re PR target/27060 (divide libcall size has increased)

2006-04-12  J"orn Rennecke <joern.rennecke@st.com>

	PR target/27060
        * config/sh/lib1funcs.h: New file, broken out of:
        * config/sh/lib1funcs.asm.
        * config/sh/lib1funcs.h (DR00, DR01, DR20, DR21, DR40, DR41):
	New macros.
	* config/sh/lib1funcs.asm (udivsi3_i4): Use them.
        * config/sh/lib1funcs-Os-4-200.asm: New file.
	* config/sh/embed-elf.h (LIBGCC_SPEC): Use -lgcc-Os-4-200.
	* config/sh/t-sh (OPT_EXTRA_PARTS): New variable.
	(EXTRA_MULTILIB_PARTS): Include it.
	($(T)sdivsi3_i4i-Os-4-200.o): New rule.
	($(T)udivsi3_i4i-Os-4-200.o, $(T)libgcc-Os-4-200.a): Likewise.
	* config/sh/t-superh (EXTRA_MULTILIB_PARTS): Include OPT_EXTRA_PARTS.
	* config/sh/sh.h (OVERRIDE_OPTIONS): Recognize -mdiv=call-table for
	TARGET_SH2.

From-SVN: r112898

gcc/ChangeLog

+2006-04-12  J"orn Rennecke <joern.rennecke@st.com>
+
+	PR target/27060
+        * config/sh/lib1funcs.h: New file, broken out of:
+        * config/sh/lib1funcs.asm.
+        * config/sh/lib1funcs.h (DR00, DR01, DR20, DR21, DR40, DR41):
+	New macros.
+	* config/sh/lib1funcs.asm (udivsi3_i4): Use them.
+        * config/sh/lib1funcs-Os-4-200.asm: New file.
+	* config/sh/embed-elf.h (LIBGCC_SPEC): Use -lgcc-Os-4-200.
+	* config/sh/t-sh (OPT_EXTRA_PARTS): New variable.
+	(EXTRA_MULTILIB_PARTS): Include it.
+	($(T)sdivsi3_i4i-Os-4-200.o): New rule.
+	($(T)udivsi3_i4i-Os-4-200.o, $(T)libgcc-Os-4-200.a): Likewise.
+	* config/sh/t-superh (EXTRA_MULTILIB_PARTS): Include OPT_EXTRA_PARTS.
+	* config/sh/sh.h (OVERRIDE_OPTIONS): Recognize -mdiv=call-table for
+	TARGET_SH2.
+
 2006-04-12  Kazu Hirata  <kazu@codesourcery.com>
 
 	* Makefile.in (TREE_SSA_LIVE_H): Add vecprim.h.

gcc/config/sh/embed-elf.h

@@ -25,7 +25,14 @@ Boston, MA 02110-1301, USA.  */
 
 #undef TARGET_POSIX_IO
 
+/* While the speed-optimized implementations of udivsi3_i4i / sdivsi3_i4i
+   in libgcc are not available for SH2, the space-optimized ones in
+   libgcc-Os-4-200 are.  Thus, when not optimizing for space, link
+   libgcc-Os-4-200 after libgcc, so that -mdiv=call-table works for -m2.  */
 #define LIBGCC_SPEC "%{!shared: \
   %{m4-100*:-lic_invalidate_array_4-100} \
   %{m4-200*:-lic_invalidate_array_4-200} \
-  %{m4a*:-lic_invalidate_array_4a}} -lgcc"
+  %{m4a*:-lic_invalidate_array_4a}} \
+  %{Os: -lgcc-Os-4-200} \
+  -lgcc \
+  %{!Os: -lgcc-Os-4-200}"

gcc/config/sh/lib1funcs-Os-4-200.asm

+/* Copyright (C) 2006 Free Software Foundation, Inc.
+
+This file is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 2, or (at your option) any
+later version.
+
+In addition to the permissions in the GNU General Public License, the
+Free Software Foundation gives you unlimited permission to link the
+compiled version of this file into combinations with other programs,
+and to distribute those combinations without any restriction coming
+from the use of this file.  (The General Public License restrictions
+do apply in other respects; for example, they cover modification of
+the file, and distribution when not linked into a combine
+executable.)
+
+This file is distributed in the hope that it will be useful, but
+WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program; see the file COPYING.  If not, write to
+the Free Software Foundation, 51 Franklin Street, Fifth Floor,
+Boston, MA 02110-1301, USA.  */
+
+/* Moderately Space-optimized libgcc routines for the Renesas SH /
+   STMicroelectronics ST40 CPUs.
+   Contributed by J"orn Rennecke joern.rennecke@st.com.  */
+
+#include "lib1funcs.h"
+
+#ifdef L_udivsi3_i4i
+
+/* 88 bytes; sh4-200 cycle counts:
+   divisor  >= 2G: 11 cycles
+   dividend <  2G: 48 cycles
+   dividend >= 2G: divisor != 1: 54 cycles
+   dividend >= 2G, divisor == 1: 22 cycles */
+#if defined (__SH_FPU_DOUBLE__) || defined (__SH4_SINGLE_ONLY__)
+!! args in r4 and r5, result in r0, clobber r1
+
+	.global GLOBAL(udivsi3_i4i)
+	FUNC(GLOBAL(udivsi3_i4i))
+GLOBAL(udivsi3_i4i):
+	mova L1,r0
+	cmp/pz r5
+	sts fpscr,r1
+	lds.l @r0+,fpscr
+	sts.l fpul,@-r15
+	bf LOCAL(huge_divisor)
+	mov.l r1,@-r15
+	lds r4,fpul
+	cmp/pz r4
+#ifdef FMOVD_WORKS
+	fmov.d dr0,@-r15
+	float fpul,dr0
+	fmov.d dr2,@-r15
+	bt LOCAL(dividend_adjusted)
+	mov #1,r1
+	fmov.d @r0,dr2
+	cmp/eq r1,r5
+	bt LOCAL(div_by_1)
+	fadd dr2,dr0
+LOCAL(dividend_adjusted):
+	lds r5,fpul
+	float fpul,dr2
+	fdiv dr2,dr0
+LOCAL(div_by_1):
+	fmov.d @r15+,dr2
+	ftrc dr0,fpul
+	fmov.d @r15+,dr0
+#else /* !FMOVD_WORKS */
+	fmov.s DR01,@-r15
+	mov #1,r1
+	fmov.s DR00,@-r15
+	float fpul,dr0
+	fmov.s DR21,@-r15
+	bt/s LOCAL(dividend_adjusted)
+	fmov.s DR20,@-r15
+	cmp/eq r1,r5
+	bt LOCAL(div_by_1)
+	fmov.s @r0+,DR20
+	fmov.s @r0,DR21
+	fadd dr2,dr0
+LOCAL(dividend_adjusted):
+	lds r5,fpul
+	float fpul,dr2
+	fdiv dr2,dr0
+LOCAL(div_by_1):
+	fmov.s @r15+,DR20
+	fmov.s @r15+,DR21
+	ftrc dr0,fpul
+	fmov.s @r15+,DR00
+	fmov.s @r15+,DR01
+#endif /* !FMOVD_WORKS */
+	lds.l @r15+,fpscr
+	sts fpul,r0
+	rts
+	lds.l @r15+,fpul
+
+#ifdef FMOVD_WORKS
+	.p2align 3        ! make double below 8 byte aligned.
+#endif
+LOCAL(huge_divisor):
+	lds r1,fpscr
+	add #4,r15
+	cmp/hs r5,r4
+	rts
+	movt r0
+
+	.p2align 2
+L1:
+#ifndef FMOVD_WORKS
+	.long 0x80000
+#else
+	.long 0x180000
+#endif
+	.double 4294967296
+
+	ENDFUNC(GLOBAL(udivsi3_i4i))
+#elif !defined (__sh1__)  /* !__SH_FPU_DOUBLE__ */
+
+#if 0
+/* With 36 bytes, the following would probably be the most compact
+   implementation, but with 139 cycles on an sh4-200, it is extremely slow.  */
+GLOBAL(udivsi3_i4i):
+	mov.l r2,@-r15
+	mov #0,r1
+	div0u
+	mov r1,r2
+	mov.l r3,@-r15
+	mov r1,r3
+	sett
+	mov r4,r0
+LOCAL(loop):
+	rotcr r2
+	;
+	bt/s LOCAL(end)
+	cmp/gt r2,r3
+	rotcl r0
+	bra LOCAL(loop)
+	div1 r5,r1
+LOCAL(end):
+	rotcl r0
+	mov.l @r15+,r3
+	rts
+	mov.l @r15+,r2
+#endif /* 0 */
+
+/* Size: 186 bytes jointly for udivsi3_i4i and sdivsi3_i4i
+   sh4-200 run times:
+   udiv small divisor: 55 cycles
+   udiv large divisor: 52 cycles
+   sdiv small divisor, positive result: 59 cycles
+   sdiv large divisor, positive result: 56 cycles
+   sdiv small divisor, negative result: 65 cycles (*)
+   sdiv large divisor, negative result: 62 cycles (*)
+   (*): r2 is restored in the rts delay slot and has a lingering latency
+        of two more cycles.  */
+	.balign 4
+	.global	GLOBAL(udivsi3_i4i)
+	FUNC(GLOBAL(udivsi3_i4i))
+	FUNC(GLOBAL(sdivsi3_i4i))
+GLOBAL(udivsi3_i4i):
+	sts pr,r1
+	mov.l r4,@-r15
+	extu.w r5,r0
+	cmp/eq r5,r0
+	swap.w r4,r0
+	shlr16 r4
+	bf/s LOCAL(large_divisor)
+	div0u
+	mov.l r5,@-r15
+	shll16 r5
+LOCAL(sdiv_small_divisor):
+	div1 r5,r4
+	bsr LOCAL(div6)
+	div1 r5,r4
+	div1 r5,r4
+	bsr LOCAL(div6)
+	div1 r5,r4
+	xtrct r4,r0
+	xtrct r0,r4
+	bsr LOCAL(div7)
+	swap.w r4,r4
+	div1 r5,r4
+	bsr LOCAL(div7)
+	div1 r5,r4
+	xtrct r4,r0
+	mov.l @r15+,r5
+	swap.w r0,r0
+	mov.l @r15+,r4
+	jmp @r1
+	rotcl r0
+LOCAL(div7):
+	div1 r5,r4
+LOCAL(div6):
+	            div1 r5,r4; div1 r5,r4; div1 r5,r4
+	div1 r5,r4; div1 r5,r4; rts;        div1 r5,r4
+
+LOCAL(divx3):
+	rotcl r0
+	div1 r5,r4
+	rotcl r0
+	div1 r5,r4
+	rotcl r0
+	rts
+	div1 r5,r4
+
+LOCAL(large_divisor):
+	mov.l r5,@-r15
+LOCAL(sdiv_large_divisor):
+	xor r4,r0
+	.rept 4
+	rotcl r0
+	bsr LOCAL(divx3)
+	div1 r5,r4
+	.endr
+	mov.l @r15+,r5
+	mov.l @r15+,r4
+	jmp @r1
+	rotcl r0
+	ENDFUNC(GLOBAL(udivsi3_i4i))
+
+	.global	GLOBAL(sdivsi3_i4i)
+GLOBAL(sdivsi3_i4i):
+	mov.l r4,@-r15
+	cmp/pz r5
+	mov.l r5,@-r15
+	bt/s LOCAL(pos_divisor)
+	cmp/pz r4
+	neg r5,r5
+	extu.w r5,r0
+	bt/s LOCAL(neg_result)
+	cmp/eq r5,r0
+	neg r4,r4
+LOCAL(pos_result):
+	swap.w r4,r0
+	bra LOCAL(sdiv_check_divisor)
+	sts pr,r1
+LOCAL(pos_divisor):
+	extu.w r5,r0
+	bt/s LOCAL(pos_result)
+	cmp/eq r5,r0
+	neg r4,r4
+LOCAL(neg_result):
+	mova LOCAL(negate_result),r0
+	;
+	mov r0,r1
+	swap.w r4,r0
+	lds r2,macl
+	sts pr,r2
+LOCAL(sdiv_check_divisor):
+	shlr16 r4
+	bf/s LOCAL(sdiv_large_divisor)
+	div0u
+	bra LOCAL(sdiv_small_divisor)
+	shll16 r5
+	.balign 4
+LOCAL(negate_result):
+	neg r0,r0
+	jmp @r2
+	sts macl,r2
+	ENDFUNC(GLOBAL(sdivsi3_i4i))
+#endif /* !__SH_FPU_DOUBLE__ */
+#endif /* L_udivsi3_i4i */
+
+#ifdef L_sdivsi3_i4i
+#if defined (__SH_FPU_DOUBLE__) || defined (__SH4_SINGLE_ONLY__)
+/* 48 bytes, 45 cycles on sh4-200  */
+!! args in r4 and r5, result in r0, clobber r1
+
+	.global GLOBAL(sdivsi3_i4i)
+	FUNC(GLOBAL(sdivsi3_i4i))
+GLOBAL(sdivsi3_i4i):
+	sts.l fpscr,@-r15
+	sts fpul,r1
+	mova L1,r0
+	lds.l @r0+,fpscr
+	lds r4,fpul
+#ifdef FMOVD_WORKS
+	fmov.d dr0,@-r15
+	float fpul,dr0
+	lds r5,fpul
+	fmov.d dr2,@-r15
+#else
+	fmov.s DR01,@-r15
+	fmov.s DR00,@-r15
+	float fpul,dr0
+	lds r5,fpul
+	fmov.s DR21,@-r15
+	fmov.s DR20,@-r15
+#endif
+	float fpul,dr2
+	fdiv dr2,dr0
+#ifdef FMOVD_WORKS
+	fmov.d @r15+,dr2
+#else
+	fmov.s @r15+,DR20
+	fmov.s @r15+,DR21
+#endif
+	ftrc dr0,fpul
+#ifdef FMOVD_WORKS
+	fmov.d @r15+,dr0
+#else
+	fmov.s @r15+,DR00
+	fmov.s @r15+,DR01
+#endif
+	lds.l @r15+,fpscr
+	sts fpul,r0
+	rts
+	lds r1,fpul
+
+	.p2align 2
+L1:
+#ifndef FMOVD_WORKS
+	.long 0x80000
+#else
+	.long 0x180000
+#endif
+
+	ENDFUNC(GLOBAL(sdivsi3_i4i))
+#endif /* __SH_FPU_DOUBLE__ */
+#endif /* L_sdivsi3_i4i */

gcc/config/sh/lib1funcs.asm

@@ -38,31 +38,7 @@ Boston, MA 02110-1301, USA.  */
    ELF local label prefixes by J"orn Rennecke
    amylaar@cygnus.com  */
 
-#ifdef __ELF__
-#define LOCAL(X)	.L_##X
-#define FUNC(X)		.type X,@function
-#define HIDDEN_FUNC(X)	FUNC(X); .hidden X
-#define HIDDEN_ALIAS(X,Y) ALIAS (X,Y); .hidden GLOBAL(X)
-#define ENDFUNC0(X)	.Lfe_##X: .size X,.Lfe_##X-X
-#define ENDFUNC(X)	ENDFUNC0(X)
-#else
-#define LOCAL(X)	L_##X
-#define FUNC(X)
-#define HIDDEN_FUNC(X)
-#define HIDDEN_ALIAS(X,Y) ALIAS (X,Y)
-#define ENDFUNC(X)
-#endif
-
-#define	CONCAT(A,B)	A##B
-#define	GLOBAL0(U,X)	CONCAT(U,__##X)
-#define	GLOBAL(X)	GLOBAL0(__USER_LABEL_PREFIX__,X)
-
-#define ALIAS(X,Y)	.global GLOBAL(X); .set GLOBAL(X),GLOBAL(Y)
-
-#ifdef __SH2A__
-#undef FMOVD_WORKS
-#define FMOVD_WORKS
-#endif
+#include "lib1funcs.h"
 
 #if ! __SH5__
 #ifdef L_ashiftrt
@@ -1375,13 +1351,8 @@ GLOBAL(udivsi3_i4):
 #ifdef FMOVD_WORKS
 	fmov.d @r0+,dr4
 #else
-#ifdef __LITTLE_ENDIAN__
-	fmov.s @r0+,fr5
-	fmov.s @r0,fr4
-#else
-	fmov.s @r0+,fr4
-	fmov.s @r0,fr5
-#endif
+	fmov.s @r0+,DR40
+	fmov.s @r0,DR41
 #endif
 	float fpul,dr0
 	xor r1,r5
@@ -1444,13 +1415,8 @@ GLOBAL(udivsi3_i4):
 #ifdef FMOVD_WORKS
 	fmov.d @r0+,dr4
 #else
-#ifdef __LITTLE_ENDIAN__
-	fmov.s @r0+,fr5
-	fmov.s @r0,fr4
-#else
-	fmov.s @r0+,fr4
-	fmov.s @r0,fr5
-#endif
+	fmov.s @r0+,DR40
+	fmov.s @r0,DR41
 #endif
 	float fpul,dr0
 	xor r1,r5

gcc/config/sh/lib1funcs.h

+/* Copyright (C) 1994, 1995, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
+   2004, 2005, 2006
+   Free Software Foundation, Inc.
+
+This file is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 2, or (at your option) any
+later version.
+
+In addition to the permissions in the GNU General Public License, the
+Free Software Foundation gives you unlimited permission to link the
+compiled version of this file into combinations with other programs,
+and to distribute those combinations without any restriction coming
+from the use of this file.  (The General Public License restrictions
+do apply in other respects; for example, they cover modification of
+the file, and distribution when not linked into a combine
+executable.)
+
+This file is distributed in the hope that it will be useful, but
+WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with this program; see the file COPYING.  If not, write to
+the Free Software Foundation, 51 Franklin Street, Fifth Floor,
+Boston, MA 02110-1301, USA.  */
+
+#ifdef __ELF__
+#define LOCAL(X)	.L_##X
+#define FUNC(X)		.type X,@function
+#define HIDDEN_FUNC(X)	FUNC(X); .hidden X
+#define HIDDEN_ALIAS(X,Y) ALIAS (X,Y); .hidden GLOBAL(X)
+#define ENDFUNC0(X)	.Lfe_##X: .size X,.Lfe_##X-X
+#define ENDFUNC(X)	ENDFUNC0(X)
+#else
+#define LOCAL(X)	L_##X
+#define FUNC(X)
+#define HIDDEN_FUNC(X)
+#define HIDDEN_ALIAS(X,Y) ALIAS (X,Y)
+#define ENDFUNC(X)
+#endif
+
+#define	CONCAT(A,B)	A##B
+#define	GLOBAL0(U,X)	CONCAT(U,__##X)
+#define	GLOBAL(X)	GLOBAL0(__USER_LABEL_PREFIX__,X)
+
+#define ALIAS(X,Y)	.global GLOBAL(X); .set GLOBAL(X),GLOBAL(Y)
+
+#ifdef __SH2A__
+#undef FMOVD_WORKS
+#define FMOVD_WORKS
+#endif
+
+#ifdef __LITTLE_ENDIAN__
+#define DR00 fr1
+#define DR01 fr0
+#define DR20 fr3
+#define DR21 fr2
+#define DR40 fr5
+#define DR41 fr4
+#else /* !__LITTLE_ENDIAN__ */
+#define DR00 fr0
+#define DR01 fr1
+#define DR20 fr2
+#define DR21 fr3
+#define DR40 fr4
+#define DR41 fr5
+#endif /* !__LITTLE_ENDIAN__ */

gcc/config/sh/sh.h

@@ -631,7 +631,7 @@ do {									\
 		   || (TARGET_HARD_SH4 && TARGET_SH2E)			\
 		   || (TARGET_SHCOMPACT && TARGET_FPU_ANY)))		\
 	sh_div_strategy = SH_DIV_CALL_FP;				\
-      else if (! strcmp (sh_div_str, "call-table") && TARGET_SH3)	\
+      else if (! strcmp (sh_div_str, "call-table") && TARGET_SH2)	\
 	sh_div_strategy = SH_DIV_CALL_TABLE;				\
       else								\
 	/* Pick one that makes most sense for the target in general.	\
@@ -651,6 +651,8 @@ do {									\
 	  sh_div_strategy = SH_DIV_CALL_FP;				\
         /* SH1 .. SH3 cores often go into small-footprint systems, so	\
 	   default to the smallest implementation available.  */	\
+	else if (TARGET_SH2)	/* ??? EXPERIMENTAL */			\
+	  sh_div_strategy = SH_DIV_CALL_TABLE;				\
 	else								\
 	  sh_div_strategy = SH_DIV_CALL_DIV1;				\
     }									\

gcc/config/sh/t-sh

@@ -76,7 +76,8 @@ gt-sh.h : s-gtype ; @true
 
 IC_EXTRA_PARTS= libic_invalidate_array_4-100.a libic_invalidate_array_4-200.a \
 libic_invalidate_array_4a.a
-EXTRA_MULTILIB_PARTS= $(IC_EXTRA_PARTS)
+OPT_EXTRA_PARTS= libgcc-Os-4-200.a
+EXTRA_MULTILIB_PARTS= $(IC_EXTRA_PARTS) $(OPT_EXTRA_PARTS)
 
 $(T)ic_invalidate_array_4-100.o: $(srcdir)/config/sh/lib1funcs.asm $(GCC_PASSES)
 	$(GCC_FOR_TARGET) $(MULTILIB_CFLAGS) -c -o $(T)ic_invalidate_array_4-100.o -DL_ic_invalidate_array -DWAYS=1 -DWAY_SIZE=0x2000 -x assembler-with-cpp $(srcdir)/config/sh/lib1funcs.asm
@@ -93,6 +94,13 @@ $(T)ic_invalidate_array_4a.o: $(srcdir)/config/sh/lib1funcs.asm $(GCC_PASSES)
 $(T)libic_invalidate_array_4a.a: $(T)ic_invalidate_array_4a.o $(GCC_PASSES)
 	$(AR_CREATE_FOR_TARGET) $(T)libic_invalidate_array_4a.a $(T)ic_invalidate_array_4a.o
 
+$(T)sdivsi3_i4i-Os-4-200.o: $(srcdir)/config/sh/lib1funcs-Os-4-200.asm $(GCC_PASSES)
+	$(GCC_FOR_TARGET) $(MULTILIB_CFLAGS) -c -o $@ -DL_sdivsi3_i4i -x assembler-with-cpp $<
+$(T)udivsi3_i4i-Os-4-200.o: $(srcdir)/config/sh/lib1funcs-Os-4-200.asm $(GCC_PASSES)
+	$(GCC_FOR_TARGET) $(MULTILIB_CFLAGS) -c -o $@ -DL_udivsi3_i4i -x assembler-with-cpp $<
+$(T)libgcc-Os-4-200.a: $(T)sdivsi3_i4i-Os-4-200.o $(T)udivsi3_i4i-Os-4-200.o $(GCC_PASSES)
+	$(AR_CREATE_FOR_TARGET) $@ $(T)sdivsi3_i4i-Os-4-200.o $(T)udivsi3_i4i-Os-4-200.o
+
 # Local Variables:
 # mode: Makefile
 # End:

gcc/config/sh/t-superh

 EXTRA_MULTILIB_PARTS= crt1.o crti.o crtn.o \
 	crtbegin.o crtend.o crtbeginS.o crtendS.o \
-	crt1-mmu.o gcrt1-mmu.o gcrt1.o $(IC_EXTRA_PARTS)
+	crt1-mmu.o gcrt1-mmu.o gcrt1.o $(IC_EXTRA_PARTS) $(OPT_EXTRA_PARTS)
 
 # Compile crt1-mmu.o as crt1.o with -DMMU_SUPPORT
 $(T)crt1-mmu.o: $(srcdir)/config/sh/crt1.asm $(GCC_PASSES)