i386: Add V2SFmode copysign, xorsign and signbit expanders [PR95046]

gcc/ChangeLog:

	PR target/95046
	* config/i386/mmx.md (copysignv2sf3): New expander.
	(xorsignv2sf3): Ditto.
	(signbitv2sf3): Ditto.

testsuite/ChangeLog:

	PR target/95046
	* gcc.target/i386/pr95046-4.c: New test.

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index de4cbd42e168a87239eef2ccda9febb9d3691381..6e4d3df3768f38eb5860a580763870e7e8cec395 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,10 @@
+2020-05-12  Uroš Bizjak  <ubizjak@gmail.com>
+
+       PR target/95046
+       * config/i386/mmx.md (copysignv2sf3): New expander.
+       (xorsignv2sf3): Ditto.
+       (signbitv2sf3): Ditto.
+
 2020-05-12  Uroš Bizjak  <ubizjak@gmail.com>
 
        PR target/95046

diff --git a/gcc/config/i386/mmx.md b/gcc/config/i386/mmx.md
index f6f302eb7ff47ebaad8b46c590a06de611af7091..d159134e0fb7c3abe48f731bbd60f95e10286ea1 100644 (file)
--- a/gcc/config/i386/mmx.md
+++ b/gcc/config/i386/mmx.md
@@ -639,8 +639,8 @@
          (match_operand:V2SF 2 "register_operand" "x,x")))]
   "TARGET_MMX_WITH_SSE"
   "@
-   andps\t{%2, %0|%0, %2}
-   vandps\t{%2, %1, %0|%0, %1, %2}"
+   andnps\t{%2, %0|%0, %2}
+   vandnps\t{%2, %1, %0|%0, %1, %2}"
   [(set_attr "isa" "noavx,avx")
    (set_attr "type" "sselog")
    (set_attr "prefix" "orig,vex")
@@ -660,6 +660,47 @@
    (set_attr "prefix" "orig,vex")
    (set_attr "mode" "V4SF")])
 
+(define_expand "copysignv2sf3"
+  [(set (match_dup 4)
+       (and:V2SF
+         (not:V2SF (match_dup 3))
+         (match_operand:V2SF 1 "register_operand")))
+   (set (match_dup 5)
+       (and:V2SF (match_dup 3)
+                 (match_operand:V2SF 2 "register_operand")))
+   (set (match_operand:V2SF 0 "register_operand")
+       (ior:V2SF (match_dup 4) (match_dup 5)))]
+  "TARGET_MMX_WITH_SSE"
+{
+  operands[3] = ix86_build_signbit_mask (V2SFmode, true, false);
+
+  operands[4] = gen_reg_rtx (V2SFmode);
+  operands[5] = gen_reg_rtx (V2SFmode);
+})
+
+(define_expand "xorsignv2sf3"
+  [(set (match_dup 4)
+       (and:V2SF (match_dup 3)
+                 (match_operand:V2SF 2 "register_operand")))
+   (set (match_operand:V2SF 0 "register_operand")
+       (xor:V2SF (match_dup 4)
+                 (match_operand:V2SF 1 "register_operand")))]
+  "TARGET_MMX_WITH_SSE"
+{
+  operands[3] = ix86_build_signbit_mask (V2SFmode, true, false);
+
+  operands[4] = gen_reg_rtx (V2SFmode);
+})
+
+(define_expand "signbitv2sf2"
+  [(set (match_operand:V2SI 0 "register_operand")
+       (lshiftrt:V2SI
+         (subreg:V2SI
+           (match_operand:V2SF 1 "register_operand") 0)
+         (match_dup 2)))]
+  "TARGET_MMX_WITH_SSE"
+  "operands[2] = GEN_INT (GET_MODE_UNIT_BITSIZE (V2SFmode)-1);")
+
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;;
 ;; Parallel single-precision FMA multiply/accumulate instructions.

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 99ebf82025c4798d3977afcb8f4a444bb7819e2e..8935bee0fbe122c0f375322f38b0a1cfb0f126b9 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2020-05-12  Uroš Bizjak  <ubizjak@gmail.com>
+
+       PR target/95046
+       * gcc.target/i386/pr95046-4.c: New test.
+
 2020-05-12  Patrick Palka  <ppalka@redhat.com>
 
        PR c++/78752

diff --git a/gcc/testsuite/gcc.target/i386/pr95046-4.c b/gcc/testsuite/gcc.target/i386/pr95046-4.c
new file mode 100644 (file)
index 0000000..5a85045
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr95046-4.c
@@ -0,0 +1,39 @@
+/* PR target/95046 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O3 -msse2" } */
+
+
+float r[2], a[2], b[2];
+
+float copysignf (float, float);
+
+void
+test_copysign (void)
+{
+  for (int i = 0; i < 2; i++)
+    r[i] = copysignf (a[i], b[i]);
+}
+
+/* { dg-final { scan-assembler "\tv?andnps" } } */
+
+void
+test_xorsign (void)
+{
+  for (int i = 0; i < 2; i++)
+    r[i] = a[i] * copysignf (1.0f, b[i]);
+}
+
+/* { dg-final { scan-assembler "\tv?xorps" } } */
+
+int s[2];
+
+int signbitf (float);
+
+void
+test_signbitf (void)
+{
+  for (int i = 0; i < 2; i++)
+    s[i] = signbitf (a[i]);
+}
+
+/* { dg-final { scan-assembler "\tv?psrld" } } */