i386: Implement integer vector compares for 64bit vectors [PR98218]

Implement integer vector compares for 64bit vectors for TARGET_MMX_WITH_SSE.

2021-05-05  Uroš Bizjak  <ubizjak@gmail.com>

gcc/
	PR target/98218
	* config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
	Handle V8QI, V4HI and V2SI modes.
	* config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
	(ix86_build_signbit_mask): Ditto.
	* config/i386/mmx.md (MMXMODE14): New mode iterator.
	(<smaxmin:code><MMXMODE14:mode>3): New expander.
	(*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
	(<umaxmin:code><MMXMODE24:mode>3): New expander.
	(*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
	(vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
	(vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
	(vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
	(vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
	(vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.

gcc/testsuite/

	PR target/98218
	* gcc.target/i386/pr98218-1.c: New test.
	* gcc.target/i386/pr98218-1a.c: Ditto.
	* gcc.target/i386/pr98218-2.c: Ditto.
	* gcc.target/i386/pr98218-2a.c: Ditto.
	* gcc.target/i386/pr98218-3.c: Ditto.
	* gcc.target/i386/pr98218-3a.c: Ditto.
	* gcc.dg/vect/vect-bool-cmp.c (dg-final):
	Scan vect tree dump for "LOOP VECTORIZED", not VECTORIZED.

10 files changed, 260 insertions, 1 deletions

diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
index fee4d07..4dfe7d6 100644
--- a/gcc/config/i386/i386-expand.c
+++ b/gcc/config/i386/i386-expand.c
@@ -4204,16 +4204,32 @@ ix86_expand_int_sse_cmp (rtx dest, enum rtx_code code, rtx cop0, rtx cop1,
 	      else if (code == GT && TARGET_SSE4_1)
 		gen = gen_sminv16qi3;
 	      break;
+	    case E_V8QImode:
+	      if (code == GTU && TARGET_SSE2)
+		gen = gen_uminv8qi3;
+	      else if (code == GT && TARGET_SSE4_1)
+		gen = gen_sminv8qi3;
+	      break;
 	    case E_V8HImode:
 	      if (code == GTU && TARGET_SSE4_1)
 		gen = gen_uminv8hi3;
 	      else if (code == GT && TARGET_SSE2)
 		gen = gen_sminv8hi3;
 	      break;
+	    case E_V4HImode:
+	      if (code == GTU && TARGET_SSE4_1)
+		gen = gen_uminv4hi3;
+	      else if (code == GT && TARGET_SSE2)
+		gen = gen_sminv4hi3;
+	      break;
 	    case E_V4SImode:
 	      if (TARGET_SSE4_1)
 		gen = (code == GTU) ? gen_uminv4si3 : gen_sminv4si3;
 	      break;
+	    case E_V2SImode:
+	      if (TARGET_SSE4_1)
+		gen = (code == GTU) ? gen_uminv2si3 : gen_sminv2si3;
+	      break;
 	    case E_V2DImode:
 	      if (TARGET_AVX512VL)
 		{
@@ -4254,6 +4270,7 @@ ix86_expand_int_sse_cmp (rtx dest, enum rtx_code code, rtx cop0, rtx cop1,
 	    case E_V8SImode:
 	    case E_V4DImode:
 	    case E_V4SImode:
+	    case E_V2SImode:
 	    case E_V2DImode:
 		{
 		  rtx t1, t2, mask;
@@ -4278,7 +4295,9 @@ ix86_expand_int_sse_cmp (rtx dest, enum rtx_code code, rtx cop0, rtx cop1,
 	    case E_V32QImode:
 	    case E_V16HImode:
 	    case E_V16QImode:
+	    case E_V8QImode:
 	    case E_V8HImode:
+	    case E_V4HImode:
 	      /* Perform a parallel unsigned saturating subtraction.  */
 	      x = gen_reg_rtx (mode);
 	      emit_insn (gen_rtx_SET
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 780da10..06b0f58 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -15284,6 +15284,7 @@ ix86_build_const_vector (machine_mode mode, bool vect, rtx value)
     case E_V16SImode:
     case E_V8SImode:
     case E_V4SImode:
+    case E_V2SImode:
     case E_V8DImode:
     case E_V4DImode:
     case E_V2DImode:
@@ -15334,6 +15335,7 @@ ix86_build_signbit_mask (machine_mode mode, bool vect, bool invert)
     case E_V8SFmode:
     case E_V4SFmode:
     case E_V2SFmode:
+    case E_V2SImode:
       vec_mode = mode;
       imode = SImode;
       break;
diff --git a/gcc/config/i386/mmx.md b/gcc/config/i386/mmx.md
index 4c2b724..347295a 100644
--- a/gcc/config/i386/mmx.md
+++ b/gcc/config/i386/mmx.md
@@ -52,6 +52,7 @@
 
 ;; Mix-n-match
 (define_mode_iterator MMXMODE12 [V8QI V4HI])
+(define_mode_iterator MMXMODE14 [V8QI V2SI])
 (define_mode_iterator MMXMODE24 [V4HI V2SI])
 (define_mode_iterator MMXMODE248 [V4HI V2SI V1DI])
 
@@ -1417,6 +1418,31 @@
    (set_attr "type" "mmxmul,ssemul,ssemul")
    (set_attr "mode" "DI,TI,TI")])
 
+(define_expand "<code><mode>3"
+  [(set (match_operand:MMXMODE14 0 "register_operand")
+        (smaxmin:MMXMODE14
+	  (match_operand:MMXMODE14 1 "register_operand")
+	  (match_operand:MMXMODE14 2 "register_operand")))]
+  "TARGET_MMX_WITH_SSE && TARGET_SSE4_1"
+  "ix86_fixup_binary_operands_no_copy (<CODE>, <MODE>mode, operands);")
+
+(define_insn "*mmx_<code><mode>3"
+  [(set (match_operand:MMXMODE14 0 "register_operand" "=Yr,*x,Yv")
+	(smaxmin:MMXMODE14
+	  (match_operand:MMXMODE14 1 "register_operand" "%0,0,Yv")
+	  (match_operand:MMXMODE14 2 "register_operand" "Yr,*x,Yv")))]
+  "TARGET_MMX_WITH_SSE && TARGET_SSE4_1
+   && ix86_binary_operator_ok (<CODE>, <MODE>mode, operands)"
+  "@
+   p<maxmin_int><mmxvecsize>\t{%2, %0|%0, %2}
+   p<maxmin_int><mmxvecsize>\t{%2, %0|%0, %2}
+   vp<maxmin_int><mmxvecsize>\t{%2, %1, %0|%0, %1, %2}"
+  [(set_attr "isa" "noavx,noavx,avx")
+   (set_attr "type" "sseiadd")
+   (set_attr "prefix_extra" "1,1,*")
+   (set_attr "prefix" "orig,orig,vex")
+   (set_attr "mode" "TI")])
+
 (define_expand "mmx_<code>v4hi3"
   [(set (match_operand:V4HI 0 "register_operand")
         (smaxmin:V4HI
@@ -1451,6 +1477,31 @@
    (set_attr "type" "mmxadd,sseiadd,sseiadd")
    (set_attr "mode" "DI,TI,TI")])
 
+(define_expand "<code><mode>3"
+  [(set (match_operand:MMXMODE24 0 "register_operand")
+        (umaxmin:MMXMODE24
+	  (match_operand:MMXMODE24 1 "register_operand")
+	  (match_operand:MMXMODE24 2 "register_operand")))]
+  "TARGET_MMX_WITH_SSE && TARGET_SSE4_1"
+  "ix86_fixup_binary_operands_no_copy (<CODE>, <MODE>mode, operands);")
+
+(define_insn "*mmx_<code><mode>3"
+  [(set (match_operand:MMXMODE24 0 "register_operand" "=Yr,*x,Yv")
+	(umaxmin:MMXMODE24
+	  (match_operand:MMXMODE24 1 "register_operand" "%0,0,Yv")
+	  (match_operand:MMXMODE24 2 "register_operand" "Yr,*x,Yv")))]
+  "TARGET_MMX_WITH_SSE && TARGET_SSE4_1
+   && ix86_binary_operator_ok (<CODE>, <MODE>mode, operands)"
+  "@
+   p<maxmin_int><mmxvecsize>\t{%2, %0|%0, %2}
+   p<maxmin_int><mmxvecsize>\t{%2, %0|%0, %2}
+   vp<maxmin_int><mmxvecsize>\t{%2, %1, %0|%0, %1, %2}"
+  [(set_attr "isa" "noavx,noavx,avx")
+   (set_attr "type" "sseiadd")
+   (set_attr "prefix_extra" "1,1,*")
+   (set_attr "prefix" "orig,orig,vex")
+   (set_attr "mode" "TI")])
+
 (define_expand "mmx_<code>v8qi3"
   [(set (match_operand:V8QI 0 "register_operand")
         (umaxmin:V8QI
@@ -1582,6 +1633,73 @@
    (set_attr "type" "mmxcmp,ssecmp,ssecmp")
    (set_attr "mode" "DI,TI,TI")])
 
+(define_expand "vec_cmp<mode><mode>"
+  [(set (match_operand:MMXMODEI 0 "register_operand")
+	(match_operator:MMXMODEI 1 ""
+	  [(match_operand:MMXMODEI 2 "register_operand")
+	   (match_operand:MMXMODEI 3 "register_operand")]))]
+  "TARGET_MMX_WITH_SSE"
+{
+  bool ok = ix86_expand_int_vec_cmp (operands);
+  gcc_assert (ok);
+  DONE;
+})
+
+(define_expand "vec_cmpu<mode><mode>"
+  [(set (match_operand:MMXMODEI 0 "register_operand")
+	(match_operator:MMXMODEI 1 ""
+	  [(match_operand:MMXMODEI 2 "register_operand")
+	   (match_operand:MMXMODEI 3 "register_operand")]))]
+  "TARGET_MMX_WITH_SSE"
+{
+  bool ok = ix86_expand_int_vec_cmp (operands);
+  gcc_assert (ok);
+  DONE;
+})
+
+(define_expand "vcond<mode><mode>"
+  [(set (match_operand:MMXMODEI 0 "register_operand")
+	(if_then_else:MMXMODEI
+	  (match_operator 3 ""
+	    [(match_operand:MMXMODEI 4 "register_operand")
+	     (match_operand:MMXMODEI 5 "register_operand")])
+	  (match_operand:MMXMODEI 1)
+	  (match_operand:MMXMODEI 2)))]
+  "TARGET_MMX_WITH_SSE"
+{
+  bool ok = ix86_expand_int_vcond (operands);
+  gcc_assert (ok);
+  DONE;
+})
+
+(define_expand "vcondu<mode><mode>"
+  [(set (match_operand:MMXMODEI 0 "register_operand")
+	(if_then_else:MMXMODEI
+	  (match_operator 3 ""
+	    [(match_operand:MMXMODEI 4 "register_operand")
+	     (match_operand:MMXMODEI 5 "register_operand")])
+	  (match_operand:MMXMODEI 1)
+	  (match_operand:MMXMODEI 2)))]
+  "TARGET_MMX_WITH_SSE"
+{
+  bool ok = ix86_expand_int_vcond (operands);
+  gcc_assert (ok);
+  DONE;
+})
+
+(define_expand "vcond_mask_<mode><mode>"
+  [(set (match_operand:MMXMODEI 0 "register_operand")
+	(vec_merge:MMXMODEI
+	  (match_operand:MMXMODEI 1 "register_operand")
+	  (match_operand:MMXMODEI 2 "register_operand")
+	  (match_operand:MMXMODEI 3 "register_operand")))]
+  "TARGET_MMX_WITH_SSE"
+{
+  ix86_expand_sse_movcc (operands[0], operands[3],
+			 operands[1], operands[2]);
+  DONE;
+})
+
 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 ;;
 ;; Parallel integral logical operations
diff --git a/gcc/testsuite/gcc.dg/vect/vect-bool-cmp.c b/gcc/testsuite/gcc.dg/vect/vect-bool-cmp.c
index 35d2a3c..c97da52 100644
--- a/gcc/testsuite/gcc.dg/vect/vect-bool-cmp.c
+++ b/gcc/testsuite/gcc.dg/vect/vect-bool-cmp.c
@@ -253,4 +253,4 @@ main (int argc, char **argv)
   check (res, ne);
 }
 
-/* { dg-final { scan-tree-dump-times "VECTORIZED" 18 "vect" { target sse4_runtime } } } */
+/* { dg-final { scan-tree-dump-times "LOOP VECTORIZED" 18 "vect" { target sse4_runtime } } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr98218-1.c b/gcc/testsuite/gcc.target/i386/pr98218-1.c
new file mode 100644
index 0000000..48407da
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98218-1.c
@@ -0,0 +1,21 @@
+/* PR target/98522 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -msse2" } */
+
+typedef char vec __attribute__((vector_size(8)));
+
+vec lt (vec a, vec b) { return a < b; }
+vec le (vec a, vec b) { return a <= b; }
+vec eq (vec a, vec b) { return a == b; }
+vec ne (vec a, vec b) { return a != b; }
+vec ge (vec a, vec b) { return a >= b; }
+vec gt (vec a, vec b) { return a > b; }
+
+typedef unsigned char uvec __attribute__((vector_size(8)));
+
+vec ltu (uvec a, uvec b) { return a < b; }
+vec leu (uvec a, uvec b) { return a <= b; }
+vec geu (uvec a, uvec b) { return a >= b; }
+vec gtu (uvec a, uvec b) { return a > b; }
+
+/* { dg-final { scan-assembler-not "cmpb" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr98218-1a.c b/gcc/testsuite/gcc.target/i386/pr98218-1a.c
new file mode 100644
index 0000000..3470c87
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98218-1a.c
@@ -0,0 +1,19 @@
+/* PR target/98522 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -ftree-vectorize -msse2" } */
+
+typedef char T;
+
+#define M 8
+
+extern T a[M], b[M], s1[M], s2[M], r[M];
+
+void foo (void)
+{
+  int j;
+
+  for (j = 0; j < M; j++)
+    r[j] = (a[j] < b[j]) ? s1[j] : s2[j];
+}
+
+/* { dg-final { scan-assembler "pcmpgtb" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr98218-2.c b/gcc/testsuite/gcc.target/i386/pr98218-2.c
new file mode 100644
index 0000000..0b71612
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98218-2.c
@@ -0,0 +1,21 @@
+/* PR target/98522 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -msse2" } */
+
+typedef short vec __attribute__((vector_size(8)));
+
+vec lt (vec a, vec b) { return a < b; }
+vec le (vec a, vec b) { return a <= b; }
+vec eq (vec a, vec b) { return a == b; }
+vec ne (vec a, vec b) { return a != b; }
+vec ge (vec a, vec b) { return a >= b; }
+vec gt (vec a, vec b) { return a > b; }
+
+typedef unsigned short uvec __attribute__((vector_size(8)));
+
+vec ltu (uvec a, uvec b) { return a < b; }
+vec leu (uvec a, uvec b) { return a <= b; }
+vec geu (uvec a, uvec b) { return a >= b; }
+vec gtu (uvec a, uvec b) { return a > b; }
+
+/* { dg-final { scan-assembler-not "cmpw" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr98218-2a.c b/gcc/testsuite/gcc.target/i386/pr98218-2a.c
new file mode 100644
index 0000000..6afd0a4
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98218-2a.c
@@ -0,0 +1,19 @@
+/* PR target/98522 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -ftree-vectorize -msse2" } */
+
+typedef short T;
+
+#define M 4
+
+extern T a[M], b[M], s1[M], s2[M], r[M];
+
+void foo (void)
+{
+  int j;
+
+  for (j = 0; j < M; j++)
+    r[j] = (a[j] < b[j]) ? s1[j] : s2[j];
+}
+
+/* { dg-final { scan-assembler "pcmpgtw" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr98218-3.c b/gcc/testsuite/gcc.target/i386/pr98218-3.c
new file mode 100644
index 0000000..83a8c29
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98218-3.c
@@ -0,0 +1,21 @@
+/* PR target/98522 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -msse2" } */
+
+typedef int vec __attribute__((vector_size(8)));
+
+vec lt (vec a, vec b) { return a < b; }
+vec le (vec a, vec b) { return a <= b; }
+vec eq (vec a, vec b) { return a == b; }
+vec ne (vec a, vec b) { return a != b; }
+vec ge (vec a, vec b) { return a >= b; }
+vec gt (vec a, vec b) { return a > b; }
+
+typedef unsigned int uvec __attribute__((vector_size(8)));
+
+vec ltu (uvec a, uvec b) { return a < b; }
+vec leu (uvec a, uvec b) { return a <= b; }
+vec geu (uvec a, uvec b) { return a >= b; }
+vec gtu (uvec a, uvec b) { return a > b; }
+
+/* { dg-final { scan-assembler-not "cmpl" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr98218-3a.c b/gcc/testsuite/gcc.target/i386/pr98218-3a.c
new file mode 100644
index 0000000..272d54e
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr98218-3a.c
@@ -0,0 +1,19 @@
+/* PR target/98522 */
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -ftree-vectorize -msse2" } */
+
+typedef int T;
+
+#define M 2
+
+extern T a[M], b[M], s1[M], s2[M], r[M];
+
+void foo (void)
+{
+  int j;
+
+  for (j = 0; j < M; j++)
+    r[j] = (a[j] < b[j]) ? s1[j] : s2[j];
+}
+
+/* { dg-final { scan-assembler "pcmpgtd" { xfail *-*-* } } } */