aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Be more careful about using the right mask when emitting gathers.

Browse Source 
Specifically, we had been using the full mask for all gathers, rather than
using the internal mask when we were loading from locally-declared arrays.
Thus, given code like:

  uniform float x[programCount] = { .. . };
  float xx = x[programIndex];

Previously we weren't generating a plain vector load to initialize xx, when
this code was in a function where it wasn't known that the mask was all on,
even though it should have.  Now it does.
This commit is contained in:
Matt Pharr
2011-10-17 20:25:52 -04:00
parent 19087e4761
commit 290032f4f5
4 changed files with 83 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
ctx.h
View File

@@ -342,13 +342,13 @@ public:
llvm::Value *GetElementPtrInst(llvm::Value *basePtr, int v0, int v1,
const char *name = NULL);
/** Load from the memory location(s) given by lvalue. The lvalue may
be varying, in which case this corresponds to a gather from the
multiple memory locations given by the array of pointer values
given by the lvalue. If the lvalue is not varying, then the type
parameter may be NULL. */
llvm::Value *LoadInst(llvm::Value *lvalue, const Type *type,
const char *name = NULL);
/** Load from the memory location(s) given by lvalue, using the given
mask. The lvalue may be varying, in which case this corresponds to
a gather from the multiple memory locations given by the array of
pointer values given by the lvalue. If the lvalue is not varying,
then both the mask pointer and the type pointer may be NULL. */
llvm::Value *LoadInst(llvm::Value *lvalue, llvm::Value *mask,
const Type *type, const char *name = NULL);
/** Emits an alloca instruction to allocate stack storage for the given
type. If a non-zero alignment is specified, the object is also
@@ -521,8 +521,8 @@ private:
void scatter(llvm::Value *rvalue, llvm::Value *lvalue,
llvm::Value *maskPtr, const Type *rvalueType);
llvm::Value *gather(llvm::Value *lvalue, const Type *type,
const char *name);
llvm::Value *gather(llvm::Value *lvalue, llvm::Value *mask,
const Type *type, const char *name);
void maskedStore(llvm::Value *rvalue, llvm::Value *lvalue,
const Type *rvalueType, llvm::Value *maskPtr);
};