This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [Patch, MIPS] Add Octeon3 support


Hi Catherine,

>> Would you please add some testcases and resubmit your patch?

Thanks for the review and suggestions.
Added the testcase "gcc.target/mips/octeon3-pipe-1.c"

Please review the modified patch and let us know if its good.

Thanks,
Naveen

2014-10-31  Andrew Pinski  <apinski@cavium.com>

        * config/mips/mips-cpus.def (octeon3): New cpu.
        * config/mips/mips.c (mips_rtx_cost_data): Add octeon3.
        (mips_print_operand <case 'T', case 't'>): Fix a bug as the mode
        of the comparison no longer matches mode of the operands.
        (mips_issue_rate): Handle PROCESSOR_OCTEON3.
        * config/mips/mips.h (TARGET_OCTEON):  Add Octeon3.
        (TARGET_OCTEON2): Likewise.
        (TUNE_OCTEON): Add Octeon3.
        * config/mips/mips.md (processor): Add octeon3.
        * config/mips/octeon.md (octeon_fpu): New automaton and cpu_unit.
        (octeon_arith): Add octeon3.
        (octeon_condmove): Remove.
        (octeon_condmove_o1): New reservation.
        (octeon_condmove_o2): New reservation.
        (octeon_condmove_o3_int_on_cc): New reservation.
        (octeon_load_o2): Add octeon3.
        (octeon_cop_o2): Likewise.
        (octeon_store): Likewise.
        (octeon_brj_o2): Likewise.
        (octeon_imul3_o2): Likewise.
        (octeon_imul_o2): Likewise.
        (octeon_mfhilo_o2): Likewise.
        (octeon_imadd_o2): Likewise.
        (octeon_idiv_o2_si): Likewise.
        (octeon_idiv_o2_di): Likewise.
        (octeon_fpu): Add to the automaton.
        (octeon_fpu): New cpu unit.
        (octeon_condmove_o2): Check for non floating point modes.
        (octeon_load_o2): Add prefetchx.
        (octeon_cop_o2): Don't check for octeon3.
        (octeon3_faddsubcvt): New reservation.
        (octeon3_fmul): Likewise.
        (octeon3_fmadd): Likewise.
        (octeon3_div_sf): Likewise.
        (octeon3_div_df): Likewise.
        (octeon3_sqrt_sf): Likewise.
        (octeon3_sqrt_df): Likewise.
        (octeon3_rsqrt_sf): Likewise.
        (octeon3_rsqrt_df): Likewise.
        (octeon3_fabsnegmov): Likewise.
        (octeon_fcond): Likewise.
        (octeon_fcondmov): Likewise.
        (octeon_fpmtc1): Likewise.
        (octeon_fpmfc1): Likewise.
        (octeon_fpload): Likewise.
        (octeon_fpstore): Likewise.
        * config/mips/mips-tables.opt: Regenerate.
        * doc/invoke.texi (-march=@var{arch}): Add octeon3.

2014-10-31  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>

        * gcc.target/mips/octeon3-pipe-1.c: New test.
diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 28ae552..77455a2 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,54 @@
+2014-10-31  Andrew Pinski  <apinski@cavium.com>
+
+	* config/mips/mips-cpus.def (octeon3): New cpu.
+	* config/mips/mips.c (mips_rtx_cost_data): Add octeon3.
+	(mips_print_operand <case 'T', case 't'>): Fix a bug as the mode
+	of the comparison no longer matches mode of the operands.
+	(mips_issue_rate): Handle PROCESSOR_OCTEON3.
+	* config/mips/mips.h (TARGET_OCTEON):  Add Octeon3.
+	(TARGET_OCTEON2): Likewise.
+	(TUNE_OCTEON): Add Octeon3.
+	* config/mips/mips.md (processor): Add octeon3.
+	* config/mips/octeon.md (octeon_fpu): New automaton and cpu_unit.
+	(octeon_arith): Add octeon3.
+	(octeon_condmove): Remove.
+	(octeon_condmove_o1): New reservation.
+	(octeon_condmove_o2): New reservation.
+	(octeon_condmove_o3_int_on_cc): New reservation.
+	(octeon_load_o2): Add octeon3.
+	(octeon_cop_o2): Likewise.
+	(octeon_store): Likewise.
+	(octeon_brj_o2): Likewise.
+	(octeon_imul3_o2): Likewise.
+	(octeon_imul_o2): Likewise.
+	(octeon_mfhilo_o2): Likewise.
+	(octeon_imadd_o2): Likewise.
+	(octeon_idiv_o2_si): Likewise.
+	(octeon_idiv_o2_di): Likewise.
+	(octeon_fpu): Add to the automaton.
+	(octeon_fpu): New cpu unit.
+	(octeon_condmove_o2): Check for non floating point modes.
+	(octeon_load_o2): Add prefetchx.
+	(octeon_cop_o2): Don't check for octeon3.
+	(octeon3_faddsubcvt): New reservation.
+	(octeon3_fmul): Likewise.
+	(octeon3_fmadd): Likewise.
+	(octeon3_div_sf): Likewise.
+	(octeon3_div_df): Likewise.
+	(octeon3_sqrt_sf): Likewise.
+	(octeon3_sqrt_df): Likewise.
+	(octeon3_rsqrt_sf): Likewise.
+	(octeon3_rsqrt_df): Likewise.
+	(octeon3_fabsnegmov): Likewise.
+	(octeon_fcond): Likewise.
+	(octeon_fcondmov): Likewise.
+	(octeon_fpmtc1): Likewise.
+	(octeon_fpmfc1): Likewise.
+	(octeon_fpload): Likewise.
+	(octeon_fpstore): Likewise.
+	* config/mips/mips-tables.opt: Regenerate.
+	* doc/invoke.texi (-march=@var{arch}): Add octeon3.
+
 2014-10-10  Felix Yang  <felix.yang@huawei.com>
 
 	* config/xtensa/xtensa.h (TARGET_LOOPS): New Macro.
diff --git a/gcc/config/mips/mips-cpus.def b/gcc/config/mips/mips-cpus.def
index d5528d3..e2985b8 100644
--- a/gcc/config/mips/mips-cpus.def
+++ b/gcc/config/mips/mips-cpus.def
@@ -162,4 +162,5 @@ MIPS_CPU ("loongson3a", PROCESSOR_LOONGSON_3A, 65, PTF_AVOID_BRANCHLIKELY)
 MIPS_CPU ("octeon", PROCESSOR_OCTEON, 65, PTF_AVOID_BRANCHLIKELY)
 MIPS_CPU ("octeon+", PROCESSOR_OCTEON, 65, PTF_AVOID_BRANCHLIKELY)
 MIPS_CPU ("octeon2", PROCESSOR_OCTEON2, 65, PTF_AVOID_BRANCHLIKELY)
+MIPS_CPU ("octeon3", PROCESSOR_OCTEON3, 65, PTF_AVOID_BRANCHLIKELY)
 MIPS_CPU ("xlp", PROCESSOR_XLP, 65, PTF_AVOID_BRANCHLIKELY)
diff --git a/gcc/config/mips/mips-tables.opt b/gcc/config/mips/mips-tables.opt
index 5791b41..99d2ed8 100644
--- a/gcc/config/mips/mips-tables.opt
+++ b/gcc/config/mips/mips-tables.opt
@@ -667,5 +667,8 @@ EnumValue
 Enum(mips_arch_opt_value) String(octeon2) Value(94) Canonical
 
 EnumValue
-Enum(mips_arch_opt_value) String(xlp) Value(95) Canonical
+Enum(mips_arch_opt_value) String(octeon3) Value(95) Canonical
+
+EnumValue
+Enum(mips_arch_opt_value) String(xlp) Value(96) Canonical
 
diff --git a/gcc/config/mips/mips.c b/gcc/config/mips/mips.c
index 3d9db92..3deb6e0 100644
--- a/gcc/config/mips/mips.c
+++ b/gcc/config/mips/mips.c
@@ -975,6 +975,20 @@ static const struct mips_rtx_cost_data
                      4,		  /* branch_cost */
                      4		  /* memory_latency */
   },
+    /* Octeon III */
+  {
+    COSTS_N_INSNS (6),            /* fp_add */
+    COSTS_N_INSNS (6),            /* fp_mult_sf */
+    COSTS_N_INSNS (7),            /* fp_mult_df */
+    COSTS_N_INSNS (25),           /* fp_div_sf */
+    COSTS_N_INSNS (48),           /* fp_div_df */
+    COSTS_N_INSNS (6),            /* int_mult_si */
+    COSTS_N_INSNS (6),            /* int_mult_di */
+    COSTS_N_INSNS (18),           /* int_div_si */
+    COSTS_N_INSNS (35),           /* int_div_di */
+                     4,		  /* branch_cost */
+                     4		  /* memory_latency */
+  },
   { /* R3900 */
     COSTS_N_INSNS (2),            /* fp_add */
     COSTS_N_INSNS (4),            /* fp_mult_sf */
@@ -13172,6 +13186,7 @@ mips_issue_rate (void)
     case PROCESSOR_R9000:
     case PROCESSOR_OCTEON:
     case PROCESSOR_OCTEON2:
+    case PROCESSOR_OCTEON3:
       return 2;
 
     case PROCESSOR_SB1:
diff --git a/gcc/config/mips/mips.h b/gcc/config/mips/mips.h
index c7b998b..726cd48 100644
--- a/gcc/config/mips/mips.h
+++ b/gcc/config/mips/mips.h
@@ -230,8 +230,10 @@ struct mips_cpu_info {
 #define TARGET_MIPS7000             (mips_arch == PROCESSOR_R7000)
 #define TARGET_MIPS9000             (mips_arch == PROCESSOR_R9000)
 #define TARGET_OCTEON		    (mips_arch == PROCESSOR_OCTEON	\
-				     || mips_arch == PROCESSOR_OCTEON2)
-#define TARGET_OCTEON2		    (mips_arch == PROCESSOR_OCTEON2)
+				     || mips_arch == PROCESSOR_OCTEON2	\
+				     || mips_arch == PROCESSOR_OCTEON3)
+#define TARGET_OCTEON2		    (mips_arch == PROCESSOR_OCTEON2	\
+				     || mips_arch == PROCESSOR_OCTEON3)
 #define TARGET_SB1                  (mips_arch == PROCESSOR_SB1		\
 				     || mips_arch == PROCESSOR_SB1A)
 #define TARGET_SR71K                (mips_arch == PROCESSOR_SR71000)
@@ -261,7 +263,8 @@ struct mips_cpu_info {
 #define TUNE_MIPS7000               (mips_tune == PROCESSOR_R7000)
 #define TUNE_MIPS9000               (mips_tune == PROCESSOR_R9000)
 #define TUNE_OCTEON		    (mips_tune == PROCESSOR_OCTEON	\
-				     || mips_tune == PROCESSOR_OCTEON2)
+				     || mips_tune == PROCESSOR_OCTEON2	\
+				     || mips_tune == PROCESSOR_OCTEON3)
 #define TUNE_SB1                    (mips_tune == PROCESSOR_SB1		\
 				     || mips_tune == PROCESSOR_SB1A)
 #define TUNE_P5600                  (mips_tune == PROCESSOR_P5600)
diff --git a/gcc/config/mips/mips.md b/gcc/config/mips/mips.md
index d47bb78..4b72546 100644
--- a/gcc/config/mips/mips.md
+++ b/gcc/config/mips/mips.md
@@ -41,6 +41,7 @@
   m4k
   octeon
   octeon2
+  octeon3
   r3900
   r6000
   r4000
diff --git a/gcc/config/mips/octeon.md b/gcc/config/mips/octeon.md
index 1d6251c..960894f 100644
--- a/gcc/config/mips/octeon.md
+++ b/gcc/config/mips/octeon.md
@@ -22,41 +22,55 @@
 ;; Octeon is a dual-issue processor that can issue all instructions on
 ;; pipe0 and a subset on pipe1.
 
-(define_automaton "octeon_main, octeon_mult")
+(define_automaton "octeon_main, octeon_mult, octeon_fpu")
 
 (define_cpu_unit "octeon_pipe0" "octeon_main")
 (define_cpu_unit "octeon_pipe1" "octeon_main")
 (define_cpu_unit "octeon_mult" "octeon_mult")
+(define_cpu_unit "octeon_fpu" "octeon_fpu")
 
 (define_insn_reservation "octeon_arith" 1
-  (and (eq_attr "cpu" "octeon,octeon2")
+  (and (eq_attr "cpu" "octeon,octeon2,octeon3")
        (eq_attr "type" "arith,const,logical,move,shift,signext,slt,nop"))
   "octeon_pipe0 | octeon_pipe1")
 
-(define_insn_reservation "octeon_condmove" 2
-  (and (eq_attr "cpu" "octeon,octeon2")
+(define_insn_reservation "octeon_condmove_o1" 2
+  (and (eq_attr "cpu" "octeon")
        (eq_attr "type" "condmove"))
   "octeon_pipe0 | octeon_pipe1")
 
+(define_insn_reservation "octeon_condmove_o2" 3
+  (and (eq_attr "cpu" "octeon2,octeon3")
+       (eq_attr "type" "condmove")
+       (not (eq_attr "mode" "SF, DF")))
+  "octeon_pipe0 | octeon_pipe1")
+
+;; movt/movf can only issue in pipe1
+(define_insn_reservation "octeon_condmove_o3_int_on_cc" 3
+  (and (eq_attr "cpu" "octeon2,octeon3")
+       (eq_attr "type" "condmove")
+       (not (eq_attr "mode" "SF, DF")))
+  "octeon_pipe1")
+
 (define_insn_reservation "octeon_load_o1" 2
   (and (eq_attr "cpu" "octeon")
        (eq_attr "type" "load,prefetch,mtc,mfc"))
   "octeon_pipe0")
 
 (define_insn_reservation "octeon_load_o2" 3
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "load,prefetch"))
   "octeon_pipe0")
 
 ;; ??? memory-related cop0 reads are pipe0 with 3-cycle latency.
 ;; Front-end-related ones are 1-cycle on pipe1.  Assume front-end for now.
 (define_insn_reservation "octeon_cop_o2" 1
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "mtc,mfc"))
   "octeon_pipe1")
 
 (define_insn_reservation "octeon_store" 1
-  (and (eq_attr "cpu" "octeon,octeon2")
+  (and (eq_attr "cpu" "octeon,octeon2,octeon3")
        (eq_attr "type" "store"))
   "octeon_pipe0")
 
@@ -66,7 +80,7 @@
   "octeon_pipe0")
 
 (define_insn_reservation "octeon_brj_o2" 2
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "branch,jump,call,trap"))
   "octeon_pipe1")
 
@@ -76,7 +90,7 @@
   "(octeon_pipe0 | octeon_pipe1) + octeon_mult")
 
 (define_insn_reservation "octeon_imul3_o2" 6
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "imul3,pop,clz"))
   "octeon_pipe1 + octeon_mult")
 
@@ -86,7 +100,7 @@
   "(octeon_pipe0 | octeon_pipe1) + octeon_mult, octeon_mult")
 
 (define_insn_reservation "octeon_imul_o2" 1
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "imul,mthi,mtlo"))
   "octeon_pipe1 + octeon_mult")
 
@@ -96,7 +110,7 @@
   "(octeon_pipe0 | octeon_pipe1) + octeon_mult")
 
 (define_insn_reservation "octeon_mfhilo_o2" 6
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "mfhi,mflo"))
   "octeon_pipe1 + octeon_mult")
 
@@ -106,7 +120,7 @@
   "(octeon_pipe0 | octeon_pipe1) + octeon_mult, octeon_mult*3")
 
 (define_insn_reservation "octeon_imadd_o2" 1
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "type" "imadd"))
   "octeon_pipe1 + octeon_mult")
 
@@ -116,13 +130,13 @@
   "(octeon_pipe0 | octeon_pipe1) + octeon_mult, octeon_mult*71")
 
 (define_insn_reservation "octeon_idiv_o2_si" 18
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "mode" "SI")
        (eq_attr "type" "idiv"))
   "octeon_pipe1 + octeon_mult, octeon_mult*17")
 
 (define_insn_reservation "octeon_idiv_o2_di" 35
-  (and (eq_attr "cpu" "octeon2")
+  (and (eq_attr "cpu" "octeon2,octeon3")
        (eq_attr "mode" "DI")
        (eq_attr "type" "idiv"))
   "octeon_pipe1 + octeon_mult, octeon_mult*34")
@@ -131,6 +145,95 @@
 ;; patterns.
 
 (define_insn_reservation "octeon_unknown" 1
-  (and (eq_attr "cpu" "octeon,octeon2")
+  (and (eq_attr "cpu" "octeon,octeon2,octeon3")
        (eq_attr "type" "unknown,multi,atomic,syncloop"))
   "octeon_pipe0 + octeon_pipe1")
+
+;; Octeon3 FPU
+
+(define_insn_reservation "octeon3_faddsubcvt" 4
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fadd, fcvt"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon3_fmul" 5
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fmul"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon3_fmadd" 9
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fmadd"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu")
+
+(define_insn_reservation "octeon3_div_sf" 12
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fdiv, frdiv")
+       (eq_attr "mode" "SF"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu*8")
+
+(define_insn_reservation "octeon3_div_df" 22
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fdiv, frdiv")
+       (eq_attr "mode" "SF"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu*18")
+
+(define_insn_reservation "octeon3_sqrt_sf" 16
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fsqrt")
+       (eq_attr "mode" "SF"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu*12")
+
+(define_insn_reservation "octeon3_sqrt_df" 30
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fsqrt")
+       (eq_attr "mode" "DF"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu*26")
+
+(define_insn_reservation "octeon3_rsqrt_sf" 27
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "frsqrt")
+       (eq_attr "mode" "SF"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu*23")
+
+(define_insn_reservation "octeon3_rsqrt_df" 51
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "frsqrt")
+       (eq_attr "mode" "DF"))
+  "octeon_pipe1 + octeon_fpu, octeon_fpu*47")
+
+(define_insn_reservation "octeon3_fabsnegmov" 2
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fabs, fneg, fmove"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon_fcond" 1
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fcmp"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon_fcondmov" 2
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "condmove")
+       (eq_attr "mode" "SF,DF"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon_fpmtc1" 2
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "mtc"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon_fpmfc1" 6
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "mtc"))
+  "octeon_pipe1 + octeon_fpu")
+
+(define_insn_reservation "octeon_fpload" 3
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fpload,fpidxload"))
+  "octeon_pipe0 + octeon_fpu")
+
+(define_insn_reservation "octeon_fpstore" 3
+  (and (eq_attr "cpu" "octeon3")
+       (eq_attr "type" "fpstore,fpidxstore"))
+  "octeon_pipe0 + octeon_pipe1")
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index 792f25b..b75ebc7 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -17531,7 +17531,7 @@ The processor names are:
 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
 @samp{m4k},
 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
-@samp{octeon}, @samp{octeon+}, @samp{octeon2},
+@samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
 @samp{orion},
 @samp{p5600},
 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 3aa9b6b..f3ebee5 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,7 @@
+2014-10-31  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
+
+	* gcc.target/mips/octeon3-pipe-1.c: New test.
+
 2014-10-30  Marek Polacek  <polacek@redhat.com>
 
 	* gcc.dg/diag-aka-1.c: New test.
diff --git a/gcc/testsuite/gcc.target/mips/octeon3-pipe-1.c b/gcc/testsuite/gcc.target/mips/octeon3-pipe-1.c
new file mode 100644
index 0000000..3b83266
--- /dev/null
+++ b/gcc/testsuite/gcc.target/mips/octeon3-pipe-1.c
@@ -0,0 +1,12 @@
+/* Check that we use the octeon3 pipeline description.  */
+/* { dg-do compile } */
+/* { dg-options "-fschedule-insns2 -fdump-rtl-sched2 -march=octeon3" } */
+/* { dg-skip-if "code quality test" { *-*-* } { "-O0" } { "" } } */
+
+NOMIPS16 int f (int a, int b)
+{
+  return a / b;
+}
+
+/* { dg-final { scan-rtl-dump "octeon_mult\\*17" "sched2" } }  */
+/* { dg-final { cleanup-rtl-dump "sched2" } }  */

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]