535e6c5fa4
Fourth of five patches adding RISC-V to GEM5. This patch adds the RV64A extension, which includes atomic memory instructions. These instructions atomically read a value from memory, modify it with a value contained in a source register, and store the original memory value in the destination register and modified value back into memory. Because this requires two memory accesses and GEM5 does not support two timing memory accesses in a single instruction, each of these instructions is split into two micro- ops: A "load" micro-op, which reads the memory, and a "store" micro-op, which modifies and writes it back. Each atomic memory instruction also has two bits that acquire and release a lock on its memory location. Additionally, there are atomic load and store instructions that only either load or store, but not both, and can acquire or release memory locks. Note that because the current implementation of RISC-V only supports one core and one thread, it doesn't make sense to make use of AMO instructions. However, they do form a standard extension of the RISC-V ISA, so they are included mostly as a placeholder for when multithreaded execution is implemented. As a result, any tests for their correctness in a future patch may be abbreviated. Patch 1 introduced RISC-V and implemented the base instruction set, RV64I; patch 2 implemented the integer multiply extension, RV64M; and patch 3 implemented the single- and double-precision floating point extensions, RV64FD. Patch 5 will add support for timing, minor, and detailed CPU models that isn't present in patches 1-4. [Added missing file amo.isa] [Replaced information removed from initial patch that was missed during division into multiple patches.] [Fixed some minor formatting issues.] [Fixed oversight where LR and SC didn't have both AQ and RL flags.] Signed-off by: Alec Roelke Signed-off by: Jason Lowe-Power <jason@lowepower.com>
1384 lines
52 KiB
C++
1384 lines
52 KiB
C++
// -*- mode:c++ -*-
|
|
|
|
// Copyright (c) 2015 RISC-V Foundation
|
|
// Copyright (c) 2016 The University of Virginia
|
|
// All rights reserved.
|
|
//
|
|
// Redistribution and use in source and binary forms, with or without
|
|
// modification, are permitted provided that the following conditions are
|
|
// met: redistributions of source code must retain the above copyright
|
|
// notice, this list of conditions and the following disclaimer;
|
|
// redistributions in binary form must reproduce the above copyright
|
|
// notice, this list of conditions and the following disclaimer in the
|
|
// documentation and/or other materials provided with the distribution;
|
|
// neither the name of the copyright holders nor the names of its
|
|
// contributors may be used to endorse or promote products derived from
|
|
// this software without specific prior written permission.
|
|
//
|
|
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
|
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
|
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
|
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
|
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
|
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
|
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
//
|
|
// Authors: Alec Roelke
|
|
|
|
////////////////////////////////////////////////////////////////////
|
|
//
|
|
// The RISC-V ISA decoder
|
|
//
|
|
|
|
decode OPCODE default Unknown::unknown() {
|
|
0x03: decode FUNCT3 {
|
|
format Load {
|
|
0x0: lb({{
|
|
Rd_sd = Mem_sb;
|
|
}});
|
|
0x1: lh({{
|
|
Rd_sd = Mem_sh;
|
|
}});
|
|
0x2: lw({{
|
|
Rd_sd = Mem_sw;
|
|
}});
|
|
0x3: ld({{
|
|
Rd_sd = Mem_sd;
|
|
}});
|
|
0x4: lbu({{
|
|
Rd = Mem_ub;
|
|
}});
|
|
0x5: lhu({{
|
|
Rd = Mem_uh;
|
|
}});
|
|
0x6: lwu({{
|
|
Rd = Mem_uw;
|
|
}});
|
|
}
|
|
}
|
|
|
|
0x07: decode FUNCT3 {
|
|
format Load {
|
|
0x2: flw({{
|
|
Fd_bits = (uint64_t)Mem_uw;
|
|
}});
|
|
0x3: fld({{
|
|
Fd_bits = Mem;
|
|
}});
|
|
}
|
|
}
|
|
|
|
0x0f: decode FUNCT3 {
|
|
format IOp {
|
|
0x0: fence({{
|
|
}}, IsNonSpeculative, IsMemBarrier, No_OpClass);
|
|
0x1: fence_i({{
|
|
}}, IsNonSpeculative, IsSerializeAfter, No_OpClass);
|
|
}
|
|
}
|
|
|
|
0x13: decode FUNCT3 {
|
|
format IOp {
|
|
0x0: addi({{
|
|
Rd_sd = Rs1_sd + imm;
|
|
}});
|
|
0x1: slli({{
|
|
Rd = Rs1 << SHAMT6;
|
|
}});
|
|
0x2: slti({{
|
|
Rd = (Rs1_sd < imm) ? 1 : 0;
|
|
}});
|
|
0x3: sltiu({{
|
|
Rd = (Rs1 < (uint64_t)imm) ? 1 : 0;
|
|
}});
|
|
0x4: xori({{
|
|
Rd = Rs1 ^ (uint64_t)imm;
|
|
}});
|
|
0x5: decode SRTYPE {
|
|
0x0: srli({{
|
|
Rd = Rs1 >> SHAMT6;
|
|
}});
|
|
0x1: srai({{
|
|
Rd_sd = Rs1_sd >> SHAMT6;
|
|
}});
|
|
}
|
|
0x6: ori({{
|
|
Rd = Rs1 | (uint64_t)imm;
|
|
}});
|
|
0x7: andi({{
|
|
Rd = Rs1 & (uint64_t)imm;
|
|
}});
|
|
}
|
|
}
|
|
|
|
0x17: UOp::auipc({{
|
|
Rd = PC + imm;
|
|
}});
|
|
|
|
0x1b: decode FUNCT3 {
|
|
format IOp {
|
|
0x0: addiw({{
|
|
Rd_sd = (int32_t)Rs1 + (int32_t)imm;
|
|
}});
|
|
0x1: slliw({{
|
|
Rd_sd = Rs1_sw << SHAMT5;
|
|
}});
|
|
0x5: decode SRTYPE {
|
|
0x0: srliw({{
|
|
Rd = Rs1_uw >> SHAMT5;
|
|
}});
|
|
0x1: sraiw({{
|
|
Rd_sd = Rs1_sw >> SHAMT5;
|
|
}});
|
|
}
|
|
}
|
|
}
|
|
|
|
0x23: decode FUNCT3 {
|
|
format Store {
|
|
0x0: sb({{
|
|
Mem_ub = Rs2_ub;
|
|
}});
|
|
0x1: sh({{
|
|
Mem_uh = Rs2_uh;
|
|
}});
|
|
0x2: sw({{
|
|
Mem_uw = Rs2_uw;
|
|
}});
|
|
0x3: sd({{
|
|
Mem_ud = Rs2_ud;
|
|
}});
|
|
}
|
|
}
|
|
|
|
0x27: decode FUNCT3 {
|
|
format Store {
|
|
0x2: fsw({{
|
|
Mem_uw = (uint32_t)Fs2_bits;
|
|
}});
|
|
0x3: fsd({{
|
|
Mem_ud = Fs2_bits;
|
|
}});
|
|
}
|
|
}
|
|
|
|
0x2f: decode FUNCT3 {
|
|
0x2: decode AMOFUNCT {
|
|
0x2: LoadReserved::lr_w({{
|
|
Rd_sd = Mem_sw;
|
|
}}, mem_flags=LLSC, aq=AQ, rl=RL);
|
|
0x3: StoreCond::sc_w({{
|
|
Mem_uw = Rs2_uw;
|
|
}}, {{
|
|
Rd = result;
|
|
}}, inst_flags=IsStoreConditional, mem_flags=LLSC, aq=AQ, rl=RL);
|
|
format AtomicMemOp {
|
|
0x0: amoadd_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = Rs2_sw + Rt_sd;
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x1: amoswap_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = Rs2_uw;
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x4: amoxor_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = Rs2_uw^Rt_sd;
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x8: amoor_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = Rs2_uw | Rt_sd;
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0xc: amoand_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = Rs2_uw&Rt_sd;
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x10: amomin_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = std::min<int32_t>(Rs2_sw, Rt_sd);
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x14: amomax_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = std::max<int32_t>(Rs2_sw, Rt_sd);
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x18: amominu_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = std::min<uint32_t>(Rs2_uw, Rt_sd);
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x1c: amomaxu_w({{Rt_sd = Mem_sw;}}, {{
|
|
Mem_sw = std::max<uint32_t>(Rs2_uw, Rt_sd);
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
}
|
|
}
|
|
0x3: decode AMOFUNCT {
|
|
0x2: LoadReserved::lr_d({{
|
|
Rd_sd = Mem_sd;
|
|
}}, aq=AQ, rl=RL);
|
|
0x3: StoreCond::sc_d({{
|
|
Mem = Rs2;
|
|
}}, {{
|
|
Rd = result;
|
|
}}, aq=AQ, rl=RL);
|
|
format AtomicMemOp {
|
|
0x0: amoadd_d({{Rt_sd = Mem_sd;}}, {{
|
|
Mem_sd = Rs2_sd + Rt_sd;
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x1: amoswap_d({{Rt = Mem;}}, {{
|
|
Mem = Rs2;
|
|
Rd = Rt;
|
|
}}, {{EA = Rs1;}});
|
|
0x4: amoxor_d({{Rt = Mem;}}, {{
|
|
Mem = Rs2^Rt;
|
|
Rd = Rt;
|
|
}}, {{EA = Rs1;}});
|
|
0x8: amoor_d({{Rt = Mem;}}, {{
|
|
Mem = Rs2 | Rt;
|
|
Rd = Rt;
|
|
}}, {{EA = Rs1;}});
|
|
0xc: amoand_d({{Rt = Mem;}}, {{
|
|
Mem = Rs2&Rt;
|
|
Rd = Rt;
|
|
}}, {{EA = Rs1;}});
|
|
0x10: amomin_d({{Rt_sd = Mem_sd;}}, {{
|
|
Mem_sd = std::min(Rs2_sd, Rt_sd);
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x14: amomax_d({{Rt_sd = Mem_sd;}}, {{
|
|
Mem_sd = std::max(Rs2_sd, Rt_sd);
|
|
Rd_sd = Rt_sd;
|
|
}}, {{EA = Rs1;}});
|
|
0x18: amominu_d({{Rt = Mem;}}, {{
|
|
Mem = std::min(Rs2, Rt);
|
|
Rd = Rt;
|
|
}}, {{EA = Rs1;}});
|
|
0x1c: amomaxu_d({{Rt = Mem;}}, {{
|
|
Mem = std::max(Rs2, Rt);
|
|
Rd = Rt;
|
|
}}, {{EA = Rs1;}});
|
|
}
|
|
}
|
|
}
|
|
0x33: decode FUNCT3 {
|
|
format ROp {
|
|
0x0: decode FUNCT7 {
|
|
0x0: add({{
|
|
Rd = Rs1_sd + Rs2_sd;
|
|
}});
|
|
0x1: mul({{
|
|
Rd = Rs1_sd*Rs2_sd;
|
|
}}, IntMultOp);
|
|
0x20: sub({{
|
|
Rd = Rs1_sd - Rs2_sd;
|
|
}});
|
|
}
|
|
0x1: decode FUNCT7 {
|
|
0x0: sll({{
|
|
Rd = Rs1 << Rs2<5:0>;
|
|
}});
|
|
0x1: mulh({{
|
|
bool negate = (Rs1_sd < 0) != (Rs2_sd < 0);
|
|
|
|
uint64_t Rs1_lo = (uint32_t)std::abs(Rs1_sd);
|
|
uint64_t Rs1_hi = (uint64_t)std::abs(Rs1_sd) >> 32;
|
|
uint64_t Rs2_lo = (uint32_t)std::abs(Rs2_sd);
|
|
uint64_t Rs2_hi = (uint64_t)std::abs(Rs2_sd) >> 32;
|
|
|
|
uint64_t hi = Rs1_hi*Rs2_hi;
|
|
uint64_t mid1 = Rs1_hi*Rs2_lo;
|
|
uint64_t mid2 = Rs1_lo*Rs2_hi;
|
|
uint64_t lo = Rs2_lo*Rs1_lo;
|
|
uint64_t carry = ((uint64_t)(uint32_t)mid1
|
|
+ (uint64_t)(uint32_t)mid2 + (lo >> 32)) >> 32;
|
|
|
|
uint64_t res = hi + (mid1 >> 32) + (mid2 >> 32) + carry;
|
|
Rd = negate ? ~res + (Rs1_sd*Rs2_sd == 0 ? 1 : 0) : res;
|
|
}}, IntMultOp);
|
|
}
|
|
0x2: decode FUNCT7 {
|
|
0x0: slt({{
|
|
Rd = (Rs1_sd < Rs2_sd) ? 1 : 0;
|
|
}});
|
|
0x1: mulhsu({{
|
|
bool negate = Rs1_sd < 0;
|
|
uint64_t Rs1_lo = (uint32_t)std::abs(Rs1_sd);
|
|
uint64_t Rs1_hi = (uint64_t)std::abs(Rs1_sd) >> 32;
|
|
uint64_t Rs2_lo = (uint32_t)Rs2;
|
|
uint64_t Rs2_hi = Rs2 >> 32;
|
|
|
|
uint64_t hi = Rs1_hi*Rs2_hi;
|
|
uint64_t mid1 = Rs1_hi*Rs2_lo;
|
|
uint64_t mid2 = Rs1_lo*Rs2_hi;
|
|
uint64_t lo = Rs1_lo*Rs2_lo;
|
|
uint64_t carry = ((uint64_t)(uint32_t)mid1
|
|
+ (uint64_t)(uint32_t)mid2 + (lo >> 32)) >> 32;
|
|
|
|
uint64_t res = hi + (mid1 >> 32) + (mid2 >> 32) + carry;
|
|
Rd = negate ? ~res + (Rs1_sd*Rs2 == 0 ? 1 : 0) : res;
|
|
}}, IntMultOp);
|
|
}
|
|
0x3: decode FUNCT7 {
|
|
0x0: sltu({{
|
|
Rd = (Rs1 < Rs2) ? 1 : 0;
|
|
}});
|
|
0x1: mulhu({{
|
|
uint64_t Rs1_lo = (uint32_t)Rs1;
|
|
uint64_t Rs1_hi = Rs1 >> 32;
|
|
uint64_t Rs2_lo = (uint32_t)Rs2;
|
|
uint64_t Rs2_hi = Rs2 >> 32;
|
|
|
|
uint64_t hi = Rs1_hi*Rs2_hi;
|
|
uint64_t mid1 = Rs1_hi*Rs2_lo;
|
|
uint64_t mid2 = Rs1_lo*Rs2_hi;
|
|
uint64_t lo = Rs1_lo*Rs2_lo;
|
|
uint64_t carry = ((uint64_t)(uint32_t)mid1
|
|
+ (uint64_t)(uint32_t)mid2 + (lo >> 32)) >> 32;
|
|
|
|
Rd = hi + (mid1 >> 32) + (mid2 >> 32) + carry;
|
|
}}, IntMultOp);
|
|
}
|
|
0x4: decode FUNCT7 {
|
|
0x0: xor({{
|
|
Rd = Rs1 ^ Rs2;
|
|
}});
|
|
0x1: div({{
|
|
if (Rs2_sd == 0) {
|
|
Rd_sd = -1;
|
|
} else if (Rs1_sd == std::numeric_limits<int64_t>::min()
|
|
&& Rs2_sd == -1) {
|
|
Rd_sd = std::numeric_limits<int64_t>::min();
|
|
} else {
|
|
Rd_sd = Rs1_sd/Rs2_sd;
|
|
}
|
|
}}, IntDivOp);
|
|
}
|
|
0x5: decode FUNCT7 {
|
|
0x0: srl({{
|
|
Rd = Rs1 >> Rs2<5:0>;
|
|
}});
|
|
0x1: divu({{
|
|
if (Rs2 == 0) {
|
|
Rd = std::numeric_limits<uint64_t>::max();
|
|
} else {
|
|
Rd = Rs1/Rs2;
|
|
}
|
|
}}, IntDivOp);
|
|
0x20: sra({{
|
|
Rd_sd = Rs1_sd >> Rs2<5:0>;
|
|
}});
|
|
}
|
|
0x6: decode FUNCT7 {
|
|
0x0: or({{
|
|
Rd = Rs1 | Rs2;
|
|
}});
|
|
0x1: rem({{
|
|
if (Rs2_sd == 0) {
|
|
Rd = Rs1_sd;
|
|
} else if (Rs1_sd == std::numeric_limits<int64_t>::min()
|
|
&& Rs2_sd == -1) {
|
|
Rd = 0;
|
|
} else {
|
|
Rd = Rs1_sd%Rs2_sd;
|
|
}
|
|
}}, IntDivOp);
|
|
}
|
|
0x7: decode FUNCT7 {
|
|
0x0: and({{
|
|
Rd = Rs1 & Rs2;
|
|
}});
|
|
0x1: remu({{
|
|
if (Rs2 == 0) {
|
|
Rd = Rs1;
|
|
} else {
|
|
Rd = Rs1%Rs2;
|
|
}
|
|
}}, IntDivOp);
|
|
}
|
|
}
|
|
}
|
|
|
|
0x37: UOp::lui({{
|
|
Rd = (uint64_t)imm;
|
|
}});
|
|
|
|
0x3b: decode FUNCT3 {
|
|
format ROp {
|
|
0x0: decode FUNCT7 {
|
|
0x0: addw({{
|
|
Rd_sd = Rs1_sw + Rs2_sw;
|
|
}});
|
|
0x1: mulw({{
|
|
Rd_sd = (int32_t)(Rs1_sw*Rs2_sw);
|
|
}}, IntMultOp);
|
|
0x20: subw({{
|
|
Rd_sd = Rs1_sw - Rs2_sw;
|
|
}});
|
|
}
|
|
0x1: sllw({{
|
|
Rd_sd = Rs1_sw << Rs2<4:0>;
|
|
}});
|
|
0x4: divw({{
|
|
if (Rs2_sw == 0) {
|
|
Rd_sd = -1;
|
|
} else if (Rs1_sw == std::numeric_limits<int32_t>::min()
|
|
&& Rs2_sw == -1) {
|
|
Rd_sd = std::numeric_limits<int32_t>::min();
|
|
} else {
|
|
Rd_sd = Rs1_sw/Rs2_sw;
|
|
}
|
|
}}, IntDivOp);
|
|
0x5: decode FUNCT7 {
|
|
0x0: srlw({{
|
|
Rd_uw = Rs1_uw >> Rs2<4:0>;
|
|
}});
|
|
0x1: divuw({{
|
|
if (Rs2_uw == 0) {
|
|
Rd_sd = std::numeric_limits<IntReg>::max();
|
|
} else {
|
|
Rd_sd = (int32_t)(Rs1_uw/Rs2_uw);
|
|
}
|
|
}}, IntDivOp);
|
|
0x20: sraw({{
|
|
Rd_sd = Rs1_sw >> Rs2<4:0>;
|
|
}});
|
|
}
|
|
0x6: remw({{
|
|
if (Rs2_sw == 0) {
|
|
Rd_sd = Rs1_sw;
|
|
} else if (Rs1_sw == std::numeric_limits<int32_t>::min()
|
|
&& Rs2_sw == -1) {
|
|
Rd_sd = 0;
|
|
} else {
|
|
Rd_sd = Rs1_sw%Rs2_sw;
|
|
}
|
|
}}, IntDivOp);
|
|
0x7: remuw({{
|
|
if (Rs2_uw == 0) {
|
|
Rd_sd = (int32_t)Rs1_uw;
|
|
} else {
|
|
Rd_sd = (int32_t)(Rs1_uw%Rs2_uw);
|
|
}
|
|
}}, IntDivOp);
|
|
}
|
|
}
|
|
|
|
format FPR4Op {
|
|
0x43: decode FUNCT2 {
|
|
0x0: fmadd_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fs3 = reinterpret_cast<float&>(temp = Fs3_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2) || std::isnan(fs3)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)
|
|
|| issignalingnan(fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else if (std::isinf(fs1) || std::isinf(fs2) ||
|
|
std::isinf(fs3)) {
|
|
if (std::signbit(fs1) == std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = std::numeric_limits<float>::infinity();
|
|
} else if (std::signbit(fs1) != std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = -std::numeric_limits<float>::infinity();
|
|
} else { // Fs3_sf is infinity
|
|
fd = fs3;
|
|
}
|
|
} else {
|
|
fd = fs1*fs2 + fs3;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatMultOp);
|
|
0x1: fmadd_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2) || std::isnan(Fs3)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)
|
|
|| issignalingnan(Fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else if (std::isinf(Fs1) || std::isinf(Fs2) ||
|
|
std::isinf(Fs3)) {
|
|
if (std::signbit(Fs1) == std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = std::numeric_limits<double>::infinity();
|
|
} else if (std::signbit(Fs1) != std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = -std::numeric_limits<double>::infinity();
|
|
} else {
|
|
Fd = Fs3;
|
|
}
|
|
} else {
|
|
Fd = Fs1*Fs2 + Fs3;
|
|
}
|
|
}}, FloatMultOp);
|
|
}
|
|
0x47: decode FUNCT2 {
|
|
0x0: fmsub_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fs3 = reinterpret_cast<float&>(temp = Fs3_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2) || std::isnan(fs3)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)
|
|
|| issignalingnan(fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else if (std::isinf(fs1) || std::isinf(fs2) ||
|
|
std::isinf(fs3)) {
|
|
if (std::signbit(fs1) == std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = std::numeric_limits<float>::infinity();
|
|
} else if (std::signbit(fs1) != std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = -std::numeric_limits<float>::infinity();
|
|
} else { // Fs3_sf is infinity
|
|
fd = -fs3;
|
|
}
|
|
} else {
|
|
fd = fs1*fs2 - fs3;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatMultOp);
|
|
0x1: fmsub_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2) || std::isnan(Fs3)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)
|
|
|| issignalingnan(Fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else if (std::isinf(Fs1) || std::isinf(Fs2) ||
|
|
std::isinf(Fs3)) {
|
|
if (std::signbit(Fs1) == std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = std::numeric_limits<double>::infinity();
|
|
} else if (std::signbit(Fs1) != std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = -std::numeric_limits<double>::infinity();
|
|
} else {
|
|
Fd = -Fs3;
|
|
}
|
|
} else {
|
|
Fd = Fs1*Fs2 - Fs3;
|
|
}
|
|
}}, FloatMultOp);
|
|
}
|
|
0x4b: decode FUNCT2 {
|
|
0x0: fnmsub_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fs3 = reinterpret_cast<float&>(temp = Fs3_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2) || std::isnan(fs3)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)
|
|
|| issignalingnan(fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else if (std::isinf(fs1) || std::isinf(fs2) ||
|
|
std::isinf(fs3)) {
|
|
if (std::signbit(fs1) == std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = -std::numeric_limits<float>::infinity();
|
|
} else if (std::signbit(fs1) != std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = std::numeric_limits<float>::infinity();
|
|
} else { // Fs3_sf is infinity
|
|
fd = fs3;
|
|
}
|
|
} else {
|
|
fd = -(fs1*fs2 - fs3);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatMultOp);
|
|
0x1: fnmsub_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2) || std::isnan(Fs3)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)
|
|
|| issignalingnan(Fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else if (std::isinf(Fs1) || std::isinf(Fs2)
|
|
|| std::isinf(Fs3)) {
|
|
if (std::signbit(Fs1) == std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = -std::numeric_limits<double>::infinity();
|
|
} else if (std::signbit(Fs1) != std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = std::numeric_limits<double>::infinity();
|
|
} else {
|
|
Fd = Fs3;
|
|
}
|
|
} else {
|
|
Fd = -(Fs1*Fs2 - Fs3);
|
|
}
|
|
}}, FloatMultOp);
|
|
}
|
|
0x4f: decode FUNCT2 {
|
|
0x0: fnmadd_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fs3 = reinterpret_cast<float&>(temp = Fs3_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2) || std::isnan(fs3)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)
|
|
|| issignalingnan(fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else if (std::isinf(fs1) || std::isinf(fs2) ||
|
|
std::isinf(fs3)) {
|
|
if (std::signbit(fs1) == std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = -std::numeric_limits<float>::infinity();
|
|
} else if (std::signbit(fs1) != std::signbit(fs2)
|
|
&& !std::isinf(fs3)) {
|
|
fd = std::numeric_limits<float>::infinity();
|
|
} else { // Fs3_sf is infinity
|
|
fd = -fs3;
|
|
}
|
|
} else {
|
|
fd = -(fs1*fs2 + fs3);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatMultOp);
|
|
0x1: fnmadd_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2) || std::isnan(Fs3)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)
|
|
|| issignalingnan(Fs3)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else if (std::isinf(Fs1) || std::isinf(Fs2) ||
|
|
std::isinf(Fs3)) {
|
|
if (std::signbit(Fs1) == std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = -std::numeric_limits<double>::infinity();
|
|
} else if (std::signbit(Fs1) != std::signbit(Fs2)
|
|
&& !std::isinf(Fs3)) {
|
|
Fd = std::numeric_limits<double>::infinity();
|
|
} else {
|
|
Fd = -Fs3;
|
|
}
|
|
} else {
|
|
Fd = -(Fs1*Fs2 + Fs3);
|
|
}
|
|
}}, FloatMultOp);
|
|
}
|
|
}
|
|
|
|
0x53: decode FUNCT7 {
|
|
format FPROp {
|
|
0x0: fadd_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else {
|
|
fd = fs1 + fs2;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatAddOp);
|
|
0x1: fadd_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else {
|
|
Fd = Fs1 + Fs2;
|
|
}
|
|
}}, FloatAddOp);
|
|
0x4: fsub_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else {
|
|
fd = fs1 - fs2;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatAddOp);
|
|
0x5: fsub_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else {
|
|
Fd = Fs1 - Fs2;
|
|
}
|
|
}}, FloatAddOp);
|
|
0x8: fmul_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else {
|
|
fd = fs1*fs2;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatMultOp);
|
|
0x9: fmul_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else {
|
|
Fd = Fs1*Fs2;
|
|
}
|
|
}}, FloatMultOp);
|
|
0xc: fdiv_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2)) {
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
} else {
|
|
fd = fs1/fs2;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatDivOp);
|
|
0xd: fdiv_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2)) {
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
} else {
|
|
Fd = Fs1/Fs2;
|
|
}
|
|
}}, FloatDivOp);
|
|
0x10: decode ROUND_MODE {
|
|
0x0: fsgnj_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (issignalingnan(fs1)) {
|
|
fd = std::numeric_limits<float>::signaling_NaN();
|
|
std::feclearexcept(FE_INVALID);
|
|
} else {
|
|
fd = std::copysign(fs1, fs2);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}});
|
|
0x1: fsgnjn_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (issignalingnan(fs1)) {
|
|
fd = std::numeric_limits<float>::signaling_NaN();
|
|
std::feclearexcept(FE_INVALID);
|
|
} else {
|
|
fd = std::copysign(fs1, -fs2);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}});
|
|
0x2: fsgnjx_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (issignalingnan(fs1)) {
|
|
fd = std::numeric_limits<float>::signaling_NaN();
|
|
std::feclearexcept(FE_INVALID);
|
|
} else {
|
|
fd = fs1*(std::signbit(fs2) ? -1.0 : 1.0);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}});
|
|
}
|
|
0x11: decode ROUND_MODE {
|
|
0x0: fsgnj_d({{
|
|
if (issignalingnan(Fs1)) {
|
|
Fd = std::numeric_limits<double>::signaling_NaN();
|
|
std::feclearexcept(FE_INVALID);
|
|
} else {
|
|
Fd = std::copysign(Fs1, Fs2);
|
|
}
|
|
}});
|
|
0x1: fsgnjn_d({{
|
|
if (issignalingnan(Fs1)) {
|
|
Fd = std::numeric_limits<double>::signaling_NaN();
|
|
std::feclearexcept(FE_INVALID);
|
|
} else {
|
|
Fd = std::copysign(Fs1, -Fs2);
|
|
}
|
|
}});
|
|
0x2: fsgnjx_d({{
|
|
if (issignalingnan(Fs1)) {
|
|
Fd = std::numeric_limits<double>::signaling_NaN();
|
|
std::feclearexcept(FE_INVALID);
|
|
} else {
|
|
Fd = Fs1*(std::signbit(Fs2) ? -1.0 : 1.0);
|
|
}
|
|
}});
|
|
}
|
|
0x14: decode ROUND_MODE {
|
|
0x0: fmin_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (issignalingnan(fs2)) {
|
|
fd = fs1;
|
|
FFLAGS |= FloatInvalid;
|
|
} else if (issignalingnan(fs1)) {
|
|
fd = fs2;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
fd = std::fmin(fs1, fs2);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatCmpOp);
|
|
0x1: fmax_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
float fd;
|
|
|
|
if (issignalingnan(fs2)) {
|
|
fd = fs1;
|
|
FFLAGS |= FloatInvalid;
|
|
} else if (issignalingnan(fs1)) {
|
|
fd = fs2;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
fd = std::fmax(fs1, fs2);
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatCmpOp);
|
|
}
|
|
0x15: decode ROUND_MODE {
|
|
0x0: fmin_d({{
|
|
if (issignalingnan(Fs2)) {
|
|
Fd = Fs1;
|
|
FFLAGS |= FloatInvalid;
|
|
} else if (issignalingnan(Fs1)) {
|
|
Fd = Fs2;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Fd = std::fmin(Fs1, Fs2);
|
|
}
|
|
}}, FloatCmpOp);
|
|
0x1: fmax_d({{
|
|
if (issignalingnan(Fs2)) {
|
|
Fd = Fs1;
|
|
FFLAGS |= FloatInvalid;
|
|
} else if (issignalingnan(Fs1)) {
|
|
Fd = Fs2;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Fd = std::fmax(Fs1, Fs2);
|
|
}
|
|
}}, FloatCmpOp);
|
|
}
|
|
0x20: fcvt_s_d({{
|
|
assert(CONV_SGN == 1);
|
|
float fd;
|
|
if (issignalingnan(Fs1)) {
|
|
fd = std::numeric_limits<float>::quiet_NaN();
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
fd = (float)Fs1;
|
|
}
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatCvtOp);
|
|
0x21: fcvt_d_s({{
|
|
assert(CONV_SGN == 0);
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
|
|
if (issignalingnan(fs1)) {
|
|
Fd = std::numeric_limits<double>::quiet_NaN();
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Fd = (double)fs1;
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x2c: fsqrt_s({{
|
|
assert(RS2 == 0);
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fd;
|
|
|
|
if (issignalingnan(Fs1_sf)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
fd = std::sqrt(fs1);
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(fd);
|
|
}}, FloatSqrtOp);
|
|
0x2d: fsqrt_d({{
|
|
assert(RS2 == 0);
|
|
Fd = std::sqrt(Fs1);
|
|
}}, FloatSqrtOp);
|
|
0x50: decode ROUND_MODE {
|
|
0x0: fle_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
Rd = 0;
|
|
} else {
|
|
Rd = fs1 <= fs2 ? 1 : 0;
|
|
}
|
|
}}, FloatCmpOp);
|
|
0x1: flt_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
|
|
if (std::isnan(fs1) || std::isnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
Rd = 0;
|
|
} else {
|
|
Rd = fs1 < fs2 ? 1 : 0;
|
|
}
|
|
}}, FloatCmpOp);
|
|
0x2: feq_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
float fs2 = reinterpret_cast<float&>(temp = Fs2_bits);
|
|
|
|
if (issignalingnan(fs1) || issignalingnan(fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Rd = fs1 == fs2 ? 1 : 0;
|
|
}}, FloatCmpOp);
|
|
}
|
|
0x51: decode ROUND_MODE {
|
|
0x0: fle_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
Rd = 0;
|
|
} else {
|
|
Rd = Fs1 <= Fs2 ? 1 : 0;
|
|
}
|
|
}}, FloatCmpOp);
|
|
0x1: flt_d({{
|
|
if (std::isnan(Fs1) || std::isnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
Rd = 0;
|
|
} else {
|
|
Rd = Fs1 < Fs2 ? 1 : 0;
|
|
}
|
|
}}, FloatCmpOp);
|
|
0x2: feq_d({{
|
|
if (issignalingnan(Fs1) || issignalingnan(Fs2)) {
|
|
FFLAGS |= FloatInvalid;
|
|
}
|
|
Rd = Fs1 == Fs2 ? 1 : 0;
|
|
}}, FloatCmpOp);
|
|
}
|
|
0x60: decode CONV_SGN {
|
|
0x0: fcvt_w_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
|
|
if (std::isnan(fs1)) {
|
|
Rd_sd = std::numeric_limits<int32_t>::max();
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Rd_sd = (int32_t)fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
if (std::signbit(fs1)) {
|
|
Rd_sd = std::numeric_limits<int32_t>::min();
|
|
} else {
|
|
Rd_sd = std::numeric_limits<int32_t>::max();
|
|
}
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x1: fcvt_wu_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
|
|
if (fs1 < 0.0) {
|
|
Rd = 0;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Rd = (uint32_t)fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
Rd = std::numeric_limits<uint64_t>::max();
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x2: fcvt_l_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
|
|
if (std::isnan(fs1)) {
|
|
Rd_sd = std::numeric_limits<int64_t>::max();
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Rd_sd = (int64_t)fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
if (std::signbit(fs1)) {
|
|
Rd_sd = std::numeric_limits<int64_t>::min();
|
|
} else {
|
|
Rd_sd = std::numeric_limits<int64_t>::max();
|
|
}
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x3: fcvt_lu_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
|
|
if (fs1 < 0.0) {
|
|
Rd = 0;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Rd = (uint64_t)fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
Rd = std::numeric_limits<uint64_t>::max();
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}
|
|
}}, FloatCvtOp);
|
|
}
|
|
0x61: decode CONV_SGN {
|
|
0x0: fcvt_w_d({{
|
|
Rd_sd = (int32_t)Fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
if (Fs1 < 0.0) {
|
|
Rd_sd = std::numeric_limits<int32_t>::min();
|
|
} else {
|
|
Rd_sd = std::numeric_limits<int32_t>::max();
|
|
}
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x1: fcvt_wu_d({{
|
|
if (Fs1 < 0.0) {
|
|
Rd = 0;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Rd = (uint32_t)Fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
Rd = std::numeric_limits<uint64_t>::max();
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x2: fcvt_l_d({{
|
|
Rd_sd = Fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
if (Fs1 < 0.0) {
|
|
Rd_sd = std::numeric_limits<int64_t>::min();
|
|
} else {
|
|
Rd_sd = std::numeric_limits<int64_t>::max();
|
|
}
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x3: fcvt_lu_d({{
|
|
if (Fs1 < 0.0) {
|
|
Rd = 0;
|
|
FFLAGS |= FloatInvalid;
|
|
} else {
|
|
Rd = (uint64_t)Fs1;
|
|
if (std::fetestexcept(FE_INVALID)) {
|
|
Rd = std::numeric_limits<uint64_t>::max();
|
|
std::feclearexcept(FE_INEXACT);
|
|
}
|
|
}
|
|
}}, FloatCvtOp);
|
|
}
|
|
0x68: decode CONV_SGN {
|
|
0x0: fcvt_s_w({{
|
|
float temp = (float)Rs1_sw;
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(temp);
|
|
}}, FloatCvtOp);
|
|
0x1: fcvt_s_wu({{
|
|
float temp = (float)Rs1_uw;
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(temp);
|
|
}}, FloatCvtOp);
|
|
0x2: fcvt_s_l({{
|
|
float temp = (float)Rs1_sd;
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(temp);
|
|
}}, FloatCvtOp);
|
|
0x3: fcvt_s_lu({{
|
|
float temp = (float)Rs1;
|
|
Fd_bits = (uint64_t)reinterpret_cast<uint32_t&>(temp);
|
|
}}, FloatCvtOp);
|
|
}
|
|
0x69: decode CONV_SGN {
|
|
0x0: fcvt_d_w({{
|
|
Fd = (double)Rs1_sw;
|
|
}}, FloatCvtOp);
|
|
0x1: fcvt_d_wu({{
|
|
Fd = (double)Rs1_uw;
|
|
}}, FloatCvtOp);
|
|
0x2: fcvt_d_l({{
|
|
Fd = (double)Rs1_sd;
|
|
}}, FloatCvtOp);
|
|
0x3: fcvt_d_lu({{
|
|
Fd = (double)Rs1;
|
|
}}, FloatCvtOp);
|
|
}
|
|
0x70: decode ROUND_MODE {
|
|
0x0: fmv_x_s({{
|
|
Rd = (uint32_t)Fs1_bits;
|
|
if ((Rd&0x80000000) != 0) {
|
|
Rd |= (0xFFFFFFFFULL << 32);
|
|
}
|
|
}}, FloatCvtOp);
|
|
0x1: fclass_s({{
|
|
uint32_t temp;
|
|
float fs1 = reinterpret_cast<float&>(temp = Fs1_bits);
|
|
switch (std::fpclassify(fs1)) {
|
|
case FP_INFINITE:
|
|
if (std::signbit(fs1)) {
|
|
Rd = 1 << 0;
|
|
} else {
|
|
Rd = 1 << 7;
|
|
}
|
|
break;
|
|
case FP_NAN:
|
|
if (issignalingnan(fs1)) {
|
|
Rd = 1 << 8;
|
|
} else {
|
|
Rd = 1 << 9;
|
|
}
|
|
break;
|
|
case FP_ZERO:
|
|
if (std::signbit(fs1)) {
|
|
Rd = 1 << 3;
|
|
} else {
|
|
Rd = 1 << 4;
|
|
}
|
|
break;
|
|
case FP_SUBNORMAL:
|
|
if (std::signbit(fs1)) {
|
|
Rd = 1 << 2;
|
|
} else {
|
|
Rd = 1 << 5;
|
|
}
|
|
break;
|
|
case FP_NORMAL:
|
|
if (std::signbit(fs1)) {
|
|
Rd = 1 << 1;
|
|
} else {
|
|
Rd = 1 << 6;
|
|
}
|
|
break;
|
|
default:
|
|
panic("Unknown classification for operand.");
|
|
break;
|
|
}
|
|
}});
|
|
}
|
|
0x71: decode ROUND_MODE {
|
|
0x0: fmv_x_d({{
|
|
Rd = Fs1_bits;
|
|
}}, FloatCvtOp);
|
|
0x1: fclass_d({{
|
|
switch (std::fpclassify(Fs1)) {
|
|
case FP_INFINITE:
|
|
if (std::signbit(Fs1)) {
|
|
Rd = 1 << 0;
|
|
} else {
|
|
Rd = 1 << 7;
|
|
}
|
|
break;
|
|
case FP_NAN:
|
|
if (issignalingnan(Fs1)) {
|
|
Rd = 1 << 8;
|
|
} else {
|
|
Rd = 1 << 9;
|
|
}
|
|
break;
|
|
case FP_ZERO:
|
|
if (std::signbit(Fs1)) {
|
|
Rd = 1 << 3;
|
|
} else {
|
|
Rd = 1 << 4;
|
|
}
|
|
break;
|
|
case FP_SUBNORMAL:
|
|
if (std::signbit(Fs1)) {
|
|
Rd = 1 << 2;
|
|
} else {
|
|
Rd = 1 << 5;
|
|
}
|
|
break;
|
|
case FP_NORMAL:
|
|
if (std::signbit(Fs1)) {
|
|
Rd = 1 << 1;
|
|
} else {
|
|
Rd = 1 << 6;
|
|
}
|
|
break;
|
|
default:
|
|
panic("Unknown classification for operand.");
|
|
break;
|
|
}
|
|
}});
|
|
}
|
|
0x78: fmv_s_x({{
|
|
Fd_bits = (uint64_t)Rs1_uw;
|
|
}}, FloatCvtOp);
|
|
0x79: fmv_d_x({{
|
|
Fd_bits = Rs1;
|
|
}}, FloatCvtOp);
|
|
}
|
|
}
|
|
0x63: decode FUNCT3 {
|
|
format SBOp {
|
|
0x0: beq({{
|
|
if (Rs1 == Rs2) {
|
|
NPC = PC + imm;
|
|
} else {
|
|
NPC = NPC;
|
|
}
|
|
}}, IsDirectControl, IsCondControl);
|
|
0x1: bne({{
|
|
if (Rs1 != Rs2) {
|
|
NPC = PC + imm;
|
|
} else {
|
|
NPC = NPC;
|
|
}
|
|
}}, IsDirectControl, IsCondControl);
|
|
0x4: blt({{
|
|
if (Rs1_sd < Rs2_sd) {
|
|
NPC = PC + imm;
|
|
} else {
|
|
NPC = NPC;
|
|
}
|
|
}}, IsDirectControl, IsCondControl);
|
|
0x5: bge({{
|
|
if (Rs1_sd >= Rs2_sd) {
|
|
NPC = PC + imm;
|
|
} else {
|
|
NPC = NPC;
|
|
}
|
|
}}, IsDirectControl, IsCondControl);
|
|
0x6: bltu({{
|
|
if (Rs1 < Rs2) {
|
|
NPC = PC + imm;
|
|
} else {
|
|
NPC = NPC;
|
|
}
|
|
}}, IsDirectControl, IsCondControl);
|
|
0x7: bgeu({{
|
|
if (Rs1 >= Rs2) {
|
|
NPC = PC + imm;
|
|
} else {
|
|
NPC = NPC;
|
|
}
|
|
}}, IsDirectControl, IsCondControl);
|
|
}
|
|
}
|
|
|
|
0x67: decode FUNCT3 {
|
|
0x0: Jump::jalr({{
|
|
Rd = NPC;
|
|
NPC = (imm + Rs1) & (~0x1);
|
|
}}, IsIndirectControl, IsUncondControl, IsCall);
|
|
}
|
|
|
|
0x6f: UJOp::jal({{
|
|
Rd = NPC;
|
|
NPC = PC + imm;
|
|
}}, IsDirectControl, IsUncondControl, IsCall);
|
|
|
|
0x73: decode FUNCT3 {
|
|
format IOp {
|
|
0x0: decode FUNCT12 {
|
|
0x0: ecall({{
|
|
fault = std::make_shared<SyscallFault>();
|
|
}}, IsSerializeAfter, IsNonSpeculative, IsSyscall, No_OpClass);
|
|
0x1: ebreak({{
|
|
fault = std::make_shared<BreakpointFault>();
|
|
}}, IsSerializeAfter, IsNonSpeculative, No_OpClass);
|
|
0x100: eret({{
|
|
fault = std::make_shared<UnimplementedFault>("eret");
|
|
}}, No_OpClass);
|
|
}
|
|
0x1: csrrw({{
|
|
Rd = xc->readMiscReg(FUNCT12);
|
|
xc->setMiscReg(FUNCT12, Rs1);
|
|
}}, IsNonSpeculative, No_OpClass);
|
|
0x2: csrrs({{
|
|
Rd = xc->readMiscReg(FUNCT12);
|
|
if (Rs1 != 0) {
|
|
xc->setMiscReg(FUNCT12, Rd | Rs1);
|
|
}
|
|
}}, IsNonSpeculative, No_OpClass);
|
|
0x3: csrrc({{
|
|
Rd = xc->readMiscReg(FUNCT12);
|
|
if (Rs1 != 0) {
|
|
xc->setMiscReg(FUNCT12, Rd & ~Rs1);
|
|
}
|
|
}}, IsNonSpeculative, No_OpClass);
|
|
0x5: csrrwi({{
|
|
Rd = xc->readMiscReg(FUNCT12);
|
|
xc->setMiscReg(FUNCT12, ZIMM);
|
|
}}, IsNonSpeculative, No_OpClass);
|
|
0x6: csrrsi({{
|
|
Rd = xc->readMiscReg(FUNCT12);
|
|
if (ZIMM != 0) {
|
|
xc->setMiscReg(FUNCT12, Rd | ZIMM);
|
|
}
|
|
}}, IsNonSpeculative, No_OpClass);
|
|
0x7: csrrci({{
|
|
Rd = xc->readMiscReg(FUNCT12);
|
|
if (ZIMM != 0) {
|
|
xc->setMiscReg(FUNCT12, Rd & ~ZIMM);
|
|
}
|
|
}}, IsNonSpeculative, No_OpClass);
|
|
}
|
|
}
|
|
}
|