llvm-svn: 18467

llvm-svn: 18466

correctly link globals whose LLVM types do not match.

This fixes several of the F2C SPEC FP benchmarks, which were failing this
due to the implementation of common blocks used by f2c.

llvm-svn: 18465

PR472

llvm-svn: 18459

gccas/gccld more than just a noop.

llvm-svn: 18456

This fixes the ugly darwin "cannot find symbols starting with __" issue.

Thanks for Owen/resistor for testing this out for me.

llvm-svn: 18454

to rewrite this to use relocations.

llvm-svn: 18453

When writing kernel, save the branches til the end. They are still put in the "right place" in the schedule, but sometimes when folding to make a kernel instructions are added between branches. This is wrong. To avoid this, we handle branches special.

llvm-svn: 18450

llvm-svn: 18449

-S

llvm-svn: 18440

llvm-svn: 18439

llvm-svn: 18436

Including alloca.h on Solaris brings in the prototype of strftime(), which
breaks compilation of CBE generated code.

llvm-svn: 18435

instead of 80-bits of precision.  This fixes PR467.

This change speeds up fldry on X86 with LLC from 7.32s on apoc to 4.68s.

llvm-svn: 18433

llvm-svn: 18432

* Make the linker find lib*.bca files now instead of lib*.bc since those
  are what the makefiles now generate for bytecode archives.
* Make sure the linker only links archives when LinkLibraries is called.
  Previously if it found a lib*.bc file and that file was a bytecode file,
  it would link in the entire bytecode. This could make -lc -lc fail with
  duplicate symbols error but it shouldn't as searching multiple libraries,
  even the same one more than once, is permitted.
* Now that the above problems are corrected, implement the dependent libs
  feature. After the module is linked with all specified libraries, the
  LinkLibraries function will obtain the set of dependent libraries from
  the linked modules and attemp to find and link against those libraries.

llvm-svn: 18428

llvm-svn: 18422

llvm-svn: 18419

in scary and unknown ways before we promote it.  This fixes the miscompilation
of 188.ammp that has been plauging us since a globalopt patch went in.

Thanks a ton to Tanya for helping me diagnose the problem!

llvm-svn: 18418

This doesn't fix anything that I'm aware of, just noticed it by inspection

llvm-svn: 18417

http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041122/021428.html

It broke Mutlisource/Applications/obsequi

llvm-svn: 18407

llvm-svn: 18406

llvm-svn: 18405

llvm-svn: 18404

llvm-svn: 18399

llvm-svn: 18398

llvm-svn: 18397

llvm-svn: 18391

llvm-svn: 18387

llvm-svn: 18386

llvm-svn: 18384

  %f0 = fmovs %f0
  %f1 = fmovs %f1

then just delete the FpMOVD pseudo-instruction instead.  Also, add
statistics and debug printouts.

llvm-svn: 18383

llvm-svn: 18381

   addi r3, r3, -1
instead of
   addi r3, r3, 1

for 'sub int X, 1'.

Secondarily, this fixes several cases where we could crash given an unsigned
constant.  And fixes a couple of minor missed optimization cases, such as
xor X, ~0U -> not X

llvm-svn: 18379

bytecode files. This should help linking substantially.

llvm-svn: 18378

This only fails on darwin or on X86 under valgrind.

llvm-svn: 18377

llvm-svn: 18376

llvm-svn: 18374

llvm-svn: 18371

llvm-svn: 18369