From: Peter Bergner <bergner@linux.ibm.com>
To: Segher Boessenkool <segher@kernel.crashing.org>
Cc: GCC Patches <gcc-patches@gcc.gnu.org>,
Bill Schmidt <wschmidt@linux.ibm.com>
Subject: [PATCH] rs6000: inefficient 64-bit constant generation for consecutive 1-bits
Date: Thu, 10 Sep 2020 16:58:03 -0500 [thread overview]
Message-ID: <838b2e97-dfa9-3ca0-c3c6-1767d60ddf05@linux.ibm.com> (raw)
Generating arbitrary 64-bit constants on POWER can take up to 5 instructions.
However, some special constants can be generated in fewer instructions.
One special class of constants we don't handle, is constants that have one
set of consecutive 1-bits. These can be generated with a "li rT,-1"
followed by a "rldic rX,rT,SH,MB" instruction. The following patch
implements this idea.
This has passed bootstrap and regtesting on powerpc64le-linux with no
regressions. Testing on powerpc64-linux is still running.
Ok for trunk if the BE testing comes back clean too?
Peter
gcc/
PR target/93176
* config/rs6000/rs6000.c (has_consecutive_ones): New function.
(num_insns_constant_gpr): Use it.
(rs6000_emit_set_long_const): Likewise.
* config/rs6000/rs6000.md UNSPEC_RLDIC: New unspec.
(rldic): New.
gcc/testsuite/
PR target/93176
* gcc.target/powerpc/pr93176.c: New test.
diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
index ca5b71ecdd3..273cab14138 100644
--- a/gcc/config/rs6000/rs6000.c
+++ b/gcc/config/rs6000/rs6000.c
@@ -5728,6 +5728,47 @@ direct_return (void)
return 0;
}
+/* Helper for num_insns_constant_gpr and rs6000_emit_set_long_const.
+ Return TRUE if VALUE contains one set of consecutive 1-bits. Also set
+ *SH and *MB to values needed to generate VALUE with the rldic instruction.
+ We accept consecutive 1-bits that wrap from MSB to LSB, ex: 0xff00...00ff.
+ Otherwise, return FALSE. */
+
+static bool
+has_consecutive_ones (unsigned HOST_WIDE_INT value, int *sh, int *mb)
+{
+ unsigned HOST_WIDE_INT nlz, ntz, mask;
+ unsigned HOST_WIDE_INT allones = -1;
+
+ ntz = ctz_hwi (value);
+ nlz = clz_hwi (value);
+ mask = (allones >> nlz) & (allones << ntz);
+ if (value == mask)
+ {
+ /* Compute beginning and ending bit numbers, using IBM bit numbering. */
+ *mb = nlz;
+ *sh = ntz;
+ return true;
+ }
+
+ /* Check if the inverted value contains consecutive ones. We can create
+ that constant by basically swapping the MB and ME bit numbers. */
+ value = ~value;
+ ntz = ctz_hwi (value);
+ nlz = clz_hwi (value);
+ mask = (allones >> nlz) & (allones << ntz);
+ if (value == mask)
+ {
+ /* Compute beginning and ending bit numbers, using IBM bit numbering. */
+ *mb = GET_MODE_BITSIZE (DImode) - ntz;
+ *sh = GET_MODE_BITSIZE (DImode) - nlz;
+ return true;
+ }
+
+ *sh = *mb = 0;
+ return false;
+}
+
/* Helper for num_insns_constant. Calculate number of instructions to
load VALUE to a single gpr using combinations of addi, addis, ori,
oris and sldi instructions. */
@@ -5752,10 +5793,14 @@ num_insns_constant_gpr (HOST_WIDE_INT value)
{
HOST_WIDE_INT low = ((value & 0xffffffff) ^ 0x80000000) - 0x80000000;
HOST_WIDE_INT high = value >> 31;
+ int sh, mb;
if (high == 0 || high == -1)
return 2;
+ if (has_consecutive_ones (value, &sh, &mb))
+ return 2;
+
high >>= 1;
if (low == 0)
@@ -9427,7 +9472,8 @@ static void
rs6000_emit_set_long_const (rtx dest, HOST_WIDE_INT c)
{
rtx temp;
- HOST_WIDE_INT ud1, ud2, ud3, ud4;
+ HOST_WIDE_INT ud1, ud2, ud3, ud4, value = c;
+ int sh, mb;
ud1 = c & 0xffff;
c = c >> 16;
@@ -9453,6 +9499,12 @@ rs6000_emit_set_long_const (rtx dest, HOST_WIDE_INT c)
gen_rtx_IOR (DImode, copy_rtx (temp),
GEN_INT (ud1)));
}
+ else if (has_consecutive_ones (value, &sh, &mb))
+ {
+ temp = !can_create_pseudo_p () ? dest : gen_reg_rtx (DImode);
+ emit_insn (gen_rtx_SET (copy_rtx (temp), CONSTM1_RTX (DImode)));
+ emit_insn (gen_rldic (dest, copy_rtx (temp), GEN_INT (sh), GEN_INT (mb)));
+ }
else if (ud3 == 0 && ud4 == 0)
{
temp = !can_create_pseudo_p () ? dest : gen_reg_rtx (DImode);
diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md
index 43b620ae1c0..feb5884505c 100644
--- a/gcc/config/rs6000/rs6000.md
+++ b/gcc/config/rs6000/rs6000.md
@@ -154,6 +154,7 @@
UNSPEC_CNTTZDM
UNSPEC_PDEPD
UNSPEC_PEXTD
+ UNSPEC_RLDIC
])
;;
@@ -9173,6 +9174,14 @@
DONE;
})
+(define_insn "rldic"
+ [(set (match_operand:DI 0 "gpc_reg_operand" "=r")
+ (unspec:DI [(match_operand:DI 1 "gpc_reg_operand" "r")
+ (match_operand:DI 2 "u6bit_cint_operand" "n")
+ (match_operand:DI 3 "u6bit_cint_operand" "n")]
+ UNSPEC_RLDIC))]
+ "TARGET_POWERPC64"
+ "rldic %0,%1,%2,%3")
\f
;; TImode/PTImode is similar, except that we usually want to compute the
;; address into a register and use lsi/stsi (the exception is during reload).
diff --git a/gcc/testsuite/gcc.target/powerpc/pr93176.c b/gcc/testsuite/gcc.target/powerpc/pr93176.c
new file mode 100644
index 00000000000..d4d93f8f1b3
--- /dev/null
+++ b/gcc/testsuite/gcc.target/powerpc/pr93176.c
@@ -0,0 +1,49 @@
+/* PR target/93176 */
+/* { dg-do compile } */
+/* { dg-require-effective-target lp64 } */
+/* { dg-options "-O2" } */
+
+/* Verify we generate the correct 2 instruction sequence:
+ li rT,-1; rldic rX,rT,SH,MB for the constants below. */
+
+unsigned long
+test0 (void)
+{
+ return 0x00ffffffffffff00UL;
+}
+
+unsigned long
+test1 (void)
+{
+ return 0x00ffffffff000000UL;
+}
+
+unsigned long
+test2 (void)
+{
+ return 0x00ffff0000000000UL;
+}
+
+unsigned long
+test3 (void)
+{
+ return 0xffffff0000000000UL;
+}
+
+unsigned long
+test4 (void)
+{
+ return 0xffffff000000ffffUL;
+}
+
+unsigned long
+test5 (void)
+{
+ return 0x0000010000000000UL;
+}
+
+/* { dg-final { scan-assembler "rldic r?\[0-9\]+,r?\[0-9\]+,8,8" } } */
+/* { dg-final { scan-assembler "rldic r?\[0-9\]+,r?\[0-9\]+,24,8" } } */
+/* { dg-final { scan-assembler "rldic r?\[0-9\]+,r?\[0-9\]+,40,8" } } */
+/* { dg-final { scan-assembler "rldic r?\[0-9\]+,r?\[0-9\]+,40,48" } } */
+/* { dg-final { scan-assembler "rldic r?\[0-9\]+,r?\[0-9\]+,40,23" } } */
next reply other threads:[~2020-09-10 21:58 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-10 21:58 Peter Bergner [this message]
2020-09-15 4:53 ` Alan Modra
2020-09-15 14:01 ` Segher Boessenkool
2020-09-15 13:56 ` Segher Boessenkool
2020-09-15 15:48 ` Peter Bergner
2020-09-15 19:54 ` Segher Boessenkool
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=838b2e97-dfa9-3ca0-c3c6-1767d60ddf05@linux.ibm.com \
--to=bergner@linux.ibm.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=segher@kernel.crashing.org \
--cc=wschmidt@linux.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).