public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc/devel/c++-modules] RISC-V: Optimize si to di zero-extend followed by left shift.
@ 2020-06-11 12:59 Nathan Sidwell
  0 siblings, 0 replies; only message in thread
From: Nathan Sidwell @ 2020-06-11 12:59 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:d5cdcd5cf2b2920b44836005baceb59d046b6e5a

commit d5cdcd5cf2b2920b44836005baceb59d046b6e5a
Author: Jim Wilson <jimw@sifive.com>
Date:   Sat May 30 17:04:17 2020 -0700

    RISC-V: Optimize si to di zero-extend followed by left shift.
    
    This is potentially a sequence of 3 shifts, we which optimize to a sequence
    of 2 shifts.  This can happen when unsigned int is used for array indexing.
    
            gcc/
            * config/riscv/riscv.md (zero_extendsidi2_shifted): New.
    
            gcc/testsuite/
            * gcc.target/riscv/zero-extend-5.c: New.

Diff:
---
 gcc/config/riscv/riscv.md                      | 22 ++++++++++++++++++++++
 gcc/testsuite/gcc.target/riscv/zero-extend-5.c |  8 ++++++++
 2 files changed, 30 insertions(+)

diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md
index 434e064e095..f4bdb7d8cfe 100644
--- a/gcc/config/riscv/riscv.md
+++ b/gcc/config/riscv/riscv.md
@@ -1808,6 +1808,28 @@
   operands[2] = GEN_INT (ctz_hwi (INTVAL (operands[2])));
 })
 
+;; Handle SImode to DImode zero-extend combined with a left shift.  This can
+;; occur when unsigned int is used for array indexing.  Split this into two
+;; shifts.  Otherwise we can get 3 shifts.
+
+(define_insn_and_split "zero_extendsidi2_shifted"
+  [(set (match_operand:DI 0 "register_operand" "=r")
+	(and:DI (ashift:DI (match_operand:DI 1 "register_operand" "r")
+			   (match_operand:QI 2 "immediate_operand" "I"))
+		(match_operand 3 "immediate_operand" "")))
+   (clobber (match_scratch:DI 4 "=&r"))]
+  "TARGET_64BIT
+   && ((INTVAL (operands[3]) >> INTVAL (operands[2])) == 0xffffffff)"
+  "#"
+  "&& reload_completed"
+  [(set (match_dup 4)
+	(ashift:DI (match_dup 1) (const_int 32)))
+   (set (match_dup 0)
+	(lshiftrt:DI (match_dup 4) (match_dup 5)))]
+  "operands[5] = GEN_INT (32 - (INTVAL (operands [2])));"
+  [(set_attr "type" "shift")
+   (set_attr "mode" "DI")])
+
 ;;
 ;;  ....................
 ;;
diff --git a/gcc/testsuite/gcc.target/riscv/zero-extend-5.c b/gcc/testsuite/gcc.target/riscv/zero-extend-5.c
new file mode 100644
index 00000000000..1a135b8c097
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/zero-extend-5.c
@@ -0,0 +1,8 @@
+/* { dg-do compile { target { riscv64*-*-* } } } */
+/* { dg-options "-march=rv64gc -mabi=lp64 -O2" } */
+int
+sub (unsigned int i, unsigned int j, unsigned int k, int *array)
+{
+  return array[i] + array[j] + array[k];
+}
+/* { dg-final { scan-assembler-times "slli" 3 } } */


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2020-06-11 12:59 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-06-11 12:59 [gcc/devel/c++-modules] RISC-V: Optimize si to di zero-extend followed by left shift Nathan Sidwell

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).