config.gcc (powerpc*-*-*): Handle a2.

	* config.gcc (powerpc*-*-*): Handle a2.
	* config/rs6000/rs6000.md (cpu): Add ppca2.  Include "a2.md".
	* config/rs6000/a2.md: New file.
	* config/rs6000/rs6000.opt (mno-update): New.
	(mupdate): Return to using a mask, not a var.
	* config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for a2.
	(enum processor_type): Add PROCESSOR_PPCA2.
	* config/rs6000/rs6000.c (ppca2_cost): New costs.
	(rs6000_override_options): Add "a2" to processor_target_table.
	Update rs6000_always_hint logic.  Correctly set rs6000_cost for
	a2.
	* doc/invoke.texi (RS/6000 and PowerPC Options): Document
	-mcpu=a2.

From-SVN: r152499

8 files changed, 296 insertions, 8 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 684352f..b7c4f54 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,19 @@
+2009-10-07  Ben Elliston  <bje@au.ibm.com>
+
+	* config.gcc (powerpc*-*-*): Handle a2.
+	* config/rs6000/rs6000.md (cpu): Add ppca2.  Include "a2.md".
+	* config/rs6000/a2.md: New file.
+	* config/rs6000/rs6000.opt (mno-update): New.
+	(mupdate): Return to using a mask, not a var.
+	* config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for a2.
+	(enum processor_type): Add PROCESSOR_PPCA2.
+	* config/rs6000/rs6000.c (ppca2_cost): New costs.
+	(rs6000_override_options): Add "a2" to processor_target_table.
+	Update rs6000_always_hint logic.  Correctly set rs6000_cost for
+	a2.
+	* doc/invoke.texi (RS/6000 and PowerPC Options): Document
+	-mcpu=a2.
+
 2009-10-06  Uros Bizjak  <ubizjak@gmail.com>
 
 	* config/i386/i386.md (float<SSEMODEI24:mode><X87MODEF:mode>2):
diff --git a/gcc/config.gcc b/gcc/config.gcc
index 1b0194b..9e3ada7 100644
--- a/gcc/config.gcc
+++ b/gcc/config.gcc
@@ -327,7 +327,7 @@ powerpc*-*-*)
 	extra_headers="ppc-asm.h altivec.h spe.h ppu_intrinsics.h paired.h spu2vmx.h vec_types.h si2vmx.h"
 	need_64bit_hwint=yes
 	case x$with_cpu in
-	    xpowerpc64|xdefault64|x6[23]0|x970|xG5|xpower[34567]|xpower6x|xrs64a|xcell)
+	    xpowerpc64|xdefault64|x6[23]0|x970|xG5|xpower[34567]|xpower6x|xrs64a|xcell|xa2)
 		cpu_is_64bit=yes
 		;;
 	esac
@@ -3070,7 +3070,7 @@ case "${target}" in
 			| 401 | 403 | 405 | 405fp | 440 | 440fp | 464 | 464fp \
 			| 505 | 601 | 602 | 603 | 603e | ec603e | 604 \
 			| 604e | 620 | 630 | 740 | 750 | 7400 | 7450 \
-			| e300c[23] | 854[08] | e500mc \
+			| a2 | e300c[23] | 854[08] | e500mc \
 			| 801 | 821 | 823 | 860 | 970 | G3 | G4 | G5 | cell)
 				# OK
 				;;
diff --git a/gcc/config/rs6000/a2.md b/gcc/config/rs6000/a2.md
new file mode 100644
index 0000000..5707536
--- /dev/null
+++ b/gcc/config/rs6000/a2.md
@@ -0,0 +1,238 @@
+;; Scheduling description for PowerPC A2 processors.
+;; Copyright (C) 2009 Free Software Foundation, Inc.
+;;
+;; Contributed by Ben Elliston (bje@au.ibm.com).
+
+(define_automaton "ppca2")
+
+;; CPU units
+
+;; The multiplier pipeline.
+(define_cpu_unit "mult" "ppca2")
+
+;; The auxillary processor unit (FP/vector unit).
+(define_cpu_unit "axu" "ppca2")
+
+;; D.4.6
+;; Some peculiarities for certain SPRs
+
+(define_insn_reservation "ppca2-mfcr" 1
+  (and (eq_attr "type" "mfcr")
+       (eq_attr "cpu" "ppca2"))
+   "nothing")
+
+(define_insn_reservation "ppca2-mfjmpr" 5
+  (and (eq_attr "type" "mfjmpr")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+(define_insn_reservation "ppca2-mtjmpr" 5
+  (and (eq_attr "type" "mtjmpr")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+;; D.4.8
+(define_insn_reservation "ppca2-imul" 1
+  (and (eq_attr "type" "imul,imul2,imul3,imul_compare")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+;; FIXME: latency and multiplier reservation for 64-bit multiply?
+(define_insn_reservation "ppca2-lmul" 6
+  (and (eq_attr "type" "lmul,lmul_compare")
+       (eq_attr "cpu" "ppca2"))
+  "mult*3")
+
+;; D.4.9
+(define_insn_reservation "ppca2-idiv" 32
+  (and (eq_attr "type" "idiv")
+       (eq_attr "cpu" "ppca2"))
+  "mult*32")
+
+(define_insn_reservation "ppca2-ldiv" 65
+  (and (eq_attr "type" "ldiv")
+       (eq_attr "cpu" "ppca2"))
+  "mult*65")
+
+;; D.4.13
+(define_insn_reservation "ppca2-load" 5
+  (and (eq_attr "type" "load,load_ext,load_ext_u,load_ext_ux,load_ux,load_u")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+;; D.8.1
+(define_insn_reservation "ppca2-fp" 6
+  (and (eq_attr "type" "fp")     	   ;; Ignore fpsimple insn types (SPE only).
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+;; D.8.4
+(define_insn_reservation "ppca2-fp-load" 6
+  (and (eq_attr "type" "fpload,fpload_u,fpload_ux")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+;; D.8.5
+(define_insn_reservation "ppca2-fp-store" 2
+  (and (eq_attr "type" "fpstore,fpstore_u,fpstore_ux")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+;; D.8.6
+(define_insn_reservation "ppca2-fpcompare" 5
+  (and (eq_attr "type" "fpcompare")
+       (eq_attr "cpu" "ppca2"))
+ "axu")
+
+;; D.8.7
+;;
+;; Instructions from the same thread succeeding the floating-point
+;; divide cannot be executed until the floating-point divide has
+;; completed.  Since there is nothing else we can do, this thread will
+;; just have to stall.
+
+(define_insn_reservation "ppca2-ddiv" 72
+  (and (eq_attr "type" "ddiv")
+       (eq_attr "cpu" "ppca2"))
+   "axu")
+
+(define_insn_reservation "ppca2-sdiv" 59
+  (and (eq_attr "type" "sdiv")
+       (eq_attr "cpu" "ppca2"))
+   "axu")
+
+;; D.8.8
+;; 
+;; Instructions from the same thread succeeding the floating-point
+;; divide cannot be executed until the floating-point divide has
+;; completed.  Since there is nothing else we can do, this thread will
+;; just have to stall.
+
+(define_insn_reservation "ppca2-dsqrt" 69
+  (and (eq_attr "type" "dsqrt")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+(define_insn_reservation "ppca2-ssqrt" 65
+  (and (eq_attr "type" "ssqrt")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+;; Scheduling description for PowerPC A2 processors.
+;; Copyright (C) 2008 Free Software Foundation, Inc.
+;;
+;; Contributed by Ben Elliston (bje@au.ibm.com).
+
+(define_automaton "a2")
+
+;; CPU units
+
+;; The multiplier pipeline.
+(define_cpu_unit "mult" "a2")
+
+;; The auxillary processor unit (FP/vector unit).
+(define_cpu_unit "axu" "a2")
+
+;; D.4.6
+;; Some peculiarities for certain SPRs
+
+(define_insn_reservation "ppca2-mfcr" 1
+  (and (eq_attr "type" "mfcr")
+       (eq_attr "cpu" "ppca2"))
+   "nothing")
+
+(define_insn_reservation "ppca2-mfjmpr" 5
+  (and (eq_attr "type" "mfjmpr")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+(define_insn_reservation "ppca2-mtjmpr" 5
+  (and (eq_attr "type" "mtjmpr")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+;; D.4.8
+(define_insn_reservation "ppca2-imul" 1
+  (and (eq_attr "type" "imul,imul2,imul3,imul_compare")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+;; FIXME: latency and multiplier reservation for 64-bit multiply?
+(define_insn_reservation "ppca2-lmul" 6
+  (and (eq_attr "type" "lmul,lmul_compare")
+       (eq_attr "cpu" "ppca2"))
+  "mult*3")
+
+;; D.4.9
+(define_insn_reservation "ppca2-idiv" 32
+  (and (eq_attr "type" "idiv")
+       (eq_attr "cpu" "ppca2"))
+  "mult*32")
+
+(define_insn_reservation "ppca2-ldiv" 65
+  (and (eq_attr "type" "ldiv")
+       (eq_attr "cpu" "ppca2"))
+  "mult*65")
+
+;; D.4.13
+(define_insn_reservation "pcca2-load" 5
+  (and (eq_attr "type" "load,load_ext,load_ext_u,load_ext_ux,load_ux,load_u")
+       (eq_attr "cpu" "ppca2"))
+  "nothing")
+
+;; D.8.1
+(define_insn_reservation "ppca2-fp" 6
+  (and (eq_attr "type" "fp")     	   ;; Ignore fpsimple insn types (SPE only).
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+;; D.8.4
+(define_insn_reservation "ppca2-fp-load" 6
+  (and (eq_attr "type" "fpload,fpload_u,fpload_ux")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+;; D.8.5
+(define_insn_reservation "ppca2-fp-store" 2
+  (and (eq_attr "type" "fpstore,fpstore_u,fpstore_ux")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+;; D.8.6
+(define_insn_reservation "ppca2-fpcompare" 5
+  (and (eq_attr "type" "fpcompare")
+       (eq_attr "cpu" "ppca2"))
+ "axu")
+
+;; D.8.7
+;;
+;; Instructions from the same thread succeeding the floating-point
+;; divide cannot be executed until the floating-point divide has
+;; completed.  Since there is nothing else we can do, this thread will
+;; just have to stall.
+
+(define_insn_reservation "ppca2-ddiv" 72
+  (and (eq_attr "type" "ddiv")
+       (eq_attr "cpu" "ppca2"))
+   "axu")
+
+(define_insn_reservation "ppca2-sdiv" 59
+  (and (eq_attr "type" "sdiv")
+       (eq_attr "cpu" "ppca2"))
+   "axu")
+
+;; D.8.8
+;; 
+;; Instructions from the same thread succeeding the floating-point
+;; divide cannot be executed until the floating-point divide has
+;; completed.  Since there is nothing else we can do, this thread will
+;; just have to stall.
+
+(define_insn_reservation "ppca2-dsqrt" 69
+  (and (eq_attr "type" "dsqrt")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
+
+(define_insn_reservation "ppca2-ssqrt" 65
+  (and (eq_attr "type" "ssqrt")
+       (eq_attr "cpu" "ppca2"))
+  "axu")
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index f44bc99..1215a9f 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -797,6 +797,25 @@ struct processor_costs power7_cost = {
   12,			/* prefetch streams */
 };
 
+/* Instruction costs on POWER A2 processors.  */
+static const
+struct processor_costs ppca2_cost = {
+  COSTS_N_INSNS (16),    /* mulsi */
+  COSTS_N_INSNS (16),    /* mulsi_const */
+  COSTS_N_INSNS (16),    /* mulsi_const9 */
+  COSTS_N_INSNS (16),   /* muldi */
+  COSTS_N_INSNS (22),   /* divsi */
+  COSTS_N_INSNS (28),   /* divdi */
+  COSTS_N_INSNS (3),    /* fp */
+  COSTS_N_INSNS (3),    /* dmul */
+  COSTS_N_INSNS (59),   /* sdiv */
+  COSTS_N_INSNS (72),   /* ddiv */
+  64,
+  16,			/* l1 cache */
+  2048,			/* l2 cache */
+  16,			/* prefetch streams */
+};
+
 
 static bool rs6000_function_ok_for_sibcall (tree, tree);
 static const char *rs6000_invalid_within_doloop (const_rtx);
@@ -2149,6 +2168,9 @@ rs6000_override_options (const char *default_cpu)
 	 /* 8548 has a dummy entry for now.  */
 	 {"8548", PROCESSOR_PPC8540, POWERPC_BASE_MASK | MASK_STRICT_ALIGN
 	  | MASK_ISEL},
+ 	 {"a2", PROCESSOR_PPCA2,
+ 	  POWERPC_BASE_MASK | MASK_PPC_GFXOPT | MASK_POWERPC64 | MASK_POPCNTB
+ 	  | MASK_CMPB | MASK_NO_UPDATE },
 	 {"e300c2", PROCESSOR_PPCE300C2, POWERPC_BASE_MASK | MASK_SOFT_FLOAT},
 	 {"e300c3", PROCESSOR_PPCE300C3, POWERPC_BASE_MASK},
 	 {"e500mc", PROCESSOR_PPCE500MC, POWERPC_BASE_MASK | MASK_PPC_GFXOPT
@@ -2216,7 +2238,7 @@ rs6000_override_options (const char *default_cpu)
 		     | MASK_PPC_GFXOPT | MASK_POWERPC64 | MASK_ALTIVEC
 		     | MASK_MFCRF | MASK_POPCNTB | MASK_FPRND | MASK_MULHW
 		     | MASK_DLMZB | MASK_CMPB | MASK_MFPGPR | MASK_DFP
-		     | MASK_POPCNTD | MASK_VSX | MASK_ISEL)
+		     | MASK_POPCNTD | MASK_VSX | MASK_ISEL | MASK_NO_UPDATE)
   };
 
   /* Set the pointer size.  */
@@ -2495,6 +2517,7 @@ rs6000_override_options (const char *default_cpu)
 			&& rs6000_cpu != PROCESSOR_POWER5
 			&& rs6000_cpu != PROCESSOR_POWER6
 			&& rs6000_cpu != PROCESSOR_POWER7
+			&& rs6000_cpu != PROCESSOR_PPCA2
 			&& rs6000_cpu != PROCESSOR_CELL);
   rs6000_sched_groups = (rs6000_cpu == PROCESSOR_POWER4
 			 || rs6000_cpu == PROCESSOR_POWER5
@@ -2713,6 +2736,10 @@ rs6000_override_options (const char *default_cpu)
 	rs6000_cost = &power7_cost;
 	break;
 
+      case PROCESSOR_PPCA2:
+	rs6000_cost = &ppca2_cost;
+	break;
+
       default:
 	gcc_unreachable ();
       }
diff --git a/gcc/config/rs6000/rs6000.h b/gcc/config/rs6000/rs6000.h
index 52d9a59..a5ce9dd 100644
--- a/gcc/config/rs6000/rs6000.h
+++ b/gcc/config/rs6000/rs6000.h
@@ -107,6 +107,7 @@
 %{mcpu=power6: %(asm_cpu_power6) -maltivec} \
 %{mcpu=power6x: %(asm_cpu_power6) -maltivec} \
 %{mcpu=power7: %(asm_cpu_power7)} \
+%{mcpu=a2: -ma2} \
 %{mcpu=powerpc: -mppc} \
 %{mcpu=rios: -mpwr} \
 %{mcpu=rios1: -mpwr} \
@@ -334,7 +335,8 @@ enum processor_type
    PROCESSOR_POWER5,
    PROCESSOR_POWER6,
    PROCESSOR_POWER7,
-   PROCESSOR_CELL
+   PROCESSOR_CELL,
+   PROCESSOR_PPCA2
 };
 
 /* FPU operations supported. 
diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md
index ba51f1c..7b3de2a 100644
--- a/gcc/config/rs6000/rs6000.md
+++ b/gcc/config/rs6000/rs6000.md
@@ -139,7 +139,7 @@
 ;; Processor type -- this attribute must exactly match the processor_type
 ;; enumeration in rs6000.h.
 
-(define_attr "cpu" "rios1,rios2,rs64a,mpccore,ppc403,ppc405,ppc440,ppc601,ppc603,ppc604,ppc604e,ppc620,ppc630,ppc750,ppc7400,ppc7450,ppc8540,ppce300c2,ppce300c3,ppce500mc,power4,power5,power6,power7,cell"
+(define_attr "cpu" "rios1,rios2,rs64a,mpccore,ppc403,ppc405,ppc440,ppc601,ppc603,ppc604,ppc604e,ppc620,ppc630,ppc750,ppc7400,ppc7450,ppc8540,ppce300c2,ppce300c3,ppce500mc,power4,power5,power6,power7,cell,ppca2"
   (const (symbol_ref "rs6000_cpu_attr")))
 
 
@@ -171,6 +171,7 @@
 (include "power7.md")
 (include "cell.md")
 (include "xfpu.md")
+(include "a2.md")
 
 (include "predicates.md")
 (include "constraints.md")
diff --git a/gcc/config/rs6000/rs6000.opt b/gcc/config/rs6000/rs6000.opt
index 90af9dc..63f0f8c 100644
--- a/gcc/config/rs6000/rs6000.opt
+++ b/gcc/config/rs6000/rs6000.opt
@@ -155,8 +155,12 @@ mvectorize-builtins
 Target Undocumented Report Var(TARGET_VECTORIZE_BUILTINS) Init(-1)
 ; Explicitly control whether we vectorize the builtins or not.
 
+mno-update
+Target Report RejectNegative Mask(NO_UPDATE)
+Do not generate load/store with update instructions
+
 mupdate
-Target Report Var(TARGET_UPDATE) Init(1)
+Target Report RejectNegative InverseMask(NO_UPDATE, UPDATE)
 Generate load/store with update instructions
 
 mavoid-indexed-addresses
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index 75622b3..a805f2b 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -14607,8 +14607,8 @@ Supported values for @var{cpu_type} are @samp{401}, @samp{403},
 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
-@samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
-@samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
+@samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
+@samp{e300c3}, @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7},
 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},