Backport from mainline

	2014-08-19  H.J. Lu  <hongjiu.lu@intel.com>

	* config/i386/i386.md (*ctz<mode>2_falsedep_1): Don't clear
	destination if it is used in source.
	(*clz<mode>2_lzcnt_falsedep_1): Likewise.
	(*popcount<mode>2_falsedep_1): Likewise.

	Backport from mainline
	2014-08-18  Uros Bizjak  <ubizjak@gmail.com>

	PR target/62011
	* config/i386/x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI):
	New tune flag.
	* config/i386/i386.h (TARGET_AVOID_FALSE_DEP_FOR_BMI): New define.
	* config/i386/i386.md (unspec) <UNSPEC_INSN_FALSE_DEP>: New unspec.
	(ffs<mode>2): Do not expand with tzcnt for
	TARGET_AVOID_FALSE_DEP_FOR_BMI.
	(ffssi2_no_cmove): Ditto.
	(*tzcnt<mode>_1): Disable for TARGET_AVOID_FALSE_DEP_FOR_BMI.
	(ctz<mode>2): New expander.
	(*ctz<mode>2_falsedep_1): New insn_and_split pattern.
	(*ctz<mode>2_falsedep): New insn.
	(*ctz<mode>2): Rename from ctz<mode>2.
	(clz<mode>2_lzcnt): New expander.
	(*clz<mode>2_lzcnt_falsedep_1): New insn_and_split pattern.
	(*clz<mode>2_lzcnt_falsedep): New insn.
	(*clz<mode>2): Rename from ctz<mode>2.
	(popcount<mode>2): New expander.
	(*popcount<mode>2_falsedep_1): New insn_and_split pattern.
	(*popcount<mode>2_falsedep): New insn.
	(*popcount<mode>2): Rename from ctz<mode>2.
	(*popcount<mode>2_cmp): Remove.
	(*popcountsi2_cmp_zext): Ditto.



git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/branches/gcc-4_9-branch@214279 138bc75d-0d04-0410-961f-82ee72b054a4

4 files changed, 182 insertions, 37 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index eb05afa9ce7..1a2e5068440 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,40 @@
+2014-08-21  Uros Bizjak  <ubizjak@gmail.com>
+
+	Backport from mainline
+	2014-08-19  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* config/i386/i386.md (*ctz<mode>2_falsedep_1): Don't clear
+	destination if it is used in source.
+	(*clz<mode>2_lzcnt_falsedep_1): Likewise.
+	(*popcount<mode>2_falsedep_1): Likewise.
+
+	Backport from mainline
+	2014-08-18  Uros Bizjak  <ubizjak@gmail.com>
+
+	PR target/62011
+	* config/i386/x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI):
+	New tune flag.
+	* config/i386/i386.h (TARGET_AVOID_FALSE_DEP_FOR_BMI): New define.
+	* config/i386/i386.md (unspec) <UNSPEC_INSN_FALSE_DEP>: New unspec.
+	(ffs<mode>2): Do not expand with tzcnt for
+	TARGET_AVOID_FALSE_DEP_FOR_BMI.
+	(ffssi2_no_cmove): Ditto.
+	(*tzcnt<mode>_1): Disable for TARGET_AVOID_FALSE_DEP_FOR_BMI.
+	(ctz<mode>2): New expander.
+	(*ctz<mode>2_falsedep_1): New insn_and_split pattern.
+	(*ctz<mode>2_falsedep): New insn.
+	(*ctz<mode>2): Rename from ctz<mode>2.
+	(clz<mode>2_lzcnt): New expander.
+	(*clz<mode>2_lzcnt_falsedep_1): New insn_and_split pattern.
+	(*clz<mode>2_lzcnt_falsedep): New insn.
+	(*clz<mode>2): Rename from ctz<mode>2.
+	(popcount<mode>2): New expander.
+	(*popcount<mode>2_falsedep_1): New insn_and_split pattern.
+	(*popcount<mode>2_falsedep): New insn.
+	(*popcount<mode>2): Rename from ctz<mode>2.
+	(*popcount<mode>2_cmp): Remove.
+	(*popcountsi2_cmp_zext): Ditto.
+
 2014-08-20  Martin Jambor  <mjambor@suse.cz>
 	    Wei Mi  <wmi@google.com>
 
diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h
index 51659deb309..b3b7c8d3090 100644
--- a/gcc/config/i386/i386.h
+++ b/gcc/config/i386/i386.h
@@ -457,6 +457,8 @@ extern unsigned char ix86_tune_features[X86_TUNE_LAST];
 	ix86_tune_features[X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS]
 #define TARGET_ADJUST_UNROLL \
     ix86_tune_features[X86_TUNE_ADJUST_UNROLL]
+#define TARGET_AVOID_FALSE_DEP_FOR_BMI \
+	ix86_tune_features[X86_TUNE_AVOID_FALSE_DEP_FOR_BMI]
 
 /* Feature tests against the various architecture variations.  */
 enum ix86_arch_indices {
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index f677a7ef05a..e63be9bf7fe 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -111,6 +111,7 @@
   UNSPEC_LEA_ADDR
   UNSPEC_XBEGIN_ABORT
   UNSPEC_STOS
+  UNSPEC_INSN_FALSE_DEP
 
   ;; For SSE/MMX support:
   UNSPEC_FIX_NOTRUNC
@@ -11856,7 +11857,8 @@
       DONE;
     }
 
-  flags_mode = TARGET_BMI ? CCCmode : CCZmode;
+  flags_mode
+    = (TARGET_BMI && !TARGET_AVOID_FALSE_DEP_FOR_BMI) ? CCCmode : CCZmode;
 
   operands[2] = gen_reg_rtx (<MODE>mode);
   operands[3] = gen_rtx_REG (flags_mode, FLAGS_REG);
@@ -11882,7 +11884,8 @@
    (parallel [(set (match_dup 0) (plus:SI (match_dup 0) (const_int 1)))
 	      (clobber (reg:CC FLAGS_REG))])]
 {
-  enum machine_mode flags_mode = TARGET_BMI ? CCCmode : CCZmode;
+  enum machine_mode flags_mode
+    = (TARGET_BMI && !TARGET_AVOID_FALSE_DEP_FOR_BMI) ? CCCmode : CCZmode;
 
   operands[3] = gen_lowpart (QImode, operands[2]);
   operands[4] = gen_rtx_REG (flags_mode, FLAGS_REG);
@@ -11897,7 +11900,7 @@
 		     (const_int 0)))
    (set (match_operand:SWI48 0 "register_operand" "=r")
 	(ctz:SWI48 (match_dup 1)))]
-  "TARGET_BMI"
+  "TARGET_BMI && !TARGET_AVOID_FALSE_DEP_FOR_BMI"
   "tzcnt{<imodesuffix>}\t{%1, %0|%0, %1}"
   [(set_attr "type" "alu1")
    (set_attr "prefix_0f" "1")
@@ -11918,7 +11921,58 @@
    (set_attr "btver2_decode" "double")
    (set_attr "mode" "<MODE>")])
 
-(define_insn "ctz<mode>2"
+(define_expand "ctz<mode>2"
+  [(parallel
+    [(set (match_operand:SWI248 0 "register_operand")
+	  (ctz:SWI248
+	    (match_operand:SWI248 1 "nonimmediate_operand")))
+     (clobber (reg:CC FLAGS_REG))])])
+
+; False dependency happens when destination is only updated by tzcnt,
+; lzcnt or popcnt.  There is no false dependency when destination is
+; also used in source.
+(define_insn_and_split "*ctz<mode>2_falsedep_1"
+  [(set (match_operand:SWI48 0 "register_operand" "=r")
+	(ctz:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand" "rm")))
+   (clobber (reg:CC FLAGS_REG))]
+  "(TARGET_BMI || TARGET_GENERIC)
+   && TARGET_AVOID_FALSE_DEP_FOR_BMI && optimize_function_for_speed_p (cfun)"
+  "#"
+  "&& reload_completed"
+  [(parallel
+    [(set (match_dup 0)
+	  (ctz:SWI48 (match_dup 1)))
+     (unspec [(match_dup 0)] UNSPEC_INSN_FALSE_DEP)
+     (clobber (reg:CC FLAGS_REG))])]
+{
+  if (!reg_mentioned_p (operands[0], operands[1]))
+    ix86_expand_clear (operands[0]);
+})
+
+(define_insn "*ctz<mode>2_falsedep"
+  [(set (match_operand:SWI48 0 "register_operand" "=r")
+	(ctz:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand" "rm")))
+   (unspec [(match_operand:SWI48 2 "register_operand" "0")]
+	   UNSPEC_INSN_FALSE_DEP)
+   (clobber (reg:CC FLAGS_REG))]
+  ""
+{
+  if (TARGET_BMI)
+    return "tzcnt{<imodesuffix>}\t{%1, %0|%0, %1}";
+  else if (TARGET_GENERIC)
+    /* tzcnt expands to 'rep bsf' and we can use it even if !TARGET_BMI.  */
+    return "rep%; bsf{<imodesuffix>}\t{%1, %0|%0, %1}";
+  else
+    gcc_unreachable ();
+}
+  [(set_attr "type" "alu1")
+   (set_attr "prefix_0f" "1")
+   (set_attr "prefix_rep" "1")
+   (set_attr "mode" "<MODE>")])
+
+(define_insn "*ctz<mode>2"
   [(set (match_operand:SWI248 0 "register_operand" "=r")
 	(ctz:SWI248 (match_operand:SWI248 1 "nonimmediate_operand" "rm")))
    (clobber (reg:CC FLAGS_REG))]
@@ -11965,7 +12019,47 @@
   operands[2] = GEN_INT (GET_MODE_BITSIZE (<MODE>mode)-1);
 })
 
-(define_insn "clz<mode>2_lzcnt"
+(define_expand "clz<mode>2_lzcnt"
+  [(parallel
+    [(set (match_operand:SWI248 0 "register_operand")
+	  (clz:SWI248
+	    (match_operand:SWI248 1 "nonimmediate_operand")))
+     (clobber (reg:CC FLAGS_REG))])]
+  "TARGET_LZCNT")
+
+(define_insn_and_split "*clz<mode>2_lzcnt_falsedep_1"
+  [(set (match_operand:SWI48 0 "register_operand" "=r")
+	(clz:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand" "rm")))
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_LZCNT
+   && TARGET_AVOID_FALSE_DEP_FOR_BMI && optimize_function_for_speed_p (cfun)"
+  "#"
+  "&& reload_completed"
+  [(parallel
+    [(set (match_dup 0)
+	  (clz:SWI48 (match_dup 1)))
+     (unspec [(match_dup 0)] UNSPEC_INSN_FALSE_DEP)
+     (clobber (reg:CC FLAGS_REG))])]
+{
+  if (!reg_mentioned_p (operands[0], operands[1]))
+    ix86_expand_clear (operands[0]);
+})
+
+(define_insn "*clz<mode>2_lzcnt_falsedep"
+  [(set (match_operand:SWI48 0 "register_operand" "=r")
+	(clz:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand" "rm")))
+   (unspec [(match_operand:SWI48 2 "register_operand" "0")]
+	   UNSPEC_INSN_FALSE_DEP)
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_LZCNT"
+  "lzcnt{<imodesuffix>}\t{%1, %0|%0, %1}"
+  [(set_attr "prefix_rep" "1")
+   (set_attr "type" "bitmanip")
+   (set_attr "mode" "<MODE>")])
+
+(define_insn "*clz<mode>2_lzcnt"
   [(set (match_operand:SWI248 0 "register_operand" "=r")
 	(clz:SWI248 (match_operand:SWI248 1 "nonimmediate_operand" "rm")))
    (clobber (reg:CC FLAGS_REG))]
@@ -12248,10 +12342,39 @@
    (set_attr "prefix_0f" "1")
    (set_attr "mode" "HI")])
 
-(define_insn "popcount<mode>2"
-  [(set (match_operand:SWI248 0 "register_operand" "=r")
-	(popcount:SWI248
-	  (match_operand:SWI248 1 "nonimmediate_operand" "rm")))
+(define_expand "popcount<mode>2"
+  [(parallel
+    [(set (match_operand:SWI248 0 "register_operand")
+	  (popcount:SWI248
+	    (match_operand:SWI248 1 "nonimmediate_operand")))
+     (clobber (reg:CC FLAGS_REG))])]
+  "TARGET_POPCNT")
+
+(define_insn_and_split "*popcount<mode>2_falsedep_1"
+  [(set (match_operand:SWI48 0 "register_operand" "=r")
+	(popcount:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand" "rm")))
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_POPCNT
+   && TARGET_AVOID_FALSE_DEP_FOR_BMI && optimize_function_for_speed_p (cfun)"
+  "#"
+  "&& reload_completed"
+  [(parallel
+    [(set (match_dup 0)
+	  (popcount:SWI48 (match_dup 1)))
+     (unspec [(match_dup 0)] UNSPEC_INSN_FALSE_DEP)
+     (clobber (reg:CC FLAGS_REG))])]
+{
+  if (!reg_mentioned_p (operands[0], operands[1]))
+    ix86_expand_clear (operands[0]);
+})
+
+(define_insn "*popcount<mode>2_falsedep"
+  [(set (match_operand:SWI48 0 "register_operand" "=r")
+	(popcount:SWI48
+	  (match_operand:SWI48 1 "nonimmediate_operand" "rm")))
+   (unspec [(match_operand:SWI48 2 "register_operand" "0")]
+	   UNSPEC_INSN_FALSE_DEP)
    (clobber (reg:CC FLAGS_REG))]
   "TARGET_POPCNT"
 {
@@ -12265,15 +12388,12 @@
    (set_attr "type" "bitmanip")
    (set_attr "mode" "<MODE>")])
 
-(define_insn "*popcount<mode>2_cmp"
-  [(set (reg FLAGS_REG)
-	(compare
-	  (popcount:SWI248
-	    (match_operand:SWI248 1 "nonimmediate_operand" "rm"))
-	  (const_int 0)))
-   (set (match_operand:SWI248 0 "register_operand" "=r")
-	(popcount:SWI248 (match_dup 1)))]
-  "TARGET_POPCNT && ix86_match_ccmode (insn, CCZmode)"
+(define_insn "*popcount<mode>2"
+  [(set (match_operand:SWI248 0 "register_operand" "=r")
+	(popcount:SWI248
+	  (match_operand:SWI248 1 "nonimmediate_operand" "rm")))
+   (clobber (reg:CC FLAGS_REG))]
+  "TARGET_POPCNT"
 {
 #if TARGET_MACHO
   return "popcnt\t{%1, %0|%0, %1}";
@@ -12285,25 +12405,6 @@
    (set_attr "type" "bitmanip")
    (set_attr "mode" "<MODE>")])
 
-(define_insn "*popcountsi2_cmp_zext"
-  [(set (reg FLAGS_REG)
-        (compare
-          (popcount:SI (match_operand:SI 1 "nonimmediate_operand" "rm"))
-          (const_int 0)))
-   (set (match_operand:DI 0 "register_operand" "=r")
-        (zero_extend:DI(popcount:SI (match_dup 1))))]
-  "TARGET_64BIT && TARGET_POPCNT && ix86_match_ccmode (insn, CCZmode)"
-{
-#if TARGET_MACHO
-  return "popcnt\t{%1, %0|%0, %1}";
-#else
-  return "popcnt{l}\t{%1, %0|%0, %1}";
-#endif
-}
-  [(set_attr "prefix_rep" "1")
-   (set_attr "type" "bitmanip")
-   (set_attr "mode" "SI")])
-
 (define_expand "bswapdi2"
   [(set (match_operand:DI 0 "register_operand")
 	(bswap:DI (match_operand:DI 1 "nonimmediate_operand")))]
diff --git a/gcc/config/i386/x86-tune.def b/gcc/config/i386/x86-tune.def
index 8399102671c..c36174855cb 100644
--- a/gcc/config/i386/x86-tune.def
+++ b/gcc/config/i386/x86-tune.def
@@ -500,6 +500,11 @@ DEF_TUNE (X86_TUNE_NOT_VECTORMODE, "not_vectormode", m_K6)
 DEF_TUNE (X86_TUNE_AVOID_VECTOR_DECODE, "avoid_vector_decode",
           m_K8)
 
+/* X86_TUNE_AVOID_FALSE_DEP_FOR_BMI: Avoid false dependency
+   for bit-manipulation instructions.  */
+DEF_TUNE (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI, "avoid_false_dep_for_bmi",
+	  m_SANDYBRIDGE | m_HASWELL | m_INTEL | m_GENERIC)
+
 /*****************************************************************************/
 /* This never worked well before.                                            */
 /*****************************************************************************/