From 267bc92d207d999ea5f7a9cb9df9e15b45e72bad Mon Sep 17 00:00:00 2001
From: Ryan Oldenburg <ryan@guzba.com>
Date: Fri, 22 Nov 2024 18:55:53 -0600
Subject: [PATCH] more neon

---
 src/nimsimd/neon.nim | 186 +++++++++++++++++++++++++++++++++++++++----
 1 file changed, 170 insertions(+), 16 deletions(-)

diff --git a/src/nimsimd/neon.nim b/src/nimsimd/neon.nim
index 816c455..08a3fc7 100644
--- a/src/nimsimd/neon.nim
+++ b/src/nimsimd/neon.nim
@@ -3,6 +3,9 @@
 {.push header: "arm_neon.h".}
 
 type
+  int32x2* {.importc: "int32x2_t".} = object
+  int32x4* {.importc: "int32x4_t".} = object
+
   uint8x16* {.importc: "uint8x16_t".} = object
   uint16x8* {.importc: "uint16x8_t".} = object
   uint32x4* {.importc: "uint32x4_t".} = object
@@ -12,6 +15,14 @@ type
   uint32x2* {.importc: "uint32x2_t".} = object
   uint64x1* {.importc: "uint64x1_t".} = object
 
+  float32x4* {.importc: "float32x4_t".} = object
+
+  float32x2* {.importc: "float32x2_t".} = object
+  float64x2* {.importc: "float64x2_t".} = object
+
+  int32x4x2* {.importc: "int32x4x2_t".} = object
+    val*: array[2, int32x4]
+
   uint8x16x2* {.importc: "uint8x16x2_t".} = object
     val*: array[2, uint8x16]
   uint16x8x2* {.importc: "uint16x8x2_t".} = object
@@ -66,26 +77,38 @@ type
   uint64x1x4* {.importc: "uint64x1x4_t".} = object
     val*: array[4, uint64x1]
 
+func vmov_n_f32*(a: float32): float32x2
+
 func vmovq_n_u8*(a: uint8): uint8x16
 func vmovq_n_u16*(a: uint16): uint16x8
 func vmovq_n_u32*(a: uint32): uint32x4
 func vmovq_n_u64*(a: uint64): uint64x2
+func vmovq_n_f32*(a: float32): float32x4
+func vmovq_n_f64*(a: float64): float64x2
+
+func vmovq_n_s32*(a: int32): int32x4
 
 func vmov_n_u8*(a: uint8): uint8x8
 func vmov_n_u16*(a: uint16): uint16x4
 func vmov_n_u32*(a: uint32): uint32x2
 func vmov_n_u64*(a: uint64): uint64x1
 
+func vld1q_s32*(p: pointer): int32x4
+
 func vld1q_u8*(p: pointer): uint8x16
 func vld1q_u16*(p: pointer): uint16x8
 func vld1q_u32*(p: pointer): uint32x4
 func vld1q_u64*(p: pointer): uint64x2
 
+func vld1q_f32*(p: pointer): float32x4
+
 func vld1_u8*(p: pointer): uint8x8
 func vld1_u16*(p: pointer): uint16x4
 func vld1_u32*(p: pointer): uint32x2
 func vld1_u64*(p: pointer): uint64x1
 
+func vld1_f32*(p: pointer): float32x2
+
 func vceq_u8*(a, b: uint8x8): uint8x8
 func vceq_u16*(a, b: uint16x4): uint16x4
 func vceq_u32*(a, b: uint32x2): uint32x2
@@ -95,16 +118,31 @@ func vceqq_u8*(a, b: uint8x16): uint8x16
 func vceqq_u16*(a, b: uint16x8): uint16x8
 func vceqq_u32*(a, b: uint32x4): uint32x4
 func vceqq_u64*(a, b: uint64x2): uint64x2
+func vceqq_f32*(a, b: float32x4): uint32x4
 
 func vcltq_u8*(a, b: uint8x16): uint8x16
 func vcltq_u16*(a, b: uint16x8): uint16x8
 func vcltq_u32*(a, b: uint32x4): uint32x4
 func vcltq_u64*(a, b: uint64x2): uint64x2
+func vcltq_f32*(a, b: float32x4): uint32x4
+
+func vcgtq_f32*(a, b: float32x4): uint32x4
+func vcgtq_u32*(a, b: uint32x4): uint32x4
+
+func vcleq_f32*(a, b: float32x4): uint32x4
+func vcleq_s32*(a, b: int32x4): uint32x4
+
+func vcgeq_f32*(a, b: float32x4): uint32x4
+func vcgeq_s32*(a, b: int32x4): uint32x4
+
+func vdivq_f32*(a, b: float32x4): float32x4
 
 func vpaddlq_u8*(a: uint8x16): uint16x8
 func vpaddlq_u16*(a: uint16x8): uint32x4
 func vpaddlq_u32*(a: uint32x4): uint64x2
 
+func vpaddd_u64*(a: uint64x2): uint64
+
 func vadd_u8*(a, b: uint8x8): uint8x8
 func vadd_u16*(a, b: uint16x4): uint16x4
 func vadd_u32*(a, b: uint32x2): uint32x2
@@ -118,6 +156,14 @@ func vaddq_u8*(a, b: uint8x16): uint8x16
 func vaddq_u16*(a, b: uint16x8): uint16x8
 func vaddq_u32*(a, b: uint32x4): uint32x4
 func vaddq_u64*(a, b: uint64x2): uint64x2
+func vaddq_f32*(a, b: float32x4): float32x4
+
+func vadd_f32*(a, b: float32x2): float32x2
+
+func vsub_f32*(a, b: float32x2): float32x2
+
+func vmul_n_f32*(a: float32x2, b: float32): float32x2
+func vmulq_n_f32*(a: float32x4, b: float32): float32x4
 
 func vpadalq_u8*(a: uint16x8, b: uint8x16): uint16x8
 func vpadalq_u16*(a: uint32x4, b: uint16x8): uint32x4
@@ -131,26 +177,33 @@ func vmlal_u8*(a: uint16x8, b, c: uint8x8): uint16x8
 func vmlal_u16*(a: uint32x4, b, c: uint16x4): uint32x4
 func vmlal_u32*(a: uint64x2, b, c: uint32x2): uint64x2
 
-func vst1q_lane_u8*(p: pointer, v: uint8x16, lane: int)
-func vst1q_lane_u16*(p: pointer, v: uint16x8, lane: int)
-func vst1q_lane_u32*(p: pointer, v: uint32x4, lane: int)
-func vst1q_lane_u64*(p: pointer, v: uint64x2, lane: int)
+func vst1q_lane_u8*(p: pointer, v: uint8x16, lane: int32)
+func vst1q_lane_u16*(p: pointer, v: uint16x8, lane: int32)
+func vst1q_lane_u32*(p: pointer, v: uint32x4, lane: int32)
+func vst1q_lane_u64*(p: pointer, v: uint64x2, lane: int32)
 
-func vld1q_lane_u8*(p: pointer, v: uint8x16, lane: int): uint8x16
-func vld1q_lane_u16*(p: pointer, v: uint16x8, lane: int): uint16x8
-func vld1q_lane_u32*(p: pointer, v: uint32x4, lane: int): uint32x4
-func vld1q_lane_u64*(p: pointer, v: uint64x2, lane: int): uint64x2
+func vld1q_lane_u8*(p: pointer, v: uint8x16, lane: int32): uint8x16
+func vld1q_lane_u16*(p: pointer, v: uint16x8, lane: int32): uint16x8
+func vld1q_lane_u32*(p: pointer, v: uint32x4, lane: int32): uint32x4
+func vld1q_lane_u64*(p: pointer, v: uint64x2, lane: int32): uint64x2
+
+func vld1q_dup_f64*(p: pointer): float64x2
+
+func vst1q_s32*(p: pointer, v: int32x4)
 
 func vst1q_u8*(p: pointer, v: uint8x16)
 func vst1q_u16*(p: pointer, v: uint16x8)
 func vst1q_u32*(p: pointer, v: uint32x4)
 func vst1q_u64*(p: pointer, v: uint64x2)
+func vst1q_f32*(p: pointer, v: float32x4)
 
 func vst1_u8*(p: pointer, v: uint8x8)
 func vst1_u16*(p: pointer, v: uint16x4)
 func vst1_u32*(p: pointer, v: uint32x2)
 func vst1_u64*(p: pointer, v: uint64x1)
 
+func vst1_f32*(p: pointer, v: float32x2)
+
 func vst2_u8*(p: pointer, v: uint8x8x2)
 func vst2_u16*(p: pointer, v: uint16x4x2)
 func vst2_u32*(p: pointer, v: uint32x2x2)
@@ -190,11 +243,13 @@ func vget_low_u8*(a: uint8x16): uint8x8
 func vget_low_u16*(a: uint16x8): uint16x4
 func vget_low_u32*(a: uint32x4): uint32x2
 func vget_low_u64*(a: uint64x2): uint64x1
+func vget_low_f32*(a: float32x4): float32x2
 
 func vget_high_u8*(a: uint8x16): uint8x8
 func vget_high_u16*(a: uint16x8): uint16x4
 func vget_high_u32*(a: uint32x4): uint32x2
 func vget_high_u64*(a: uint64x2): uint64x1
+func vget_high_f32*(a: float32x4): float32x2
 
 func vld4_u8*(p: pointer): uint8x8x4
 func vld4_u16*(p: pointer): uint16x4x4
@@ -206,6 +261,8 @@ func vld4q_u16*(p: pointer): uint16x8x4
 func vld4q_u32*(p: pointer): uint32x4x4
 func vld4q_u64*(p: pointer): uint64x2x4
 
+func vst2q_s32*(p: pointer, a: int32x4x2)
+
 func vst4q_u8*(p: pointer, a: uint8x16x4)
 func vst4q_u16*(p: pointer, a: uint16x8x4)
 func vst4q_u32*(p: pointer, a: uint32x4x4)
@@ -231,15 +288,20 @@ func vzip1q_u16*(a, b: uint16x8): uint16x8
 func vzip1q_u32*(a, b: uint32x4): uint32x4
 func vzip1q_u64*(a, b: uint64x2): uint64x2
 
-func vget_lane_u8*(a: uint8x8, lane: int): uint8
-func vget_lane_u16*(a: uint16x4, lane: int): uint16
-func vget_lane_u32*(a: uint32x2, lane: int): uint32
-func vget_lane_u64*(a: uint64x1, lane: int): uint64
+func vget_lane_u8*(a: uint8x8, lane: int32): uint8
+func vget_lane_u16*(a: uint16x4, lane: int32): uint16
+func vget_lane_u32*(a: uint32x2, lane: int32): uint32
+func vget_lane_u64*(a: uint64x1, lane: int32): uint64
+
+func vget_lane_s32*(a: int32x2, lane: int32): int32
+
+func vgetq_lane_s32*(a: int32x4, lane: int32): int32
 
-func vgetq_lane_u8*(a: uint8x16, lane: int): uint8
-func vgetq_lane_u16*(a: uint16x8, lane: int): uint16
-func vgetq_lane_u32*(a: uint32x4, lane: int): uint32
-func vgetq_lane_u64*(a: uint64x2, lane: int): uint64
+func vgetq_lane_u8*(a: uint8x16, lane: int32): uint8
+func vgetq_lane_u16*(a: uint16x8, lane: int32): uint16
+func vgetq_lane_u32*(a: uint32x4, lane: int32): uint32
+func vgetq_lane_u64*(a: uint64x2, lane: int32): uint64
+func vgetq_lane_f32*(a: float32x4, lane: int32): float32
 
 func vaddl_u8*(a, b: uint8x8): uint16x8
 func vaddl_u16*(a, b: uint16x4): uint32x4
@@ -299,6 +361,7 @@ func vsubq_u8*(a, b: uint8x16): uint8x16
 func vsubq_u16*(a, b: uint16x8): uint16x8
 func vsubq_u32*(a, b: uint32x4): uint32x4
 func vsubq_u64*(a, b: uint64x2): uint64x2
+func vsubq_f32*(a, b: float32x4): float32x4
 
 func vzip_u8*(a, b: uint8x8): uint8x8x2
 func vzip_u16*(a, b: uint16x4): uint16x4x2
@@ -310,4 +373,95 @@ func vmovl_u32*(a: uint32x2): uint64x2
 
 func vtbl1_u8*(a, idx: uint8x8): uint8x8
 
+func vbslq_s32*(a: uint32x4, b, c: int32x4): int32x4
+func vbslq_u8*(a, b, c: uint8x16): uint8x16
+func vbslq_u32*(a, b, c: uint32x4): uint32x4
+func vbslq_f32*(a: uint32x4, b, c: float32x4): float32x4
+
+func vminq_u32*(a, b: uint32x4): uint32x4
+func vminq_f32*(a, b: float32x4): float32x4
+func vmaxq_f32*(a, b: float32x4): float32x4
+
+func vmul_f32*(a, b: float32x2): float32x2
+func vmulq_f32*(a, b: float32x4): float32x4
+
+# func vcvtq_f32_u32*(a: uint32x4): float32x4
+# func vcvtq_u32_f32*(a: float32x4): uint32x4
+func vcvtq_s32_f32*(a: float32x4): int32x4
+func vcvtq_f32_s32*(a: int32x4): float32x4
+
+func vextq_u32*(a, b: uint32x4, n: int): uint32x4
+func vextq_u64*(a, b: uint64x2, n: int): uint64x2
+func vextq_f64*(a, b: float64x2, n: int): float64x2
+
+func vminvq_u32*(a: uint32x4): uint32
+func vminvq_f32*(a: float32x4): float32
+
+func vmaxq_u8*(a, b: uint8x16): uint8x16
+
+func vmaxvq_u8*(a: uint8x16): uint8
+func vmaxvq_u32*(a: uint32x4): uint32
+func vmaxvq_f32*(a: float32x4): float32
+
+func vmvnq_u8*(a: uint8x16): uint8x16
+
+func vmvnq_u32*(a: uint32x4): uint32x4
+
+func vornq_u8*(a, b: uint8x16): uint8x16
+
+func vqtbl1q_u8*(t: uint8x16, idx: uint8x16): uint8x16
+
+func vcntq_u8*(a: uint8x16): uint8x16
+
+func vpaddq_u32*(a, b: uint32x4): uint32x4
+
+func vsetq_lane_u64*(a: uint64, b: uint64x2, lane: int32): uint64x2
+func vsetq_lane_u8*(a: uint8, b: uint8x16, lane: int32): uint8x16
+
+func vshlq_u32*(a, b: uint32x4): uint32x4
+
+func vaddlvq_u32*(a: uint32x4): uint64
+
+func vdupq_laneq_u32*(a: uint32x4, lane: int32): uint32x4
+
+func vshrq_n_s32*(a: int32x4, n: int32): int32x4
+
+func vtstq_u32*(a, b: uint32x4): uint32x4
+
+func vcombine_f32*(a, b: float32x2): float32x4
+
+func vfmaq_f32*(a, b, c: float32x4): float32x4
+func vmlaq_f32*(a, b, c: float32x4): float32x4
+
+func vmla_f32*(a, b, c: float32x2): float32x2
+func vmla_n_f32*(a, b: float32x2, c: float32): float32x2
+func vmlaq_n_f32*(a, b: float32x4, c: float32): float32x4
+
+func vpadds_f32*(a: float32x2): float32
+
+func vreinterpretq_u8_f32*(a: float32x4): uint8x16
+func vreinterpretq_u32_f32*(a: float32x4): uint32x4
+func vreinterpretq_f32_f64*(a: float64x2): float32x4
+func vreinterpretq_u32_u64*(a: uint64x2): uint32x4
+func vreinterpretq_u64_u32*(a: uint32x4): uint64x2
+func vreinterpretq_u8_u32*(a: uint32x4): uint8x16
+func vreinterpretq_u32_u8*(a: uint8x16): uint32x4
+func vreinterpretq_u8_s32*(a: int32x4): uint8x16
+func vreinterpretq_s32_u8*(a: uint8x16): int32x4
+func vreinterpretq_s32_u32*(a: uint32x4): int32x4
+func vreinterpretq_u16_u8*(a: uint8x16): uint16x8
+
+func vreinterpret_u64_u8*(a: uint8x8): uint64x1
+func vreinterpret_s32_u32*(a: uint32x2): int32x2
+func vreinterpret_f32_u64*(a: uint64x1): float32x2
+
 {.pop.}
+
+func uint64x2_immediate*(hi, lo: static uint64): uint64x2 {.inline.} =
+  {.emit: [result, " = (uint64x2_t){", lo, ",", hi, "};"].}
+
+func uint32x4_immediate*(v3, v2, v1, v0: static uint32): uint32x4 {.inline.} =
+  {.emit: [result, " = (uint32x4_t){", v0, ",", v1, ",", v2, ",", v3, "};"].}
+
+func uint8x16_immediate*(v15, v14, v13, v12, v11, v10, v9, v8, v7, v6, v5, v4, v3, v2, v1, v0: static uint8): uint8x16 {.inline.} =
+  {.emit: [result, " = (uint8x16_t){", v0, ",", v1, ",", v2, ",", v3, ",", v4, ",", v5, ",", v6, ",", v7, ",", v8, ",", v9, ",", v10, ",", v11, ",", v12, ",", v13, ",", v14, ",", v15, "};"].}