[17/31] VAX: Actually produce QImode and HImode `ctz' operations

Message ID alpine.LFD.2.21.2011200252070.656242@eddie.linux-mips.org
State New
Headers show
Series
  • VAX: Bring the port up to date (yes, MODE_CC conversion is included)
Related show

Commit Message

Maciej W. Rozycki Nov. 20, 2020, 3:35 a.m.
The middle end does not refer to `ctzqi2'/`ctzhi2' or `ffsqi2'/`ffshi2'
patterns by name where `__builtin_ctz' or `__builtin_ffs' respectively
is invoked for an argument of the QImode or HImode type, and instead it
extends the data type before passing it to `ctzsi2' or `ffssi2'.

Avoid the redundant operation and use a peephole2 to convert it to the
right RTL expression that will collapse the two operations into a single
machine instruction instead unless we need the extended intermediate
result for another purpose.

	gcc/
	* config/vax/builtins.md: Add a peephole2 for QImode and HImode
	`ctz' operations.
	(any_extend): New code iterator.

	gcc/testsuite/
	* gcc.target/vax/ctzhi.c: New test.
	* gcc.target/vax/ctzqi.c: New test.
	* gcc.target/vax/ffshi.c: New test.
	* gcc.target/vax/ffsqi.c: New test.
---
 gcc/config/vax/builtins.md           | 22 ++++++++++++++++++++++
 gcc/testsuite/gcc.target/vax/ctzhi.c | 20 ++++++++++++++++++++
 gcc/testsuite/gcc.target/vax/ctzqi.c | 20 ++++++++++++++++++++
 gcc/testsuite/gcc.target/vax/ffshi.c | 24 ++++++++++++++++++++++++
 gcc/testsuite/gcc.target/vax/ffsqi.c | 24 ++++++++++++++++++++++++
 5 files changed, 110 insertions(+)
 create mode 100644 gcc/testsuite/gcc.target/vax/ctzhi.c
 create mode 100644 gcc/testsuite/gcc.target/vax/ctzqi.c
 create mode 100644 gcc/testsuite/gcc.target/vax/ffshi.c
 create mode 100644 gcc/testsuite/gcc.target/vax/ffsqi.c

-- 
2.11.0

Comments

Kees Cook via Gcc-patches Nov. 20, 2020, 11:24 p.m. | #1
On 11/19/20 8:35 PM, Maciej W. Rozycki wrote:
> The middle end does not refer to `ctzqi2'/`ctzhi2' or `ffsqi2'/`ffshi2'

> patterns by name where `__builtin_ctz' or `__builtin_ffs' respectively

> is invoked for an argument of the QImode or HImode type, and instead it

> extends the data type before passing it to `ctzsi2' or `ffssi2'.

>

> Avoid the redundant operation and use a peephole2 to convert it to the

> right RTL expression that will collapse the two operations into a single

> machine instruction instead unless we need the extended intermediate

> result for another purpose.

>

> 	gcc/

> 	* config/vax/builtins.md: Add a peephole2 for QImode and HImode

> 	`ctz' operations.

> 	(any_extend): New code iterator.

>

> 	gcc/testsuite/

> 	* gcc.target/vax/ctzhi.c: New test.

> 	* gcc.target/vax/ctzqi.c: New test.

> 	* gcc.target/vax/ffshi.c: New test.

> 	* gcc.target/vax/ffsqi.c: New test.

OK
jeff

Patch

diff --git a/gcc/config/vax/builtins.md b/gcc/config/vax/builtins.md
index b7ed9762c23..e96ac3f52ab 100644
--- a/gcc/config/vax/builtins.md
+++ b/gcc/config/vax/builtins.md
@@ -29,6 +29,8 @@  (define_mode_attr bb_mem [(QI "m") (HI "Q") (SI "Q")])
 (define_int_iterator bit [0 1])
 (define_int_attr ccss [(0 "cc") (1 "ss")])
 
+(define_code_iterator any_extend [sign_extend zero_extend])
+
 (define_expand "ffs<mode>2"
   [(set (match_operand:SI 0 "nonimmediate_operand" "")
 	(ffs:SI (match_operand:VAXint 1 "general_operand" "")))]
@@ -57,6 +59,26 @@  (define_insn "ctz<mode>2"
   ""
   "ffs $0,$<width>,%1,%0")
 
+;; Our FFS hardware instruction supports any field width,
+;; so handle narrower inputs directly as well.
+(define_peephole2
+  [(set (match_operand:SI 0 "register_operand")
+        (any_extend:SI (match_operand:VAXintQH 1 "general_operand")))
+   (parallel
+     [(set (match_operand:SI 2 "nonimmediate_operand")
+	   (ctz:SI (match_dup 0)))
+      (set (cc0)
+	   (compare (match_dup 2)
+		    (const_int 0)))])]
+  "rtx_equal_p (operands[0], operands[2]) || peep2_reg_dead_p (2, operands[0])"
+  [(parallel
+     [(set (match_dup 2)
+	   (ctz:SI (match_dup 1)))
+      (set (cc0)
+	   (compare (match_dup 1)
+		    (const_int 0)))])]
+  "")
+
 (define_expand "sync_lock_test_and_set<mode>"
   [(match_operand:VAXint 0 "nonimmediate_operand" "=&g")
    (match_operand:VAXint 1 "memory_operand" "+m")
diff --git a/gcc/testsuite/gcc.target/vax/ctzhi.c b/gcc/testsuite/gcc.target/vax/ctzhi.c
new file mode 100644
index 00000000000..fcc9f06f7d2
--- /dev/null
+++ b/gcc/testsuite/gcc.target/vax/ctzhi.c
@@ -0,0 +1,20 @@ 
+/* { dg-do compile } */
+/* { dg-options "-fdump-rtl-peephole2" } */
+/* { dg-skip-if "code quality test" { *-*-* } { "-O0" "-O1" } { "" } } */
+
+typedef unsigned int __attribute__ ((mode (HI))) int_t;
+
+int
+ctzhi (int_t *x)
+{
+  return __builtin_ctz (*x);
+}
+
+/* Expect assembly like:
+
+	ffs $0,$16,*4(%ap),%r0
+
+ */
+
+/* { dg-final { scan-rtl-dump-times "Splitting with gen_peephole2" 1 "peephole2" } } */
+/* { dg-final { scan-assembler "\tffs \\\$0,\\\$16," } } */
diff --git a/gcc/testsuite/gcc.target/vax/ctzqi.c b/gcc/testsuite/gcc.target/vax/ctzqi.c
new file mode 100644
index 00000000000..067334b09e2
--- /dev/null
+++ b/gcc/testsuite/gcc.target/vax/ctzqi.c
@@ -0,0 +1,20 @@ 
+/* { dg-do compile } */
+/* { dg-options "-fdump-rtl-peephole2" } */
+/* { dg-skip-if "code quality test" { *-*-* } { "-O0" "-O1" } { "" } } */
+
+typedef unsigned int __attribute__ ((mode (QI))) int_t;
+
+int
+ctzqi (int_t *x)
+{
+  return __builtin_ctz (*x);
+}
+
+/* Expect assembly like:
+
+	ffs $0,$8,*4(%ap),%r0
+
+ */
+
+/* { dg-final { scan-rtl-dump-times "Splitting with gen_peephole2" 1 "peephole2" } } */
+/* { dg-final { scan-assembler "\tffs \\\$0,\\\$8," } } */
diff --git a/gcc/testsuite/gcc.target/vax/ffshi.c b/gcc/testsuite/gcc.target/vax/ffshi.c
new file mode 100644
index 00000000000..db592fb5724
--- /dev/null
+++ b/gcc/testsuite/gcc.target/vax/ffshi.c
@@ -0,0 +1,24 @@ 
+/* { dg-do compile } */
+/* { dg-options "-fdump-rtl-peephole2" } */
+/* { dg-skip-if "code quality test" { *-*-* } { "-O0" "-O1" } { "" } } */
+
+typedef int __attribute__ ((mode (HI))) int_t;
+
+int
+ffshi (int_t *x)
+{
+  return __builtin_ffs (*x);
+}
+
+/* Expect assembly like:
+
+	ffs $0,$16,*4(%ap),%r0
+	jneq .L2
+	mnegl $1,%r0
+.L2:
+	incl %r0
+
+ */
+
+/* { dg-final { scan-rtl-dump-times "Splitting with gen_peephole2" 1 "peephole2" } } */
+/* { dg-final { scan-assembler "\tffs \\\$0,\\\$16," } } */
diff --git a/gcc/testsuite/gcc.target/vax/ffsqi.c b/gcc/testsuite/gcc.target/vax/ffsqi.c
new file mode 100644
index 00000000000..ebcd9460754
--- /dev/null
+++ b/gcc/testsuite/gcc.target/vax/ffsqi.c
@@ -0,0 +1,24 @@ 
+/* { dg-do compile } */
+/* { dg-options "-fdump-rtl-peephole2" } */
+/* { dg-skip-if "code quality test" { *-*-* } { "-O0" "-O1" } { "" } } */
+
+typedef int __attribute__ ((mode (QI))) int_t;
+
+int
+ffsqi (int_t *x)
+{
+  return __builtin_ffs (*x);
+}
+
+/* Expect assembly like:
+
+	ffs $0,$8,*4(%ap),%r0
+	jneq .L2
+	mnegl $1,%r0
+.L2:
+	incl %r0
+
+ */
+
+/* { dg-final { scan-rtl-dump-times "Splitting with gen_peephole2" 1 "peephole2" } } */
+/* { dg-final { scan-assembler "\tffs \\\$0,\\\$8," } } */