# HG changeset patch
# User bellard
# Date 1043620187 0
# Node ID b4172ff70d27643e7e6ad19108790ec43c93a601
# Parent  8f440ca8e0b0c898579148bfd9e082b27843caaf
Altivec on non darwin systems patch by Romain Dolbeau

diff -r 8f440ca8e0b0 -r b4172ff70d27 Makefile
--- a/Makefile	Thu Jan 23 23:18:42 2003 +0000
+++ b/Makefile	Sun Jan 26 22:29:47 2003 +0000
@@ -86,7 +86,11 @@
 endif
 
 ifeq ($(TARGET_ALTIVEC),yes)
+ifeq ($(TARGET_OS),Darwin)
 CFLAGS += -faltivec
+else
+CFLAGS += -maltivec -mabi=altivec
+endif
 OBJS += ppc/dsputil_altivec.o ppc/mpegvideo_altivec.o ppc/idct_altivec.o \
         ppc/fft_altivec.o ppc/gmc_altivec.o
 endif
diff -r 8f440ca8e0b0 -r b4172ff70d27 dsputil.h
--- a/dsputil.h	Thu Jan 23 23:18:42 2003 +0000
+++ b/dsputil.h	Sun Jan 26 22:29:47 2003 +0000
@@ -224,6 +224,10 @@
 
 extern int mm_flags;
 
+#if defined(HAVE_ALTIVEC) && !defined(CONFIG_DARWIN)
+#include <altivec.h>
+#endif
+
 #define __align8 __attribute__ ((aligned (16)))
 
 void dsputil_init_ppc(DSPContext* c, unsigned mask);
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/dsputil_altivec.c
--- a/ppc/dsputil_altivec.c	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/dsputil_altivec.c	Sun Jan 26 22:29:47 2003 +0000
@@ -21,22 +21,39 @@
 #include "../dsputil.h"
 #include "dsputil_altivec.h"
 
-#if CONFIG_DARWIN
+#ifdef CONFIG_DARWIN
 #include <sys/sysctl.h>
-#endif
+#else /* CONFIG_DARWIN */
+#include <signal.h>
+#include <setjmp.h>
+
+static sigjmp_buf jmpbuf;
+static volatile sig_atomic_t canjump = 0;
+
+static void sigill_handler (int sig)
+{
+    if (!canjump) {
+        signal (sig, SIG_DFL);
+        raise (sig);
+    }
+    
+    canjump = 0;
+    siglongjmp (jmpbuf, 1);
+}
+#endif /* CONFIG_DARWIN */
 
 int pix_abs16x16_x2_altivec(uint8_t *pix1, uint8_t *pix2, int line_size)
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned char zero = (const vector unsigned char)(0);
+    const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     vector unsigned char *tv;
     vector unsigned char pix1v, pix2v, pix2iv, avgv, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
 
     s = 0;
-    sad = (vector unsigned int)(0);
+    sad = (vector unsigned int)vec_splat_u32(0);
     for(i=0;i<16;i++) {
         /*
            Read unaligned pixels into our vectors. The vectors are as follows:
@@ -76,7 +93,7 @@
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned char zero = (const vector unsigned char)(0);
+    const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     vector unsigned char *tv;
     vector unsigned char pix1v, pix2v, pix3v, avgv, t5;
     vector unsigned int sad;
@@ -84,7 +101,7 @@
     uint8_t *pix3 = pix2 + line_size;
 
     s = 0;
-    sad = (vector unsigned int)(0);
+    sad = (vector unsigned int)vec_splat_u32(0);
 
     /*
        Due to the fact that pix3 = pix2 + line_size, the pix3 of one
@@ -137,8 +154,8 @@
     int i;
     int s __attribute__((aligned(16)));
     uint8_t *pix3 = pix2 + line_size;
-    const vector unsigned char zero = (const vector unsigned char)(0);
-    const vector unsigned short two = (const vector unsigned short)(2);
+    const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
+    const vector unsigned short two = (const vector unsigned short)vec_splat_u16(2);
     vector unsigned char *tv, avgv, t5;
     vector unsigned char pix1v, pix2v, pix3v, pix2iv, pix3iv;
     vector unsigned short pix2lv, pix2hv, pix2ilv, pix2ihv;
@@ -148,7 +165,7 @@
     vector unsigned int sad;
     vector signed int sumdiffs;
 
-    sad = (vector unsigned int)(0);
+    sad = (vector unsigned int)vec_splat_u32(0);
     
     s = 0;
 
@@ -237,13 +254,13 @@
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned int zero = (const vector unsigned int)(0);
+    const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
     vector unsigned char perm1, perm2, *pix1v, *pix2v;
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
     
-    sad = (vector unsigned int) (0);
+    sad = (vector unsigned int)vec_splat_u32(0);
 
 
     for(i=0;i<16;i++) {
@@ -279,14 +296,18 @@
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned int zero = (const vector unsigned int)(0);
+    const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
     vector unsigned char perm1, perm2, permclear, *pix1v, *pix2v;
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
 
-    sad = (vector unsigned int)(0);
-    permclear = (vector unsigned char) (255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0);
+    sad = (vector unsigned int)vec_splat_u32(0);
+#ifdef CONFIG_DARWIN
+    permclear = (vector unsigned char)(255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0);
+#else
+    permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
+#endif
 
     for(i=0;i<8;i++) {
 	/* Read potentially unaligned pixels into t1 and t2
@@ -323,13 +344,13 @@
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned int zero = (const vector unsigned int)(0);
+    const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
     vector unsigned char *tv;
     vector unsigned char pixv;
     vector unsigned int sv;
     vector signed int sum;
     
-    sv = (vector unsigned int)(0);
+    sv = (vector unsigned int)vec_splat_u32(0);
     
     s = 0;
     for (i = 0; i < 16; i++) {
@@ -359,14 +380,18 @@
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned int zero = (const vector unsigned int)(0);
+    const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
     vector unsigned char perm1, perm2, permclear, *pix1v, *pix2v;
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sum;
     vector signed int sumsqr;
     
-    sum = (vector unsigned int)(0);
-    permclear = (vector unsigned char)(0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00);
+    sum = (vector unsigned int)vec_splat_u32(0);
+#ifdef CONFIG_DARWIN
+    permclear = (vector unsigned char)(255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0);
+#else
+    permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
+#endif
     
     for(i=0;i<8;i++) {
 	/* Read potentially unaligned pixels into t1 and t2
@@ -413,13 +438,13 @@
 {
     int i;
     int s __attribute__((aligned(16)));
-    const vector unsigned int zero = (const vector unsigned int)(0);
+    const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
     vector unsigned char perm1, perm2, *pix1v, *pix2v;
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sum;
     vector signed int sumsqr;
     
-    sum = (vector unsigned int)(0);
+    sum = (vector unsigned int)vec_splat_u32(0);
     
     for(i=0;i<16;i++) {
 	/* Read potentially unaligned pixels into t1 and t2 */
@@ -457,7 +482,7 @@
 
 int pix_sum_altivec(UINT8 * pix, int line_size)
 {
-    const vector unsigned int zero = (const vector unsigned int)(0);
+    const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
     vector unsigned char perm, *pixv;
     vector unsigned char t1;
     vector unsigned int sad;
@@ -466,7 +491,7 @@
     int i;
     int s __attribute__((aligned(16)));
     
-    sad = (vector unsigned int) (0);
+    sad = (vector unsigned int)vec_splat_u32(0);
     
     for (i = 0; i < 16; i++) {
 	/* Read the potentially unaligned 16 pixels into t1 */
@@ -492,7 +517,7 @@
 {
     int i;
     vector unsigned char perm, bytes, *pixv;
-    const vector unsigned char zero = (const vector unsigned char) (0);
+    const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     vector signed short shorts;
 
     for(i=0;i<8;i++)
@@ -519,7 +544,7 @@
 {
     int i;
     vector unsigned char perm, bytes, *pixv;
-    const vector unsigned char zero = (const vector unsigned char) (0);
+    const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     vector signed short shorts1, shorts2;
 
     for(i=0;i<4;i++)
@@ -830,8 +855,8 @@
      blockv, temp1, temp2;
    register vector unsigned short
      pixelssum1, pixelssum2, temp3;
-   register const vector unsigned char vczero = (const vector unsigned char)(0);
-   register const vector unsigned short vctwo = (const vector unsigned short)(2);
+   register const vector unsigned char vczero = (const vector unsigned char)vec_splat_u8(0);
+   register const vector unsigned short vctwo = (const vector unsigned short)vec_splat_u16(2);
    
    temp1 = vec_ld(0, pixels);
    temp2 = vec_ld(16, pixels);
@@ -945,9 +970,9 @@
      blockv, temp1, temp2;
    register vector unsigned short
      pixelssum1, pixelssum2, temp3;
-   register const vector unsigned char vczero = (const vector unsigned char)(0);
-   register const vector unsigned short vcone = (const vector unsigned short)(1);
-   register const vector unsigned short vctwo = (const vector unsigned short)(2);
+   register const vector unsigned char vczero = (const vector unsigned char)vec_splat_u8(0);
+   register const vector unsigned short vcone = (const vector unsigned short)vec_splat_u16(1);
+   register const vector unsigned short vctwo = (const vector unsigned short)vec_splat_u16(2);
    
    temp1 = vec_ld(0, pixels);
    temp2 = vec_ld(16, pixels);
@@ -1061,8 +1086,8 @@
    register vector unsigned short
      pixelssum1, pixelssum2, temp3,
      pixelssum3, pixelssum4, temp4;
-   register const vector unsigned char vczero = (const vector unsigned char)(0);
-   register const vector unsigned short vctwo = (const vector unsigned short)(2);
+   register const vector unsigned char vczero = (const vector unsigned char)vec_splat_u8(0);
+   register const vector unsigned short vctwo = (const vector unsigned short)vec_splat_u16(2);
    
    temp1 = vec_ld(0, pixels);
    temp2 = vec_ld(16, pixels);
@@ -1181,9 +1206,9 @@
    register vector unsigned short
      pixelssum1, pixelssum2, temp3,
      pixelssum3, pixelssum4, temp4;
-   register const vector unsigned char vczero = (const vector unsigned char)(0);
-   register const vector unsigned short vcone = (const vector unsigned short)(1);
-   register const vector unsigned short vctwo = (const vector unsigned short)(2);
+   register const vector unsigned char vczero = (const vector unsigned char)vec_splat_u8(0);
+   register const vector unsigned short vcone = (const vector unsigned short)vec_splat_u16(1);
+   register const vector unsigned short vctwo = (const vector unsigned short)vec_splat_u16(2);
    
    temp1 = vec_ld(0, pixels);
    temp2 = vec_ld(16, pixels);
@@ -1254,7 +1279,7 @@
 
 int has_altivec(void)
 {
-#if CONFIG_DARWIN
+#ifdef CONFIG_DARWIN
     int sels[2] = {CTL_HW, HW_VECTORUNIT};
     int has_vu = 0;
     size_t len = sizeof(has_vu);
@@ -1263,6 +1288,25 @@
     err = sysctl(sels, 2, &has_vu, &len, NULL, 0);
 
     if (err == 0) return (has_vu != 0);
-#endif
+#else /* CONFIG_DARWIN */
+/* no Darwin, do it the brute-force way */
+/* this is borrowed from the libmpeg2 library */
+    {
+      signal (SIGILL, sigill_handler);
+      if (sigsetjmp (jmpbuf, 1)) {
+        signal (SIGILL, SIG_DFL);
+      } else {
+        canjump = 1;
+        
+        asm volatile ("mtspr 256, %0\n\t"
+                      "vand %%v0, %%v0, %%v0"
+                      :
+                      : "r" (-1));
+        
+        signal (SIGILL, SIG_DFL);
+        return 1;
+      }
+    }
+#endif /* CONFIG_DARWIN */
     return 0;
 }
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/dsputil_altivec.h
--- a/ppc/dsputil_altivec.h	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/dsputil_altivec.h	Sun Jan 26 22:29:47 2003 +0000
@@ -63,7 +63,11 @@
 #define WORD_s2 0x18,0x19,0x1a,0x1b
 #define WORD_s3 0x1c,0x1d,0x1e,0x1f
 
+#ifdef CONFIG_DARWIN
 #define vcprm(a,b,c,d) (const vector unsigned char)(WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d)
+#else
+#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
+#endif
 
 // vcprmle is used to keep the same index as in the SSE version.
 // it's the same as vcprm, with the index inversed
@@ -75,7 +79,12 @@
 #define FLOAT_n -1.
 #define FLOAT_p 1.
 
+
+#ifdef CONFIG_DARWIN
 #define vcii(a,b,c,d) (const vector float)(FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d)
+#else
+#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
+#endif
 
 #else /* HAVE_ALTIVEC */
 #ifdef ALTIVEC_USE_REFERENCE_C_CODE
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/dsputil_ppc.c
--- a/ppc/dsputil_ppc.c	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/dsputil_ppc.c	Sun Jan 26 22:29:47 2003 +0000
@@ -120,7 +120,6 @@
       ((unsigned long*)blocks)[1] = 0L;
       ((unsigned long*)blocks)[2] = 0L;
       ((unsigned long*)blocks)[3] = 0L;
-      vec_st((vector short)(0), 0, blocks);
       i += 16;
     }
     for ( ; i < sizeof(DCTELEM)*6*64 ; i += 32) {
@@ -142,18 +141,17 @@
 /* check dcbz report how many bytes are set to 0 by dcbz */
 long check_dcbz_effect(void)
 {
-  register char *fakedata = (char*)malloc(1024);
+  register char *fakedata = (char*)av_malloc(1024);
   register char *fakedata_middle;
   register long zero = 0;
   register long i = 0;
   long count = 0;
 
-  if (fakedata == NULL)
+  if (!fakedata)
   {
     return 0L;
   }
 
-
   fakedata_middle = (fakedata + 512);
 
   memset(fakedata, 0xFF, 1024);
@@ -166,7 +164,7 @@
       count++;
   }
 
-  free(fakedata);
+  av_free(fakedata);
   
   return count;
 }
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/fft_altivec.c
--- a/ppc/fft_altivec.c	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/fft_altivec.c	Sun Jan 26 22:29:47 2003 +0000
@@ -138,7 +138,11 @@
 POWERPC_TBL_STOP_COUNT(altivec_fft_num, s->nbits >= 6);
 
 #else /* ALTIVEC_USE_REFERENCE_C_CODE */
+#ifdef CONFIG_DARWIN
     register const vector float vczero = (const vector float)(0.);
+#else
+    register const vector float vczero = (const vector float){0.,0.,0.,0.};
+#endif
     
     int ln = s->nbits;
     int	j, np, np2;
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/gmc_altivec.c
--- a/ppc/gmc_altivec.c	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/gmc_altivec.c	Sun Jan 26 22:29:47 2003 +0000
@@ -66,8 +66,8 @@
         (   x16)*(   y16), /* D */
         0, 0, 0, 0         /* padding */
       };
-    register const vector unsigned char vczero = (const vector unsigned char)(0);
-    register const vector unsigned short vcsr8 = (const vector unsigned short)(8);
+    register const vector unsigned char vczero = (const vector unsigned char)vec_splat_u8(0);
+    register const vector unsigned short vcsr8 = (const vector unsigned short)vec_splat_u16(8);
     register vector unsigned char dstv, dstv2, src_0, src_1, srcvA, srcvB, srcvC, srcvD;
     register vector unsigned short Av, Bv, Cv, Dv, rounderV, tempA, tempB, tempC, tempD;
     int i;
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/idct_altivec.c
--- a/ppc/idct_altivec.c	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/idct_altivec.c	Sun Jan 26 22:29:47 2003 +0000
@@ -151,6 +151,8 @@
     vx6 = vec_sra (vy6, shift);						\
     vx7 = vec_sra (vy7, shift);
 
+
+#ifdef CONFIG_DARWIN
 static const vector_s16_t constants[5] = {
     (vector_s16_t)(23170, 13573, 6518, 21895, -23170, -21895, 32, 31),
     (vector_s16_t)(16384, 22725, 21407, 19266, 16384, 19266, 21407, 22725),
@@ -158,6 +160,16 @@
     (vector_s16_t)(21407, 29692, 27969, 25172, 21407, 25172, 27969, 29692),
     (vector_s16_t)(19266, 26722, 25172, 22654, 19266, 22654, 25172, 26722)
 };
+#else
+// broken gcc
+static const vector_s16_t constants[5] = {
+    (vector_s16_t){23170, 13573, 6518, 21895, -23170, -21895, 32, 31},
+    (vector_s16_t){16384, 22725, 21407, 19266, 16384, 19266, 21407, 22725},
+    (vector_s16_t){22725, 31521, 29692, 26722, 22725, 26722, 29692, 31521},
+    (vector_s16_t){21407, 29692, 27969, 25172, 21407, 25172, 27969, 29692},
+    (vector_s16_t){19266, 26722, 25172, 22654, 19266, 22654, 25172, 26722}
+};
+#endif
 
 void idct_put_altivec(uint8_t* dest, int stride, vector_s16_t* block)
 {
diff -r 8f440ca8e0b0 -r b4172ff70d27 ppc/mpegvideo_altivec.c
--- a/ppc/mpegvideo_altivec.c	Thu Jan 23 23:18:42 2003 +0000
+++ b/ppc/mpegvideo_altivec.c	Sun Jan 26 22:29:47 2003 +0000
@@ -90,6 +90,13 @@
     vec = vec_splat(vec, 0); \
 }
 
+
+#ifdef CONFIG_DARWIN
+#define FOUROF(a) (a)
+#else
+// slower, for dumb non-apple GCC
+#define FOUROF(a) {a,a,a,a}
+#endif
 int dct_quantize_altivec(MpegEncContext* s, 
                         DCTELEM* data, int n,
                         int qscale, int* overflow)
@@ -97,7 +104,7 @@
     int lastNonZero;
     vector float row0, row1, row2, row3, row4, row5, row6, row7;
     vector float alt0, alt1, alt2, alt3, alt4, alt5, alt6, alt7;
-    const vector float zero = (const vector float)(0.0f);
+    const vector float zero = (const vector float)FOUROF(0.);
 
     // Load the data into the row/alt vectors
     {
@@ -141,18 +148,18 @@
 		// in the vector local variables, as floats, which we'll use during the
 		// quantize step...
     {
-        const vector float vec_0_298631336 = (vector float)(0.298631336f);
-        const vector float vec_0_390180644 = (vector float)(-0.390180644f);
-        const vector float vec_0_541196100 = (vector float)(0.541196100f);
-        const vector float vec_0_765366865 = (vector float)(0.765366865f);
-        const vector float vec_0_899976223 = (vector float)(-0.899976223f);
-        const vector float vec_1_175875602 = (vector float)(1.175875602f);
-        const vector float vec_1_501321110 = (vector float)(1.501321110f);
-        const vector float vec_1_847759065 = (vector float)(-1.847759065f);
-        const vector float vec_1_961570560 = (vector float)(-1.961570560f);
-        const vector float vec_2_053119869 = (vector float)(2.053119869f);
-        const vector float vec_2_562915447 = (vector float)(-2.562915447f);
-        const vector float vec_3_072711026 = (vector float)(3.072711026f);
+        const vector float vec_0_298631336 = (vector float)FOUROF(0.298631336f);
+        const vector float vec_0_390180644 = (vector float)FOUROF(-0.390180644f);
+        const vector float vec_0_541196100 = (vector float)FOUROF(0.541196100f);
+        const vector float vec_0_765366865 = (vector float)FOUROF(0.765366865f);
+        const vector float vec_0_899976223 = (vector float)FOUROF(-0.899976223f);
+        const vector float vec_1_175875602 = (vector float)FOUROF(1.175875602f);
+        const vector float vec_1_501321110 = (vector float)FOUROF(1.501321110f);
+        const vector float vec_1_847759065 = (vector float)FOUROF(-1.847759065f);
+        const vector float vec_1_961570560 = (vector float)FOUROF(-1.961570560f);
+        const vector float vec_2_053119869 = (vector float)FOUROF(2.053119869f);
+        const vector float vec_2_562915447 = (vector float)FOUROF(-2.562915447f);
+        const vector float vec_3_072711026 = (vector float)FOUROF(3.072711026f);
 
 
         int whichPass, whichHalf;
@@ -306,7 +313,7 @@
 				// rounding when we convert to int, instead of flooring.)
         {
             vector signed int biasInt;
-            const vector float negOneFloat = (vector float)(-1.0f);
+            const vector float negOneFloat = (vector float)FOUROF(-1.0f);
             LOAD4(biasInt, biasAddr);
             bias = vec_ctf(biasInt, QUANT_BIAS_SHIFT);
             negBias = vec_madd(bias, negOneFloat, zero);
@@ -503,6 +510,7 @@
 
     return lastNonZero;
 }
+#undef FOUROF
 
 /*
   AltiVec version of dct_unquantize_h263
@@ -551,7 +559,7 @@
     }
 #else /* ALTIVEC_USE_REFERENCE_C_CODE */
     {
-      register const vector short vczero = (const vector short)(0);
+      register const vector short vczero = (const vector short)vec_splat_s16(0);
       short __attribute__ ((aligned(16))) qmul8[] =
           {
             qmul, qmul, qmul, qmul,