dynarmic/src/backend_x64/a64_jitstate.cpp

/* This file is part of the dynarmic project.
 * Copyright (c) 2016 MerryMage
 * This software may be used and distributed according to the terms of the GNU
 * General Public License version 2 or any later version.
 */

#include "backend_x64/a64_jitstate.h"
#include "common/bit_util.h"
#include "frontend/A64/location_descriptor.h"

namespace Dynarmic::BackendX64 {

u64 A64JitState::GetUniqueHash() const {
    u64 fpcr_u64 = static_cast<u64>(fpcr & A64::LocationDescriptor::FPCR_MASK) << 37;
    u64 pc_u64 = pc & A64::LocationDescriptor::PC_MASK;
    return pc_u64 | fpcr_u64;
}

/**
 * Comparing MXCSR and FPCR
 * ========================
 *
 * SSE MSCSR exception masks
 * -------------------------
 * PM   bit 12  Precision Mask
 * UM   bit 11  Underflow Mask
 * OM   bit 10  Overflow Mask
 * ZM   bit 9   Divide By Zero Mask
 * DM   bit 8   Denormal Mask
 * IM   bit 7   Invalid Operation Mask
 *
 * A64 FPCR exception trap enables
 * -------------------------------
 * IDE  bit 15  Input Denormal exception trap enable
 * IXE  bit 12  Inexact exception trap enable
 * UFE  bit 11  Underflow exception trap enable
 * OFE  bit 10  Overflow exception trap enable
 * DZE  bit 9   Division by Zero exception trap enable
 * IOE  bit 8   Invalid Operation exception trap enable
 *
 * SSE MXCSR mode bits
 * -------------------
 * FZ   bit 15  Flush To Zero
 * DAZ  bit 6   Denormals Are Zero
 * RN   bits 13-14  Round to {0 = Nearest, 1 = Negative, 2 = Positive, 3 = Zero}
 *
 * A64 FPCR mode bits
 * ------------------
 * AHP  bit 26  Alternative half-precision
 * DN   bit 25  Default NaN
 * FZ   bit 24  Flush to Zero
 * RMode    bits 22-23  Round to {0 = Nearest, 1 = Positive, 2 = Negative, 3 = Zero}
 * FZ16 bit 19  Flush to Zero for half-precision
 */

constexpr u32 FPCR_MASK = 0x07C89F00;

u32 A64JitState::GetFpcr() const {
    return fpcr;
}

void A64JitState::SetFpcr(u32 value) {
    fpcr = value & FPCR_MASK;

    guest_MXCSR &= 0x0000003D;
    guest_MXCSR |= 0x00001f80; // Mask all exceptions

    // RMode
    const std::array<u32, 4> MXCSR_RMode {0x0, 0x4000, 0x2000, 0x6000};
    guest_MXCSR |= MXCSR_RMode[(value >> 22) & 0x3];

    if (Common::Bit<24>(value)) {
        guest_MXCSR |= (1 << 15); // SSE Flush to Zero
        guest_MXCSR |= (1 << 6);  // SSE Denormals are Zero
    }
}

/**
 * Comparing MXCSR and FPSR
 * ========================
 *
 * SSE MXCSR exception flags
 * -------------------------
 * PE   bit 5   Precision Flag
 * UE   bit 4   Underflow Flag
 * OE   bit 3   Overflow Flag
 * ZE   bit 2   Divide By Zero Flag
 * DE   bit 1   Denormal Flag                                 // Appears to only be set when MXCSR.DAZ = 0
 * IE   bit 0   Invalid Operation Flag
 *
 * A64 FPSR cumulative exception bits
 * ----------------------------------
 * QC   bit 27  Cumulative saturation bit
 * IDC  bit 7   Input Denormal cumulative exception bit       // Only ever set when FPCR.FTZ = 1
 * IXC  bit 4   Inexact cumulative exception bit
 * UFC  bit 3   Underflow cumulative exception bit
 * OFC  bit 2   Overflow cumulative exception bit
 * DZC  bit 1   Division by Zero cumulative exception bit
 * IOC  bit 0   Invalid Operation cumulative exception bit
 */

u32 A64JitState::GetFpsr() const {
    u32 fpsr = 0;
    fpsr |= (guest_MXCSR & 0b0000000000001);       // IOC = IE
    fpsr |= (guest_MXCSR & 0b0000000111100) >> 1;  // IXC, UFC, OFC, DZC = PE, UE, OE, ZE
    fpsr |= FPSCR_IDC;
    fpsr |= FPSCR_UFC;
    return fpsr;
}

void A64JitState::SetFpsr(u32 value) {
    guest_MXCSR &= ~0x0000003D;
    guest_MXCSR |= ( value     ) & 0b0000000000001;  // IE = IOC
    guest_MXCSR |= ( value << 1) & 0b0000000111100;  // PE, UE, OE, ZE = IXC, UFC, OFC, DZC

    FPSCR_IDC = value & (1 << 7);
    FPSCR_UFC = value & (1 << 3);
}

} // namespace Dynarmic::BackendX64