2003-04-02 Vladimir Makarov <vmakarov@redhat.com>

	* config/rs6000/rs6000.c
	(TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD): Redefine the
	macros.
	(rs6000_issue_rate): Add case for 8540.
	(rs6000_use_sched_lookahead): New function.

	* config/rs6000/8540.md: Rename SIU units into SU ones and MIU
	units into MU ones.
	(ppc8540_branch, ppc8540_cr_logical): Add one cycle in the
	reservation before retirement.
	(ppc8540_multiply, ppc8540_load, ppc8540_store,
	ppc8540_simple_float, ppc8540_vector_load, ppc8540_vector_store):
	Remove additional cycle in the reservation before retirement.
	(ppc8540_mfcr, ppc8540_mtcrf, ppc8540_mtjmpr): Add missed
	reservation of ppc8540_issue.

From-SVN: r65167

3 files changed, 91 insertions, 47 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 3ec6e47..7a1b00f 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,21 @@
+2003-04-02  Vladimir Makarov  <vmakarov@redhat.com>
+
+	* config/rs6000/rs6000.c
+	(TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD): Redefine the
+	macros.
+	(rs6000_issue_rate): Add case for 8540.
+	(rs6000_use_sched_lookahead): New function.
+
+	* config/rs6000/8540.md: Rename SIU units into SU ones and MIU
+	units into MU ones.
+	(ppc8540_branch, ppc8540_cr_logical): Add one cycle in the
+	reservation before retirement.
+	(ppc8540_multiply, ppc8540_load, ppc8540_store,
+	ppc8540_simple_float, ppc8540_vector_load, ppc8540_vector_store):
+	Remove additional cycle in the reservation before retirement.
+	(ppc8540_mfcr, ppc8540_mtcrf, ppc8540_mtjmpr): Add missed
+	reservation of ppc8540_issue.
+	
 2003-04-02  Andreas Schwab  <schwab@suse.de>
 
 	* real.c (decode_ieee_single): Fix decoding of SNaN bit.
diff --git a/gcc/config/rs6000/8540.md b/gcc/config/rs6000/8540.md
index 3e4d97f..b8651bb 100644
--- a/gcc/config/rs6000/8540.md
+++ b/gcc/config/rs6000/8540.md
@@ -20,6 +20,13 @@
 
 (define_automaton "ppc8540_most,ppc8540_long,ppc8540_retire")
 (define_cpu_unit "ppc8540_decode_0,ppc8540_decode_1" "ppc8540_most")
+
+;; We don't simulate general issue queue (GIC).  If we have SU insn
+;; and then SU1 insn, they can not be issued on the same cycle
+;; (although SU1 insn and then SU insn can be issued) because the SU
+;; insn will go to SU1 from GIC0 entry.  Fortunately, the first cycle
+;; multipass insn scheduling will find the situation and issue the SU1
+;; insn and then the SU insn.
 (define_cpu_unit "ppc8540_issue_0,ppc8540_issue_1"   "ppc8540_most")
 
 ;; We could describe completion buffers slots in combination with the
@@ -37,19 +44,19 @@
 ;; Branch unit:
 (define_cpu_unit "ppc8540_bu" "ppc8540_most")
 
-;; SIU:
-(define_cpu_unit "ppc8540_siu0_stage0,ppc8540_siu1_stage0" "ppc8540_most")
+;; SU:
+(define_cpu_unit "ppc8540_su0_stage0,ppc8540_su1_stage0" "ppc8540_most")
 
-;; We could describe here MIU subunits for float multiply, float add
+;; We could describe here MU subunits for float multiply, float add
 ;; etc.  But the result automaton would behave the same way as the
-;; described one pipeline below because MIU can start only one insn
+;; described one pipeline below because MU can start only one insn
 ;; per cycle.  Actually we could simplify the automaton more not
 ;; describing stages 1-3, the result automata would be the same.
-(define_cpu_unit "ppc8540_miu_stage0,ppc8540_miu_stage1" "ppc8540_most")
-(define_cpu_unit "ppc8540_miu_stage2,ppc8540_miu_stage3" "ppc8540_most")
+(define_cpu_unit "ppc8540_mu_stage0,ppc8540_mu_stage1" "ppc8540_most")
+(define_cpu_unit "ppc8540_mu_stage2,ppc8540_mu_stage3" "ppc8540_most")
 
 ;; The following unit is used to describe non-pipelined division.
-(define_cpu_unit "ppc8540_miu_div" "ppc8540_long")
+(define_cpu_unit "ppc8540_mu_div" "ppc8540_long")
 
 ;; Here we simplified LSU unit description not describing the stages.
 (define_cpu_unit "ppc8540_lsu" "ppc8540_most")
@@ -58,13 +65,13 @@
 (define_cpu_unit "present_ppc8540_decode_0" "ppc8540_most")
 (define_cpu_unit "present_ppc8540_issue_0" "ppc8540_most")
 (define_cpu_unit "present_ppc8540_retire_0" "ppc8540_retire")
-(define_cpu_unit "present_ppc8540_siu0_stage0" "ppc8540_most")
+(define_cpu_unit "present_ppc8540_su0_stage0" "ppc8540_most")
 
 ;; The following sets to make automata deterministic when option ndfa is used.
 (presence_set "present_ppc8540_decode_0" "ppc8540_decode_0")
 (presence_set "present_ppc8540_issue_0" "ppc8540_issue_0")
 (presence_set "present_ppc8540_retire_0" "ppc8540_retire_0")
-(presence_set "present_ppc8540_siu0_stage0" "ppc8540_siu0_stage0")
+(presence_set "present_ppc8540_su0_stage0" "ppc8540_su0_stage0")
 
 ;; Some useful abbreviations.
 (define_reservation "ppc8540_decode"
@@ -73,153 +80,156 @@
     "ppc8540_issue_0|ppc8540_issue_1+present_ppc8540_issue_0")
 (define_reservation "ppc8540_retire"
    "ppc8540_retire_0|ppc8540_retire_1+present_ppc8540_retire_0")
-(define_reservation "ppc8540_siu_stage0"
-   "ppc8540_siu0_stage0|ppc8540_siu1_stage0+present_ppc8540_siu0_stage0")
+(define_reservation "ppc8540_su_stage0"
+   "ppc8540_su0_stage0|ppc8540_su1_stage0+present_ppc8540_su0_stage0")
 
-;; Simple SIU insns
-(define_insn_reservation "ppc8540_siu" 1
+;; Simple SU insns
+(define_insn_reservation "ppc8540_su" 1
   (and (eq_attr "type" "integer,cmp,compare,delayed_compare,fast_compare")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su_stage0+ppc8540_retire")
 
 ;; Branch.  Actually this latency time is not used by the scheduler.
 (define_insn_reservation "ppc8540_branch" 1
   (and (eq_attr "type" "jmpreg,branch")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_bu+ppc8540_retire")
+  "ppc8540_decode,ppc8540_bu,ppc8540_retire")
 
 ;; Multiply
 (define_insn_reservation "ppc8540_multiply" 4
   (and (eq_attr "type" "imul,imul2,imul3,imul_compare")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0,ppc8540_miu_stage1,\
-   ppc8540_miu_stage2,ppc8540_miu_stage3,ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0,ppc8540_mu_stage1,\
+   ppc8540_mu_stage2,ppc8540_mu_stage3+ppc8540_retire")
 
 ;; Divide.  We use the average latency time here.  We omit reserving a
-;; retire unit because of the result automata will be huge.
+;; retire unit because of the result automata will be huge.  We ignore
+;; reservation of miu_stage3 here because we use the average latency
+;; time.
 (define_insn_reservation "ppc8540_divide" 14
   (and (eq_attr "type" "idiv")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0+ppc8540_miu_div,\
-   ppc8540_miu_div*13")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0+ppc8540_mu_div,\
+   ppc8540_mu_div*13")
 
 ;; CR logical
 (define_insn_reservation "ppc8540_cr_logical" 1
   (and (eq_attr "type" "cr_logical,delayed_cr")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_bu+ppc8540_retire")
+  "ppc8540_decode,ppc8540_bu,ppc8540_retire")
 
 ;; Mfcr
 (define_insn_reservation "ppc8540_mfcr" 1
   (and (eq_attr "type" "mfcr")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_siu1_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su1_stage0+ppc8540_retire")
 
 ;; Mtcrf
 (define_insn_reservation "ppc8540_mtcrf" 1
   (and (eq_attr "type" "mtcr")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_siu1_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su1_stage0+ppc8540_retire")
 
 ;; Mtjmpr
 (define_insn_reservation "ppc8540_mtjmpr" 1
   (and (eq_attr "type" "mtjmpr")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_siu_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su_stage0+ppc8540_retire")
 
 ;; Loads
 (define_insn_reservation "ppc8540_load" 3
   (and (eq_attr "type" "load,load_ext,load_ext_u,load_ext_ux,load_ux,load_u")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing*2,ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing,ppc8540_retire")
 
 ;; Stores.
 (define_insn_reservation "ppc8540_store" 3
   (and (eq_attr "type" "store,store_ux,store_u")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing*2,ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing,ppc8540_retire")
 
 ;; Simple FP
 (define_insn_reservation "ppc8540_simple_float" 1
   (and (eq_attr "type" "fpsimple")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu_stage0,ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su_stage0+ppc8540_retire")
 
 ;; FP
 (define_insn_reservation "ppc8540_float" 4
   (and (eq_attr "type" "fp")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0,ppc8540_miu_stage1,\
-   ppc8540_miu_stage2,ppc8540_miu_stage3+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0,ppc8540_mu_stage1,\
+   ppc8540_mu_stage2,ppc8540_mu_stage3+ppc8540_retire")
 
-;; float divides.  We omit reserving a retire unit because of the
-;; result automata will be huge.
+;; float divides.  We omit reserving a retire unit and miu_stage3
+;; because of the result automata will be huge.
 (define_insn_reservation "ppc8540_float_vector_divide" 29
   (and (eq_attr "type" "vecfdiv")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0+ppc8540_miu_div,\
-   ppc8540_miu_div*28")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0+ppc8540_mu_div,\
+   ppc8540_mu_div*28")
 
 ;; Brinc
 (define_insn_reservation "ppc8540_brinc" 1
   (and (eq_attr "type" "brinc")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su_stage0+ppc8540_retire")
 
 ;; Simple vector
 (define_insn_reservation "ppc8540_simple_vector" 1
   (and (eq_attr "type" "vecsimple")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu1_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su1_stage0+ppc8540_retire")
 
 ;; Simple vector compare
 (define_insn_reservation "ppc8540_simple_vector_compare" 1
   (and (eq_attr "type" "veccmpsimple")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su_stage0+ppc8540_retire")
 
 ;; Vector compare
 (define_insn_reservation "ppc8540_vector_compare" 1
   (and (eq_attr "type" "veccmp")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu1_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su1_stage0+ppc8540_retire")
 
 ;; evsplatfi evsplati
 (define_insn_reservation "ppc8540_vector_perm" 1
   (and (eq_attr "type" "vecperm")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_siu1_stage0+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_su1_stage0+ppc8540_retire")
 
 ;; Vector float
 (define_insn_reservation "ppc8540_float_vector" 4
   (and (eq_attr "type" "vecfloat")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0,ppc8540_miu_stage1,\
-   ppc8540_miu_stage2,ppc8540_miu_stage3+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0,ppc8540_mu_stage1,\
+   ppc8540_mu_stage2,ppc8540_mu_stage3+ppc8540_retire")
 
 ;; Vector divides: Use the average.  We omit reserving a retire unit
-;; because of the result automata will be huge.
+;; because of the result automata will be huge.  We ignore reservation
+;; of miu_stage3 here because we use the average latency time.
 (define_insn_reservation "ppc8540_vector_divide" 14
   (and (eq_attr "type" "vecdiv")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0+ppc8540_miu_div,\
-   ppc8540_miu_div*13")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0+ppc8540_mu_div,\
+   ppc8540_mu_div*13")
 
 ;; Complex vector.
 (define_insn_reservation "ppc8540_complex_vector" 4
   (and (eq_attr "type" "veccomplex")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_miu_stage0,ppc8540_miu_stage1,\
-   ppc8540_miu_stage2,ppc8540_miu_stage3+ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0,ppc8540_mu_stage1,\
+   ppc8540_mu_stage2,ppc8540_mu_stage3+ppc8540_retire")
 
 ;; Vector load
 (define_insn_reservation "ppc8540_vector_load" 3
   (and (eq_attr "type" "vecload")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing*2,ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing,ppc8540_retire")
 
 ;; Vector store
 (define_insn_reservation "ppc8540_vector_store" 3
   (and (eq_attr "type" "vecstore")
        (eq_attr "cpu" "ppc8540"))
-  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing*2,ppc8540_retire")
+  "ppc8540_decode,ppc8540_issue+ppc8540_lsu,nothing,ppc8540_retire")
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index 0fed3ac..f5d6a7e 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -241,6 +241,7 @@ static bool rs6000_rtx_costs PARAMS ((rtx, int, int, int *));
 static int rs6000_adjust_cost PARAMS ((rtx, rtx, rtx, int));
 static int rs6000_adjust_priority PARAMS ((rtx, int));
 static int rs6000_issue_rate PARAMS ((void));
+static int rs6000_use_sched_lookahead PARAMS ((void));
 
 static void rs6000_init_builtins PARAMS ((void));
 static rtx rs6000_expand_unop_builtin PARAMS ((enum insn_code, tree, rtx));
@@ -407,6 +408,9 @@ static const char alt_reg_names[][8] =
 #undef TARGET_SCHED_ADJUST_PRIORITY
 #define TARGET_SCHED_ADJUST_PRIORITY rs6000_adjust_priority
 
+#undef TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
+#define TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD rs6000_use_sched_lookahead
+
 #undef TARGET_INIT_BUILTINS
 #define TARGET_INIT_BUILTINS rs6000_init_builtins
 
@@ -12661,6 +12665,7 @@ rs6000_issue_rate ()
   case CPU_PPC603:
   case CPU_PPC750:
   case CPU_PPC7400:
+  case CPU_PPC8540:
     return 2; 
   case CPU_RIOS2:
   case CPU_PPC604:
@@ -12674,6 +12679,17 @@ rs6000_issue_rate ()
   }
 }
 
+/* Return how many instructions to look ahead for better insn
+   scheduling.  */
+
+static int
+rs6000_use_sched_lookahead ()
+{
+  if (rs6000_cpu_attr == CPU_PPC8540)
+    return 4;
+  return 0;
+}
+
 
 /* Length in units of the trampoline for entering a nested function.  */