diff mbox

Bobcat (btver1) Enablement

Message ID D4C76825A6780047854A11E93CDE84D004C1768430@SAUSEXMBP01.amd.com
State New
Headers show

Commit Message

Fang, Changpeng Jan. 5, 2011, 12:19 a.m. UTC
Hi,

This patch does the basic enablement for AMD's first APU architecture -- Bobcat.
It defines -march=btver1 and -mtune=btver1, and lets -march=native correctly
recognizes bobcat. In addition, it does limited tuning for bobcat.

The patch passed bootstrapping and gcc regression tests on x86_64-unknown-linux-gnu.
Programs (cou2000, cpu2006 and polyhedron 2005, etc) compile and run OK with 
-march=btver1 on the bobcat system.

Is it OK to commit to trunk now?

Thanks,

Changpeng

Comments

Fang, Changpeng Jan. 5, 2011, 5:29 p.m. UTC | #1
Is this Bobcat enablement good for 4.6?

Thanks,

Changpeng
Eric Botcazou Jan. 6, 2011, 11:32 p.m. UTC | #2
> Is it OK to commit to trunk now?

Please fix the formatting of the ChangeLog, see other entries in the file.
diff mbox

Patch

From 5a97154db22f3bcc06125f55cd1f632dc4374626 Mon Sep 17 00:00:00 2001
From: Changpeng Fang <chfang@houghton.(none)>
Date: Tue, 4 Jan 2011 11:45:32 -0800
Subject: [PATCH] Bobcat (btver1) Enablement

	* config.gcc (i[34567]86-*-linux* | ...): Add btver1.
	  (case ${target}): Add btver1.

	* config/i386/driver-i386.c (host_detect_local_cpu): Let
	  -march=native recognize btver1 processors.

	* config/i386/i386-c.c (ix86_target_macros_internal): Add
	  btver1 def_and_undef

	* config/i386/i386.c (struct processor_costs btver1_cost): New
	  btver1 cost table.
	  (m_BTVER1): New definition.
	  ( m_AMD_MULTIPLE): Includes m_BTVER1.
	  (initial_ix86_tune_features): Add btver1 tune.
	  (processor_target_table): Add btver1 entry.
	  (static const char *const cpu_names): Add btver1 entry.
	  (software_prefetching_beneficial_p): Add btver1.
	  (ix86_option_override_internal): Add btver1 instruction sets.
	  (ix86_issue_rate): Add btver1.
	  (ix86_adjust_cost): Add btver1.

	* config/i386/i386.h (TARGET_BTVER1): New definition.
	  (enum target_cpu_default): Add TARGET_CPU_DEFAULT_btver1.
	  (enum processor_type): Add PROCESSOR_BTVER1.

	* config/i386/i386.md (define_attr "cpu"): Add btver1.
---
 gcc/config.gcc                |   22 ++++++---
 gcc/config/i386/driver-i386.c |    5 ++
 gcc/config/i386/i386-c.c      |    7 +++
 gcc/config/i386/i386.c        |  105 ++++++++++++++++++++++++++++++++++++++---
 gcc/config/i386/i386.h        |    3 +
 gcc/config/i386/i386.md       |    2 +-
 6 files changed, 129 insertions(+), 15 deletions(-)

diff --git a/gcc/config.gcc b/gcc/config.gcc
index 81eeb84..96ff86c 100644
--- a/gcc/config.gcc
+++ b/gcc/config.gcc
@@ -1236,7 +1236,7 @@  i[34567]86-*-linux* | i[34567]86-*-kfreebsd*-gnu | i[34567]86-*-knetbsd*-gnu | i
 			need_64bit_hwint=yes
 			need_64bit_isa=yes
 			case X"${with_cpu}" in
-			Xgeneric|Xatom|Xcore2|Xcorei7|Xcorei7-avx|Xnocona|Xx86-64|Xbdver1|Xamdfam10|Xbarcelona|Xk8|Xopteron|Xathlon64|Xathlon-fx|Xathlon64-sse3|Xk8-sse3|Xopteron-sse3)
+			Xgeneric|Xatom|Xcore2|Xcorei7|Xcorei7-avx|Xnocona|Xx86-64|Xbdver1|Xbtver1|Xamdfam10|Xbarcelona|Xk8|Xopteron|Xathlon64|Xathlon-fx|Xathlon64-sse3|Xk8-sse3|Xopteron-sse3)
 				;;
 			X)
 				if test x$with_cpu_64 = x; then
@@ -1245,7 +1245,7 @@  i[34567]86-*-linux* | i[34567]86-*-kfreebsd*-gnu | i[34567]86-*-knetbsd*-gnu | i
 				;;
 			*)
 				echo "Unsupported CPU used in --with-cpu=$with_cpu, supported values:" 1>&2
-				echo "generic atom core2 corei7 corei7-avx nocona x86-64 bdver1 amdfam10 barcelona k8 opteron athlon64 athlon-fx athlon64-sse3 k8-sse3 opteron-sse3" 1>&2
+				echo "generic atom core2 corei7 corei7-avx nocona x86-64 bdver1 btver1 amdfam10 barcelona k8 opteron athlon64 athlon-fx athlon64-sse3 k8-sse3 opteron-sse3" 1>&2
 				exit 1
 				;;
 			esac
@@ -1374,7 +1374,7 @@  i[34567]86-*-solaris2*)
 		need_64bit_isa=yes
 		use_gcc_stdint=wrap
 		case X"${with_cpu}" in
-		Xgeneric|Xatom|Xcore2|Xcorei7|Xcorei7-avx|Xnocona|Xx86-64|Xbdver1|Xamdfam10|Xbarcelona|Xk8|Xopteron|Xathlon64|Xathlon-fx|Xathlon64-sse3|Xk8-sse3|Xopteron-sse3)
+		Xgeneric|Xatom|Xcore2|Xcorei7|Xcorei7-avx|Xnocona|Xx86-64|Xbdver1|Xbtver1|Xamdfam10|Xbarcelona|Xk8|Xopteron|Xathlon64|Xathlon-fx|Xathlon64-sse3|Xk8-sse3|Xopteron-sse3)
 			;;
 		X)
 			if test x$with_cpu_64 = x; then
@@ -1383,7 +1383,7 @@  i[34567]86-*-solaris2*)
 			;;
 		*)
 			echo "Unsupported CPU used in --with-cpu=$with_cpu, supported values:" 1>&2
-			echo "generic atom core2 corei7 corei7-avx nocona x86-64 bdver1 amdfam10 barcelona k8 opteron athlon64 athlon-fx athlon64-sse3 k8-sse3 opteron-sse3" 1>&2
+			echo "generic atom core2 corei7 corei7-avx nocona x86-64 bdver1 btver1 amdfam10 barcelona k8 opteron athlon64 athlon-fx athlon64-sse3 k8-sse3 opteron-sse3" 1>&2
 			exit 1
 			;;
 		esac
@@ -1454,7 +1454,7 @@  i[34567]86-*-mingw* | x86_64-*-mingw*)
 			if test x$enable_targets = xall; then
 				tm_defines="${tm_defines} TARGET_BI_ARCH=1"
 				case X"${with_cpu}" in
-				Xgeneric|Xatom|Xcore2|Xcorei7|Xcorei7-avx|Xnocona|Xx86-64|Xbdver1|Xamdfam10|Xbarcelona|Xk8|Xopteron|Xathlon64|Xathlon-fx|Xathlon64-sse3|Xk8-sse3|Xopteron-sse3)
+				Xgeneric|Xatom|Xcore2|Xcorei7|Xcorei7-avx|Xnocona|Xx86-64|Xbdver1|Xbtver1|Xamdfam10|Xbarcelona|Xk8|Xopteron|Xathlon64|Xathlon-fx|Xathlon64-sse3|Xk8-sse3|Xopteron-sse3)
 					;;
 				X)
 					if test x$with_cpu_64 = x; then
@@ -1463,7 +1463,7 @@  i[34567]86-*-mingw* | x86_64-*-mingw*)
 					;;
 				*)
 					echo "Unsupported CPU used in --with-cpu=$with_cpu, supported values:" 1>&2
-					echo "generic atom core2 corei7 Xcorei7-avx nocona x86-64 bdver1 amdfam10 barcelona k8 opteron athlon64 athlon-fx athlon64-sse3 k8-sse3 opteron-sse3" 1>&2
+					echo "generic atom core2 corei7 Xcorei7-avx nocona x86-64 bdver1 btver1 amdfam10 barcelona k8 opteron athlon64 athlon-fx athlon64-sse3 k8-sse3 opteron-sse3" 1>&2
 					exit 1
 					;;
 				esac
@@ -2797,6 +2797,10 @@  case ${target} in
 	arch=bdver1
 	cpu=bdver1
 	;;
+      btver1-*)
+	arch=btver1
+	cpu=btver1
+	;;
       amdfam10-*|barcelona-*)
 	arch=amdfam10
 	cpu=amdfam10
@@ -2886,6 +2890,10 @@  case ${target} in
 	arch=bdver1
 	cpu=bdver1
 	;;
+      btver1-*)
+	arch=btver1
+	cpu=btver1
+	;;
       amdfam10-*|barcelona-*)
 	arch=amdfam10
 	cpu=amdfam10
@@ -3302,7 +3310,7 @@  case "${target}" in
 				;;
 			"" | x86-64 | generic | native \
 			| k8 | k8-sse3 | athlon64 | athlon64-sse3 | opteron \
-			| opteron-sse3 | athlon-fx | bdver1 | amdfam10 \
+			| opteron-sse3 | athlon-fx | bdver1 | btver1 | amdfam10 \
 			| barcelona | nocona | core2 | corei7 | corei7-avx | atom)
 				# OK
 				;;
diff --git a/gcc/config/i386/driver-i386.c b/gcc/config/i386/driver-i386.c
index 618f875..35962c8 100644
--- a/gcc/config/i386/driver-i386.c
+++ b/gcc/config/i386/driver-i386.c
@@ -500,6 +500,8 @@  const char *host_detect_local_cpu (int argc, const char **argv)
 	processor = PROCESSOR_GEODE;
       else if (has_xop)
 	processor = PROCESSOR_BDVER1;
+      else if (has_sse4a && has_ssse3)
+        processor = PROCESSOR_BTVER1;
       else if (has_sse4a)
 	processor = PROCESSOR_AMDFAM10;
       else if (has_sse2 || has_longmode)
@@ -647,6 +649,9 @@  const char *host_detect_local_cpu (int argc, const char **argv)
     case PROCESSOR_BDVER1:
       cpu = "bdver1";
       break;
+    case PROCESSOR_BTVER1:
+      cpu = "btver1";
+      break;
 
     default:
       /* Use something reasonable.  */
diff --git a/gcc/config/i386/i386-c.c b/gcc/config/i386/i386-c.c
index 6adf613..1497351 100644
--- a/gcc/config/i386/i386-c.c
+++ b/gcc/config/i386/i386-c.c
@@ -110,6 +110,10 @@  ix86_target_macros_internal (int isa_flag,
       def_or_undef (parse_in, "__bdver1");
       def_or_undef (parse_in, "__bdver1__");
       break;
+    case PROCESSOR_BTVER1:
+      def_or_undef (parse_in, "__btver1");
+      def_or_undef (parse_in, "__btver1__");
+      break;
     case PROCESSOR_PENTIUM4:
       def_or_undef (parse_in, "__pentium4");
       def_or_undef (parse_in, "__pentium4__");
@@ -194,6 +198,9 @@  ix86_target_macros_internal (int isa_flag,
     case PROCESSOR_BDVER1:
       def_or_undef (parse_in, "__tune_bdver1__");
       break;
+   case PROCESSOR_BTVER1:
+      def_or_undef (parse_in, "__tune_btver1__");
+      break;
     case PROCESSOR_PENTIUM4:
       def_or_undef (parse_in, "__tune_pentium4__");
       break;
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 50dac35..59a182c 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -1231,6 +1231,88 @@  struct processor_costs bdver1_cost = {
   1,					/* cond_not_taken_branch_cost.  */
 };
 
+struct processor_costs btver1_cost = {
+  COSTS_N_INSNS (1),			/* cost of an add instruction */
+  COSTS_N_INSNS (2),			/* cost of a lea instruction */
+  COSTS_N_INSNS (1),			/* variable shift costs */
+  COSTS_N_INSNS (1),			/* constant shift costs */
+  {COSTS_N_INSNS (3),			/* cost of starting multiply for QI */
+   COSTS_N_INSNS (4),			/*				 HI */
+   COSTS_N_INSNS (3),			/*				 SI */
+   COSTS_N_INSNS (4),			/*				 DI */
+   COSTS_N_INSNS (5)},			/*			      other */
+  0,					/* cost of multiply per each bit set */
+  {COSTS_N_INSNS (19),			/* cost of a divide/mod for QI */
+   COSTS_N_INSNS (35),			/*			    HI */
+   COSTS_N_INSNS (51),			/*			    SI */
+   COSTS_N_INSNS (83),			/*			    DI */
+   COSTS_N_INSNS (83)},			/*			    other */
+  COSTS_N_INSNS (1),			/* cost of movsx */
+  COSTS_N_INSNS (1),			/* cost of movzx */
+  8,					/* "large" insn */
+  9,					/* MOVE_RATIO */
+  4,				     /* cost for loading QImode using movzbl */
+  {3, 4, 3},				/* cost of loading integer registers
+					   in QImode, HImode and SImode.
+					   Relative to reg-reg move (2).  */
+  {3, 4, 3},				/* cost of storing integer registers */
+  4,					/* cost of reg,reg fld/fst */
+  {4, 4, 12},				/* cost of loading fp registers
+		   			   in SFmode, DFmode and XFmode */
+  {6, 6, 8},				/* cost of storing fp registers
+ 		   			   in SFmode, DFmode and XFmode */
+  2,					/* cost of moving MMX register */
+  {3, 3},				/* cost of loading MMX registers
+					   in SImode and DImode */
+  {4, 4},				/* cost of storing MMX registers
+					   in SImode and DImode */
+  2,					/* cost of moving SSE register */
+  {4, 4, 3},				/* cost of loading SSE registers
+					   in SImode, DImode and TImode */
+  {4, 4, 5},				/* cost of storing SSE registers
+					   in SImode, DImode and TImode */
+  3,					/* MMX or SSE register to integer */
+  					/* On K8:
+  					    MOVD reg64, xmmreg Double FSTORE 4
+					    MOVD reg32, xmmreg Double FSTORE 4
+					   On AMDFAM10:
+					    MOVD reg64, xmmreg Double FADD 3
+							       1/1  1/1
+					    MOVD reg32, xmmreg Double FADD 3
+							       1/1  1/1 */
+  32,					/* size of l1 cache.  */
+  512,					/* size of l2 cache.  */
+  64,					/* size of prefetch block */
+  100,					/* number of parallel prefetches */
+  2,					/* Branch cost */
+  COSTS_N_INSNS (4),			/* cost of FADD and FSUB insns.  */
+  COSTS_N_INSNS (4),			/* cost of FMUL instruction.  */
+  COSTS_N_INSNS (19),			/* cost of FDIV instruction.  */
+  COSTS_N_INSNS (2),			/* cost of FABS instruction.  */
+  COSTS_N_INSNS (2),			/* cost of FCHS instruction.  */
+  COSTS_N_INSNS (35),			/* cost of FSQRT instruction.  */
+
+  /* BTVER1 has optimized REP instruction for medium sized blocks, but for
+     very small blocks it is better to use loop. For large blocks, libcall can
+     do nontemporary accesses and beat inline considerably.  */
+  {{libcall, {{6, loop}, {14, unrolled_loop}, {-1, rep_prefix_4_byte}}},
+   {libcall, {{16, loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+  {{libcall, {{8, loop}, {24, unrolled_loop},
+	      {2048, rep_prefix_4_byte}, {-1, libcall}}},
+   {libcall, {{48, unrolled_loop}, {8192, rep_prefix_8_byte}, {-1, libcall}}}},
+  4,					/* scalar_stmt_cost.  */
+  2,					/* scalar load_cost.  */
+  2,					/* scalar_store_cost.  */
+  6,					/* vec_stmt_cost.  */
+  0,					/* vec_to_scalar_cost.  */
+  2,					/* scalar_to_vec_cost.  */
+  2,					/* vec_align_load_cost.  */
+  2,					/* vec_unalign_load_cost.  */
+  2,					/* vec_store_cost.  */
+  2,					/* cond_taken_branch_cost.  */
+  1,					/* cond_not_taken_branch_cost.  */
+};
+
 static const
 struct processor_costs pentium4_cost = {
   COSTS_N_INSNS (1),			/* cost of an add instruction */
@@ -1624,7 +1706,8 @@  const struct processor_costs *ix86_cost = &pentium_cost;
 #define m_ATHLON_K8  (m_K8 | m_ATHLON)
 #define m_AMDFAM10  (1<<PROCESSOR_AMDFAM10)
 #define m_BDVER1  (1<<PROCESSOR_BDVER1)
-#define m_AMD_MULTIPLE  (m_K8 | m_ATHLON | m_AMDFAM10 | m_BDVER1)
+#define m_BTVER1  (1<<PROCESSOR_BTVER1)
+#define m_AMD_MULTIPLE  (m_K8 | m_ATHLON | m_AMDFAM10 | m_BDVER1 | m_BTVER1)
 
 #define m_GENERIC32 (1<<PROCESSOR_GENERIC32)
 #define m_GENERIC64 (1<<PROCESSOR_GENERIC64)
@@ -1670,8 +1753,8 @@  static unsigned int initial_ix86_tune_features[X86_TUNE_LAST] = {
   ~m_386,
 
   /* X86_TUNE_USE_SAHF */
-  m_ATOM | m_PPRO | m_K6_GEODE | m_K8 | m_AMDFAM10 | m_BDVER1 | m_PENT4
-  | m_NOCONA | m_CORE2I7 | m_GENERIC,
+  m_ATOM | m_PPRO | m_K6_GEODE | m_K8 | m_AMDFAM10 | m_BDVER1 | m_BTVER1
+  | m_PENT4 | m_NOCONA | m_CORE2I7 | m_GENERIC,
 
   /* X86_TUNE_MOVX: Enable to zero extend integer registers to avoid
      partial dependencies.  */
@@ -1777,7 +1860,7 @@  static unsigned int initial_ix86_tune_features[X86_TUNE_LAST] = {
   | m_AMDFAM10 | m_BDVER1,
 
   /* X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL */
-  m_AMDFAM10 | m_BDVER1 | m_COREI7,
+  m_AMDFAM10 | m_BDVER1 | m_BTVER1 | m_COREI7,
 
   /* X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL */
   m_BDVER1 | m_COREI7,
@@ -1855,11 +1938,11 @@  static unsigned int initial_ix86_tune_features[X86_TUNE_LAST] = {
 
   /* X86_TUNE_SLOW_IMUL_IMM32_MEM: Imul of 32-bit constant and memory is
      vector path on AMD machines.  */
-  m_K8 | m_CORE2I7_64 | m_GENERIC64 | m_AMDFAM10 | m_BDVER1,
+  m_K8 | m_CORE2I7_64 | m_GENERIC64 | m_AMDFAM10 | m_BDVER1 | m_BTVER1,
 
   /* X86_TUNE_SLOW_IMUL_IMM8: Imul of 8-bit constant is vector path on AMD
      machines.  */
-  m_K8 | m_CORE2I7_64 | m_GENERIC64 | m_AMDFAM10 | m_BDVER1,
+  m_K8 | m_CORE2I7_64 | m_GENERIC64 | m_AMDFAM10 | m_BDVER1 | m_BTVER1,
 
   /* X86_TUNE_MOVE_M1_VIA_OR: On pentiums, it is faster to load -1 via OR
      than a MOV.  */
@@ -2451,6 +2534,7 @@  static const struct ptt processor_target_table[PROCESSOR_max] =
   {&generic64_cost, 16, 10, 16, 10, 16},
   {&amdfam10_cost, 32, 24, 32, 7, 32},
   {&bdver1_cost, 32, 24, 32, 7, 32},
+  {&btver1_cost, 32, 24, 32, 7, 32},
   {&atom_cost, 16, 7, 16, 7, 16}
 };
 
@@ -2479,7 +2563,8 @@  static const char *const cpu_names[TARGET_CPU_DEFAULT_max] =
   "athlon-4",
   "k8",
   "amdfam10",
-  "bdver1"
+  "bdver1",
+  "btver1"
 };
 
 /* Return true if a red-zone is in use.  */
@@ -3075,6 +3160,7 @@  software_prefetching_beneficial_p (void)
     case PROCESSOR_ATHLON:
     case PROCESSOR_K8:
     case PROCESSOR_AMDFAM10:
+    case PROCESSOR_BTVER1:
       return true;
 
     default:
@@ -3260,6 +3346,9 @@  ix86_option_override_internal (bool main_args_p)
 	| PTA_SSE2 | PTA_SSE3 | PTA_SSE4A | PTA_CX16 | PTA_ABM
 	| PTA_SSSE3 | PTA_SSE4_1 | PTA_SSE4_2 | PTA_AES
 	| PTA_PCLMUL | PTA_AVX | PTA_FMA4 | PTA_XOP | PTA_LWP},
+      {"btver1", PROCESSOR_BTVER1, CPU_GENERIC64,
+        PTA_64BIT | PTA_MMX |  PTA_SSE  | PTA_SSE2 | PTA_SSE3
+        | PTA_SSSE3 | PTA_SSE4A |PTA_ABM | PTA_CX16},
       {"generic32", PROCESSOR_GENERIC32, CPU_PENTIUMPRO,
 	0 /* flags are only used for -march switch.  */ },
       {"generic64", PROCESSOR_GENERIC64, CPU_GENERIC64,
@@ -22166,6 +22255,7 @@  ix86_issue_rate (void)
     case PROCESSOR_GENERIC32:
     case PROCESSOR_GENERIC64:
     case PROCESSOR_BDVER1:
+    case PROCESSOR_BTVER1:
       return 3;
 
     default:
@@ -22353,6 +22443,7 @@  ix86_adjust_cost (rtx insn, rtx link, rtx dep_insn, int cost)
     case PROCESSOR_K8:
     case PROCESSOR_AMDFAM10:
     case PROCESSOR_BDVER1:
+    case PROCESSOR_BTVER1:
     case PROCESSOR_ATOM:
     case PROCESSOR_GENERIC32:
     case PROCESSOR_GENERIC64:
diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h
index 510506a..0a948cc 100644
--- a/gcc/config/i386/i386.h
+++ b/gcc/config/i386/i386.h
@@ -251,6 +251,7 @@  extern const struct processor_costs ix86_size_cost;
 #define TARGET_GENERIC (TARGET_GENERIC32 || TARGET_GENERIC64)
 #define TARGET_AMDFAM10 (ix86_tune == PROCESSOR_AMDFAM10)
 #define TARGET_BDVER1 (ix86_tune == PROCESSOR_BDVER1)
+#define TARGET_BTVER1 (ix86_tune == PROCESSOR_BTVER1)
 #define TARGET_ATOM (ix86_tune == PROCESSOR_ATOM)
 
 /* Feature tests against the various tunings.  */
@@ -600,6 +601,7 @@  enum target_cpu_default
   TARGET_CPU_DEFAULT_k8,
   TARGET_CPU_DEFAULT_amdfam10,
   TARGET_CPU_DEFAULT_bdver1,
+  TARGET_CPU_DEFAULT_btver1,
 
   TARGET_CPU_DEFAULT_max
 };
@@ -2060,6 +2062,7 @@  enum processor_type
   PROCESSOR_GENERIC64,
   PROCESSOR_AMDFAM10,
   PROCESSOR_BDVER1,
+  PROCESSOR_BTVER1,
   PROCESSOR_ATOM,
   PROCESSOR_max
 };
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 05937e6..de93de6 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -359,7 +359,7 @@ 
 
 ;; Processor type.
 (define_attr "cpu" "none,pentium,pentiumpro,geode,k6,athlon,k8,core2,corei7,
-		    atom,generic64,amdfam10,bdver1"
+		    atom,generic64,amdfam10,bdver1,btver1"
   (const (symbol_ref "ix86_schedule")))
 
 ;; A basic instruction type.  Refinements due to arguments to be
-- 
1.6.3.3