summaryrefslogtreecommitdiff
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/arch/x86/isa/insts/general_purpose/data_transfer/stack_operations.py23
-rw-r--r--src/arch/x86/isa/microops/ldstop.isa22
-rw-r--r--src/cpu/base_dyn_inst.hh52
-rw-r--r--src/cpu/simple/atomic.cc113
-rw-r--r--src/cpu/simple/atomic.hh5
-rw-r--r--src/cpu/simple/timing.cc40
-rw-r--r--src/cpu/simple/timing.hh6
7 files changed, 252 insertions, 9 deletions
diff --git a/src/arch/x86/isa/insts/general_purpose/data_transfer/stack_operations.py b/src/arch/x86/isa/insts/general_purpose/data_transfer/stack_operations.py
index 5771242e6..5884d68c2 100644
--- a/src/arch/x86/isa/insts/general_purpose/data_transfer/stack_operations.py
+++ b/src/arch/x86/isa/insts/general_purpose/data_transfer/stack_operations.py
@@ -58,8 +58,9 @@ def macroop POP_R {
# Make the default data size of pops 64 bits in 64 bit mode
.adjust_env oszIn64Override
- ld reg, ss, [1, t0, rsp]
+ ld t1, ss, [1, t0, rsp]
addi rsp, rsp, dsz
+ mov reg, reg, t1
};
def macroop POP_M {
@@ -67,7 +68,7 @@ def macroop POP_M {
.adjust_env oszIn64Override
ld t1, ss, [1, t0, rsp]
- # Check stack address
+ cda seg, sib, disp
addi rsp, rsp, dsz
st t1, seg, sib, disp
};
@@ -78,7 +79,7 @@ def macroop POP_P {
rdip t7
ld t1, ss, [1, t0, rsp]
- # Check stack address
+ cda seg, sib, disp
addi rsp, rsp, dsz
st t1, seg, riprel, disp
};
@@ -87,8 +88,6 @@ def macroop PUSH_R {
# Make the default data size of pops 64 bits in 64 bit mode
.adjust_env oszIn64Override
- # This needs to work slightly differently from the other versions of push
- # because the -original- version of the stack pointer is what gets pushed
stupd reg, ss, [1, t0, rsp], "-env.dataSize"
};
@@ -118,7 +117,10 @@ def macroop PUSH_P {
};
def macroop PUSHA {
- # Check all the stack addresses.
+ # Check all the stack addresses. We'll assume that if the beginning and
+ # end are ok, then the stuff in the middle should be as well.
+ cda ss, [1, t0, rsp], "-env.dataSize"
+ cda ss, [1, t0, rsp], "-8 * env.dataSize"
stupd rax, ss, [1, t0, rsp], "-env.dataSize"
stupd rcx, ss, [1, t0, rsp], "-env.dataSize"
stupd rdx, ss, [1, t0, rsp], "-env.dataSize"
@@ -130,14 +132,17 @@ def macroop PUSHA {
};
def macroop POPA {
- # Check all the stack addresses.
- ld rdi, ss, [1, t0, rsp], "0 * env.dataSize"
+ # Check all the stack addresses. We'll assume that if the beginning and
+ # end are ok, then the stuff in the middle should be as well.
+ ld t1, ss, [1, t0, rsp], "0 * env.dataSize"
+ ld t2, ss, [1, t0, rsp], "7 * env.dataSize"
+ mov rdi, rdi, t1
ld rsi, ss, [1, t0, rsp], "1 * env.dataSize"
ld rbp, ss, [1, t0, rsp], "2 * env.dataSize"
ld rbx, ss, [1, t0, rsp], "4 * env.dataSize"
ld rdx, ss, [1, t0, rsp], "5 * env.dataSize"
ld rcx, ss, [1, t0, rsp], "6 * env.dataSize"
- ld rax, ss, [1, t0, rsp], "7 * env.dataSize"
+ mov rax, rax, t2
addi rsp, rsp, "8 * env.dataSize"
};
diff --git a/src/arch/x86/isa/microops/ldstop.isa b/src/arch/x86/isa/microops/ldstop.isa
index c9ace4790..61adde8d1 100644
--- a/src/arch/x86/isa/microops/ldstop.isa
+++ b/src/arch/x86/isa/microops/ldstop.isa
@@ -445,5 +445,27 @@ let {{
self.mnemonic = "lea"
microopClasses["lea"] = LeaOp
+
+
+ iop = InstObjParams("cda", "Cda", 'X86ISA::LdStOp',
+ {"code": '''
+ Addr paddr;
+ fault = xc->translateDataWriteAddr(EA, paddr,
+ dataSize, (1 << segment));
+ ''',
+ "ea_code": calculateEA})
+ header_output += MicroLeaDeclare.subst(iop)
+ decoder_output += MicroLdStOpConstructor.subst(iop)
+ exec_output += MicroLeaExecute.subst(iop)
+
+ class CdaOp(LdStOp):
+ def __init__(self, segment, addr, disp = 0,
+ dataSize="env.dataSize", addressSize="env.addressSize"):
+ super(CdaOp, self).__init__("NUM_INTREGS", segment,
+ addr, disp, dataSize, addressSize)
+ self.className = "Cda"
+ self.mnemonic = "cda"
+
+ microopClasses["cda"] = CdaOp
}};
diff --git a/src/cpu/base_dyn_inst.hh b/src/cpu/base_dyn_inst.hh
index 0f2a90bf6..74b250207 100644
--- a/src/cpu/base_dyn_inst.hh
+++ b/src/cpu/base_dyn_inst.hh
@@ -102,6 +102,9 @@ class BaseDynInst : public FastAlloc, public RefCounted
template <class T>
Fault read(Addr addr, T &data, unsigned flags);
+ Fault translateDataReadAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags);
+
/**
* Does a write to a given address.
* @param data The data to be written.
@@ -114,6 +117,9 @@ class BaseDynInst : public FastAlloc, public RefCounted
Fault write(T data, Addr addr, unsigned flags,
uint64_t *res);
+ Fault translateDataWriteAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags);
+
void prefetch(Addr addr, unsigned flags);
void writeHint(Addr addr, int size, unsigned flags);
Fault copySrcTranslate(Addr src);
@@ -838,6 +844,29 @@ class BaseDynInst : public FastAlloc, public RefCounted
};
template<class Impl>
+Fault
+BaseDynInst<Impl>::translateDataReadAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags)
+{
+ if (traceData) {
+ traceData->setAddr(vaddr);
+ }
+
+ reqMade = true;
+ Request *req = new Request();
+ req->setVirt(asid, vaddr, size, flags, PC);
+ req->setThreadContext(thread->readCpuId(), threadNumber);
+
+ fault = cpu->translateDataReadReq(req, thread);
+
+ if (fault == NoFault)
+ paddr = req->getPaddr();
+
+ delete req;
+ return fault;
+}
+
+template<class Impl>
template<class T>
inline Fault
BaseDynInst<Impl>::read(Addr addr, T &data, unsigned flags)
@@ -889,6 +918,29 @@ BaseDynInst<Impl>::read(Addr addr, T &data, unsigned flags)
}
template<class Impl>
+Fault
+BaseDynInst<Impl>::translateDataWriteAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags)
+{
+ if (traceData) {
+ traceData->setAddr(vaddr);
+ }
+
+ reqMade = true;
+ Request *req = new Request();
+ req->setVirt(asid, vaddr, size, flags, PC);
+ req->setThreadContext(thread->readCpuId(), threadNumber);
+
+ fault = cpu->translateDataWriteReq(req, thread);
+
+ if (fault == NoFault)
+ paddr = req->getPaddr();
+
+ delete req;
+ return fault;
+}
+
+template<class Impl>
template<class T>
inline Fault
BaseDynInst<Impl>::write(T data, Addr addr, unsigned flags, uint64_t *res)
diff --git a/src/cpu/simple/atomic.cc b/src/cpu/simple/atomic.cc
index 9f574e8be..4e52f9b33 100644
--- a/src/cpu/simple/atomic.cc
+++ b/src/cpu/simple/atomic.cc
@@ -366,6 +366,61 @@ AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
}
}
+Fault
+AtomicSimpleCPU::translateDataReadAddr(Addr vaddr, Addr & paddr,
+ int size, unsigned flags)
+{
+ // use the CPU's statically allocated read request and packet objects
+ Request *req = &data_read_req;
+
+ if (traceData) {
+ traceData->setAddr(vaddr);
+ }
+
+ //The block size of our peer.
+ int blockSize = dcachePort.peerBlockSize();
+ //The size of the data we're trying to read.
+ int dataSize = size;
+
+ bool firstTimeThrough = true;
+
+ //The address of the second part of this access if it needs to be split
+ //across a cache line boundary.
+ Addr secondAddr = roundDown(vaddr + dataSize - 1, blockSize);
+
+ if(secondAddr > vaddr)
+ dataSize = secondAddr - vaddr;
+
+ while(1) {
+ req->setVirt(0, vaddr, dataSize, flags, thread->readPC());
+
+ // translate to physical address
+ Fault fault = thread->translateDataReadReq(req);
+
+ //If there's a fault, return it
+ if (fault != NoFault)
+ return fault;
+
+ if (firstTimeThrough) {
+ paddr = req->getPaddr();
+ firstTimeThrough = false;
+ }
+
+ //If we don't need to access a second cache line, stop now.
+ if (secondAddr <= vaddr)
+ return fault;
+
+ /*
+ * Set up for accessing the second cache line.
+ */
+
+ //Adjust the size to get the remaining bytes.
+ dataSize = vaddr + size - secondAddr;
+ //And access the right address.
+ vaddr = secondAddr;
+ }
+}
+
#ifndef DOXYGEN_SHOULD_SKIP_THIS
template
@@ -524,6 +579,64 @@ AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
}
}
+Fault
+AtomicSimpleCPU::translateDataWriteAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags)
+{
+ // use the CPU's statically allocated write request and packet objects
+ Request *req = &data_write_req;
+
+ if (traceData) {
+ traceData->setAddr(vaddr);
+ }
+
+ //The block size of our peer.
+ int blockSize = dcachePort.peerBlockSize();
+
+ //The address of the second part of this access if it needs to be split
+ //across a cache line boundary.
+ Addr secondAddr = roundDown(vaddr + size - 1, blockSize);
+
+ //The size of the data we're trying to read.
+ int dataSize = size;
+
+ bool firstTimeThrough = true;
+
+ if(secondAddr > vaddr)
+ dataSize = secondAddr - vaddr;
+
+ dcache_latency = 0;
+
+ while(1) {
+ req->setVirt(0, vaddr, flags, flags, thread->readPC());
+
+ // translate to physical address
+ Fault fault = thread->translateDataWriteReq(req);
+
+ //If there's a fault or we don't need to access a second cache line,
+ //stop now.
+ if (fault != NoFault)
+ return fault;
+
+ if (firstTimeThrough) {
+ paddr = req->getPaddr();
+ firstTimeThrough = false;
+ }
+
+ if (secondAddr <= vaddr)
+ return fault;
+
+ /*
+ * Set up for accessing the second cache line.
+ */
+
+ //Adjust the size to get the remaining bytes.
+ dataSize = vaddr + size - secondAddr;
+ //And access the right address.
+ vaddr = secondAddr;
+ }
+}
+
#ifndef DOXYGEN_SHOULD_SKIP_THIS
diff --git a/src/cpu/simple/atomic.hh b/src/cpu/simple/atomic.hh
index 96429e5b1..f68f41a90 100644
--- a/src/cpu/simple/atomic.hh
+++ b/src/cpu/simple/atomic.hh
@@ -151,6 +151,11 @@ class AtomicSimpleCPU : public BaseSimpleCPU
template <class T>
Fault write(T data, Addr addr, unsigned flags, uint64_t *res);
+
+ Fault translateDataReadAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags);
+ Fault translateDataWriteAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags);
};
#endif // __CPU_SIMPLE_ATOMIC_HH__
diff --git a/src/cpu/simple/timing.cc b/src/cpu/simple/timing.cc
index f1e51ac70..998d0b017 100644
--- a/src/cpu/simple/timing.cc
+++ b/src/cpu/simple/timing.cc
@@ -293,6 +293,26 @@ TimingSimpleCPU::read(Addr addr, T &data, unsigned flags)
return fault;
}
+Fault
+TimingSimpleCPU::translateDataReadAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags)
+{
+ Request *req =
+ new Request(0, vaddr, size, flags, thread->readPC(), cpuId, 0);
+
+ if (traceData) {
+ traceData->setAddr(vaddr);
+ }
+
+ Fault fault = thread->translateDataWriteReq(req);
+
+ if (fault == NoFault)
+ paddr = req->getPaddr();
+
+ delete req;
+ return fault;
+}
+
#ifndef DOXYGEN_SHOULD_SKIP_THIS
template
@@ -411,6 +431,26 @@ TimingSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
return fault;
}
+Fault
+TimingSimpleCPU::translateDataWriteAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags)
+{
+ Request *req =
+ new Request(0, vaddr, size, flags, thread->readPC(), cpuId, 0);
+
+ if (traceData) {
+ traceData->setAddr(vaddr);
+ }
+
+ Fault fault = thread->translateDataWriteReq(req);
+
+ if (fault == NoFault)
+ paddr = req->getPaddr();
+
+ delete req;
+ return fault;
+}
+
#ifndef DOXYGEN_SHOULD_SKIP_THIS
template
diff --git a/src/cpu/simple/timing.hh b/src/cpu/simple/timing.hh
index 668b6ddaf..d7554f6de 100644
--- a/src/cpu/simple/timing.hh
+++ b/src/cpu/simple/timing.hh
@@ -189,9 +189,15 @@ class TimingSimpleCPU : public BaseSimpleCPU
template <class T>
Fault read(Addr addr, T &data, unsigned flags);
+ Fault translateDataReadAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags);
+
template <class T>
Fault write(T data, Addr addr, unsigned flags, uint64_t *res);
+ Fault translateDataWriteAddr(Addr vaddr, Addr &paddr,
+ int size, unsigned flags);
+
void fetch();
void completeIfetch(PacketPtr );
void completeDataAccess(PacketPtr );