[DAG] Fix typo preventing some stores merges to truncated stores.

[niravd]

Summary

Check actual memory type not extended value size when considering if
store merge is worthwhile.

[efriedma]

LGTM

llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

if (IsConstantSrc) {
           LastLegalType = i + 1;
           // Or check whether a truncstore is legal.
         } else if (TLI.getTypeAction(Context, StoreTy) ==
                    TargetLowering::TypePromoteInteger) {
           EVT LegalizedStoredValueTy =
               TLI.getTypeToTransformTo(Context, StoredVal.getValueType());
           if (TLI.isTruncStoreLegal(LegalizedStoredValueTy, StoreTy) &&
               TLI.canMergeStoresTo(FirstStoreAS, LegalizedStoredValueTy, DAG) &&
-              TLI.allowsMemoryAccess(Context, DL, LegalizedStoredValueTy,
-                                     FirstStoreAS, FirstStoreAlign, &IsFast) &&
+              TLI.allowsMemoryAccess(Context, DL, StoreTy, FirstStoreAS,
+                                     FirstStoreAlign, &IsFast) &&
               IsFast) {
             LastIntegerTrunc = true;
             LastLegalType = i + 1;
           }
         }
 
         // We only use vectors if the constant is known to be zero or the target
         // allows it and the function is not marked with the noimplicitfloat
for (unsigned i = 1; i < LoadNodes.size(); ++i) {
         EVT LegalizedStoredValueTy = TLI.getTypeToTransformTo(Context, StoreTy);
         if (TLI.isTruncStoreLegal(LegalizedStoredValueTy, StoreTy) &&
             TLI.canMergeStoresTo(FirstStoreAS, LegalizedStoredValueTy, DAG) &&
             TLI.isLoadExtLegal(ISD::ZEXTLOAD, LegalizedStoredValueTy,
                                StoreTy) &&
             TLI.isLoadExtLegal(ISD::SEXTLOAD, LegalizedStoredValueTy,
                                StoreTy) &&
             TLI.isLoadExtLegal(ISD::EXTLOAD, LegalizedStoredValueTy, StoreTy) &&
-            TLI.allowsMemoryAccess(Context, DL, LegalizedStoredValueTy,
-                                   FirstStoreAS, FirstStoreAlign, &IsFastSt) &&
+            TLI.allowsMemoryAccess(Context, DL, StoreTy, FirstStoreAS,
+                                   FirstStoreAlign, &IsFastSt) &&
             IsFastSt &&
             TLI.allowsMemoryAccess(Context, DL, StoreTy, FirstLoadAS,
                                    FirstLoadAlign, &IsFastLd) &&
             IsFastLd) {
           LastLegalIntegerType = i + 1;
           DoIntegerTruncate = true;
         }
       }

llvm/trunk/test/CodeGen/AMDGPU/merge-stores.ll

 ; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
 ; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
 
 ; This test is mostly to test DAG store merging, so disable the vectorizer.
 ; Run with devices with different unaligned load restrictions.
 
 ; TODO: Vector element tests
 ; TODO: Non-zero base offset for load and store combinations
 ; TODO: Same base addrspacecasted
 
 
 ; GCN-LABEL: {{^}}merge_global_store_2_constants_i8:
-; GCN: buffer_store_byte
-; GCN: buffer_store_byte
+; GCN: buffer_store_short
 ; GCN: s_endpgm
 define amdgpu_kernel void @merge_global_store_2_constants_i8(i8 addrspace(1)* %out) #0 {
   %out.gep.1 = getelementptr i8, i8 addrspace(1)* %out, i32 1
 
   store i8 123, i8 addrspace(1)* %out.gep.1
   store i8 456, i8 addrspace(1)* %out, align 2
   ret void
 }
define amdgpu_kernel void @merge_global_store_4_vector_elts_loads_v4i32(i32 addrspace(1)* %out, <4 x i32> addrspace(1)* %in) #0 {
   store i32 %x, i32 addrspace(1)* %out
   store i32 %y, i32 addrspace(1)* %out.gep.1
   store i32 %z, i32 addrspace(1)* %out.gep.2
   store i32 %w, i32 addrspace(1)* %out.gep.3
   ret void
 }
 
 ; GCN-LABEL: {{^}}merge_local_store_2_constants_i8:
-; GCN: ds_write_b8
-; GCN: ds_write_b8
+; GCN: ds_write_b16
 ; GCN: s_endpgm
 define amdgpu_kernel void @merge_local_store_2_constants_i8(i8 addrspace(3)* %out) #0 {
   %out.gep.1 = getelementptr i8, i8 addrspace(3)* %out, i32 1
 
   store i8 123, i8 addrspace(3)* %out.gep.1
   store i8 456, i8 addrspace(3)* %out, align 2
   ret void
 }

llvm/trunk/test/CodeGen/BPF/undef.ll

 ; CHECK: r1 += -2
 ; CHECK: r2 = 0
 ; CHECK: *(u16 *)(r1 + 6) = r2
 ; CHECK: *(u16 *)(r1 + 4) = r2
 ; CHECK: *(u16 *)(r1 + 2) = r2
 ; EL: r1 = 134678021
 ; EB: r1 = 84281096
 ; CHECK: *(u32 *)(r10 - 8) = r1
-; CHECK: r1 = 9
-; CHECK: *(u8 *)(r10 - 4) = r1
-; CHECK: r1 = 10
-; CHECK: *(u8 *)(r10 - 3) = r1
+; EL: r1 = 2569
+; EB: r1 = 2314
+; CHECK: *(u16 *)(r10 - 4) = r1
 ; CHECK: *(u16 *)(r10 + 24) = r2
 ; CHECK: *(u16 *)(r10 + 22) = r2
 ; CHECK: *(u16 *)(r10 + 20) = r2
 ; CHECK: *(u16 *)(r10 + 18) = r2
 ; CHECK: *(u16 *)(r10 + 16) = r2
 ; CHECK: *(u16 *)(r10 + 14) = r2
 ; CHECK: *(u16 *)(r10 + 12) = r2
 ; CHECK: *(u16 *)(r10 + 10) = r2