiler-rt __divtf3 for f128

[tgross35]

Multiplying f128 0x00000000000000000000000000000001 * 0x0001ffffffffffffffffffffffffffff should have the result 0x3f8e0000000000000000000000000001, which is produced by the GCC libraries (godbolt). Clang's compiler-rt returns a value with an incorrect exponent, 0x3f8f0000000000000000000000000001.

Reproduction, needs to link against compiler-rt divtf3.c with __divtf3 renamed to __divtf3_x so the system library doesn't get linked instead.

#define __STDC_WANT_IEC_60559_TYPES_EXT__

#include <stdio.h>
#include <stdlib.h>
#include <inttypes.h>

#if defined(__clang__) && (defined(__i386) || defined(__x86_64))
#define _Float128 __float128
#endif

_Float128 __divtf3_x(_Float128, _Float128);

typedef struct {
    uint64_t lower, upper;
} u128;


void f128_print(_Float128 val) {
    u128 ival = *((u128 *)(&val));

    #ifndef __clang__
    char buf[1024];
    strfromf128(buf, sizeof(buf), "%.32g", val);
    printf("%#018" PRIx64 "%016" PRIx64 " %s\n", ival.upper, ival.lower, buf);
    #else
    printf("%#018" PRIx64 "%016" PRIx64 " %lf\n", ival.upper, ival.lower, (double)val);
    #endif
}

_Float128 new_f128(uint64_t upper, uint64_t lower) {
    u128 val;
    val.lower = lower;
    val.upper = upper;
    return *((_Float128 *)(&val));
}

int main() {
    _Float128 a = new_f128(0x0, 0x1);
    _Float128 b = new_f128(0x0001ffffffffffff, 0xffffffffffffffff);
    f128_print(a);
    f128_print(b);
    _Float128 c = __divtf3_x(a, b);
    // _Float128 c = a / b;
    f128_print(c);

    return 0;
}

[tgross35]

@atrosinenko it looks like you did this implementation, do you have any ideas for how to fix this?