i386.h (X86_TUNE_PROMOTE_HIMODE_IMUL): New tuning option.

        * config/i386/i386.h (X86_TUNE_PROMOTE_HIMODE_IMUL): New tuning
        option.
        (TARGET_TUNE_PROMOTE_HIMODE_IMUL): New define. Use new tuning option.
        * config/i386/i386.c (ix86_tune_features): Initialize new
        tuning option.
        * config/i386/predicates.md (promotable_binary_operator): Use
        TARGET_TUNE_PROMOTE_HIMODE_IMUL.

From-SVN: r123301

gcc/ChangeLog

+2007-03-28  Uros Bizjak  <ubizjak@gmail.com>
+
+	* config/i386/i386.h (X86_TUNE_PROMOTE_HIMODE_IMUL): New tuning
+	option.
+	(TARGET_TUNE_PROMOTE_HIMODE_IMUL): New define. Use new tuning option.
+	* config/i386/i386.c (ix86_tune_features): Initialize new
+	tuning option.
+	* config/i386/predicates.md (promotable_binary_operator): Use
+	TARGET_TUNE_PROMOTE_HIMODE_IMUL.
+
 2007-03-28  Grigory Zagorodnev <grigory_zagorodnev@linux.intel.com>
 
 	PR target/31380

gcc/config/i386/i386.c

@@ -1200,25 +1200,29 @@ unsigned int ix86_tune_features[X86_TUNE_LAST] = {
   /* X86_TUNE_AVOID_VECTOR_DECODE */
   m_K8 | m_GENERIC64,
 
-  /* X86_TUNE_SLOW_IMUL_IMM32_MEM (imul of 32-bit constant and memory is vector
-     path on AMD machines) */
+  /* X86_TUNE_PROMOTE_HIMODE_IMUL: Modern CPUs have same latency for HImode
+     and SImode multiply, but 386 and 486 do HImode multiply faster.  */
+  ~(m_386 | m_486),
+
+  /* X86_TUNE_SLOW_IMUL_IMM32_MEM: Imul of 32-bit constant and memory is
+     vector path on AMD machines.  */
   m_K8 | m_GENERIC64 | m_AMDFAM10,
 
-  /* X86_TUNE_SLOW_IMUL_IMM8 (imul of 8-bit constant is vector path on AMD
-     machines)  */
+  /* X86_TUNE_SLOW_IMUL_IMM8: Imul of 8-bit constant is vector path on AMD
+     machines.  */
   m_K8 | m_GENERIC64 | m_AMDFAM10,
 
-  /* X86_TUNE_MOVE_M1_VIA_OR (on pentiums, it is faster to load -1 via OR than
-     a MOV) */
+  /* X86_TUNE_MOVE_M1_VIA_OR: On pentiums, it is faster to load -1 via OR
+     than a MOV.  */
   m_PENT,
 
-  /* X86_TUNE_NOT_UNPAIRABLE (NOT is not pairable on Pentium, while XOR is, but
-     one byte longer).  */
+  /* X86_TUNE_NOT_UNPAIRABLE: NOT is not pairable on Pentium, while XOR is,
+     but one byte longer.  */
   m_PENT,
 
-  /* X86_TUNE_NOT_VECTORMODE (On AMD K6, NOT is vector decoded with memory
+  /* X86_TUNE_NOT_VECTORMODE: On AMD K6, NOT is vector decoded with memory
      operand that cannot be represented using a modRM byte.  The XOR
-     replacement is long decoded, so this split helps here as well).  */
+     replacement is long decoded, so this split helps here as well.  */
   m_K6,
 };
 

gcc/config/i386/i386.h

@@ -235,6 +235,7 @@ enum ix86_tune_indices {
   X86_TUNE_EXT_80387_CONSTANTS,
   X86_TUNE_SHORTEN_X87_SSE,
   X86_TUNE_AVOID_VECTOR_DECODE,
+  X86_TUNE_PROMOTE_HIMODE_IMUL,
   X86_TUNE_SLOW_IMUL_IMM32_MEM,
   X86_TUNE_SLOW_IMUL_IMM8,
   X86_TUNE_MOVE_M1_VIA_OR,
@@ -312,6 +313,8 @@ extern unsigned int ix86_tune_features[X86_TUNE_LAST];
 #define TARGET_SHORTEN_X87_SSE	ix86_tune_features[X86_TUNE_SHORTEN_X87_SSE]
 #define TARGET_AVOID_VECTOR_DECODE \
 	ix86_tune_features[X86_TUNE_AVOID_VECTOR_DECODE]
+#define TARGET_TUNE_PROMOTE_HIMODE_IMUL \
+	ix86_tune_features[X86_TUNE_PROMOTE_HIMODE_IMUL]
 #define TARGET_SLOW_IMUL_IMM32_MEM \
 	ix86_tune_features[X86_TUNE_SLOW_IMUL_IMM32_MEM]
 #define TARGET_SLOW_IMUL_IMM8	ix86_tune_features[X86_TUNE_SLOW_IMUL_IMM8]

gcc/config/i386/predicates.md

@@ -963,12 +963,10 @@
 	       mod,udiv,umod,ashift,rotate,ashiftrt,lshiftrt,rotatert"))
 
 ;; Return 1 if OP is a binary operator that can be promoted to wider mode.
-;; Modern CPUs have same latency for HImode and SImode multiply,
-;; but 386 and 486 do HImode multiply faster.  */
 (define_predicate "promotable_binary_operator"
   (ior (match_code "plus,and,ior,xor,ashift")
        (and (match_code "mult")
-	    (match_test "ix86_tune > PROCESSOR_I486"))))
+	    (match_test "TARGET_TUNE_PROMOTE_HIMODE_IMUL"))))
 
 ;; To avoid problems when jump re-emits comparisons like testqi_ext_ccno_0,
 ;; re-recognize the operand to avoid a copy_to_mode_reg that will fail.