draft.blender

Signed-off-by: Zhigang Gong <zhigang.gong@intel.com>

1 files changed, 7 insertions, 4 deletions

diff --git a/backend/src/llvm/llvm_to_gen.cpp b/backend/src/llvm/llvm_to_gen.cpp
index a4ce4a2b..767c3fa2 100644
--- a/backend/src/llvm/llvm_to_gen.cpp
+++ b/backend/src/llvm/llvm_to_gen.cpp
@@ -158,15 +158,16 @@ namespace gbe
     MPM.add(createIndVarSimplifyPass());        // Canonicalize indvars
     MPM.add(createLoopIdiomPass());             // Recognize idioms like memset.
     MPM.add(createLoopDeletionPass());          // Delete dead loops
-    MPM.add(createLoopUnrollPass()); //1024, 32, 1024, 512)); //Unroll loops
+//    MPM.add(createLoopUnrollPass()); //1024, 32, 1024, 512)); //Unroll loops
     if(optLevel > 0) {
-      MPM.add(createSROAPass(/*RequiresDomTree*/ false));
-      MPM.add(createGVNPass());                 // Remove redundancies
+//      MPM.add(createSROAPass(/*RequiresDomTree*/ false));
+//      MPM.add(createGVNPass());                 // Remove redundancies
     }
 #if LLVM_VERSION_MAJOR == 3 && LLVM_VERSION_MINOR >= 5
     // FIXME Workaround: we find that CustomLoopUnroll may increase register pressure greatly,
     // and it may even make som cl kernel cannot compile because of limited scratch memory for spill.
     // As we observe this under strict math. So we disable CustomLoopUnroll if strict math is enabled.
+#if 0
     if (!strictMath) {
       MPM.add(createCustomLoopUnrollPass()); //1024, 32, 1024, 512)); //Unroll loops
       MPM.add(createLoopUnrollPass()); //1024, 32, 1024, 512)); //Unroll loops
@@ -176,6 +177,7 @@ namespace gbe
       }
     }
 #endif
+#endif
     MPM.add(createMemCpyOptPass());             // Remove memcpy / form memset
     MPM.add(createSCCPPass());                  // Constant prop with SCCP
 
@@ -224,7 +226,7 @@ namespace gbe
   {
     std::string errInfo;
     std::unique_ptr<llvm::raw_fd_ostream> o = NULL;
-    if (OCL_OUTPUT_LLVM_BEFORE_LINK || OCL_OUTPUT_LLVM_AFTER_LINK || OCL_OUTPUT_LLVM_AFTER_GEN)
+    //if (OCL_OUTPUT_LLVM_BEFORE_LINK || OCL_OUTPUT_LLVM_AFTER_LINK || OCL_OUTPUT_LLVM_AFTER_GEN)
       o = std::unique_ptr<llvm::raw_fd_ostream>(new llvm::raw_fd_ostream(fileno(stdout), false));
 
     // Get the module from its file
@@ -297,6 +299,7 @@ namespace gbe
     passes.add(createCFGSimplificationPass());     // Merge & remove BBs
     passes.add(createLowerSwitchPass());           // simplify cfg will generate switch-case instruction
     passes.add(createScalarizePass());             // Expand all vector ops
+    passes.add(createPrintModulePass(*o));
 
     if(OCL_OUTPUT_CFG)
       passes.add(createCFGPrinterPass());