diff options
| author | Zhigang Gong <zhigang.gong@intel.com> | 2013-12-18 07:19:05 +0000 |
|---|---|---|
| committer | Zhigang Gong <zhigang.gong@intel.com> | 2013-12-23 18:41:35 +0800 |
| commit | 6e2d5ebf0a09590a3365302b088de20841ebd97e (patch) | |
| tree | eed6f6a61896d199fb2dbeacc3be798a42ce4963 | |
| parent | 64b4ceb4366a99ede6b1f7665166ce121f9aa2a9 (diff) | |
| download | beignet-6e2d5ebf0a09590a3365302b088de20841ebd97e.tar.gz | |
GBE: adjust instruction order for load/function call for vector.
The previous implementation generates code as below:
%33 = extractelement <4 x i8> %32, i32 0
%34 = extractelement <4 x i8> %32, i32 1
%35 = extractelement <4 x i8> %32, i32 2
%36 = extractelement <4 x i8> %32, i32 3
%32 = load <4 x i8> addrspace(1)* %31, align 4, !tbaa !3
It may bring some potential problems in the consequent optimization pass.
Now fix adjust the extractelement instruction after the load instruction.
%32 = load <4 x i8> addrspace(1)* %31, align 4, !tbaa !3
%33 = extractelement <4 x i8> %32, i32 0
%34 = extractelement <4 x i8> %32, i32 1
%35 = extractelement <4 x i8> %32, i32 2
%36 = extractelement <4 x i8> %32, i32 3
This patch also move the dead code elimination pass after the scalarize pass.
As after scalarize pass, there may be some opportunity to remove more dead
instructions.
Signed-off-by: Zhigang Gong <zhigang.gong@intel.com>
Reviewed-by: "Yang, Rong R" <rong.r.yang@intel.com>
| -rw-r--r-- | backend/src/llvm/llvm_scalarize.cpp | 8 | ||||
| -rw-r--r-- | backend/src/llvm/llvm_to_gen.cpp | 2 |
2 files changed, 9 insertions, 1 deletions
diff --git a/backend/src/llvm/llvm_scalarize.cpp b/backend/src/llvm/llvm_scalarize.cpp index 6394909f..35504f37 100644 --- a/backend/src/llvm/llvm_scalarize.cpp +++ b/backend/src/llvm/llvm_scalarize.cpp @@ -222,6 +222,12 @@ namespace gbe { return GetComponentCount(value->getType()); } + /* set to insert new instructions after the specified instruction.*/ + void setAppendPoint(Instruction *insn) { + BasicBlock::iterator next(insn); + builder->SetInsertPoint(++next); + } + DenseMap<Value*, VectorValues> vectorVals; Module* module; IRBuilder<>* builder; @@ -649,6 +655,7 @@ namespace gbe { case GEN_OCL_GET_IMAGE_WIDTH: case GEN_OCL_GET_IMAGE_HEIGHT: { + setAppendPoint(call); extractFromVector(call); break; } @@ -686,6 +693,7 @@ namespace gbe { bool Scalarize::scalarizeLoad(LoadInst* ld) { + setAppendPoint(ld); extractFromVector(ld); return false; } diff --git a/backend/src/llvm/llvm_to_gen.cpp b/backend/src/llvm/llvm_to_gen.cpp index ad07c5e3..e11e449a 100644 --- a/backend/src/llvm/llvm_to_gen.cpp +++ b/backend/src/llvm/llvm_to_gen.cpp @@ -83,11 +83,11 @@ namespace gbe passes.add(createScalarReplAggregatesPass()); // Break up allocas passes.add(createRemoveGEPPass(unit)); passes.add(createConstantPropagationPass()); - passes.add(createDeadInstEliminationPass()); // Remove simplified instructions passes.add(createLowerSwitchPass()); passes.add(createPromoteMemoryToRegisterPass()); passes.add(createGVNPass()); // Remove redundancies passes.add(createScalarizePass()); // Expand all vector ops + passes.add(createDeadInstEliminationPass()); // Remove simplified instructions passes.add(createGenPass(unit)); // Print the code extra optimization passes |