inner-kernel.ll

"for conv.s1.r$x.r$x":                            ; preds = %"for conv.s1.r$x.r$x", %"for conv.s1.r$x.r$x.preheader"
  %conv20.sroa.0.0 = phi <4 x i32> [ %303, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.4.0 = phi <4 x i32> [ %304, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.8.0 = phi <4 x i32> [ %305, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.12.0 = phi <4 x i32> [ %311, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.16.0 = phi <4 x i32> [ %312, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.20.0 = phi <4 x i32> [ %313, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.24.0 = phi <4 x i32> [ %319, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.28.0 = phi <4 x i32> [ %320, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.32.0 = phi <4 x i32> [ %321, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.36.0 = phi <4 x i32> [ %327, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.40.0 = phi <4 x i32> [ %328, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %conv20.sroa.44.0 = phi <4 x i32> [ %329, %"for conv.s1.r$x.r$x" ], [ zeroinitializer, %"for conv.s1.r$x.r$x.preheader" ]
  %"conv.s1.r$x.r$x" = phi i32 [ %330, %"for conv.s1.r$x.r$x" ], [ 0, %"for conv.s1.r$x.r$x.preheader" ]
  %264 = add nsw i32 %"conv.s1.r$x.r$x", %t398
  %265 = getelementptr inbounds i8, i8* %right_in_elwise, i32 %264
  %266 = load i8, i8* %265, align 1, !tbaa !42
  %267 = insertelement <12 x i8> undef, i8 %266, i32 0
  %268 = shufflevector <12 x i8> %267, <12 x i8> undef, <12 x i32> zeroinitializer
  %269 = zext <12 x i8> %268 to <12 x i16>
  %270 = mul nsw i32 %"conv.s1.r$x.r$x", %t337
  %271 = add nsw i32 %t413, %270
  %272 = getelementptr inbounds i8, i8* %left_global_wrapper, i32 %271
  %273 = bitcast i8* %272 to <12 x i8>*
  %274 = load <12 x i8>, <12 x i8>* %273, align 1, !tbaa !46
  %275 = zext <12 x i8> %274 to <12 x i16>
  %276 = mul nuw <12 x i16> %269, %275
  %277 = add nsw i32 %"conv.s1.r$x.r$x", %t399
  %278 = getelementptr inbounds i8, i8* %right_in_elwise, i32 %277
  %279 = load i8, i8* %278, align 1, !tbaa !42
  %280 = insertelement <12 x i8> undef, i8 %279, i32 0
  %281 = shufflevector <12 x i8> %280, <12 x i8> undef, <12 x i32> zeroinitializer
  %282 = zext <12 x i8> %281 to <12 x i16>
  %283 = mul nuw <12 x i16> %282, %275
  %284 = add nsw i32 %"conv.s1.r$x.r$x", %t400
  %285 = getelementptr inbounds i8, i8* %right_in_elwise, i32 %284
  %286 = load i8, i8* %285, align 1, !tbaa !42
  %287 = insertelement <12 x i8> undef, i8 %286, i32 0
  %288 = shufflevector <12 x i8> %287, <12 x i8> undef, <12 x i32> zeroinitializer
  %289 = zext <12 x i8> %288 to <12 x i16>
  %290 = mul nuw <12 x i16> %289, %275
  %291 = add nsw i32 %"conv.s1.r$x.r$x", %t401
  %292 = getelementptr inbounds i8, i8* %right_in_elwise, i32 %291
  %293 = load i8, i8* %292, align 1, !tbaa !42
  %294 = insertelement <12 x i8> undef, i8 %293, i32 0
  %295 = shufflevector <12 x i8> %294, <12 x i8> undef, <12 x i32> zeroinitializer
  %296 = zext <12 x i8> %295 to <12 x i16>
  %297 = mul nuw <12 x i16> %296, %275
  %298 = zext <12 x i16> %276 to <12 x i32>
  %299 = shufflevector <4 x i32> %conv20.sroa.0.0, <4 x i32> %conv20.sroa.4.0, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %300 = shufflevector <4 x i32> %conv20.sroa.8.0, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
  %301 = shufflevector <8 x i32> %299, <8 x i32> %300, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
  %302 = add <12 x i32> %301, %298
  %303 = shufflevector <12 x i32> %302, <12 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %304 = shufflevector <12 x i32> %302, <12 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %305 = shufflevector <12 x i32> %302, <12 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
  %306 = zext <12 x i16> %283 to <12 x i32>
  %307 = shufflevector <4 x i32> %conv20.sroa.12.0, <4 x i32> %conv20.sroa.16.0, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %308 = shufflevector <4 x i32> %conv20.sroa.20.0, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
  %309 = shufflevector <8 x i32> %307, <8 x i32> %308, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
  %310 = add <12 x i32> %309, %306
  %311 = shufflevector <12 x i32> %310, <12 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %312 = shufflevector <12 x i32> %310, <12 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %313 = shufflevector <12 x i32> %310, <12 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
  %314 = zext <12 x i16> %290 to <12 x i32>
  %315 = shufflevector <4 x i32> %conv20.sroa.24.0, <4 x i32> %conv20.sroa.28.0, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %316 = shufflevector <4 x i32> %conv20.sroa.32.0, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
  %317 = shufflevector <8 x i32> %315, <8 x i32> %316, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
  %318 = add <12 x i32> %317, %314
  %319 = shufflevector <12 x i32> %318, <12 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %320 = shufflevector <12 x i32> %318, <12 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %321 = shufflevector <12 x i32> %318, <12 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
  %322 = zext <12 x i16> %297 to <12 x i32>
  %323 = shufflevector <4 x i32> %conv20.sroa.36.0, <4 x i32> %conv20.sroa.40.0, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %324 = shufflevector <4 x i32> %conv20.sroa.44.0, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
  %325 = shufflevector <8 x i32> %323, <8 x i32> %324, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
  %326 = add <12 x i32> %325, %322
  %327 = shufflevector <12 x i32> %326, <12 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %328 = shufflevector <12 x i32> %326, <12 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %329 = shufflevector <12 x i32> %326, <12 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
  %330 = add nuw nsw i32 %"conv.s1.r$x.r$x", 1
  %331 = icmp eq i32 %330, %lhs.extent.1
  br i1 %331, label %"consume conv", label %"for conv.s1.r$x.r$x"