Add optimized PowerPC code

Makefile.am

@@ -28,6 +28,9 @@ endif
 if ARCH_ARM
 minerd_SOURCES += sha2-arm.S scrypt-arm.S
 endif
+if ARCH_PPC
+minerd_SOURCES += sha2-ppc.S scrypt-ppc.S
+endif
 endif
 minerd_LDFLAGS	= $(PTHREAD_FLAGS)
 minerd_LDADD	= @LIBCURL@ @JANSSON_LIBS@ @PTHREAD_LIBS@ @WS2_LIBS@

README

@@ -40,6 +40,8 @@ Architecture-specific notes:
 		but the decision whether to use them is made at compile time,
 		based on compiler-defined macros.
 		To use NEON instructions, add "-mfpu=neon" to CFLAGS.
+	PowerPC: No runtime CPU detection.
+		To use AltiVec instructions, add "-maltivec" to CFLAGS.
 	x86:	The miner checks for SSE2 instructions support at runtime,
 		and uses them if they are available.
 	x86-64:	The miner can take advantage of AVX, AVX2 and XOP instructions,

configure.ac

@@ -48,6 +48,9 @@ case $target in
   arm*-*-*)
     have_arm=true
     ;;
+  powerpc*-*-*)
+    have_ppc=true
+    ;;
 esac
 
 PTHREAD_FLAGS="-pthread"
@@ -108,6 +111,7 @@ AM_CONDITIONAL([USE_ASM], [test x$enable_assembly != xno])
 AM_CONDITIONAL([ARCH_x86], [test x$have_x86 = xtrue])
 AM_CONDITIONAL([ARCH_x86_64], [test x$have_x86_64 = xtrue])
 AM_CONDITIONAL([ARCH_ARM], [test x$have_arm = xtrue])
+AM_CONDITIONAL([ARCH_PPC], [test x$have_ppc = xtrue])
 
 if test x$request_jansson = xtrue
 then

cpu-miner.c

@@ -1456,6 +1456,12 @@ static void show_version_and_exit(void)
 #if defined(__ARM_NEON__)
 		" NEON"
 #endif
+#endif
+#if defined(USE_ASM) && (defined(__powerpc__) || defined(__ppc__) || defined(__PPC__))
+		" PowerPC"
+#if defined(__ALTIVEC__)
+		" AltiVec"
+#endif
 #endif
 		"\n");
 

miner.h

@@ -137,7 +137,7 @@ void sha256_transform(uint32_t *state, const uint32_t *block, int swap);
 void sha256d(unsigned char *hash, const unsigned char *data, int len);
 
 #ifdef USE_ASM
-#if defined(__ARM_NEON__) || defined(__i386__) || defined(__x86_64__)
+#if defined(__ARM_NEON__) || defined(__ALTIVEC__) || defined(__i386__) || defined(__x86_64__)
 #define HAVE_SHA256_4WAY 1
 int sha256_use_4way();
 void sha256_init_4way(uint32_t *state);

scrypt.c

@@ -409,6 +409,12 @@ void scrypt_core(uint32_t *X, uint32_t *V, int N);
 void scrypt_core_3way(uint32_t *X, uint32_t *V, int N);
 #endif
 
+#elif defined(USE_ASM) && (defined(__powerpc__) || defined(__ppc__) || defined(__PPC__))
+
+#define SCRYPT_MAX_WAYS 4
+#define scrypt_best_throughput() 1
+void scrypt_core(uint32_t *X, uint32_t *V, int N);
+
 #else
 
 static inline void xor_salsa8(uint32_t B[16], const uint32_t Bx[16])
@@ -513,7 +519,7 @@ static void scrypt_1024_1_1_256(const uint32_t *input, uint32_t *output,
 	uint32_t *midstate, unsigned char *scratchpad, int N)
 {
 	uint32_t tstate[8], ostate[8];
-	uint32_t X[32];
+	uint32_t X[32] __attribute__((aligned(128)));
 	uint32_t *V;
 	
 	V = (uint32_t *)(((uintptr_t)(scratchpad) + 63) & ~ (uintptr_t)(63));

sha2.c

@@ -14,7 +14,9 @@
 #include <string.h>
 #include <inttypes.h>
 
-#if defined(USE_ASM) && defined(__arm__) && defined(__APCS_32__)
+#if defined(USE_ASM) && \
+	((defined(__arm__) && defined(__APCS_32__)) || \
+	 (defined(__powerpc__) || defined(__ppc__) || defined(__PPC__)))
 #define EXTERN_SHA256
 #endif