Add the ARM32 FP register table entries, simple arith, and args.
Lower some instructions, without much guarantee of
correctness. *Running* generated code will be risky
because the register allocator isn't aware of register
aliasing.
Fill in v{add,div,mul,sub}.f{32,64}, vmov, vldr
and vsqrt.f{32,64}. I tried to make the nacl-other-intrinsics
test not explode, so added vsqrt too. That was pretty
easy for sqrt, but then fabs tests also exploded. Those are not
truly fixed but are currently "fixed" by adding a FakeDef to
satisfy liveness.
Propagate float/double arguments to the right register
in lowerArguments, lowerCall, and propagate to s0/d0/q0
for lowerReturn. May need to double check the calling convention.
Currently can't test call-ret because vpush/vpop for prologues
and epilogues isn't done.
Legalize FP immediates to make the nacl-other-intrinsics sqrt
test happy. Use the correct type of load (vldr (.32 and .64 are
optional) instead of ldr{b,h,,d}).
Whether or not the float/vector instructions can be
predicated is a bit interesting. The float/double ones
can, but the SIMD versions cannot. E.g.
vadd<cond>.f32 s0, s0, s1 is okay
vadd<cond>.f32 q0, q0, q1 is not okay.
For now, just omit conditions from instructions that may
end up being reused for SIMD.
Split up the fp.pnacl.ll test into multiple ones so that
parts of lowering can be tested incrementally.
BUG= https://code.google.com/p/nativeclient/issues/detail?id=4076
R=stichnot@chromium.org
Review URL: https://codereview.chromium.org/1266263003 .
diff --git a/src/IceInstARM32.cpp b/src/IceInstARM32.cpp
index 70b6791..15664f7 100644
--- a/src/IceInstARM32.cpp
+++ b/src/IceInstARM32.cpp
@@ -28,12 +28,13 @@
namespace {
const struct TypeARM32Attributes_ {
- const char *WidthString; // b, h, <blank>, or d
+ const char *WidthString; // b, h, <blank>, or d
+ const char *VecWidthString; // i8, i16, i32, f32, f64
int8_t SExtAddrOffsetBits;
int8_t ZExtAddrOffsetBits;
} TypeARM32Attributes[] = {
-#define X(tag, elementty, width, sbits, ubits) \
- { width, sbits, ubits } \
+#define X(tag, elementty, int_width, vec_width, sbits, ubits) \
+ { int_width, vec_width, sbits, ubits } \
,
ICETYPEARM32_TABLE
#undef X
@@ -66,6 +67,10 @@
return TypeARM32Attributes[Ty].WidthString;
}
+const char *InstARM32::getVecWidthString(Type Ty) {
+ return TypeARM32Attributes[Ty].VecWidthString;
+}
+
const char *InstARM32Pred::predString(CondARM32::Cond Pred) {
return InstARM32CondAttributes[Pred].EmitString;
}
@@ -94,6 +99,18 @@
Inst->getSrc(0)->emit(Func);
}
+void InstARM32Pred::emitUnaryopFP(const char *Opcode, const InstARM32Pred *Inst,
+ const Cfg *Func) {
+ Ostream &Str = Func->getContext()->getStrEmit();
+ assert(Inst->getSrcSize() == 1);
+ Type SrcTy = Inst->getSrc(0)->getType();
+ Str << "\t" << Opcode << Inst->getPredicate() << getVecWidthString(SrcTy)
+ << "\t";
+ Inst->getDest()->emit(Func);
+ Str << ", ";
+ Inst->getSrc(0)->emit(Func);
+}
+
void InstARM32Pred::emitTwoAddr(const char *Opcode, const InstARM32Pred *Inst,
const Cfg *Func) {
if (!BuildDefs::dump())
@@ -123,6 +140,21 @@
Inst->getSrc(1)->emit(Func);
}
+void InstARM32::emitThreeAddrFP(const char *Opcode, const InstARM32 *Inst,
+ const Cfg *Func) {
+ if (!BuildDefs::dump())
+ return;
+ Ostream &Str = Func->getContext()->getStrEmit();
+ assert(Inst->getSrcSize() == 2);
+ Str << "\t" << Opcode << getVecWidthString(Inst->getDest()->getType())
+ << "\t";
+ Inst->getDest()->emit(Func);
+ Str << ", ";
+ Inst->getSrc(0)->emit(Func);
+ Str << ", ";
+ Inst->getSrc(1)->emit(Func);
+}
+
void InstARM32Pred::emitFourAddr(const char *Opcode, const InstARM32Pred *Inst,
const Cfg *Func) {
if (!BuildDefs::dump())
@@ -304,12 +336,6 @@
return ".L" + Func->getFunctionName() + "$local$__" + std::to_string(Number);
}
-InstARM32Ldr::InstARM32Ldr(Cfg *Func, Variable *Dest, OperandARM32Mem *Mem,
- CondARM32::Cond Predicate)
- : InstARM32Pred(Func, InstARM32::Ldr, 1, Dest, Predicate) {
- addSource(Mem);
-}
-
InstARM32Pop::InstARM32Pop(Cfg *Func, const VarList &Dests)
: InstARM32(Func, InstARM32::Pop, 0, nullptr), Dests(Dests) {
// Track modifications to Dests separately via FakeDefs.
@@ -363,8 +389,14 @@
template <> const char *InstARM32Rev::Opcode = "rev";
template <> const char *InstARM32Sxt::Opcode = "sxt"; // still requires b/h
template <> const char *InstARM32Uxt::Opcode = "uxt"; // still requires b/h
+// FP
+template <> const char *InstARM32Vsqrt::Opcode = "vsqrt";
// Mov-like ops
+template <> const char *InstARM32Ldr::Opcode = "ldr";
template <> const char *InstARM32Mov::Opcode = "mov";
+// FP
+template <> const char *InstARM32Vldr::Opcode = "vldr";
+template <> const char *InstARM32Vmov::Opcode = "vmov";
// Three-addr ops
template <> const char *InstARM32Adc::Opcode = "adc";
template <> const char *InstARM32Add::Opcode = "add";
@@ -381,6 +413,11 @@
template <> const char *InstARM32Sdiv::Opcode = "sdiv";
template <> const char *InstARM32Sub::Opcode = "sub";
template <> const char *InstARM32Udiv::Opcode = "udiv";
+// FP
+template <> const char *InstARM32Vadd::Opcode = "vadd";
+template <> const char *InstARM32Vdiv::Opcode = "vdiv";
+template <> const char *InstARM32Vmul::Opcode = "vmul";
+template <> const char *InstARM32Vsub::Opcode = "vsub";
// Four-addr ops
template <> const char *InstARM32Mla::Opcode = "mla";
template <> const char *InstARM32Mls::Opcode = "mls";
@@ -403,19 +440,19 @@
assert(getSrcSize() == 1);
Variable *Dest = getDest();
if (Dest->hasReg()) {
- IceString Opcode = "mov";
+ IceString ActualOpcode = Opcode;
Operand *Src0 = getSrc(0);
if (const auto *Src0V = llvm::dyn_cast<Variable>(Src0)) {
if (!Src0V->hasReg()) {
// Always use the whole stack slot. A 32-bit load has a larger range
// of offsets than 16-bit, etc.
- Opcode = IceString("ldr");
+ ActualOpcode = IceString("ldr");
}
} else {
if (llvm::isa<OperandARM32Mem>(Src0))
- Opcode = IceString("ldr") + getWidthString(Dest->getType());
+ ActualOpcode = IceString("ldr") + getWidthString(Dest->getType());
}
- Str << "\t" << Opcode << getPredicate() << "\t";
+ Str << "\t" << ActualOpcode << getPredicate() << "\t";
getDest()->emit(Func);
Str << ", ";
getSrc(0)->emit(Func);
@@ -436,6 +473,64 @@
llvm_unreachable("Not yet implemented");
}
+template <> void InstARM32Vldr::emit(const Cfg *Func) const {
+ if (!BuildDefs::dump())
+ return;
+ Ostream &Str = Func->getContext()->getStrEmit();
+ assert(getSrcSize() == 1);
+ assert(getDest()->hasReg());
+ Str << "\t"<< Opcode << getPredicate() << "\t";
+ getDest()->emit(Func);
+ Str << ", ";
+ getSrc(0)->emit(Func);
+}
+
+template <> void InstARM32Vldr::emitIAS(const Cfg *Func) const {
+ assert(getSrcSize() == 1);
+ (void)Func;
+ llvm_unreachable("Not yet implemented");
+}
+
+template <> void InstARM32Vmov::emit(const Cfg *Func) const {
+ if (!BuildDefs::dump())
+ return;
+ assert(CondARM32::AL == getPredicate());
+ Ostream &Str = Func->getContext()->getStrEmit();
+ assert(getSrcSize() == 1);
+ Variable *Dest = getDest();
+ if (Dest->hasReg()) {
+ IceString ActualOpcode = Opcode;
+ Operand *Src0 = getSrc(0);
+ if (const auto *Src0V = llvm::dyn_cast<Variable>(Src0)) {
+ if (!Src0V->hasReg()) {
+ ActualOpcode = IceString("vldr");
+ }
+ } else {
+ if (llvm::isa<OperandARM32Mem>(Src0))
+ ActualOpcode = IceString("vldr");
+ }
+ Str << "\t" << ActualOpcode << "\t";
+ getDest()->emit(Func);
+ Str << ", ";
+ getSrc(0)->emit(Func);
+ } else {
+ Variable *Src0 = llvm::cast<Variable>(getSrc(0));
+ assert(Src0->hasReg());
+ Str << "\t"
+ "vstr"
+ "\t";
+ Src0->emit(Func);
+ Str << ", ";
+ Dest->emit(Func);
+ }
+}
+
+template <> void InstARM32Vmov::emitIAS(const Cfg *Func) const {
+ assert(getSrcSize() == 1);
+ (void)Func;
+ llvm_unreachable("Not yet implemented");
+}
+
void InstARM32Br::emit(const Cfg *Func) const {
if (!BuildDefs::dump())
return;
@@ -547,37 +642,25 @@
Str << getName(Func) << ":";
}
-void InstARM32Ldr::emit(const Cfg *Func) const {
+template <> void InstARM32Ldr::emit(const Cfg *Func) const {
if (!BuildDefs::dump())
return;
Ostream &Str = Func->getContext()->getStrEmit();
assert(getSrcSize() == 1);
assert(getDest()->hasReg());
Type Ty = getSrc(0)->getType();
- Str << "\t"
- << "ldr" << getWidthString(Ty) << getPredicate() << "\t";
+ Str << "\t"<< Opcode << getWidthString(Ty) << getPredicate() << "\t";
getDest()->emit(Func);
Str << ", ";
getSrc(0)->emit(Func);
}
-void InstARM32Ldr::emitIAS(const Cfg *Func) const {
+template <> void InstARM32Ldr::emitIAS(const Cfg *Func) const {
assert(getSrcSize() == 1);
(void)Func;
llvm_unreachable("Not yet implemented");
}
-void InstARM32Ldr::dump(const Cfg *Func) const {
- if (!BuildDefs::dump())
- return;
- Ostream &Str = Func->getContext()->getStrDump();
- dumpDest(Func);
- Str << " = ";
- dumpOpcodePred(Str, "ldr", getDest()->getType());
- Str << " ";
- dumpSources(Func);
-}
-
template <> void InstARM32Movw::emit(const Cfg *Func) const {
if (!BuildDefs::dump())
return;
