This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[AArch64][PATCH 7/7] Add NEON intrinsics vqrdmlah_lane and vqrdmlsh_lane.
- From: Matthew Wahab <matthew dot wahab at foss dot arm dot com>
- To: gcc-patches at gcc dot gnu dot org
- Date: Fri, 23 Oct 2015 13:30:46 +0100
- Subject: [AArch64][PATCH 7/7] Add NEON intrinsics vqrdmlah_lane and vqrdmlsh_lane.
- Authentication-results: sourceware.org; auth=none
- References: <562A2519 dot 3020102 at foss dot arm dot com>
The ARMv8.1 architecture extension adds two Adv.SIMD instructions,
sqrdmlah and sqrdmlsh. This patch adds the NEON intrinsics vqrdmlah_lane
and vqrdmlsh_lane for these instructions. The new intrinsics are of the
form vqrdml{as}h[q]_lane_<type>.
Tested the series for aarch64-none-linux-gnu with native bootstrap and
make check on an ARMv8 architecture. Also tested aarch64-none-elf with
cross-compiled check-gcc on an ARMv8.1 emulator.
Ok for trunk?
Matthew
gcc/
2015-10-23 Matthew Wahab <matthew.wahab@arm.com>
* gcc/config/aarch64/arm_neon.h
(vqrdmlah_laneq_s16, vqrdmlah_laneq_s32): New.
(vqrdmlahq_laneq_s16, vqrdmlahq_laneq_s32): New.
(vqrdmlsh_laneq_s16, vqrdmlsh_laneq_s32): New.
(vqrdmlshq_laneq_s16, vqrdmlshq_laneq_s32): New.
(vqrdmlah_lane_s16, vqrdmlah_lane_s32): New.
(vqrdmlahq_lane_s16, vqrdmlahq_lane_s32): New.
(vqrdmlahh_s16, vqrdmlahh_lane_s16, vqrdmlahh_laneq_s16): New.
(vqrdmlahs_s32, vqrdmlahs_lane_s32, vqrdmlahs_laneq_s32): New.
(vqrdmlsh_lane_s16, vqrdmlsh_lane_s32): New.
(vqrdmlshq_lane_s16, vqrdmlshq_lane_s32): New.
(vqrdmlshh_s16, vqrdmlshh_lane_s16, vqrdmlshh_laneq_s16): New.
(vqrdmlshs_s32, vqrdmlshs_lane_s32, vqrdmlshs_laneq_s32): New.
gcc/testsuite
2015-10-23 Matthew Wahab <matthew.wahab@arm.com>
* gcc.target/aarch64/advsimd-intrinsics/vqrdmlXh_lane.inc: New file,
support code for vqrdml{as}h_lane tests.
* gcc.target/aarch64/advsimd-intrinsics/vqrdmlah_lane.c: New.
* gcc.target/aarch64/advsimd-intrinsics/vqrdmlsh_lane.c: New.
>From a2399818dba85ff2801a28bad77ef51697990da7 Mon Sep 17 00:00:00 2001
From: Matthew Wahab <matthew.wahab@arm.com>
Date: Thu, 27 Aug 2015 14:17:26 +0100
Subject: [PATCH 7/7] Add neon intrinsics: vqrdmlah_lane, vqrdmlsh_lane.
Change-Id: I6d7a372e0a5b83ef0846ab62abbe9b24ada69fc4
---
gcc/config/aarch64/arm_neon.h | 182 +++++++++++++++++++++
.../aarch64/advsimd-intrinsics/vqrdmlXh_lane.inc | 154 +++++++++++++++++
.../aarch64/advsimd-intrinsics/vqrdmlah_lane.c | 57 +++++++
.../aarch64/advsimd-intrinsics/vqrdmlsh_lane.c | 61 +++++++
4 files changed, 454 insertions(+)
create mode 100644 gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlXh_lane.inc
create mode 100644 gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlah_lane.c
create mode 100644 gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlsh_lane.c
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 9e73809..9b68e4a 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -10675,6 +10675,59 @@ vqrdmulhq_laneq_s32 (int32x4_t __a, int32x4_t __b, const int __c)
return __builtin_aarch64_sqrdmulh_laneqv4si (__a, __b, __c);
}
+#pragma GCC push_options
+#pragma GCC target ("arch=armv8.1-a")
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vqrdmlah_laneq_s16 (int16x4_t __a, int16x4_t __b, int16x8_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_laneqv4hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vqrdmlah_laneq_s32 (int32x2_t __a, int32x2_t __b, int32x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_laneqv2si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
+vqrdmlahq_laneq_s16 (int16x8_t __a, int16x8_t __b, int16x8_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_laneqv8hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
+vqrdmlahq_laneq_s32 (int32x4_t __a, int32x4_t __b, int32x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_laneqv4si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vqrdmlsh_laneq_s16 (int16x4_t __a, int16x4_t __b, int16x8_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_laneqv4hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vqrdmlsh_laneq_s32 (int32x2_t __a, int32x2_t __b, int32x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_laneqv2si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
+vqrdmlshq_laneq_s16 (int16x8_t __a, int16x8_t __b, int16x8_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_laneqv8hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
+vqrdmlshq_laneq_s32 (int32x4_t __a, int32x4_t __b, int32x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_laneqv4si (__a, __b, __c, __d);
+}
+
+#pragma GCC pop_options
+
/* Table intrinsics. */
__extension__ static __inline poly8x8_t __attribute__ ((__always_inline__))
@@ -20014,6 +20067,135 @@ vqrdmulhs_laneq_s32 (int32_t __a, int32x4_t __b, const int __c)
return __builtin_aarch64_sqrdmulh_laneqsi (__a, __b, __c);
}
+#pragma GCC push_options
+#pragma GCC target ("arch=armv8.1-a")
+
+/* vqrdmlah. */
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vqrdmlah_lane_s16 (int16x4_t __a, int16x4_t __b, int16x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_lanev4hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vqrdmlah_lane_s32 (int32x2_t __a, int32x2_t __b, int32x2_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_lanev2si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
+vqrdmlahq_lane_s16 (int16x8_t __a, int16x8_t __b, int16x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_lanev8hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
+vqrdmlahq_lane_s32 (int32x4_t __a, int32x4_t __b, int32x2_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_lanev4si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16_t __attribute__ ((__always_inline__))
+vqrdmlahh_s16 (int16_t __a, int16_t __b, int16_t __c)
+{
+ return (int16_t) __builtin_aarch64_sqrdmlahhi (__a, __b, __c);
+}
+
+__extension__ static __inline int16_t __attribute__ ((__always_inline__))
+vqrdmlahh_lane_s16 (int16_t __a, int16_t __b, int16x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_lanehi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16_t __attribute__ ((__always_inline__))
+vqrdmlahh_laneq_s16 (int16_t __a, int16_t __b, int16x8_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_laneqhi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32_t __attribute__ ((__always_inline__))
+vqrdmlahs_s32 (int32_t __a, int32_t __b, int32_t __c)
+{
+ return (int32_t) __builtin_aarch64_sqrdmlahsi (__a, __b, __c);
+}
+
+__extension__ static __inline int32_t __attribute__ ((__always_inline__))
+vqrdmlahs_lane_s32 (int32_t __a, int32_t __b, int32x2_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_lanesi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32_t __attribute__ ((__always_inline__))
+vqrdmlahs_laneq_s32 (int32_t __a, int32_t __b, int32x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlah_laneqsi (__a, __b, __c, __d);
+}
+
+/* vqrdmlsh. */
+
+__extension__ static __inline int16x4_t __attribute__ ((__always_inline__))
+vqrdmlsh_lane_s16 (int16x4_t __a, int16x4_t __b, int16x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_lanev4hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x2_t __attribute__ ((__always_inline__))
+vqrdmlsh_lane_s32 (int32x2_t __a, int32x2_t __b, int32x2_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_lanev2si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16x8_t __attribute__ ((__always_inline__))
+vqrdmlshq_lane_s16 (int16x8_t __a, int16x8_t __b, int16x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_lanev8hi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32x4_t __attribute__ ((__always_inline__))
+vqrdmlshq_lane_s32 (int32x4_t __a, int32x4_t __b, int32x2_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_lanev4si (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16_t __attribute__ ((__always_inline__))
+vqrdmlshh_s16 (int16_t __a, int16_t __b, int16_t __c)
+{
+ return (int16_t) __builtin_aarch64_sqrdmlshhi (__a, __b, __c);
+}
+
+__extension__ static __inline int16_t __attribute__ ((__always_inline__))
+vqrdmlshh_lane_s16 (int16_t __a, int16_t __b, int16x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_lanehi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int16_t __attribute__ ((__always_inline__))
+vqrdmlshh_laneq_s16 (int16_t __a, int16_t __b, int16x8_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_laneqhi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32_t __attribute__ ((__always_inline__))
+vqrdmlshs_s32 (int32_t __a, int32_t __b, int32_t __c)
+{
+ return (int32_t) __builtin_aarch64_sqrdmlshsi (__a, __b, __c);
+}
+
+__extension__ static __inline int32_t __attribute__ ((__always_inline__))
+vqrdmlshs_lane_s32 (int32_t __a, int32_t __b, int32x2_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_lanesi (__a, __b, __c, __d);
+}
+
+__extension__ static __inline int32_t __attribute__ ((__always_inline__))
+vqrdmlshs_laneq_s32 (int32_t __a, int32_t __b, int32x4_t __c, const int __d)
+{
+ return __builtin_aarch64_sqrdmlsh_laneqsi (__a, __b, __c, __d);
+}
+
+#pragma GCC pop_options
+
/* vqrshl */
__extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
diff --git a/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlXh_lane.inc b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlXh_lane.inc
new file mode 100644
index 0000000..a855502
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlXh_lane.inc
@@ -0,0 +1,154 @@
+#define FNNAME1(NAME) exec_ ## NAME ## _lane
+#define FNNAME(NAME) FNNAME1 (NAME)
+
+void FNNAME (INSN) (void)
+{
+ /* vector_res = vqrdmlXh_lane (vector, vector2, vector3, lane),
+ then store the result. */
+#define TEST_VQRDMLXH_LANE2(INSN, Q, T1, T2, W, N, N2, L, \
+ EXPECTED_CUMULATIVE_SAT, CMT) \
+ Set_Neon_Cumulative_Sat (0, VECT_VAR (vector_res, T1, W, N)); \
+ VECT_VAR (vector_res, T1, W, N) = \
+ INSN##Q##_lane_##T2##W (VECT_VAR (vector, T1, W, N), \
+ VECT_VAR (vector2, T1, W, N), \
+ VECT_VAR (vector3, T1, W, N2), \
+ L); \
+ 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)
+
+ /* Two auxliary macros are necessary to expand INSN. */
+#define TEST_VQRDMLXH_LANE1(INSN, Q, T1, T2, W, N, N2, L, \
+ EXPECTED_CUMULATIVE_SAT, CMT) \
+ TEST_VQRDMLXH_LANE2 (INSN, Q, T1, T2, W, N, N2, L, \
+ EXPECTED_CUMULATIVE_SAT, CMT)
+
+#define TEST_VQRDMLXH_LANE(Q, T1, T2, W, N, N2, L, \
+ EXPECTED_CUMULATIVE_SAT, CMT) \
+ TEST_VQRDMLXH_LANE1 (INSN, Q, T1, T2, W, N, N2, L, \
+ EXPECTED_CUMULATIVE_SAT, CMT)
+
+
+ DECL_VARIABLE (vector, int, 16, 4);
+ DECL_VARIABLE (vector, int, 32, 2);
+ DECL_VARIABLE (vector, int, 16, 8);
+ DECL_VARIABLE (vector, int, 32, 4);
+
+ DECL_VARIABLE (vector_res, int, 16, 4);
+ DECL_VARIABLE (vector_res, int, 32, 2);
+ DECL_VARIABLE (vector_res, int, 16, 8);
+ DECL_VARIABLE (vector_res, int, 32, 4);
+
+ DECL_VARIABLE (vector2, int, 16, 4);
+ DECL_VARIABLE (vector2, int, 32, 2);
+ DECL_VARIABLE (vector2, int, 16, 8);
+ DECL_VARIABLE (vector2, int, 32, 4);
+
+ DECL_VARIABLE (vector3, int, 16, 4);
+ DECL_VARIABLE (vector3, int, 32, 2);
+ DECL_VARIABLE (vector3, int, 16, 8);
+ DECL_VARIABLE (vector3, int, 32, 4);
+
+ clean_results ();
+
+ VLOAD (vector, buffer, , int, s, 16, 4);
+ VLOAD (vector, buffer, , int, s, 32, 2);
+
+ VLOAD (vector, buffer, q, int, s, 16, 8);
+ VLOAD (vector, buffer, q, int, s, 32, 4);
+
+ /* Initialize vector2. */
+ VDUP (vector2, , int, s, 16, 4, 0x5555);
+ VDUP (vector2, , int, s, 32, 2, 0xBB);
+ VDUP (vector2, q, int, s, 16, 8, 0xBB);
+ VDUP (vector2, q, int, s, 32, 4, 0x22);
+
+ /* Initialize vector3. */
+ VDUP (vector3, , int, s, 16, 4, 0x5555);
+ VDUP (vector3, , int, s, 32, 2, 0xBB);
+ VDUP (vector3, q, int, s, 16, 8, 0x33);
+ VDUP (vector3, q, int, s, 32, 4, 0x22);
+
+ /* Choose lane arbitrarily. */
+#define CMT ""
+ TEST_VQRDMLXH_LANE (, int, s, 16, 4, 4, 2, expected_cumulative_sat, CMT);
+ TEST_VQRDMLXH_LANE (, int, s, 32, 2, 2, 1, expected_cumulative_sat, CMT);
+ TEST_VQRDMLXH_LANE (q, int, s, 16, 8, 4, 3, expected_cumulative_sat, CMT);
+ TEST_VQRDMLXH_LANE (q, int, s, 32, 4, 2, 0, expected_cumulative_sat, CMT);
+
+ CHECK (TEST_MSG, int, 16, 4, PRIx16, expected, CMT);
+ CHECK (TEST_MSG, int, 32, 2, PRIx32, expected, CMT);
+ CHECK (TEST_MSG, int, 16, 8, PRIx16, expected, CMT);
+ CHECK (TEST_MSG, int, 32, 4, PRIx32, expected, CMT);
+
+ /* Now use input values such that the multiplication causes
+ saturation. */
+#define TEST_MSG_MUL " (check mul cumulative saturation)"
+ VDUP (vector, , int, s, 16, 4, 0x8000);
+ VDUP (vector, , int, s, 32, 2, 0x80000000);
+ VDUP (vector, q, int, s, 16, 8, 0x8000);
+ VDUP (vector, q, int, s, 32, 4, 0x80000000);
+
+ VDUP (vector2, , int, s, 16, 4, 0x8000);
+ VDUP (vector2, , int, s, 32, 2, 0x80000000);
+ VDUP (vector2, q, int, s, 16, 8, 0x8000);
+ VDUP (vector2, q, int, s, 32, 4, 0x80000000);
+
+ VDUP (vector3, , int, s, 16, 4, 0x8000);
+ VDUP (vector3, , int, s, 32, 2, 0x80000000);
+ VDUP (vector3, q, int, s, 16, 8, 0x8000);
+ VDUP (vector3, q, int, s, 32, 4, 0x80000000);
+
+ TEST_VQRDMLXH_LANE (, int, s, 16, 4, 4, 2, expected_cumulative_sat_mul,
+ TEST_MSG_MUL);
+ TEST_VQRDMLXH_LANE (, int, s, 32, 2, 2, 1, expected_cumulative_sat_mul,
+ TEST_MSG_MUL);
+ TEST_VQRDMLXH_LANE (q, int, s, 16, 8, 4, 3, expected_cumulative_sat_mul,
+ TEST_MSG_MUL);
+ TEST_VQRDMLXH_LANE (q, int, s, 32, 4, 2, 0, expected_cumulative_sat_mul,
+ TEST_MSG_MUL);
+
+ CHECK (TEST_MSG, int, 16, 4, PRIx16, expected_mul, TEST_MSG_MUL);
+ CHECK (TEST_MSG, int, 32, 2, PRIx32, expected_mul, TEST_MSG_MUL);
+ CHECK (TEST_MSG, int, 16, 8, PRIx16, expected_mul, TEST_MSG_MUL);
+ CHECK (TEST_MSG, int, 32, 4, PRIx32, expected_mul, TEST_MSG_MUL);
+
+ VDUP (vector, , int, s, 16, 4, 0x8000);
+ VDUP (vector, , int, s, 32, 2, 0x80000000);
+ VDUP (vector, q, int, s, 16, 8, 0x8000);
+ VDUP (vector, q, int, s, 32, 4, 0x80000000);
+
+ VDUP (vector2, , int, s, 16, 4, 0x8001);
+ VDUP (vector2, , int, s, 32, 2, 0x80000001);
+ VDUP (vector2, q, int, s, 16, 8, 0x8001);
+ VDUP (vector2, q, int, s, 32, 4, 0x80000001);
+
+ VDUP (vector3, , int, s, 16, 4, 0x8001);
+ VDUP (vector3, , int, s, 32, 2, 0x80000001);
+ VDUP (vector3, q, int, s, 16, 8, 0x8001);
+ VDUP (vector3, q, int, s, 32, 4, 0x80000001);
+
+ /* Use input values where rounding produces a result equal to the
+ saturation value, but does not set the saturation flag. */
+#define TEST_MSG_ROUND " (check rounding)"
+ TEST_VQRDMLXH_LANE (, int, s, 16, 4, 4, 2, expected_cumulative_sat_round,
+ TEST_MSG_ROUND);
+ TEST_VQRDMLXH_LANE (, int, s, 32, 2, 2, 1, expected_cumulative_sat_round,
+ TEST_MSG_ROUND);
+ TEST_VQRDMLXH_LANE (q, int, s, 16, 8, 4, 3, expected_cumulative_sat_round,
+ TEST_MSG_ROUND);
+ TEST_VQRDMLXH_LANE (q, int, s, 32, 4, 2, 0, expected_cumulative_sat_round,
+ TEST_MSG_ROUND);
+
+ CHECK (TEST_MSG, int, 16, 4, PRIx16, expected_round, TEST_MSG_ROUND);
+ CHECK (TEST_MSG, int, 32, 2, PRIx32, expected_round, TEST_MSG_ROUND);
+ CHECK (TEST_MSG, int, 16, 8, PRIx16, expected_round, TEST_MSG_ROUND);
+ CHECK (TEST_MSG, int, 32, 4, PRIx32, expected_round, TEST_MSG_ROUND);
+}
+
+int
+main (void)
+{
+ FNNAME (INSN) ();
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlah_lane.c b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlah_lane.c
new file mode 100644
index 0000000..ed43e01
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlah_lane.c
@@ -0,0 +1,57 @@
+/* { dg-require-effective-target arm_v8_1a_neon_hw } */
+/* { dg-add-options arm_v8_1a_neon } */
+
+#include <arm_neon.h>
+#include "arm-neon-ref.h"
+#include "compute-ref-data.h"
+
+/* Expected values of cumulative_saturation flag. */
+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, 16, 8) = 0;
+int VECT_VAR (expected_cumulative_sat, int, 32, 4) = 0;
+
+/* Expected results. */
+VECT_VAR_DECL (expected, int, 16, 4) [] = { 0x38d3, 0x38d4, 0x38d5, 0x38d6 };
+VECT_VAR_DECL (expected, int, 32, 2) [] = { 0xfffffff0, 0xfffffff1 };
+VECT_VAR_DECL (expected, int, 16, 8) [] = { 0x006d, 0x006e, 0x006f, 0x0070,
+ 0x0071, 0x0072, 0x0073, 0x0074 };
+VECT_VAR_DECL (expected, int, 32, 4) [] = { 0xfffffff0, 0xfffffff1,
+ 0xfffffff2, 0xfffffff3 };
+
+/* Expected values of cumulative_saturation flag when multiplication
+ saturates. */
+int VECT_VAR (expected_cumulative_sat_mul, int, 16, 4) = 0;
+int VECT_VAR (expected_cumulative_sat_mul, int, 32, 2) = 0;
+int VECT_VAR (expected_cumulative_sat_mul, int, 16, 8) = 0;
+int VECT_VAR (expected_cumulative_sat_mul, int, 32, 4) = 0;
+
+/* Expected results when multiplication saturates. */
+VECT_VAR_DECL (expected_mul, int, 16, 4) [] = { 0x0, 0x0, 0x0, 0x0 };
+VECT_VAR_DECL (expected_mul, int, 32, 2) [] = { 0x0, 0x0 };
+VECT_VAR_DECL (expected_mul, int, 16, 8) [] = { 0x0, 0x0, 0x0, 0x0,
+ 0x0, 0x0, 0x0, 0x0 };
+VECT_VAR_DECL (expected_mul, int, 32, 4) [] = { 0x0, 0x0, 0x0, 0x0 };
+
+/* Expected values of cumulative_saturation flag when rounding
+ should not cause saturation. */
+int VECT_VAR (expected_cumulative_sat_round, int, 16, 4) = 0;
+int VECT_VAR (expected_cumulative_sat_round, int, 32, 2) = 0;
+int VECT_VAR (expected_cumulative_sat_round, int, 16, 8) = 0;
+int VECT_VAR (expected_cumulative_sat_round, int, 32, 4) = 0;
+
+/* Expected results when rounding should not cause saturation. */
+VECT_VAR_DECL (expected_round, int, 16, 4) [] = { 0xfffe, 0xfffe,
+ 0xfffe, 0xfffe };
+VECT_VAR_DECL (expected_round, int, 32, 2) [] = { 0xfffffffe, 0xfffffffe };
+VECT_VAR_DECL (expected_round,int, 16, 8) [] = { 0xfffe, 0xfffe,
+ 0xfffe, 0xfffe,
+ 0xfffe, 0xfffe,
+ 0xfffe, 0xfffe };
+VECT_VAR_DECL (expected_round, int, 32, 4) [] = { 0xfffffffe, 0xfffffffe,
+ 0xfffffffe, 0xfffffffe };
+
+#define INSN vqrdmlah
+#define TEST_MSG "VQRDMLAH_LANE"
+
+#include "vqrdmlXh_lane.inc"
diff --git a/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlsh_lane.c b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlsh_lane.c
new file mode 100644
index 0000000..6010b42
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/advsimd-intrinsics/vqrdmlsh_lane.c
@@ -0,0 +1,61 @@
+/* { dg-require-effective-target arm_v8_1a_neon_hw } */
+/* { dg-add-options arm_v8_1a_neon } */
+
+#include <arm_neon.h>
+#include "arm-neon-ref.h"
+#include "compute-ref-data.h"
+
+/* Expected values of cumulative_saturation flag. */
+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, 16, 8) = 0;
+int VECT_VAR (expected_cumulative_sat, int, 32, 4) = 0;
+
+/* Expected results. */
+VECT_VAR_DECL (expected, int, 16, 4) [] = { 0xc70d, 0xc70e, 0xc70f, 0xc710 };
+VECT_VAR_DECL (expected, int, 32, 2) [] = { 0xfffffff0, 0xfffffff1 };
+VECT_VAR_DECL (expected, int, 16, 8) [] = { 0xff73, 0xff74, 0xff75, 0xff76,
+ 0xff77, 0xff78, 0xff79, 0xff7a };
+VECT_VAR_DECL (expected, int, 32, 4) [] = { 0xfffffff0, 0xfffffff1,
+ 0xfffffff2, 0xfffffff3 };
+
+/* Expected values of cumulative_saturation flag when multiplication
+ saturates. */
+int VECT_VAR (expected_cumulative_sat_mul, int, 16, 4) = 1;
+int VECT_VAR (expected_cumulative_sat_mul, int, 32, 2) = 1;
+int VECT_VAR (expected_cumulative_sat_mul, int, 16, 8) = 1;
+int VECT_VAR (expected_cumulative_sat_mul, int, 32, 4) = 1;
+
+/* Expected results when multiplication saturates. */
+VECT_VAR_DECL (expected_mul, int, 16, 4) [] = { 0x8000, 0x8000,
+ 0x8000, 0x8000 };
+VECT_VAR_DECL (expected_mul, int, 32, 2) [] = { 0x80000000, 0x80000000 };
+VECT_VAR_DECL (expected_mul, int, 16, 8) [] = { 0x8000, 0x8000,
+ 0x8000, 0x8000,
+ 0x8000, 0x8000,
+ 0x8000, 0x8000 };
+VECT_VAR_DECL (expected_mul, int, 32, 4) [] = { 0x80000000, 0x80000000,
+ 0x80000000, 0x80000000 };
+
+/* Expected values of cumulative_saturation flag when rounding
+ should not cause saturation. */
+int VECT_VAR (expected_cumulative_sat_round, int, 16, 4) = 1;
+int VECT_VAR (expected_cumulative_sat_round, int, 32, 2) = 1;
+int VECT_VAR (expected_cumulative_sat_round, int, 16, 8) = 1;
+int VECT_VAR (expected_cumulative_sat_round, int, 32, 4) = 1;
+
+/* Expected results when rounding should not cause saturation. */
+VECT_VAR_DECL (expected_round, int, 16, 4) [] = { 0x8000, 0x8000,
+ 0x8000, 0x8000 };
+VECT_VAR_DECL (expected_round, int, 32, 2) [] = { 0x80000000, 0x80000000 };
+VECT_VAR_DECL (expected_round, int, 16, 8) [] = { 0x8000, 0x8000,
+ 0x8000, 0x8000,
+ 0x8000, 0x8000,
+ 0x8000, 0x8000 };
+VECT_VAR_DECL (expected_round, int, 32, 4) [] = { 0x80000000, 0x80000000,
+ 0x80000000, 0x80000000 };
+
+#define INSN vqrdmlsh
+#define TEST_MSG "VQRDMLSH_LANE"
+
+#include "vqrdmlXh_lane.inc"
--
2.1.4