2 other variants of how to implement the core part

benchmarks welcome ...

Originally committed as revision 8367 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavutil/sha1.c

@@ -14,6 +14,11 @@ typedef struct AVSHA1 {
 #define rol(value, bits) (((value) << (bits)) | ((value) >> (32 - (bits))))
 
 /* (R0+R1), R2, R3, R4 are the different operations used in SHA1 */
+#define R0b(v,w,x,y,z,i) z+((w&(x^y))^y)    +block[i]+0x5A827999+rol(v,5);
+#define R2b(v,w,x,y,z,i) z+( w^x     ^y)    +block[i]+0x6ED9EBA1+rol(v,5);
+#define R3b(v,w,x,y,z,i) z+(((w|x)&y)|(w&x))+block[i]+0x8F1BBCDC+rol(v,5);
+#define R4b(v,w,x,y,z,i) z+( w^x     ^y)    +block[i]+0xCA62C1D6+rol(v,5);
+
 #define R0(v,w,x,y,z,i) z+=((w&(x^y))^y)    +block[i]+0x5A827999+rol(v,5);w=rol(w,30);
 #define R2(v,w,x,y,z,i) z+=( w^x     ^y)    +block[i]+0x6ED9EBA1+rol(v,5);w=rol(w,30);
 #define R3(v,w,x,y,z,i) z+=(((w|x)&y)|(w&x))+block[i]+0x8F1BBCDC+rol(v,5);w=rol(w,30);
@@ -21,22 +26,91 @@ typedef struct AVSHA1 {
 
 /* Hash a single 512-bit block. This is the core of the algorithm. */
 
+//#define VARIANT1
+//#define VARIANT2
+
 static void transform(uint32_t state[5], uint8_t buffer[64]){
-    unsigned int a, b, c, d, e, i;
     uint32_t block[80];
+    unsigned int i;
+#ifdef VARIANT1
+    uint32_t s[85];
+#else
+    unsigned int a, b, c, d, e;
+#endif
 
     for(i=0; i<16; i++)
         block[i]= be2me_32(((uint32_t*)buffer)[i]);
     for(;i<80; i++)
         block[i]= rol(block[i-3]^block[i-8]^block[i-14]^block[i-16],1);
 
-    /* Copy context->state[] to working vars */
+
+#ifdef VARIANT1
+    s[0]= state[4];
+    s[1]= state[3];
+    s[2]= state[2];
+    s[3]= state[1];
+    s[4]= state[0];
+    for(i=0; i<20; i++){
+        s[5+i]= R0b(s[4+i], s[3+i], s[2+i], s[1+i], s[i], i);
+        s[3+i]= rol(s[3+i],30);
+    }
+    for(; i<40; i++){
+        s[5+i]= R2b(s[4+i], s[3+i], s[2+i], s[1+i], s[i], i);
+        s[3+i]= rol(s[3+i],30);
+    }
+    for(; i<60; i++){
+        s[5+i]= R3b(s[4+i], s[3+i], s[2+i], s[1+i], s[i], i);
+        s[3+i]= rol(s[3+i],30);
+    }
+    for(; i<80; i++){
+        s[5+i]= R4b(s[4+i], s[3+i], s[2+i], s[1+i], s[i], i);
+        s[3+i]= rol(s[3+i],30);
+    }
+    state[0] += s[84];
+    state[1] += s[83];
+    state[2] += s[82];
+    state[3] += s[81];
+    state[4] += s[80];
+#else
     a = state[0];
     b = state[1];
     c = state[2];
     d = state[3];
     e = state[4];
-
+#ifdef VARIANT2
+    for(i=0; i<20; i++){
+        int t= R0b(a,b,c,d,e,i);
+        e= d;
+        d= c;
+        c= rol(b,30);
+        b= a;
+        a= t;
+    }
+    for(; i<40; i++){
+        int t= R2b(a,b,c,d,e,i);
+        e= d;
+        d= c;
+        c= rol(b,30);
+        b= a;
+        a= t;
+    }
+    for(; i<60; i++){
+        int t= R3b(a,b,c,d,e,i);
+        e= d;
+        d= c;
+        c= rol(b,30);
+        b= a;
+        a= t;
+    }
+    for(; i<80; i++){
+        int t= R4b(a,b,c,d,e,i);
+        e= d;
+        d= c;
+        c= rol(b,30);
+        b= a;
+        a= t;
+    }
+#else
     for(i=0; i<20; i+=5){
         R0(a,b,c,d,e,0+i); R0(e,a,b,c,d,1+i); R0(d,e,a,b,c,2+i); R0(c,d,e,a,b,3+i); R0(b,c,d,e,a,4+i);
     }
@@ -49,12 +123,13 @@ static void transform(uint32_t state[5], uint8_t buffer[64]){
     for(; i<80; i+=5){
         R4(a,b,c,d,e,0+i); R4(e,a,b,c,d,1+i); R4(d,e,a,b,c,2+i); R4(c,d,e,a,b,3+i); R4(b,c,d,e,a,4+i);
     }
-
+#endif
     state[0] += a;
     state[1] += b;
     state[2] += c;
     state[3] += d;
     state[4] += e;
+#endif
 }
 
 void av_sha1_init(AVSHA1* context){