NEON: more fp16 using intrinsics supported by architecture v7 (skip version)

meson.build

@@ -10,6 +10,8 @@ cxx = meson.get_compiler('cpp')
 
 simde_neon_families = [
   'aba',
+  'abal',
+  'abal_high',
   'abd',
   'abdl',
   'abs',
@@ -29,6 +31,8 @@ simde_neon_families = [
   'cadd_rot90',
   'cage',
   'cagt',
+  'cale',
+  'calt',
   'ceq',
   'ceqz',
   'cge',
@@ -51,6 +55,7 @@ simde_neon_families = [
   'cmla_rot270',
   'cnt',
   'cvt',
+  'cvt_n',
   'cvtn',
   'combine',
   'create',
@@ -64,6 +69,9 @@ simde_neon_families = [
   'fma',
   'fma_lane',
   'fma_n',
+  'fms',
+  'fms_lane',
+  'fms_n',
   'get_high',
   'get_lane',
   'get_low',
@@ -79,8 +87,13 @@ simde_neon_families = [
   'ld1q_x4',
   'ld1',
   'ld2',
+  'ld2_dup',
+  'ld2_lane',
   'ld3',
+  'ld3_dup',
+  'ld3_lane',
   'ld4',
+  'ld4_dup',
   'ld4_lane',
   'max',
   'maxnm',
@@ -93,16 +106,20 @@ simde_neon_families = [
   'mla_n',
   'mlal',
   'mlal_high',
+  'mlal_high_lane',
   'mlal_high_n',
   'mlal_lane',
   'mlal_n',
   'mls',
+  'mls_lane',
   'mls_n',
   'mlsl',
   'mlsl_high',
+  'mlsl_high_lane',
   'mlsl_high_n',
   'mlsl_lane',
   'mlsl_n',
+  #'mmlaq',
   'movl',
   'movl_high',
   'movn',
@@ -125,10 +142,27 @@ simde_neon_families = [
   'pmin',
   'qadd',
   'qabs',
+  'qdmlal',
+  'qdmlal_high',
+  'qdmlal_high_lane',
+  'qdmlal_high_n',
+  'qdmlal_lane',
+  'qdmlal_n',
+  'qdmlsl',
+  'qdmlsl_high',
+  'qdmlsl_high_lane',
+  'qdmlsl_high_n',
+  'qdmlsl_lane',
+  'qdmlsl_n',
   'qdmulh',
   'qdmulh_lane',
   'qdmulh_n',
   'qdmull',
+  'qdmull_high',
+  'qdmull_high_lane',
+  'qdmull_high_n',
+  'qdmull_lane',
+  'qdmull_n',
   'qrdmulh',
   'qrdmulh_lane',
   'qrdmulh_n',

simde/arm/neon.h

@@ -31,6 +31,8 @@
 #include "neon/types.h"
 
 #include "neon/aba.h"
+#include "neon/abal.h"
+#include "neon/abal_high.h"
 #include "neon/abd.h"
 #include "neon/abdl.h"
 #include "neon/abs.h"
@@ -50,6 +52,8 @@
 #include "neon/cadd_rot90.h"
 #include "neon/cage.h"
 #include "neon/cagt.h"
+#include "neon/cale.h"
+#include "neon/calt.h"
 #include "neon/ceq.h"
 #include "neon/ceqz.h"
 #include "neon/cge.h"
@@ -72,6 +76,7 @@
 #include "neon/cmla_rot270.h"
 #include "neon/cnt.h"
 #include "neon/cvt.h"
+#include "neon/cvt_n.h"
 #include "neon/cvtn.h"
 #include "neon/combine.h"
 #include "neon/create.h"
@@ -85,6 +90,9 @@
 #include "neon/fma.h"
 #include "neon/fma_lane.h"
 #include "neon/fma_n.h"
+#include "neon/fms.h"
+#include "neon/fms_lane.h"
+#include "neon/fms_n.h"
 #include "neon/get_high.h"
 #include "neon/get_lane.h"
 #include "neon/get_low.h"
@@ -100,8 +108,13 @@
 #include "neon/ld1q_x3.h"
 #include "neon/ld1q_x4.h"
 #include "neon/ld2.h"
+#include "neon/ld2_dup.h"
+#include "neon/ld2_lane.h"
 #include "neon/ld3.h"
+#include "neon/ld3_dup.h"
+#include "neon/ld3_lane.h"
 #include "neon/ld4.h"
+#include "neon/ld4_dup.h"
 #include "neon/ld4_lane.h"
 #include "neon/max.h"
 #include "neon/maxnm.h"
@@ -114,16 +127,20 @@
 #include "neon/mla_n.h"
 #include "neon/mlal.h"
 #include "neon/mlal_high.h"
+#include "neon/mlal_high_lane.h"
 #include "neon/mlal_high_n.h"
 #include "neon/mlal_lane.h"
 #include "neon/mlal_n.h"
 #include "neon/mls.h"
+#include "neon/mls_lane.h"
 #include "neon/mls_n.h"
 #include "neon/mlsl.h"
 #include "neon/mlsl_high.h"
+#include "neon/mlsl_high_lane.h"
 #include "neon/mlsl_high_n.h"
 #include "neon/mlsl_lane.h"
 #include "neon/mlsl_n.h"
+//#include "neon/mmlaq.h"
 #include "neon/movl.h"
 #include "neon/movl_high.h"
 #include "neon/movn.h"
@@ -146,10 +163,27 @@
 #include "neon/pmin.h"
 #include "neon/qabs.h"
 #include "neon/qadd.h"
+#include "neon/qdmlal.h"
+#include "neon/qdmlal_high.h"
+#include "neon/qdmlal_high_lane.h"
+#include "neon/qdmlal_high_n.h"
+#include "neon/qdmlal_lane.h"
+#include "neon/qdmlal_n.h"
+#include "neon/qdmlsl.h"
+#include "neon/qdmlsl_high.h"
+#include "neon/qdmlsl_high_lane.h"
+#include "neon/qdmlsl_high_n.h"
+#include "neon/qdmlsl_lane.h"
+#include "neon/qdmlsl_n.h"
 #include "neon/qdmulh.h"
 #include "neon/qdmulh_lane.h"
 #include "neon/qdmulh_n.h"
 #include "neon/qdmull.h"
+#include "neon/qdmull_high.h"
+#include "neon/qdmull_high_lane.h"
+#include "neon/qdmull_high_n.h"
+#include "neon/qdmull_lane.h"
+#include "neon/qdmull_n.h"
 #include "neon/qrdmulh.h"
 #include "neon/qrdmulh_lane.h"
 #include "neon/qrdmulh_n.h"

simde/arm/neon/abal.h

@@ -0,0 +1,125 @@
+/* SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Copyright:
+ *   2023      Yi-Yen Chung <[email protected]> (Copyright owned by Andes Technology)
+ */
+
+#if !defined(SIMDE_ARM_NEON_ABAL_H)
+#define SIMDE_ARM_NEON_ABAL_H
+
+#include "abdl.h"
+#include "add.h"
+
+HEDLEY_DIAGNOSTIC_PUSH
+SIMDE_DISABLE_UNWANTED_DIAGNOSTICS
+SIMDE_BEGIN_DECLS_
+
+SIMDE_FUNCTION_ATTRIBUTES
+simde_int16x8_t
+simde_vabal_s8(simde_int16x8_t a, simde_int8x8_t b, simde_int8x8_t c) {
+  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
+    return vabal_s8(a, b, c);
+  #else
+    return simde_vaddq_s16(simde_vabdl_s8(b, c), a);
+  #endif
+}
+#if defined(SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES)
+  #undef vabal_s8
+  #define vabal_s8(a, b, c) simde_vabal_s8((a), (b), (c))
+#endif
+
+SIMDE_FUNCTION_ATTRIBUTES
+simde_int32x4_t
+simde_vabal_s16(simde_int32x4_t a, simde_int16x4_t b, simde_int16x4_t c) {
+  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
+    return vabal_s16(a, b, c);
+  #else
+    return simde_vaddq_s32(simde_vabdl_s16(b, c), a);
+  #endif
+}
+#if defined(SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES)
+  #undef vabal_s16
+  #define vabal_s16(a, b, c) simde_vabal_s16((a), (b), (c))
+#endif
+
+SIMDE_FUNCTION_ATTRIBUTES
+simde_int64x2_t
+simde_vabal_s32(simde_int64x2_t a, simde_int32x2_t b, simde_int32x2_t c) {
+  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
+    return vabal_s32(a, b, c);
+  #else
+    return simde_vaddq_s64(simde_vabdl_s32(b, c), a);
+  #endif
+}
+#if defined(SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES)
+  #undef vabal_s32
+  #define vabal_s32(a, b, c) simde_vabal_s32((a), (b), (c))
+#endif
+
+SIMDE_FUNCTION_ATTRIBUTES
+simde_uint16x8_t
+simde_vabal_u8(simde_uint16x8_t a, simde_uint8x8_t b, simde_uint8x8_t c) {
+  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
+    return vabal_u8(a, b, c);
+  #else
+    return simde_vaddq_u16(simde_vabdl_u8(b, c), a);
+  #endif
+}
+#if defined(SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES)
+  #undef vabal_u8
+  #define vabal_u8(a, b, c) simde_vabal_u8((a), (b), (c))
+#endif
+
+SIMDE_FUNCTION_ATTRIBUTES
+simde_uint32x4_t
+simde_vabal_u16(simde_uint32x4_t a, simde_uint16x4_t b, simde_uint16x4_t c) {
+  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
+    return vabal_u16(a, b, c);
+  #else
+    return simde_vaddq_u32(simde_vabdl_u16(b, c), a);
+  #endif
+}
+#if defined(SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES)
+  #undef vabal_u16
+  #define vabal_u16(a, b, c) simde_vabal_u16((a), (b), (c))
+#endif
+
+SIMDE_FUNCTION_ATTRIBUTES
+simde_uint64x2_t
+simde_vabal_u32(simde_uint64x2_t a, simde_uint32x2_t b, simde_uint32x2_t c) {
+  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
+    return vabal_u32(a, b, c);
+  #else
+    return simde_vaddq_u64(simde_vabdl_u32(b, c), a);
+  #endif
+}
+#if defined(SIMDE_ARM_NEON_A32V7_ENABLE_NATIVE_ALIASES)
+  #undef vabal_u32
+  #define vabal_u32(a, b, c) simde_vabal_u32((a), (b), (c))
+#endif
+
+
+SIMDE_END_DECLS_
+HEDLEY_DIAGNOSTIC_POP
+
+#endif /* !defined(SIMDE_ARM_NEON_abal_H) */