From 21f7f6059a2da4a1a21c544ea2ee958c9f1af78e Mon Sep 17 00:00:00 2001
From: twinaphex <libretro@gmail.com>
Date: Sat, 22 Apr 2017 12:25:54 +0200
Subject: [PATCH] (ffmpeg FFT) No longer needs GLM

---
 cores/libretro-ffmpeg/fft/fft.cpp           | 14 ++--
 libretro-common/include/gfx/math/vector_2.h | 84 +++++++++++++++++++++
 2 files changed, 91 insertions(+), 7 deletions(-)

diff --git a/cores/libretro-ffmpeg/fft/fft.cpp b/cores/libretro-ffmpeg/fft/fft.cpp
index c5320b7300..0ebde3f4af 100644
--- a/cores/libretro-ffmpeg/fft/fft.cpp
+++ b/cores/libretro-ffmpeg/fft/fft.cpp
@@ -10,10 +10,7 @@
 #include <filters.h>
 #include <math/complex.h>
 #include <gfx/math/matrix_4x4.h>
-
-#define GLM_SWIZZLE
-#define GLM_FORCE_RADIANS
-#include <glm/packing.hpp>
+#include <gfx/math/vector_2.h>
 
 #if GL_DEBUG
 #define GL_CHECK_ERROR() do { \
@@ -373,6 +370,7 @@ void fft_build_params(glfft_t *fft, GLuint *buffer,
    {
       for (j = i; j < i + step_size; j++)
       {
+         vec2_t tmp;
          int s                 = j - i;
          float phase           = -1.0f * (float)s / step_size;
          unsigned a            = j;
@@ -381,12 +379,14 @@ void fft_build_params(glfft_t *fft, GLuint *buffer,
 
          unsigned read_a       = (step == 0) ? bitinverse(a, size) : a;
          unsigned read_b       = (step == 0) ? bitinverse(b, size) : b;
-         glm::vec2 tmp         = glm::vec2(twiddle.real, twiddle.imag);
+
+         tmp[0]                = twiddle.real;
+         tmp[1]                = twiddle.imag;
 
          buffer[2 * a + 0]     = (read_a << 16) | read_b;
-         buffer[2 * a + 1]     = glm::packHalf2x16(tmp);
+         buffer[2 * a + 1]     = vec2_packHalf2x16(tmp[0], tmp[1]);
          buffer[2 * b + 0]     = (read_a << 16) | read_b;
-         buffer[2 * b + 1]     = glm::packHalf2x16(-tmp);
+         buffer[2 * b + 1]     = vec2_packHalf2x16(-tmp[0], -tmp[1]);
       }
    }
 }
diff --git a/libretro-common/include/gfx/math/vector_2.h b/libretro-common/include/gfx/math/vector_2.h
index 5248693b06..bf6b20ae86 100644
--- a/libretro-common/include/gfx/math/vector_2.h
+++ b/libretro-common/include/gfx/math/vector_2.h
@@ -27,6 +27,7 @@
 #include <math.h>
 
 #include <retro_common_api.h>
+#include <retro_inline.h>
 
 RETRO_BEGIN_DECLS
 
@@ -48,6 +49,89 @@ typedef float vec2_t[2];
    dst[0] = src[0]; \
    dst[1] = src[1]
 
+static INLINE float overflow(void)
+{
+   volatile float f = 1e10;
+
+   for(int i = 0; i < 10; ++i)	
+      f *= f;
+   return f;
+}
+
+static INLINE int16_t tofloat16(float f)
+{
+	union uif32
+   {
+      float f;
+      uint32_t i;
+   };
+
+   int i, s, e, m;
+   uif32 Entry;
+   Entry.f = f;
+   i       = (int)Entry.i;
+   s       =  (i >> 16) & 0x00008000;
+   e       = ((i >> 23) & 0x000000ff) - (127 - 15);
+   m       =   i        & 0x007fffff;
+
+   if(e <= 0)
+   {
+      if(e < -10)
+         return (int16_t)(s);
+
+      m = (m | 0x00800000) >> (1 - e);
+
+      if(m & 0x00001000) 
+         m += 0x00002000;
+
+      return (int16_t)(s | (m >> 13));
+   }
+
+   if(e == 0xff - (127 - 15))
+   {
+      if(m == 0)
+         return (int16_t)(s | 0x7c00);
+
+      m >>= 13;
+
+      return (int16_t)(s | 0x7c00 | m | (m == 0));
+   }
+
+   if(m &  0x00001000)
+   {
+      m += 0x00002000;
+
+      if(m & 0x00800000)
+      {
+         m =  0;
+         e += 1;
+      }
+   }
+
+   if (e > 30)
+   {
+      overflow();
+
+      return (int16_t)(s | 0x7c00);
+   }
+
+   return (int16_t)(s | (e << 10) | (m >> 13));
+}
+
+static INLINE unsigned int vec2_packHalf2x16(float vec0, float vec1)
+{
+   union
+   {
+      int16_t in[2];
+      unsigned int out;
+   } u;
+
+   u.in[0] = tofloat16(vec0);
+   u.in[1] = tofloat16(vec1);
+
+   return u.out;
+}
+
 RETRO_END_DECLS
 
 #endif