diff mbox

[ARM/testsuite,07/22] Add binary saturating operators: vqadd, vqsub.

Message ID 1402005882-31597-8-git-send-email-christophe.lyon@linaro.org
State New
Headers show

Commit Message

Christophe Lyon June 5, 2014, 10:04 p.m. UTC

diff mbox

Patch

diff --git a/gcc/testsuite/gcc.target/arm/neon-intrinsics/binary_sat_op.inc b/gcc/testsuite/gcc.target/arm/neon-intrinsics/binary_sat_op.inc
new file mode 100644
index 0000000..35d7701
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/neon-intrinsics/binary_sat_op.inc
@@ -0,0 +1,91 @@ 
+/* Template file for saturating binary operator validation.
+
+   This file is meant to be included by the relevant test files, which
+   have to define the intrinsic family to test. If a given intrinsic
+   supports variants which are not supported by all the other
+   saturating binary operators, these can be tested by providing a
+   definition for EXTRA_TESTS.  */
+
+#include <arm_neon.h>
+#include "arm-neon-ref.h"
+#include "compute-ref-data.h"
+
+#define FNNAME1(NAME) exec_ ## NAME
+#define FNNAME(NAME) FNNAME1(NAME)
+
+void FNNAME (INSN_NAME) (void)
+{
+  /* vector_res = OP(vector1,vector2), then store the result.  */
+
+#define TEST_BINARY_SAT_OP1(INSN, Q, T1, T2, W, N, EXPECTED_CUMULATIVE_SAT, CMT) \
+  Set_Neon_Cumulative_Sat(0);						\
+  VECT_VAR(vector_res, T1, W, N) =					\
+    INSN##Q##_##T2##W(VECT_VAR(vector1, T1, W, N),			\
+		      VECT_VAR(vector2, T1, W, N));			\
+    vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
+		      VECT_VAR(vector_res, T1, W, N));			\
+      CHECK_CUMULATIVE_SAT(TEST_MSG, T1, W, N, EXPECTED_CUMULATIVE_SAT, CMT)
+
+#define TEST_BINARY_SAT_OP(INSN, Q, T1, T2, W, N, EXPECTED_CUMULATIVE_SAT, CMT) \
+  TEST_BINARY_SAT_OP1(INSN, Q, T1, T2, W, N, EXPECTED_CUMULATIVE_SAT, CMT)
+
+  DECL_VARIABLE_ALL_VARIANTS(vector1);
+  DECL_VARIABLE_ALL_VARIANTS(vector2);
+  DECL_VARIABLE_ALL_VARIANTS(vector_res);
+
+  clean_results ();
+
+  /* Initialize input "vector1" from "buffer".  */
+  TEST_MACRO_ALL_VARIANTS_2_5(VLOAD, vector1, buffer);
+
+  /* Choose arbitrary initialization values.  */
+  VDUP(vector2, , int, s, 8, 8, 0x11);
+  VDUP(vector2, , int, s, 16, 4, 0x22);
+  VDUP(vector2, , int, s, 32, 2, 0x33);
+  VDUP(vector2, , int, s, 64, 1, 0x44);
+  VDUP(vector2, , uint, u, 8, 8, 0x55);
+  VDUP(vector2, , uint, u, 16, 4, 0x66);
+  VDUP(vector2, , uint, u, 32, 2, 0x77);
+  VDUP(vector2, , uint, u, 64, 1, 0x88);
+
+  VDUP(vector2, q, int, s, 8, 16, 0x11);
+  VDUP(vector2, q, int, s, 16, 8, 0x22);
+  VDUP(vector2, q, int, s, 32, 4, 0x33);
+  VDUP(vector2, q, int, s, 64, 2, 0x44);
+  VDUP(vector2, q, uint, u, 8, 16, 0x55);
+  VDUP(vector2, q, uint, u, 16, 8, 0x66);
+  VDUP(vector2, q, uint, u, 32, 4, 0x77);
+  VDUP(vector2, q, uint, u, 64, 2, 0x88);
+
+  /* Apply a saturating binary operator named INSN_NAME.  */
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 8, 8, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 16, 4, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 32, 2, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 8, 8, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 16, 4, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 32, 2, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat, "");
+
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 8, 16, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 16, 8, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 32, 4, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 8, 16, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 16, 8, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 32, 4, expected_cumulative_sat, "");
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat, "");
+
+  CHECK_RESULTS (TEST_MSG, "");
+
+#ifdef EXTRA_TESTS
+  EXTRA_TESTS();
+#endif
+}
+
+int main (void)
+{
+  FNNAME (INSN_NAME) ();
+
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.target/arm/neon-intrinsics/vqadd.c b/gcc/testsuite/gcc.target/arm/neon-intrinsics/vqadd.c
new file mode 100644
index 0000000..c07f5ff
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/neon-intrinsics/vqadd.c
@@ -0,0 +1,278 @@ 
+#define INSN_NAME vqadd
+#define TEST_MSG "VQADD/VQADDQ"
+
+/* Extra tests for special cases:
+   - some requiring intermediate types larger than 64 bits to
+   compute saturation flag.
+   - corner case saturations with types smaller than 64 bits.
+*/
+void vqadd_extras(void);
+#define EXTRA_TESTS vqadd_extras
+
+#include "binary_sat_op.inc"
+
+/* Expected values of cumulative_saturation flag.  */
+int VECT_VAR(expected_cumulative_sat,int,8,8) = 0;
+int VECT_VAR(expected_cumulative_sat,int,16,4) = 0;
+int VECT_VAR(expected_cumulative_sat,int,32,2) = 0;
+int VECT_VAR(expected_cumulative_sat,int,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1;
+int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1;
+int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1;
+int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1;
+int VECT_VAR(expected_cumulative_sat,int,8,16) = 0;
+int VECT_VAR(expected_cumulative_sat,int,16,8) = 0;
+int VECT_VAR(expected_cumulative_sat,int,32,4) = 0;
+int VECT_VAR(expected_cumulative_sat,int,64,2) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1;
+int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1;
+int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1;
+int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1;
+/* Expected results.  */
+VECT_VAR_DECL(expected,int,8,8) [] = { 0x1, 0x2, 0x3, 0x4,
+				       0x5, 0x6, 0x7, 0x8 };
+VECT_VAR_DECL(expected,int,16,4) [] = { 0x12, 0x13, 0x14, 0x15 };
+VECT_VAR_DECL(expected,int,32,2) [] = { 0x23, 0x24 };
+VECT_VAR_DECL(expected,int,64,1) [] = { 0x34 };
+VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff,
+					0xff, 0xff, 0xff, 0xff };
+VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff };
+VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff };
+VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff };
+VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33,
+					0x33, 0x33, 0x33, 0x33 };
+VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 };
+VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 };
+VECT_VAR_DECL(expected,int,8,16) [] = { 0x1, 0x2, 0x3, 0x4,
+					0x5, 0x6, 0x7, 0x8,
+					0x9, 0xa, 0xb, 0xc,
+					0xd, 0xe, 0xf, 0x10 };
+VECT_VAR_DECL(expected,int,16,8) [] = { 0x12, 0x13, 0x14, 0x15,
+					0x16, 0x17, 0x18, 0x19 };
+VECT_VAR_DECL(expected,int,32,4) [] = { 0x23, 0x24, 0x25, 0x26 };
+VECT_VAR_DECL(expected,int,64,2) [] = { 0x34, 0x35 };
+VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff,
+					 0xff, 0xff, 0xff, 0xff,
+					 0xff, 0xff, 0xff, 0xff,
+					 0xff, 0xff, 0xff, 0xff };
+VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff,
+					 0xffff, 0xffff, 0xffff, 0xffff };
+VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff,
+					 0xffffffff, 0xffffffff };
+VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff,
+					 0xffffffffffffffff };
+VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33,
+					 0x33, 0x33, 0x33, 0x33,
+					 0x33, 0x33, 0x33, 0x33,
+					 0x33, 0x33, 0x33, 0x33 };
+VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333,
+					 0x3333, 0x3333, 0x3333, 0x3333 };
+VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333,
+					   0x33333333, 0x33333333 };
+
+
+/* 64-bits types, with 0 as second input.  */
+int VECT_VAR(expected_cumulative_sat_64,int,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64,uint,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64,int,64,2) = 0;
+int VECT_VAR(expected_cumulative_sat_64,uint,64,2) = 0;
+VECT_VAR_DECL(expected_64,int,64,1) [] = { 0xfffffffffffffff0 };
+VECT_VAR_DECL(expected_64,uint,64,1) [] = { 0xfffffffffffffff0 };
+VECT_VAR_DECL(expected_64,int,64,2) [] = { 0xfffffffffffffff0,
+					   0xfffffffffffffff1 };
+VECT_VAR_DECL(expected_64,uint,64,2) [] = { 0xfffffffffffffff0,
+					    0xfffffffffffffff1 };
+
+/* 64-bits types, some cases causing cumulative saturation.  */
+int VECT_VAR(expected_cumulative_sat_64_2,int,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64_2,uint,64,1) = 1;
+int VECT_VAR(expected_cumulative_sat_64_2,int,64,2) = 0;
+int VECT_VAR(expected_cumulative_sat_64_2,uint,64,2) = 1;
+VECT_VAR_DECL(expected_64_2,int,64,1) [] = { 0x34 };
+VECT_VAR_DECL(expected_64_2,uint,64,1) [] = { 0xffffffffffffffff };
+VECT_VAR_DECL(expected_64_2,int,64,2) [] = { 0x34, 0x35 };
+VECT_VAR_DECL(expected_64_2,uint,64,2) [] = { 0xffffffffffffffff,
+					      0xffffffffffffffff };
+
+/* 64-bits types, all causing cumulative saturation.  */
+int VECT_VAR(expected_cumulative_sat_64_3,int,64,1) = 1;
+int VECT_VAR(expected_cumulative_sat_64_3,uint,64,1) = 1;
+int VECT_VAR(expected_cumulative_sat_64_3,int,64,2) = 1;
+int VECT_VAR(expected_cumulative_sat_64_3,uint,64,2) = 1;
+VECT_VAR_DECL(expected_64_3,int,64,1) [] = { 0x8000000000000000 };
+VECT_VAR_DECL(expected_64_3,uint,64,1) [] = { 0xffffffffffffffff };
+VECT_VAR_DECL(expected_64_3,int,64,2) [] = { 0x7fffffffffffffff,
+					     0x7fffffffffffffff };
+VECT_VAR_DECL(expected_64_3,uint,64,2) [] = { 0xffffffffffffffff,
+					      0xffffffffffffffff };
+
+/* smaller types, corner cases causing cumulative saturation. (1)  */
+int VECT_VAR(expected_csat_lt_64_1,int,8,8) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,16,4) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,32,2) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,8,16) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,16,8) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,32,4) = 1;
+VECT_VAR_DECL(expected_lt_64_1,int,8,8) [] = { 0x80, 0x80, 0x80, 0x80,
+					       0x80, 0x80, 0x80, 0x80 };
+VECT_VAR_DECL(expected_lt_64_1,int,16,4) [] = { 0x8000, 0x8000,
+						0x8000, 0x8000 };
+VECT_VAR_DECL(expected_lt_64_1,int,32,2) [] = { 0x80000000, 0x80000000 };
+VECT_VAR_DECL(expected_lt_64_1,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80,
+						0x80, 0x80, 0x80, 0x80,
+						0x80, 0x80, 0x80, 0x80,
+						0x80, 0x80, 0x80, 0x80 };
+VECT_VAR_DECL(expected_lt_64_1,int,16,8) [] = { 0x8000, 0x8000,
+						0x8000, 0x8000,
+						0x8000, 0x8000,
+						0x8000, 0x8000 };
+VECT_VAR_DECL(expected_lt_64_1,int,32,4) [] = { 0x80000000, 0x80000000,
+						0x80000000, 0x80000000 };
+
+/* smaller types, corner cases causing cumulative saturation. (2)  */
+int VECT_VAR(expected_csat_lt_64_2,uint,8,8) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,16,4) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,32,2) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,8,16) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,16,8) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,32,4) = 1;
+VECT_VAR_DECL(expected_lt_64_2,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff,
+						0xff, 0xff, 0xff, 0xff };
+VECT_VAR_DECL(expected_lt_64_2,uint,16,4) [] = { 0xffff, 0xffff,
+						 0xffff, 0xffff };
+VECT_VAR_DECL(expected_lt_64_2,uint,32,2) [] = { 0xffffffff,
+						 0xffffffff };
+VECT_VAR_DECL(expected_lt_64_2,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff,
+						 0xff, 0xff, 0xff, 0xff,
+						 0xff, 0xff, 0xff, 0xff,
+						 0xff, 0xff, 0xff, 0xff };
+VECT_VAR_DECL(expected_lt_64_2,uint,16,8) [] = { 0xffff, 0xffff,
+						 0xffff, 0xffff,
+						 0xffff, 0xffff,
+						 0xffff, 0xffff };
+VECT_VAR_DECL(expected_lt_64_2,uint,32,4) [] = { 0xffffffff, 0xffffffff,
+						 0xffffffff, 0xffffffff };
+
+void vqadd_extras(void)
+{
+  DECL_VARIABLE_ALL_VARIANTS(vector1);
+  DECL_VARIABLE_ALL_VARIANTS(vector2);
+  DECL_VARIABLE_ALL_VARIANTS(vector_res);
+
+  /* Initialize input "vector1" from "buffer".  */
+  TEST_MACRO_ALL_VARIANTS_2_5(VLOAD, vector1, buffer);
+
+  /* Use a second vector full of 0.  */
+  VDUP(vector2, , int, s, 64, 1, 0);
+  VDUP(vector2, , uint, u, 64, 1, 0);
+  VDUP(vector2, q, int, s, 64, 2, 0);
+  VDUP(vector2, q, uint, u, 64, 2, 0);
+
+#define MSG "64 bits saturation adding zero"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat_64, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat_64, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat_64, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat_64, MSG);
+
+  CHECK(TEST_MSG, int, 64, 1, PRIx64, expected_64, MSG);
+  CHECK(TEST_MSG, uint, 64, 1, PRIx64, expected_64, MSG);
+  CHECK(TEST_MSG, int, 64, 2, PRIx64, expected_64, MSG);
+  CHECK(TEST_MSG, uint, 64, 2, PRIx64, expected_64, MSG);
+
+  /* Another set of tests with non-zero values, some chosen to create
+     overflow.  */
+  VDUP(vector2, , int, s, 64, 1, 0x44);
+  VDUP(vector2, , uint, u, 64, 1, 0x88);
+  VDUP(vector2, q, int, s, 64, 2, 0x44);
+  VDUP(vector2, q, uint, u, 64, 2, 0x88);
+
+#undef MSG
+#define MSG "64 bits saturation cumulative_sat (2)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat_64_2, MSG);
+
+  CHECK(TEST_MSG, int, 64, 1, PRIx64, expected_64_2, MSG);
+  CHECK(TEST_MSG, uint, 64, 1, PRIx64, expected_64_2, MSG);
+  CHECK(TEST_MSG, int, 64, 2, PRIx64, expected_64_2, MSG);
+  CHECK(TEST_MSG, uint, 64, 2, PRIx64, expected_64_2, MSG);
+
+  /* Another set of tests, with input values chosen to set
+     cumulative_sat in all cases.  */
+  VDUP(vector2, , int, s, 64, 1, 0x8000000000000003LL);
+  VDUP(vector2, , uint, u, 64, 1, 0x88);
+  /* To check positive saturation, we need to write a positive value
+     in vector1.  */
+  VDUP(vector1, q, int, s, 64, 2, 0x4000000000000000LL);
+  VDUP(vector2, q, int, s, 64, 2, 0x4000000000000000LL);
+  VDUP(vector2, q, uint, u, 64, 2, 0x22);
+
+#undef MSG
+#define MSG "64 bits saturation cumulative_sat (3)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat_64_3, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat_64_3, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat_64_3, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat_64_3, MSG);
+
+  CHECK(TEST_MSG, int, 64, 1, PRIx64, expected_64_3, MSG);
+  CHECK(TEST_MSG, uint, 64, 1, PRIx64, expected_64_3, MSG);
+  CHECK(TEST_MSG, int, 64, 2, PRIx64, expected_64_3, MSG);
+  CHECK(TEST_MSG, uint, 64, 2, PRIx64, expected_64_3, MSG);
+
+  /* To improve coverage, check saturation with less than 64 bits
+     too.  */
+  VDUP(vector2, , int, s, 8, 8, 0x81);
+  VDUP(vector2, , int, s, 16, 4, 0x8001);
+  VDUP(vector2, , int, s, 32, 2, 0x80000001);
+  VDUP(vector2, q, int, s, 8, 16, 0x81);
+  VDUP(vector2, q, int, s, 16, 8, 0x8001);
+  VDUP(vector2, q, int, s, 32, 4, 0x80000001);
+
+#undef MSG
+#define MSG "less than 64 bits saturation cumulative_sat (1)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 8, 8, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 16, 4, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 32, 2, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 8, 16, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 16, 8, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 32, 4, expected_csat_lt_64_1, MSG);
+
+  CHECK(TEST_MSG, int, 8, 8, PRIx8, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 16, 4, PRIx16, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 32, 2, PRIx32, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 8, 16, PRIx8, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 16, 8, PRIx16, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 32, 4, PRIx32, expected_lt_64_1, MSG);
+
+  /* Another set of tests with large vector1 values.  */
+  VDUP(vector1, , uint, u, 8, 8, 0xF0);
+  VDUP(vector1, , uint, u, 16, 4, 0xFFF0);
+  VDUP(vector1, , uint, u, 32, 2, 0xFFFFFFF0);
+  VDUP(vector1, q, uint, u, 8, 16, 0xF0);
+  VDUP(vector1, q, uint, u, 16, 8, 0xFFF0);
+  VDUP(vector1, q, uint, u, 32, 4, 0xFFFFFFF0);
+
+  VDUP(vector2, , uint, u, 8, 8, 0x20);
+  VDUP(vector2, , uint, u, 16, 4, 0x20);
+  VDUP(vector2, , uint, u, 32, 2, 0x20);
+  VDUP(vector2, q, uint, u, 8, 16, 0x20);
+  VDUP(vector2, q, uint, u, 16, 8, 0x20);
+  VDUP(vector2, q, uint, u, 32, 4, 0x20);
+
+#undef MSG
+#define MSG "less than 64 bits saturation cumulative_sat (2)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 8, 8, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 16, 4, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 32, 2, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 8, 16, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 16, 8, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 32, 4, expected_csat_lt_64_2, MSG);
+
+  CHECK(TEST_MSG, uint, 8, 8, PRIx8, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 16, 4, PRIx16, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 32, 2, PRIx32, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 8, 16, PRIx8, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 16, 8, PRIx16, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 32, 4, PRIx32, expected_lt_64_2, MSG);
+}
diff --git a/gcc/testsuite/gcc.target/arm/neon-intrinsics/vqsub.c b/gcc/testsuite/gcc.target/arm/neon-intrinsics/vqsub.c
new file mode 100644
index 0000000..04df5fe
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/neon-intrinsics/vqsub.c
@@ -0,0 +1,278 @@ 
+#define INSN_NAME vqsub
+#define TEST_MSG "VQSUB/VQSUBQ"
+
+/* Extra tests for special cases:
+   - some requiring intermediate types larger than 64 bits to
+   compute saturation flag.
+   - corner case saturations with types smaller than 64 bits.
+*/
+void vqsub_extras(void);
+#define EXTRA_TESTS vqsub_extras
+
+#include "binary_sat_op.inc"
+
+
+/* Expected results.  */
+VECT_VAR_DECL(expected,int,8,8) [] = { 0xdf, 0xe0, 0xe1, 0xe2,
+				       0xe3, 0xe4, 0xe5, 0xe6 };
+VECT_VAR_DECL(expected,int,16,4) [] = { 0xffce, 0xffcf,
+					0xffd0, 0xffd1 };
+VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffbd, 0xffffffbe };
+VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffac };
+VECT_VAR_DECL(expected,uint,8,8) [] = { 0x9b, 0x9c, 0x9d, 0x9e,
+					0x9f, 0xa0, 0xa1, 0xa2 };
+VECT_VAR_DECL(expected,uint,16,4) [] = { 0xff8a, 0xff8b,
+					 0xff8c, 0xff8d };
+VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffff79, 0xffffff7a };
+VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffff68 };
+VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33,
+					0x33, 0x33, 0x33, 0x33 };
+VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 };
+VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 };
+VECT_VAR_DECL(expected,int,8,16) [] = { 0xdf, 0xe0, 0xe1, 0xe2,
+					0xe3, 0xe4, 0xe5, 0xe6,
+					0xe7, 0xe8, 0xe9, 0xea,
+					0xeb, 0xec, 0xed, 0xee };
+VECT_VAR_DECL(expected,int,16,8) [] = { 0xffce, 0xffcf, 0xffd0, 0xffd1,
+					0xffd2, 0xffd3, 0xffd4, 0xffd5 };
+VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffbd, 0xffffffbe,
+					0xffffffbf, 0xffffffc0 };
+VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffac,
+					0xffffffffffffffad };
+VECT_VAR_DECL(expected,uint,8,16) [] = { 0x9b, 0x9c, 0x9d, 0x9e,
+					 0x9f, 0xa0, 0xa1, 0xa2,
+					 0xa3, 0xa4, 0xa5, 0xa6,
+					 0xa7, 0xa8, 0xa9, 0xaa };
+VECT_VAR_DECL(expected,uint,16,8) [] = { 0xff8a, 0xff8b, 0xff8c, 0xff8d,
+					 0xff8e, 0xff8f, 0xff90, 0xff91 };
+VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffff79, 0xffffff7a,
+					 0xffffff7b, 0xffffff7c };
+VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffff68,
+					 0xffffffffffffff69 };
+VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33,
+					 0x33, 0x33, 0x33, 0x33,
+					 0x33, 0x33, 0x33, 0x33,
+					 0x33, 0x33, 0x33, 0x33 };
+VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333,
+					 0x3333, 0x3333, 0x3333, 0x3333 };
+VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333,
+					   0x33333333, 0x33333333 };
+
+/* Expected values of cumulative saturation flag.  */
+int VECT_VAR(expected_cumulative_sat,int,8,8) = 0;
+int VECT_VAR(expected_cumulative_sat,int,16,4) = 0;
+int VECT_VAR(expected_cumulative_sat,int,32,2) = 0;
+int VECT_VAR(expected_cumulative_sat,int,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat,int,8,16) = 0;
+int VECT_VAR(expected_cumulative_sat,int,16,8) = 0;
+int VECT_VAR(expected_cumulative_sat,int,32,4) = 0;
+int VECT_VAR(expected_cumulative_sat,int,64,2) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0;
+int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0;
+
+/* 64-bits types, with 0 as second input.  */
+VECT_VAR_DECL(expected_64,int,64,1) [] = { 0xfffffffffffffff0 };
+VECT_VAR_DECL(expected_64,uint,64,1) [] = { 0xfffffffffffffff0 };
+VECT_VAR_DECL(expected_64,int,64,2) [] = { 0xfffffffffffffff0,
+					   0xfffffffffffffff1 };
+VECT_VAR_DECL(expected_64,uint,64,2) [] = { 0xfffffffffffffff0,
+					    0xfffffffffffffff1 };
+int VECT_VAR(expected_cumulative_sat_64,int,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64,uint,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64,int,64,2) = 0;
+int VECT_VAR(expected_cumulative_sat_64,uint,64,2) = 0;
+
+/* 64-bits types, other cases.  */
+VECT_VAR_DECL(expected_64_2,int,64,1) [] = { 0xffffffffffffffac };
+VECT_VAR_DECL(expected_64_2,uint,64,1) [] = { 0xffffffffffffff68 };
+VECT_VAR_DECL(expected_64_2,int,64,2) [] = { 0xffffffffffffffac,
+					     0xffffffffffffffad };
+VECT_VAR_DECL(expected_64_2,uint,64,2) [] = { 0xffffffffffffff68,
+					      0xffffffffffffff69 };
+int VECT_VAR(expected_cumulative_sat_64_2,int,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64_2,uint,64,1) = 0;
+int VECT_VAR(expected_cumulative_sat_64_2,int,64,2) = 0;
+int VECT_VAR(expected_cumulative_sat_64_2,uint,64,2) = 0;
+
+/* 64-bits types, all causing cumulative saturation.  */
+VECT_VAR_DECL(expected_64_3,int,64,1) [] = { 0x8000000000000000 };
+VECT_VAR_DECL(expected_64_3,uint,64,1) [] = { 0x0 };
+VECT_VAR_DECL(expected_64_3,int,64,2) [] = { 0x7fffffffffffffff,
+					     0x7fffffffffffffff };
+VECT_VAR_DECL(expected_64_3,uint,64,2) [] = { 0x0, 0x0 };
+int VECT_VAR(expected_cumulative_sat_64_3,int,64,1) = 1;
+int VECT_VAR(expected_cumulative_sat_64_3,uint,64,1) = 1;
+int VECT_VAR(expected_cumulative_sat_64_3,int,64,2) = 1;
+int VECT_VAR(expected_cumulative_sat_64_3,uint,64,2) = 1;
+
+/* smaller types, corner cases causing cumulative saturation. (1)  */
+VECT_VAR_DECL(expected_lt_64_1,int,8,8) [] = { 0x80, 0x80, 0x80, 0x80,
+					       0x80, 0x80, 0x80, 0x80 };
+VECT_VAR_DECL(expected_lt_64_1,int,16,4) [] = { 0x8000, 0x8000,
+						0x8000, 0x8000 };
+VECT_VAR_DECL(expected_lt_64_1,int,32,2) [] = { 0x80000000, 0x80000000 };
+VECT_VAR_DECL(expected_lt_64_1,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80,
+						0x80, 0x80, 0x80, 0x80,
+						0x80, 0x80, 0x80, 0x80,
+						0x80, 0x80, 0x80, 0x80 };
+VECT_VAR_DECL(expected_lt_64_1,int,16,8) [] = { 0x8000, 0x8000,
+						0x8000, 0x8000,
+						0x8000, 0x8000,
+						0x8000, 0x8000 };
+VECT_VAR_DECL(expected_lt_64_1,int,32,4) [] = { 0x80000000, 0x80000000,
+						0x80000000, 0x80000000 };
+int VECT_VAR(expected_csat_lt_64_1,int,8,8) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,16,4) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,32,2) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,8,16) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,16,8) = 1;
+int VECT_VAR(expected_csat_lt_64_1,int,32,4) = 1;
+
+/* smaller types, corner cases causing cumulative saturation. (2)  */
+VECT_VAR_DECL(expected_lt_64_2,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0,
+						0x0, 0x0, 0x0, 0x0 };
+VECT_VAR_DECL(expected_lt_64_2,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 };
+VECT_VAR_DECL(expected_lt_64_2,uint,32,2) [] = { 0x0, 0x0 };
+VECT_VAR_DECL(expected_lt_64_2,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0,
+						 0x0, 0x0, 0x0, 0x0,
+						 0x0, 0x0, 0x0, 0x0,
+						 0x0, 0x0, 0x0, 0x0 };
+VECT_VAR_DECL(expected_lt_64_2,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0,
+						 0x0, 0x0, 0x0, 0x0 };
+VECT_VAR_DECL(expected_lt_64_2,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 };
+int VECT_VAR(expected_csat_lt_64_2,uint,8,8) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,16,4) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,32,2) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,8,16) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,16,8) = 1;
+int VECT_VAR(expected_csat_lt_64_2,uint,32,4) = 1;
+
+void vqsub_extras(void)
+{
+  DECL_VARIABLE_ALL_VARIANTS(vector1);
+  DECL_VARIABLE_ALL_VARIANTS(vector2);
+  DECL_VARIABLE_ALL_VARIANTS(vector_res);
+
+  /* Initialize input "vector1" from "buffer".  */
+  TEST_MACRO_ALL_VARIANTS_2_5(VLOAD, vector1, buffer);
+
+  /* Use a second vector full of 0.  */
+  VDUP(vector2, , int, s, 64, 1, 0x0);
+  VDUP(vector2, , uint, u, 64, 1, 0x0);
+  VDUP(vector2, q, int, s, 64, 2, 0x0);
+  VDUP(vector2, q, uint, u, 64, 2, 0x0);
+
+#define MSG "64 bits saturation when adding zero"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat_64, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat_64, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat_64, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat_64, MSG);
+
+  CHECK(TEST_MSG, int, 64, 1, PRIx64, expected_64, MSG);
+  CHECK(TEST_MSG, uint, 64, 1, PRIx64, expected_64, MSG);
+  CHECK(TEST_MSG, int, 64, 2, PRIx64, expected_64, MSG);
+  CHECK(TEST_MSG, uint, 64, 2, PRIx64, expected_64, MSG);
+
+  /* Another set of tests with non-zero values.  */
+  VDUP(vector2, , int, s, 64, 1, 0x44);
+  VDUP(vector2, , uint, u, 64, 1, 0x88);
+  VDUP(vector2, q, int, s, 64, 2, 0x44);
+  VDUP(vector2, q, uint, u, 64, 2, 0x88);
+
+#undef MSG
+#define MSG "64 bits saturation cumulative_sat (2)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat_64_2, MSG);
+
+  CHECK(TEST_MSG, int, 64, 1, PRIx64, expected_64_2, MSG);
+  CHECK(TEST_MSG, uint, 64, 1, PRIx64, expected_64_2, MSG);
+  CHECK(TEST_MSG, int, 64, 2, PRIx64, expected_64_2, MSG);
+  CHECK(TEST_MSG, uint, 64, 2, PRIx64, expected_64_2, MSG);
+
+  /* Another set of tests, with input values chosen to set
+     cumulative_sat in all cases.  */
+  VDUP(vector2, , int, s, 64, 1, 0x7fffffffffffffffLL);
+  VDUP(vector2, , uint, u, 64, 1, 0xffffffffffffffffULL);
+  /* To check positive saturation, we need to write a positive value
+     in vector1.  */
+  VDUP(vector1, q, int, s, 64, 2, 0x3fffffffffffffffLL);
+  VDUP(vector2, q, int, s, 64, 2, 0x8000000000000000LL);
+  VDUP(vector2, q, uint, u, 64, 2, 0xffffffffffffffffULL);
+
+#undef MSG
+#define MSG "64 bits saturation cumulative_sat (3)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 64, 1, expected_cumulative_sat_64_3, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 64, 1, expected_cumulative_sat_64_3, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 64, 2, expected_cumulative_sat_64_3, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 64, 2, expected_cumulative_sat_64_3, MSG);
+
+  CHECK(TEST_MSG, int, 64, 1, PRIx64, expected_64_3, MSG);
+  CHECK(TEST_MSG, uint, 64, 1, PRIx64, expected_64_3, MSG);
+  CHECK(TEST_MSG, int, 64, 2, PRIx64, expected_64_3, MSG);
+  CHECK(TEST_MSG, uint, 64, 2, PRIx64, expected_64_3, MSG);
+
+  /* To improve coverage, check saturation with less than 64 bits
+     too.  */
+  VDUP(vector2, , int, s, 8, 8, 0x7F);
+  VDUP(vector2, , int, s, 16, 4, 0x7FFF);
+  VDUP(vector2, , int, s, 32, 2, 0x7FFFFFFF);
+  VDUP(vector2, q, int, s, 8, 16, 0x7F);
+  VDUP(vector2, q, int, s, 16, 8, 0x7FFF);
+  VDUP(vector2, q, int, s, 32, 4, 0x7FFFFFFF);
+
+#undef MSG
+#define MSG "less than 64 bits saturation cumulative_sat (1)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 8, 8, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 16, 4, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , int, s, 32, 2, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 8, 16, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 16, 8, expected_csat_lt_64_1, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, int, s, 32, 4, expected_csat_lt_64_1, MSG);
+
+  CHECK(TEST_MSG, int, 8, 8, PRIx8, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 16, 4, PRIx16, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 32, 2, PRIx32, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 8, 16, PRIx8, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 16, 8, PRIx16, expected_lt_64_1, MSG);
+  CHECK(TEST_MSG, int, 32, 4, PRIx32, expected_lt_64_1, MSG);
+
+  /* Another set of tests with vector1 values smaller than
+     vector2.  */
+  VDUP(vector1, , uint, u, 8, 8, 0x10);
+  VDUP(vector1, , uint, u, 16, 4, 0x10);
+  VDUP(vector1, , uint, u, 32, 2, 0x10);
+  VDUP(vector1, q, uint, u, 8, 16, 0x10);
+  VDUP(vector1, q, uint, u, 16, 8, 0x10);
+  VDUP(vector1, q, uint, u, 32, 4, 0x10);
+
+  VDUP(vector2, , uint, u, 8, 8, 0x20);
+  VDUP(vector2, , uint, u, 16, 4, 0x20);
+  VDUP(vector2, , uint, u, 32, 2, 0x20);
+  VDUP(vector2, q, uint, u, 8, 16, 0x20);
+  VDUP(vector2, q, uint, u, 16, 8, 0x20);
+  VDUP(vector2, q, uint, u, 32, 4, 0x20);
+
+#undef MSG
+#define MSG "less than 64 bits saturation cumulative_sat (2)"
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 8, 8, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 16, 4, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, , uint, u, 32, 2, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 8, 16, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 16, 8, expected_csat_lt_64_2, MSG);
+  TEST_BINARY_SAT_OP(INSN_NAME, q, uint, u, 32, 4, expected_csat_lt_64_2, MSG);
+
+  CHECK(TEST_MSG, uint, 8, 8, PRIx8, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 16, 4, PRIx16, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 32, 2, PRIx32, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 8, 16, PRIx8, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 16, 8, PRIx16, expected_lt_64_2, MSG);
+  CHECK(TEST_MSG, uint, 32, 4, PRIx32, expected_lt_64_2, MSG);
+}