fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-25 21:00:22 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

swr/rast: Fix alloca usage in jitter

Browse source 
Fix issue where temporary allocas were getting hoisted to function entry
unnecessarily. We now explicitly mark temporary allocas and skip hoisting
during the hoist pass. Shuold reduce stack usage.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
This commit is contained in:
George Kyriazis 2018-04-05 15:59:54 -05:00
parent 81371a5909
commit aa482014e5
3 changed files with 20 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

17
src/gallium/drivers/swr/rasterizer/jitter/builder.cpp
View file

@ -111,4 +111,21 @@ namespace SwrJit
mSimdVectorIntTy = ArrayType::get(mSimdInt32Ty, 4);
mSimdVectorTRTy = ArrayType::get(mSimdFP32Ty, 5);
}
/// @brief Mark this alloca as temporary to avoid hoisting later on
void Builder::SetTempAlloca(Value* inst)
{
AllocaInst* pAlloca = dyn_cast<AllocaInst>(inst);
SWR_ASSERT(pAlloca, "Unexpected non-alloca instruction");
MDNode* N = MDNode::get(JM()->mContext, MDString::get(JM()->mContext, "is_temp_alloca"));
pAlloca->setMetadata("is_temp_alloca", N);
}
bool Builder::IsTempAlloca(Value* inst)
{
AllocaInst* pAlloca = dyn_cast<AllocaInst>(inst);
SWR_ASSERT(pAlloca, "Unexpected non-alloca instruction");
return (pAlloca->getMetadata("is_temp_alloca") != nullptr);
}
}

2
src/gallium/drivers/swr/rasterizer/jitter/builder.h
View file

@ -96,6 +96,8 @@ namespace SwrJit
Type* mSimd32Int8Ty;
void SetTargetWidth(uint32_t width);
void SetTempAlloca(Value* inst);
bool IsTempAlloca(Value* inst);
#include "gen_builder.hpp"
#include "gen_builder_meta.hpp"

1
src/gallium/drivers/swr/rasterizer/jitter/builder_mem.cpp
View file

@ -229,6 +229,7 @@ namespace SwrJit
// store vSrc on the stack. this way we can select between a valid load address and the vSrc address
Value* vSrcPtr = ALLOCA(vSrc->getType());
SetTempAlloca(vSrcPtr);
STORE(vSrc, vSrcPtr);
vGather = UndefValue::get(VectorType::get(mDoubleTy, 4));