Re: [SH] PR 50751 - add HImode displacement addressing support
Oleg Endo oleg.e...@t-online.de wrote: BTW, do you have the numbers of CSiBE with this? Only for -m4-single -ml -O2 -mpretend-cmove so far. Not so spectacular :T I'll also do a comparison of more variants to see if something went really bad. It's a bit difficult to isolate the degradations because there's quite some code reordering happening after the patch... Thanks for numbers! Looks good considering that HImode would be less frequently used than QImode in the usual working set. Regards, kaz
Re: [SH] PR 50751 - add HImode displacement addressing support
On Wed, 2012-04-11 at 22:10 +0900, Kaz Kojima wrote: Oleg Endo oleg.e...@t-online.de wrote: BTW, do you have the numbers of CSiBE with this? Only for -m4-single -ml -O2 -mpretend-cmove so far. Not so spectacular :T I'll also do a comparison of more variants to see if something went really bad. It's a bit difficult to isolate the degradations because there's quite some code reordering happening after the patch... Thanks for numbers! Looks good considering that HImode would be less frequently used than QImode in the usual working set. Sure, no problem. I think there is some room for improvement in the 'sh_find_mov_disp_adjust' function. If it was a bit smarter, all the displacement move insns would instantly benefit from it. Cheers, Oleg
Re: [SH] PR 50751 - add HImode displacement addressing support
Oleg Endo oleg.e...@t-online.de wrote:
The attached patch adds HImode addressing support.
Tested against rev. 186243 with
sudo make -k check RUNTESTFLAGS=--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a/-mb,-m2a-single/-mb,-m4/-ml,-m4/-mb,
-m4-single/-ml,-m4-single/-mb,-m4a-single/-ml,-m4a-single/-mb}
and no new failures.
Test cases will follow soon.
The patch is OK for trunk.
BTW, do you have the numbers of CSiBE with this?
Regards,
kaz
Re: [SH] PR 50751 - add HImode displacement addressing support
On Tue, 2012-04-10 at 22:42 +0900, Kaz Kojima wrote:
Oleg Endo oleg.e...@t-online.de wrote:
The attached patch adds HImode addressing support.
Tested against rev. 186243 with
sudo make -k check RUNTESTFLAGS=--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a/-mb,-m2a-single/-mb,-m4/-ml,-m4/-mb,
-m4-single/-ml,-m4-single/-mb,-m4a-single/-ml,-m4a-single/-mb}
and no new failures.
Test cases will follow soon.
The patch is OK for trunk.
BTW, do you have the numbers of CSiBE with this?
Only for -m4-single -ml -O2 -mpretend-cmove so far.
Not so spectacular :T
I'll also do a comparison of more variants to see if something went
really bad. It's a bit difficult to isolate the degradations because
there's quite some code reordering happening after the patch...
Cheers,
Oleg
OpenTCP-1.0.4
arp 2089 - 2061 -28 / -1.340354 %
bootp/bootp 740 - 704 -36 / -4.864865 %
demo/main_demo 372 - 368 -4 / -1.075269 %
demo/tcp_client_demo 396 - 396
demo/tcp_server_demo 468 - 468
demo/udp_demo268 - 268
dhcp/dhcpc 1588 - 1588
dns/dns 1340 - 1328 -12 / -0.895522 %
ethernet1240 - 1228 -12 / -0.967742 %
http/http_server1784 - 1764 -20 / -1.121076 %
http/https_callbacks 656 - 656
icmp 348 - 352 +4 / +1.149425 %
ip 1780 - 1668-112 / -6.292135 %
pop3/pop3_client3796 - 3728 -68 / -1.791359 %
pop3/pop3c_callbacks 28 - 28
smtp/smtp_client2284 - 2212 -72 / -3.152364 %
smtp/smtpc_callbacks 32 - 32
system 980 - 980
tcp 4966 - 4970 +4 / +0.080548 %
tftp/tftps 1024 - 920 -104 / -10.156250 %
timers 228 - 228
udp 1362 - 1322 -40 / -2.936858 %
total: 27769 - 27269 -500 / -1.800569 %
bzip2-1.0.2
blocksort 7262 - 7258 -4 / -0.055081 %
bzip2 17288 - 17248-40 / -0.231374 %
bzip2recover3868 - 3868
bzlib 10668 - 10636-32 / -0.299963 %
compress 14208 - 14192-16 / -0.112613 %
crctable1024 - 1024
decompress 8380 - 8380
huffman 1436 - 1436
randtable 2048 - 2048
total: 66182 - 66090 -92 / -0.139011 %
cg_compiler_opensrc
atom3728 - 3728
binding 1752 - 1728 -24 / -1.369863 %
cgcmain 2388 - 2388
cgstruct 128 - 128
check 3300 - 3300
compile14052 - 14052
constfold 3996 - 3996
cpp 6776 - 6768 -8 / -0.118064 %
generic_hal 2516 - 2516
hal 2720 - 2720
inline 2724 - 2724
memory 548 - 548
parser 15796 - 15796
printutils 11888 - 11888
scanner 6716 - 6672 -44 / -0.655152 %
semantic4364 - 4364
stdlib 3224 - 3224
support24328 - 24332 +4 / +0.016442 %
support_iter1832 - 1832
symbols 6324 - 6312 -12 / -0.189753 %
tokenize 624 - 624
tokens 3112 - 3120 +8 / +0.257069 %
total:122836 - 122760 -76 / -0.061871 %
compiler
cg 2868 - 2868
main 2600 - 2600
parser 4524 - 4528 +4 / +0.088417 %
scanner3788 - 3788
vam5264 - 5264
vas6292 - 6300 +8 / +0.127146 %
total: 25336 - 25348 +12 / +0.047363 %
flex-2.5.31
buf 1100 - 1100
ccl 1220 - 1220
dfa 6952 - 6952
ecs 612 - 612
filter 2360 - 2360
gen 24512 - 24488-24 / -0.097911 %
libmain28 - 28
libyywrap 4 - 4
main19804 - 19792-12 / -0.060594 %
misc 6480 - 6480
nfa 3812 - 3812
options 2768 - 2768
parse9524 - 9528 +4 / +0.041999 %
regex 596 - 596
scan47012 - 47004 -8 / -0.017017 %
scanopt 3672 - 3672
skel96652 - 96652
sym 1012 - 1012
tables 2080 - 2064 -16 / -0.769231 %
tables_shared 28 - 28
tblcmp 3764 - 3764
yylex1292 - 1292
total:235284 - 235228 -56 / -0.023801 %
jikespg-1.3
src/ctabs 49912 - 49876-36 / -0.072127 %
src/globals 288 - 288
src/lpgparse 47340 - 47036 -304 / -0.642163 %
src/lpgutil 5152 - 5120 -32 / -0.621118 %
src/main5440 - 5440
src/mkfirst
Re: [SH] PR 50751 - add HImode displacement addressing support
- Original Message - BTW, do you have the numbers of CSiBE with this? Only for -m4-single -ml -O2 -mpretend-cmove so far. Not so spectacular :T I'll also do a comparison of more variants to see if something went really bad. It's a bit difficult to isolate the degradations because there's quite some code reordering happening after the patch... Are you looking at execution time or code size? If the latter, you could try disabling scheduling for comparison purposes (-fno-schedule-insns -fno-schedule-insns2). Even if you're looking at execution time, disabling scheduling might make it easier to see where things are going wrong in the patched code. -Nathan
Re: [SH] PR 50751 - add HImode displacement addressing support
On Tue, 2012-04-10 at 07:14 -0700, Nathan Froyd wrote: - Original Message - BTW, do you have the numbers of CSiBE with this? Only for -m4-single -ml -O2 -mpretend-cmove so far. Not so spectacular :T I'll also do a comparison of more variants to see if something went really bad. It's a bit difficult to isolate the degradations because there's quite some code reordering happening after the patch... Are you looking at execution time or code size? If the latter, you could try disabling scheduling for comparison purposes (-fno-schedule-insns -fno-schedule-insns2). Even if you're looking at execution time, disabling scheduling might make it easier to see where things are going wrong in the patched code. Code size only. Thanks for the hint, will try it out. Cheers, Oleg
[SH] PR 50751 - add HImode displacement addressing support
Hello,
The attached patch adds HImode addressing support.
Tested against rev. 186243 with
sudo make -k check RUNTESTFLAGS=--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a/-mb,-m2a-single/-mb,-m4/-ml,-m4/-mb,
-m4-single/-ml,-m4-single/-mb,-m4a-single/-ml,-m4a-single/-mb}
and no new failures.
Test cases will follow soon.
Cheers,
Oleg
ChangeLog:
PR target/50751
* config/sh/sh-protos.h (sh_legitimate_index_p): Add
new arguments consider_sh2a and allow_zero.
* config/sh/sh.c (sh_legitimate_index_p): Likewise.
(disp_addr_displacement): New function.
(sh_address_cost): Use disp_addr_displacement function instead
of DISP_ADDR_OFFSET.
(sh_legitimate_address_p): Adapt to changed
sh_legitimate_index_p declaration.
(sh_find_mov_disp_adjust): Remove HImode check.
(sh_secondary_reload): Add HImode case. Use
satisfies_constraint_Sdd, disp_addr_displacement and
max_mov_insn_displacement.
(max_mov_insn_displacement): Remove HImode check.
* config/sh/sh.h (CONST_OK_FOR_K04, CONST_OK_FOR_K12,
DISP_ADDR_P, DISP_ADDR_OFFSET): Remove.
* config/sh/constraints.md (K05, K13): New constraints.
(K12): Correct comment.
(Sdd): Do not use DISP_ADDR_P macro.
(Snd): Use satisfies_constraint_Sdd.
(Sbw): Likewise.
* config/sh/sh.md (extendhisi2): Remove constraints from
expander.
(*extendhisi2_compact, movhi_i): Remove.
(*extendhisi2_compact_reg, *extendhisi2_compact_mem_disp,
*extendhisi2_compact_mem_disp, *extendhisi2_compact_snd,
*movhi_reg_reg, *movhi_store_mem_disp05,
*movhi_store_mem_disp13, *movhi_load_mem_disp,
*movhi_load_mem_disp, *movhi): New insns.
(*extendqisi2_compact_mem_disp, *extendqisi2_compact_mem_disp,
*movqi_store_mem_disp04, *movqi_store_mem_disp12,
*movqi_load_mem_disp, *movqi_load_mem_disp): Use
sh_legitimate_index_p instead of CONST_OK_FOR_Kxx.
Add new peepholes for HImode displacement addressing.
Index: gcc/config/sh/predicates.md
===
--- gcc/config/sh/predicates.md (revision 186233)
+++ gcc/config/sh/predicates.md (working copy)
@@ -404,7 +404,7 @@
if (GET_CODE (x) == PLUS
REG_P (XEXP (x, 0))
CONST_INT_P (XEXP (x, 1)))
- return sh_legitimate_index_p (mode, XEXP (x, 1));
+ return sh_legitimate_index_p (mode, XEXP (x, 1), TARGET_SH2A, false);
}
if (TARGET_SHMEDIA
@@ -466,7 +466,7 @@
if (GET_CODE (x) == PLUS
REG_P (XEXP (x, 0))
CONST_INT_P (XEXP (x, 1)))
- return sh_legitimate_index_p (mode, XEXP (x, 1));
+ return sh_legitimate_index_p (mode, XEXP (x, 1), TARGET_SH2A, false);
}
return general_operand (op, mode);
Index: gcc/config/sh/sh-protos.h
===
--- gcc/config/sh/sh-protos.h (revision 186233)
+++ gcc/config/sh/sh-protos.h (working copy)
@@ -56,7 +56,7 @@
extern bool fp_zero_operand (rtx);
extern bool fp_one_operand (rtx);
extern rtx get_fpscr_rtx (void);
-extern bool sh_legitimate_index_p (enum machine_mode, rtx);
+extern bool sh_legitimate_index_p (enum machine_mode, rtx, bool, bool);
extern bool sh_legitimize_reload_address (rtx *, enum machine_mode, int, int);
extern rtx legitimize_pic_address (rtx, enum machine_mode, rtx);
extern bool nonpic_symbol_mentioned_p (rtx);
Index: gcc/config/sh/sh.c
===
--- gcc/config/sh/sh.c (revision 186233)
+++ gcc/config/sh/sh.c (working copy)
@@ -304,6 +304,7 @@
static int mov_insn_size (enum machine_mode, bool);
static int max_mov_insn_displacement (enum machine_mode, bool);
static int mov_insn_alignment_mask (enum machine_mode, bool);
+static HOST_WIDE_INT disp_addr_displacement (rtx);
static void sh_init_sync_libfuncs (void) ATTRIBUTE_UNUSED;
@@ -3160,11 +3161,6 @@
scale the max. displacement value accordingly. */
const int disp_scale = consider_sh2a ? (4095 / 15) : 1;
- /* FIXME: HImode with displacement addressing is not supported yet.
- Make it purposefully fail for now. */
- if (mode == HImode)
-return 0;
-
/* SH2A supports FPU move insns with 12 bit displacements.
Other variants to do not support any kind of displacements for
FPU move insns. */
@@ -3194,15 +3190,24 @@
return mov_insn_sz 0 ? (mov_insn_sz - 1) : 0;
}
+/* Return the displacement value of a displacement address. */
+
+static inline HOST_WIDE_INT
+disp_addr_displacement (rtx x)
+{
+ gcc_assert (satisfies_constraint_Sdd (x));
+ return INTVAL (XEXP (XEXP (x, 0), 1));
+}
+
/* Compute the cost of an address. */
static int
sh_address_cost (rtx x, bool speed ATTRIBUTE_UNUSED)
{
/* 'reg + disp' addressing. */
- if (DISP_ADDR_P (x))
+ if (satisfies_constraint_Sdd (x))
{
- const HOST_WIDE_INT offset =
