search for: emitmemmodrmbyt

Hi, The attached small path fixes X86InstrInfo::getMemModRMByteSize to be consistent with the emit code in X86CodeEmitter::emitMemModRMByte on amd64. Zoltan -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20090408/ad831982/attachment.html> -------------- next part -------------- A non-text attachment was scrubbed... Name: llvm.diff...

...; RegOpcode < 8 && RM < 8 && "ModRM Fields out of range!"' failed. lli((anonymous namespace)::PrintStackTrace()+0x15)[0x84ab499] /lib/tls/libc.so.6[0x3d7ec8] /lib/tls/libc.so.6(abort+0x15)[0x3d94f5] /lib/tls/libc.so.6[0x3d1619] lli((anonymous namespace)::Emitter::emitMemModRMByte(llvm::MachineInstr const&, unsigned int, unsigned int)+0x219)[0x81af7c9] Aborted Thanks, Arthur. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20060914/1b609019/attachment.html>

...; RegOpcode < 8 && RM < 8 && "ModRM Fields out of range!"' failed. lli((anonymous namespace)::PrintStackTrace()+0x15)[0x84ab499] /lib/tls/libc.so.6[0x3d7ec8] /lib/tls/libc.so.6(abort+0x15)[0x3d94f5] /lib/tls/libc.so.6[0x3d1619] lli((anonymous namespace)::Emitter::emitMemModRMByte(llvm::MachineInstr const&, unsigned int, unsigned int)+0x219)[0x81af7c9] Aborted

.... diff --git a/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp b/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp index 7952607..12a30cf 100644 --- a/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp +++ b/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp @@ -402,6 +402,56 @@ void X86MCCodeEmitter::EmitMemModRMByte(const MCInst &MI, unsigned Op, unsigned BaseRegNo = BaseReg ? GetX86RegNum(Base) : -1U; + // 16-bit addressing forms of the ModR/M byte have a different encoding for + // the R/M field and are far more limited in which registers can be used. + if (Is16BitMemOperand(MI, Op)) { + if...

...t/X86/MCTargetDesc/X86MCCodeEmitter.cpp > b/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp > index 7952607..12a30cf 100644 > --- a/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp > +++ b/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp > @@ -402,6 +402,56 @@ void X86MCCodeEmitter::EmitMemModRMByte(const MCInst > &MI, unsigned Op, > > unsigned BaseRegNo = BaseReg ? GetX86RegNum(Base) : -1U; > > + // 16-bit addressing forms of the ModR/M byte have a different encoding > for > + // the R/M field and are far more limited in which registers can be > used. > +...

...lvm/obj/tools/llvmc' 2) gcc 3.5 (mainline CVS) Mostly result using ~0 (signed int) expression as value of unsigned int function parameter. Compiling X86CodeEmitter.cpp /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp: In member function `void<unnamed>::Emitter::emitMemModRMByte(const llvm::MachineInstr&, unsigned int, unsigned int)': /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp:437: warning: converting of negative value `-0x000000001' to `unsigned int' /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp:4...

...'s either only one single byte for opcode or one additional ModR/M byte. Again, all the information is right there. 3. Some memory instructions needs a SIB byte. 4. Process the operands! Some special handling for operands that made up of the memory address. But the logic is all there in emitMemModRMByte. What do you think? Evan > > >> >> Suppose you have the size, can't you pass the desired buffer size to >> MachineCodeEmitter and have it do the right thing? > > Sure, that's what I'm doing. The emitter with the SizeEmitter class > will > give t...

...single byte for opcode or one additional ModR/M byte. Again, > all the information is right there. > 3. Some memory instructions needs a SIB byte. > 4. Process the operands! Some special handling for operands that made > up of the memory address. But the logic is all there in > emitMemModRMByte. > > What do you think? > > That's doable, but I'm afraid I'll have to duplicate many code (eg emitMemModRMByte -> sizeMemModRMByte and probably others). That's not a problem (I'm thinking bug fixes and improvements, we'll have to handle two duplicated...

...= 2; > if (CurOp != NumOps) > - emitConstant(MI.getOperand(CurOp++).getImm(), sizeOfImm(Desc)); > + emitConstant(MI.getOperand(CurOp++).getImm(), > X86InstrInfo::sizeOfImm(Desc)); > break; > } > case X86II::MRMDestMem: { > @@ -706,7 +571,7 @@ > emitMemModRMByte(MI, CurOp, getX86RegNum(MI.getOperand(CurOp > +4).getReg())); > CurOp += 5; > if (CurOp != NumOps) > - emitConstant(MI.getOperand(CurOp++).getImm(), sizeOfImm(Desc)); > + emitConstant(MI.getOperand(CurOp++).getImm(), > X86InstrInfo::sizeOfImm(Desc)); >...

...wanderer/pkg/build/llvm/src/llvm/include/llvm/PassSupport.h:171: warning: converting of negative value `-0x000000001' to `unsigned int' Compiling X86CodeEmitter.cpp /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp: In member function `void<unnamed>::Emitter::emitMemModRMByte(const llvm::MachineInstr&, unsigned int, unsigned int)': /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp:437: warning: converting of negative value `-0x000000001' to `unsigned int' /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp:4...

OK, here's a new patch that adds the infrastructure and the implementation for X86, ARM and PPC of GetInstSize and GetFunctionSize. Both functions are virtual functions defined in TargetInstrInfo.h. For X86, I moved some commodity functions from X86CodeEmitter to X86InstrInfo. What do you think? Nicolas Evan Cheng wrote: > > I think both of these belong to TargetInstrInfo. And

Evan Cheng wrote: > On Apr 1, 2008, at 12:50 AM, Nicolas Geoffray wrote: > > > That's a hack. :-) It is if you think that code emitter should only be used for actually writing somewhere the data. It is not if you find it another useful utility ;-) > Some targets already have ways to compute the exact > size of a function. See ARM::GetFunctionSize()

...rc/llvm/include/llvm/PassSupport.h:171: > warning: converting of negative value `-0x000000001' to `unsigned int' > > Compiling X86CodeEmitter.cpp > /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp: In > member function `void<unnamed>::Emitter::emitMemModRMByte(const > llvm::MachineInstr&, unsigned int, unsigned int)': > /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/X86CodeEmitter.cpp:437: > warning: converting of negative value `-0x000000001' to `unsigned int' > /home/wanderer/pkg/build/llvm/src/llvm/lib/Target/X86/...

...r opcode or one additional ModR/M byte. Again, >> all the information is right there. >> 3. Some memory instructions needs a SIB byte. >> 4. Process the operands! Some special handling for operands that made >> up of the memory address. But the logic is all there in >> emitMemModRMByte. >> >> What do you think? >> >> > > That's doable, but I'm afraid I'll have to duplicate many code (eg > emitMemModRMByte -> sizeMemModRMByte and probably others). That's > not a > problem (I'm thinking bug fixes and improvements, we&...