swr/rast: Fix alloca usage in jitter

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>

src/gallium/drivers/swr/rasterizer/jitter/builder.cpp

@@ -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);
+    }
 }

src/gallium/drivers/swr/rasterizer/jitter/builder.h

@@ -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"

src/gallium/drivers/swr/rasterizer/jitter/builder_mem.cpp

@@ -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));