fixed several AM problems (and disabled most of the new optimisation...

fixed several AM problems (and disabled most of the new optimisation possibilities again, as we have problems with heights not being recomputed)

[r15433]

ir/be/ia32/ia32_emitter.c

@@ -1749,7 +1749,7 @@ void emit_ia32_Conv_I2I(ia32_emit_env_t *env, const ir_node *node) {
 			be_emit_cstring(env, "\tmov");
 			be_emit_string(env, sign_suffix);
 			ia32_emit_mode_suffix_mode(env, smaller_mode);
-			be_emit_cstring(env, "l %");
+			be_emit_cstring(env, "l ");
 			ia32_emit_am(env, node);
 			be_emit_cstring(env, ", ");
 			ia32_emit_dest_register(env, node, 0);

ir/be/ia32/ia32_nodes_attr.h

@@ -140,8 +140,8 @@ struct ia32_attr_t {
 		ir_entity *sc;        /**< the symconst ident */
 	} cnst_val;
 
-	ir_mode *ls_mode;     /**< Load/Store mode: This is the mode of the value
-	                           that is manipulated by this node. */
+	ir_mode   *ls_mode;       /**< Load/Store mode: This is the mode of the
+	                               value that is manipulated by this node. */
 
 	ir_entity *frame_ent; /**< the frame entity attached to this node */
 

ir/be/ia32/ia32_optimize.c

@@ -406,7 +406,8 @@ static int is_addr_candidate(const ir_node *irn)
 static ia32_am_cand_t is_am_candidate(heights_t *h, const ir_node *block, ir_node *irn) {
 	ir_node *in, *load, *other, *left, *right;
 	int      is_cand = 0, cand;
-	int arity;
+	int      arity;
+	int      is_binary;
 
 	if (is_ia32_Ld(irn) || is_ia32_St(irn) ||
 		is_ia32_vfild(irn) || is_ia32_vfist(irn) ||
@@ -416,9 +417,10 @@ static ia32_am_cand_t is_am_candidate(heights_t *h, const ir_node *block, ir_nod
 	if(get_ia32_frame_ent(irn) != NULL)
 		return IA32_AM_CAND_NONE;
 
-	left  = get_irn_n(irn, 2);
-	arity = get_irn_arity(irn);
-	if(get_ia32_am_arity(irn) == ia32_am_binary) {
+	left      = get_irn_n(irn, 2);
+	arity     = get_irn_arity(irn);
+	is_binary = get_ia32_am_arity(irn) == ia32_am_binary;
+	if(is_binary) {
 		/* binary op */
 		right = get_irn_n(irn, 3);
 	} else {
@@ -449,7 +451,7 @@ static ia32_am_cand_t is_am_candidate(heights_t *h, const ir_node *block, ir_nod
 		}
 
 		/* If there is a data dependency of other irn from load: cannot use AM */
-		if (is_cand && get_nodes_block(other) == block) {
+		if (is_cand && is_binary && get_nodes_block(other) == block) {
 			other   = skip_Proj(other);
 			is_cand = heights_reachable_in_block(h, other, load) ? 0 : is_cand;
 			/* this could happen in loops */
@@ -474,11 +476,15 @@ static ia32_am_cand_t is_am_candidate(heights_t *h, const ir_node *block, ir_nod
 		other = left;
 
 		/* 8bit Loads are not supported, they cannot be used with every register */
-		if (get_mode_size_bits(get_ia32_ls_mode(load)) < 16)
+		/* 8bit Loads are not supported (for binary ops),
+		 * they cannot be used with every register */
+		if (get_ia32_am_arity(irn) == ia32_am_binary &&
+				get_mode_size_bits(get_ia32_ls_mode(load)) < 16) {
 			is_cand = 0;
+		}
 
 		/* If there is a data dependency of other irn from load: cannot use load */
-		if (is_cand && get_nodes_block(other) == block) {
+		if (is_cand && is_binary && get_nodes_block(other) == block) {
 			other   = skip_Proj(other);
 			is_cand = heights_reachable_in_block(h, other, load) ? 0 : is_cand;
 			/* this could happen in loops */
@@ -1426,6 +1432,12 @@ static void optimize_am(ir_node *irn, void *env) {
 		assert(cand & IA32_AM_CAND_RIGHT);
 		load = get_Proj_pred(right);
 
+		if(get_irn_n_edges(right) > 1) {
+			source_possible = 0;
+		}
+		/* TODO: this isn't really needed, but the code below is buggy
+		   as heights won't get recomputed when the graph is reconstructed
+		   so we can only transform loads with the result proj only */
 		if(get_irn_n_edges(load) > 1) {
 			source_possible = 0;
 		}
@@ -1433,8 +1445,7 @@ static void optimize_am(ir_node *irn, void *env) {
 
 	if (source_possible) {
 		ir_mode *ls_mode = get_ia32_ls_mode(load);
-		if(get_mode_size_bits(ls_mode) != 32
-				|| ls_mode == mode_D)
+		if(get_mode_size_bits(ls_mode) != 32 || ls_mode == mode_D)
 			source_possible = 0;
 
 	}
@@ -1451,7 +1462,12 @@ static void optimize_am(ir_node *irn, void *env) {
 		set_ia32_am_flavour(irn, get_ia32_am_flavour(load));
 		set_ia32_op_type(irn, ia32_AddrModeS);
 		set_ia32_frame_ent(irn, get_ia32_frame_ent(load));
-		set_ia32_ls_mode(irn, get_ia32_ls_mode(load));
+
+		/* set ls_mode if not already present (conv nodes already have ls_mode
+		   set) */
+		if(get_ia32_ls_mode(irn) == NULL) {
+			set_ia32_ls_mode(irn, get_ia32_ls_mode(load));
+		}
 
 		set_ia32_am_sc(irn, get_ia32_am_sc(load));
 		if (is_ia32_am_sc_sign(load))
@@ -1488,17 +1504,21 @@ static void optimize_am(ir_node *irn, void *env) {
 			ir_node *res_proj;
 			ir_mode *mode = get_irn_mode(irn);
 
-			assert(mode != mode_T);
+			if(mode != mode_T) {
+				res_proj = new_rd_Proj(get_irn_dbg_info(irn), irg,
+									   get_nodes_block(irn),
+									   new_Unknown(mode_T), mode, 0);
+				set_irn_mode(irn, mode_T);
+				edges_reroute(irn, res_proj, irg);
+				set_Proj_pred(res_proj, irn);
 
-			res_proj = new_rd_Proj(get_irn_dbg_info(irn), irg,
-			                       get_nodes_block(irn), new_Unknown(mode_T),
-			                       mode, 0);
-			set_irn_mode(irn, mode_T);
-			edges_reroute(irn, res_proj, irg);
-			set_Proj_pred(res_proj, irn);
-
-			set_Proj_pred(mem_proj, irn);
-			set_Proj_proj(mem_proj, 1);
+				set_Proj_pred(mem_proj, irn);
+				set_Proj_proj(mem_proj, 1);
+			} else {
+				/* hacky: we need some proj number which is not used yet... */
+				set_Proj_proj(mem_proj, -1);
+				set_Proj_pred(mem_proj, irn);
+			}
 		}
 
 		try_kill(load);