RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-31 20:58:11 +00:00
Code Issues Activity Actions

AK+LibX86: Generalize u128/256 to AK::UFixedBigInt

Browse source 
Doing these as custom classes might be faster, especially when writing
them in SSE, but this would cause a lot of Code duplication and due to
the nature of constexprs and the intelligence of the compiler they might
be using SSE/MMX either way
This commit is contained in:
Hendiadyoin1 2021-05-19 23:04:48 +02:00 • committed by Ali Mohammad Pur
parent 9270fb1e69
commit 5ffe23e4f3
11 changed files with 905 additions and 665 deletions
Download patch file Download diff file Expand all files Collapse all files

7
Userland/DevTools/UserspaceEmulator/SoftCPU.cpp
View file

@ -1855,9 +1855,10 @@ void SoftCPU::FLD_RM80(const X86::Instruction& insn)
// long doubles can be up to 128 bits wide in memory for reasons (alignment) and only uses 80 bits of precision
// GCC uses 12 bytes in 32 bit and 16 bytes in 64 bit mode
// so in the 32 bit case we read a bit to much, but that shouldn't be an issue.
auto new_f80 = insn.modrm().read128(*this, insn);
// FIXME: Respect shadow values
fpu_push(*(long double*)new_f80.value().bytes());
auto new_f80 = insn.modrm().read128(*this, insn).value();
fpu_push(*(long double*)new_f80.bytes().data());
}
void SoftCPU::FUCOMI(const X86::Instruction& insn)
@ -1908,7 +1909,7 @@ void SoftCPU::FSTP_RM80(const X86::Instruction& insn)
if constexpr (sizeof(long double) == 12)
f80 = insn.modrm().read128(*this, insn).value();
*(long double*)f80.bytes() = fpu_pop();
*(long double*)f80.bytes().data() = fpu_pop();
insn.modrm().write128(*this, insn, shadow_wrap_as_initialized(f80));
}