Mirror/TrinityCore
1
0
Fork 0
mirror of https://github.com/TrinityCore/TrinityCore.git synced 2026-01-16 07:30:42 +01:00
Code Issues Packages Projects Releases Wiki Activity

Dep/SFMT: Support NEON on ARM with MSVC

Browse Source
This commit is contained in:
Shauren
2024-02-17 12:01:53 +01:00
parent ddba74567c
commit ea68ead90d
2 changed files with 38 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
dep/SFMT/SFMT-neon.h
View File

@@ -10,6 +10,13 @@
#ifndef SFMT_NEON_H
#define SFMT_NEON_H
#ifdef _MSC_VER
// The .n128_u64 field is first. Combine pairs of 32-bit integers in little-endian order.
#define sfmt_neon_init_uint32x4_t(w,x,y,z) { .n128_u32 = { (w), (x), (y), (z) } }
#else
#define sfmt_neon_init_uint32x4_t(w,x,y,z) { (w), (x), (y), (z) }
#endif
inline static void neon_recursion(uint32x4_t * r, uint32x4_t a, uint32x4_t b,
uint32x4_t c, uint32x4_t d);
@@ -27,8 +34,8 @@ inline static void neon_recursion(uint32x4_t * r, uint32x4_t a, uint32x4_t b,
uint32x4_t c, uint32x4_t d)
{
uint32x4_t v, x, y, z;
static const uint32x4_t vzero = {0,0,0,0};
static const uint32x4_t vmask = {SFMT_MSK1, SFMT_MSK2, SFMT_MSK3, SFMT_MSK4};
static const uint32x4_t vzero = sfmt_neon_init_uint32x4_t(0, 0, 0, 0);
static const uint32x4_t vmask = sfmt_neon_init_uint32x4_t(SFMT_MSK1, SFMT_MSK2, SFMT_MSK3, SFMT_MSK4);
#define rotate_bytes(A, B, C) vreinterpretq_u32_u8(vextq_u8(vreinterpretq_u8_u32(A),vreinterpretq_u8_u32(B),(C)))

29
dep/SFMT/sfmt_01_msvc_arm_support.diff Normal file
View File

@@ -0,0 +1,29 @@
diff --git a/dep/SFMT/SFMT-neon.h b/dep/SFMT/SFMT-neon.h
index 8790b8b60e..cffb3a514e 100644
--- a/dep/SFMT/SFMT-neon.h
+++ b/dep/SFMT/SFMT-neon.h
@@ -10,6 +10,13 @@
#ifndef SFMT_NEON_H
#define SFMT_NEON_H
+#ifdef _MSC_VER
+// The .n128_u64 field is first. Combine pairs of 32-bit integers in little-endian order.
+#define sfmt_neon_init_uint32x4_t(w,x,y,z) { .n128_u32 = { (w), (x), (y), (z) } }
+#else
+#define sfmt_neon_init_uint32x4_t(w,x,y,z) { (w), (x), (y), (z) }
+#endif
+
inline static void neon_recursion(uint32x4_t * r, uint32x4_t a, uint32x4_t b,
uint32x4_t c, uint32x4_t d);
@@ -27,8 +34,8 @@ inline static void neon_recursion(uint32x4_t * r, uint32x4_t a, uint32x4_t b,
uint32x4_t c, uint32x4_t d)
{
uint32x4_t v, x, y, z;
- static const uint32x4_t vzero = {0,0,0,0};
- static const uint32x4_t vmask = {SFMT_MSK1, SFMT_MSK2, SFMT_MSK3, SFMT_MSK4};
+ static const uint32x4_t vzero = sfmt_neon_init_uint32x4_t(0, 0, 0, 0);
+ static const uint32x4_t vmask = sfmt_neon_init_uint32x4_t(SFMT_MSK1, SFMT_MSK2, SFMT_MSK3, SFMT_MSK4);
#define rotate_bytes(A, B, C) vreinterpretq_u32_u8(vextq_u8(vreinterpretq_u8_u32(A),vreinterpretq_u8_u32(B),(C)))