diff --git a/makefile b/makefile
index ccd6c8ada9..50f813d75b 100644
--- a/makefile
+++ b/makefile
@@ -3,7 +3,11 @@ MAKEFLAGS = -s
 name = avian
 version = 0.4
 
-build-arch := $(shell uname -m | sed 's/^i.86$$/i386/' | sed 's/^arm.*$$/arm/')
+build-arch := $(shell uname -m \
+	| sed 's/^i.86$$/i386/' \
+	| sed 's/^arm.*$$/arm/' \
+	| sed 's/ppc/powerpc/')
+
 ifeq (Power,$(filter Power,$(build-arch)))
 	build-arch = powerpc
 endif
diff --git a/src/binaryToObject/elf.cpp b/src/binaryToObject/elf.cpp
index 8af5c0cd58..44915282bb 100644
--- a/src/binaryToObject/elf.cpp
+++ b/src/binaryToObject/elf.cpp
@@ -35,6 +35,7 @@
 #define EV_CURRENT 1
 
 #define ELFDATA2LSB 1
+#define ELFDATA2MSB 2
 
 #define ELFOSABI_SYSV 0
 
@@ -43,6 +44,7 @@
 #define EM_386 3
 #define EM_X86_64 62
 #define EM_ARM 40
+#define EM_PPC 20
 
 #define SHT_PROGBITS 1
 #define SHT_SYMTAB 2
@@ -77,7 +79,6 @@
 #  error
 #endif
 
-#define Data ELFDATA2LSB
 #define OSABI ELFOSABI_SYSV
 
 namespace {
@@ -178,7 +179,7 @@ void
 writeObject(const uint8_t* data, unsigned size, FILE* out,
             const char* startName, const char* endName,
             const char* sectionName, unsigned sectionFlags,
-            unsigned alignment, int machine)
+            unsigned alignment, int machine, int encoding)
 {
   const unsigned sectionCount = 5;
   const unsigned symbolCount = 2;
@@ -222,7 +223,7 @@ writeObject(const uint8_t* data, unsigned size, FILE* out,
   fileHeader.e_ident[EI_MAG2] = ELFMAG2;
   fileHeader.e_ident[EI_MAG3] = ELFMAG3;
   fileHeader.e_ident[EI_CLASS] = Class;
-  fileHeader.e_ident[EI_DATA] = Data;
+  fileHeader.e_ident[EI_DATA] = encoding;
   fileHeader.e_ident[EI_VERSION] = EV_CURRENT;
   fileHeader.e_ident[EI_OSABI] = OSABI;
   fileHeader.e_ident[EI_ABIVERSION] = 0;
@@ -349,12 +350,19 @@ MAKE_NAME(writeElf, BITS_PER_WORD, Object)
    bool writable, bool executable)
 {
   int machine;
+  int encoding;
   if (strcmp(architecture, "x86_64") == 0) {
     machine = EM_X86_64;
+    encoding = ELFDATA2LSB;
   } else if (strcmp(architecture, "i386") == 0) {
     machine = EM_386;
+    encoding = ELFDATA2LSB;
   } else if (strcmp(architecture, "arm") == 0) {
     machine = EM_ARM;
+    encoding = ELFDATA2LSB;
+  } else if (strcmp(architecture, "powerpc") == 0) {
+    machine = EM_PPC;
+    encoding = ELFDATA2MSB;
   } else {
     fprintf(stderr, "unsupported architecture: %s\n", architecture);
     return false;
@@ -376,7 +384,7 @@ MAKE_NAME(writeElf, BITS_PER_WORD, Object)
   }
 
   writeObject(data, size, out, startName, endName, sectionName, sectionFlags,
-              alignment, machine);
+              alignment, machine, encoding);
 
   return true;
 }
diff --git a/src/binaryToObject/main.cpp b/src/binaryToObject/main.cpp
index c95f193d3d..41f2951645 100644
--- a/src/binaryToObject/main.cpp
+++ b/src/binaryToObject/main.cpp
@@ -73,8 +73,10 @@ writeObject(uint8_t* data, unsigned size, FILE* out, const char* startName,
       success = writeElf64Object
         (data, size, out, startName, endName, architecture, alignment,
          writable, executable);
-    } else if (strcmp("i386", architecture) == 0 ||
-               strcmp("arm", architecture) == 0) {
+    } else if (strcmp("i386", architecture) == 0
+               or strcmp("arm", architecture) == 0
+               or strcmp("powerpc", architecture) == 0)
+    {
       found = true;
       success = writeElf32Object
         (data, size, out, startName, endName, architecture, alignment,
diff --git a/src/common.h b/src/common.h
index 13957bedfd..5c7160e54c 100644
--- a/src/common.h
+++ b/src/common.h
@@ -88,7 +88,7 @@ alias(void* p, unsigned offset)
 #    define ARCH_x86_32
 #  elif defined __x86_64__
 #    define ARCH_x86_64
-#  elif defined __POWERPC__
+#  elif (defined __POWERPC__) || (defined __powerpc__)
 #    define ARCH_powerpc
 #  elif defined __arm__
 #    define ARCH_arm
diff --git a/src/compile-powerpc.S b/src/compile-powerpc.S
index 68e4d8acd2..546432beb3 100644
--- a/src/compile-powerpc.S
+++ b/src/compile-powerpc.S
@@ -13,16 +13,21 @@
 .text
 
 #define BYTES_PER_WORD 4
-#define LINKAGE_AREA 6
-#define ARGUMENT_BASE BYTES_PER_WORD * LINKAGE_AREA
-
-#define LOCAL(x) L##x
    
 #ifdef __APPLE__
 #  define GLOBAL(x) _##x
+#  define LOCAL(x) L##x
+#  define LINKAGE_AREA 6
+#  define RETURN_ADDRESS_OFFSET 8
 #else
-#  define GLOBAL(x) x   
+#  define GLOBAL(x) x
+#  define LOCAL(x) .L##x
+#  define LINKAGE_AREA 2
+#  define RETURN_ADDRESS_OFFSET 4
+#  include "powerpc-regs.S"
 #endif
+
+#define ARGUMENT_BASE BYTES_PER_WORD * LINKAGE_AREA
       
 #define THREAD_STACK 2148
 #define THREAD_CONTINUATION 2156
@@ -42,7 +47,7 @@
 GLOBAL(vmInvoke):
    // save return address
    mflr r0
-   stw  r0,8(r1)
+   stw  r0,RETURN_ADDRESS_OFFSET(r1)
 
    // r3: thread
    // r4: function
@@ -53,12 +58,10 @@ GLOBAL(vmInvoke):
 
    // r9: temporary
 
-   // save return type
-   stw  r8,44(r1)
-
-   // allocate stack space, adding room for callee-saved registers
+   // allocate stack space, adding room for callee-saved registers and
+   // return type
    subfic r9,r7,-80
-	 stwux r1,r1,r9
+   stwux r1,r1,r9
    
    // save callee-saved registers
    add  r9,r7,r1
@@ -83,6 +86,9 @@ GLOBAL(vmInvoke):
    stw  r30,68(r9)
    stw  r31,72(r9)
 
+   // save return type
+   stw  r8,74(r9)
+
    // we use r13 to hold the thread pointer, by convention
    mr   r13,r3
    
@@ -213,11 +219,11 @@ LOCAL(vmInvoke_exit):
    lwz  r31,72(r9)
    
    // handle return value based on expected type
-   lwz  r8,44(r1)
+   lwz  r8,74(r9)
    
 LOCAL(vmInvoke_return):
    // load return address
-   lwz  r0,8(r1)
+   lwz  r0,RETURN_ADDRESS_OFFSET(r1)
    mtlr r0
 
    // return
diff --git a/src/compiler.cpp b/src/compiler.cpp
index 18dc98e06f..2d809b1c5e 100644
--- a/src/compiler.cpp
+++ b/src/compiler.cpp
@@ -2171,9 +2171,9 @@ class MemorySite: public Site {
   }
 
   virtual Site* makeNextWord(Context* c, unsigned index) {
-    // todo: endianness?
     return memorySite
-      (c, base, offset + (index == 1 ? BytesPerWord : -BytesPerWord),
+      (c, base, offset + ((index == 1) xor c->arch->bigEndian()
+                          ? BytesPerWord : -BytesPerWord),
        this->index, scale);
   }
 
@@ -2184,12 +2184,11 @@ class MemorySite: public Site {
   }
 
   virtual SiteMask nextWordMask(Context* c, unsigned index) {
-    // todo: endianness?
     int frameIndex;
     if (base == c->arch->stack()) {
       assert(c, this->index == NoRegister);
       frameIndex = static_cast<int>(offsetToFrameIndex(c, offset))
-        + (index == 1 ? 1 : -1);
+        + ((index == 1) xor c->arch->bigEndian() ? 1 : -1);
     } else {
       frameIndex = NoFrameIndex;
     }
diff --git a/src/powerpc-regs.S b/src/powerpc-regs.S
new file mode 100644
index 0000000000..da5940f403
--- /dev/null
+++ b/src/powerpc-regs.S
@@ -0,0 +1,64 @@
+#define r0 0
+#define r1 1
+#define r2 2
+#define r3 3
+#define r4 4
+#define r5 5
+#define r6 6
+#define r7 7
+#define r8 8
+#define r9 9
+#define r10 10
+#define r11 11
+#define r12 12
+#define r13 13
+#define r14 14
+#define r15 15
+#define r16 16
+#define r17 17
+#define r18 18
+#define r19 19
+#define r20 20
+#define r21 21
+#define r22 22
+#define r23 23
+#define r24 24
+#define r25 25
+#define r26 26
+#define r27 27
+#define r28 28
+#define r29 29
+#define r30 30
+#define r31 31
+#define f0 0
+#define f1 1
+#define f2 2
+#define f3 3
+#define f4 4
+#define f5 5
+#define f6 6
+#define f7 7
+#define f8 8
+#define f9 9
+#define f10 10
+#define f11 11
+#define f12 12
+#define f13 13
+#define f14 14
+#define f15 15
+#define f16 16
+#define f17 17
+#define f18 18
+#define f19 19
+#define f20 20
+#define f21 21
+#define f22 22
+#define f23 23
+#define f24 24
+#define f25 25
+#define f26 26
+#define f27 27
+#define f28 28
+#define f29 29
+#define f30 30
+#define f31 31
diff --git a/src/powerpc.S b/src/powerpc.S
index 2b26f7985f..96bf382cfb 100644
--- a/src/powerpc.S
+++ b/src/powerpc.S
@@ -13,22 +13,28 @@
 .text
 
 #define BYTES_PER_WORD 4
-#define LINKAGE_AREA 6
 #define GPR_COUNT 8
-#define MEMORY_BASE BYTES_PER_WORD * (LINKAGE_AREA + GPR_COUNT)
-#define LOCAL(x) L##x
    
 #ifdef __APPLE__
 #  define GLOBAL(x) _##x
+#  define LOCAL(x) L##x
+#  define LINKAGE_AREA 6
+#  define MEMORY_BASE BYTES_PER_WORD * (LINKAGE_AREA + GPR_COUNT)
+#  define RETURN_ADDRESS_OFFSET 8
 #else
-#  define GLOBAL(x) x   
+#  define GLOBAL(x) x
+#  define LOCAL(x) .L##x
+#  define LINKAGE_AREA 2
+#  define MEMORY_BASE BYTES_PER_WORD * LINKAGE_AREA
+#  define RETURN_ADDRESS_OFFSET 4
+#  include "powerpc-regs.S"
 #endif
 
 .globl GLOBAL(vmNativeCall)
 GLOBAL(vmNativeCall):
    // save return address
    mflr r0
-   stw  r0,8(r1)
+   stw  r0,RETURN_ADDRESS_OFFSET(r1)
    
    // r3 aka r13: function
    // r4        : stackTotal
@@ -42,17 +48,26 @@ GLOBAL(vmNativeCall):
    // r16       : temporary
    // r17       : temporary
    // r18       : temporary
-   
-   // save registers used for local variables
-   stw  r13,24(r1)
-   stw  r14,28(r1)
-   stw  r15,32(r1)
-   stw  r16,36(r1)
-   stw  r17,40(r1)
-   stw  r18,44(r1)
 
-   // allocate stack space
-	 stwux	r1,r1,r4
+   // allocate stack space, adding room for callee-saved registers and
+   // scratch space for copying a FP return value into GPRs
+   subfic r10,r4,-48
+	 stwux r1,r1,r10
+
+   // save callee-saved registers used for local variables
+   add  r10,r4,r1
+
+   // save registers used for local variables
+   stw  r13,0(r10)
+   stw  r14,4(r10)
+   stw  r15,8(r10)
+   stw  r16,12(r10)
+   stw  r17,16(r10)
+   stw  r18,20(r10)
+   stw  r19,24(r10)
+
+   // remember where we saved the local variables
+   mr   r19,r10
 
    // save our argument registers so we can clobber them
    mr   r13,r3
@@ -84,12 +99,14 @@ LOCAL(test):
    lfd  f6,40(r8)
    lfd  f7,48(r8)
    lfd  f8,56(r8)
+#ifdef __APPLE__
    lfd  f9,64(r8)
    lfd  f10,72(r8)
    lfd  f11,80(r8)
    lfd  f12,88(r8)
    lfd  f13,96(r8)
-   
+#endif
+
 LOCAL(gpr):
    // do we need to load the general-purpose registers?
    cmpwi r7,0
@@ -128,25 +145,26 @@ LOCAL(float):
 
 LOCAL(copy):
    // move floating point return value to GPRs via memory
-   stfd f1,8(r1)
-   lwz  r3,8(r1)
-   lwz  r4,12(r1)
+   stfd f1,32(r19)
+   lwz  r3,32(r19)
+   lwz  r4,36(r19)
    b    LOCAL(exit)
 
 LOCAL(exit):
+   // restore callee-saved registers used for local variables
+   lwz  r13,0(r19)
+   lwz  r14,4(r19)
+   lwz  r15,8(r19)
+   lwz  r16,12(r19)
+   lwz  r17,16(r19)
+   lwz  r18,20(r19)
+   lwz  r19,24(r19)
+
    // restore stack pointer
 	 lwz	r1,0(r1)
 
-   // restore registers used for local variables
-   lwz  r13,24(r1)
-   lwz  r14,28(r1)
-   lwz  r15,32(r1)
-   lwz  r16,36(r1)
-   lwz  r17,40(r1)
-   lwz  r18,44(r1)
-
    // load return address
-   lwz  r0,8(r1)
+   lwz  r0,RETURN_ADDRESS_OFFSET(r1)
    mtlr r0
 
    // return
@@ -171,7 +189,7 @@ GLOBAL(vmRun):
    // r5: checkpoint
 
    mflr r0
-   stw  r0,8(r1)
+   stw  r0,RETURN_ADDRESS_OFFSET(r1)
 
    stwu r1,-(MEMORY_BASE+88)(r1)
       
@@ -226,6 +244,6 @@ GLOBAL(vmRun_returnAddress):
    lwz  r31,MEMORY_BASE+72(r1)
 
    lwz	r1,0(r1)
-   lwz  r0,8(r1)
+   lwz  r0,RETURN_ADDRESS_OFFSET(r1)
    mtlr r0
    blr
diff --git a/src/powerpc.cpp b/src/powerpc.cpp
index 6c721651ba..3644520647 100644
--- a/src/powerpc.cpp
+++ b/src/powerpc.cpp
@@ -162,7 +162,15 @@ carry16(intptr_t v)
   return static_cast<int16_t>(v) < 0 ? 1 : 0;
 }
 
+#ifdef __APPLE__
 const unsigned FrameFooterSize = 6;
+const unsigned ReturnAddressOffset = 2;
+const unsigned AlignArguments = false;
+#else
+const unsigned FrameFooterSize = 2;
+const unsigned ReturnAddressOffset = 1;
+const unsigned AlignArguments = true;
+#endif
 
 const unsigned StackAlignmentInBytes = 16;
 const unsigned StackAlignmentInWords = StackAlignmentInBytes / BytesPerWord;
@@ -1777,7 +1785,7 @@ nextFrame(ArchitectureContext* c UNUSED, int32_t* start, unsigned size,
     // todo: check for and handle tail calls
   }
 
-  *ip = static_cast<void**>(*stack)[offset + 2];
+  *ip = static_cast<void**>(*stack)[offset + ReturnAddressOffset];
   *stack = static_cast<void**>(*stack) + offset;
 }
 
@@ -1940,7 +1948,7 @@ class MyArchitecture: public Assembler::Architecture {
   }
 
   virtual bool argumentAlignment() {
-    return false;
+    return AlignArguments;
   }
 
   virtual unsigned argumentRegisterCount() {
@@ -2023,7 +2031,7 @@ class MyArchitecture: public Assembler::Architecture {
   }
 
   virtual void* frameIp(void* stack) {
-    return stack ? static_cast<void**>(stack)[2] : 0;
+    return stack ? static_cast<void**>(stack)[ReturnAddressOffset] : 0;
   }
 
   virtual unsigned frameHeaderSize() {
@@ -2039,7 +2047,7 @@ class MyArchitecture: public Assembler::Architecture {
   }
 
   virtual int returnAddressOffset() {
-    return 8 / BytesPerWord;
+    return ReturnAddressOffset;
   }
 
   virtual int framePointerOffset() {
@@ -2261,7 +2269,7 @@ class MyAssembler: public Assembler {
     Register returnAddress(0);
     emit(&c, mflr(returnAddress.low));
 
-    Memory returnAddressDst(StackRegister, 8);
+    Memory returnAddressDst(StackRegister, ReturnAddressOffset * BytesPerWord);
     moveRM(&c, BytesPerWord, &returnAddress, BytesPerWord, &returnAddressDst);
 
     Register stack(StackRegister);
@@ -2314,7 +2322,7 @@ class MyAssembler: public Assembler {
     Register returnAddress(0);
     emit(&c, mflr(returnAddress.low));
 
-    Memory returnAddressDst(StackRegister, 8);
+    Memory returnAddressDst(StackRegister, ReturnAddressOffset * BytesPerWord);
     moveRM(&c, BytesPerWord, &returnAddress, BytesPerWord, &returnAddressDst);
 
     Register stack(StackRegister);
@@ -2337,7 +2345,7 @@ class MyAssembler: public Assembler {
     moveMR(&c, BytesPerWord, &stackSrc, BytesPerWord, &stack);
 
     Register returnAddress(0);
-    Memory returnAddressSrc(StackRegister, 8);
+    Memory returnAddressSrc(StackRegister, ReturnAddressOffset * BytesPerWord);
     moveMR(&c, BytesPerWord, &returnAddressSrc, BytesPerWord, &returnAddress);
     
     emit(&c, mtlr(returnAddress.low));
@@ -2351,7 +2359,8 @@ class MyAssembler: public Assembler {
     if (TailCalls) {
       if (offset) {
         Register tmp(0);
-        Memory returnAddressSrc(StackRegister, 8 + (footprint * BytesPerWord));
+        Memory returnAddressSrc
+          (StackRegister, (ReturnAddressOffset + footprint) * BytesPerWord);
         moveMR(&c, BytesPerWord, &returnAddressSrc, BytesPerWord, &tmp);
     
         emit(&c, mtlr(tmp.low));
@@ -2366,7 +2375,8 @@ class MyAssembler: public Assembler {
           assert(&c, offset > 0);
 
           Register ras(returnAddressSurrogate);
-          Memory dst(StackRegister, 8 + (offset * BytesPerWord));
+          Memory dst
+            (StackRegister, (ReturnAddressOffset + offset) * BytesPerWord);
           moveRM(&c, BytesPerWord, &ras, BytesPerWord, &dst);
         }
 
diff --git a/src/powerpc.h b/src/powerpc.h
index b519ac8064..5fe4543cc7 100644
--- a/src/powerpc.h
+++ b/src/powerpc.h
@@ -14,8 +14,6 @@
 #include "types.h"
 #include "common.h"
 
-#define VA_LIST(x) (&(x))
-
 #ifdef __APPLE__
 #  include "mach/mach_types.h"
 #  include "mach/ppc/thread_act.h"
@@ -45,9 +43,17 @@
 #  define LINK_REGISTER(context) \
   THREAD_STATE_LINK(context->uc_mcontext->FIELD(ss))
 
-#else
-#  error "non-Apple PowerPC-based platforms not yet supported"
-#endif
+#define VA_LIST(x) (&(x))
+
+#else // not __APPLE__
+#  define IP_REGISTER(context) (context->uc_mcontext.regs->gpr[32])
+#  define STACK_REGISTER(context) (context->uc_mcontext.regs->gpr[1])
+#  define THREAD_REGISTER(context) (context->uc_mcontext.regs->gpr[13])
+#  define LINK_REGISTER(context) (context->uc_mcontext.regs->gpr[36])
+
+#define VA_LIST(x) (x)
+
+#endif // not __APPLE__
 
 extern "C" uint64_t
 vmNativeCall(void* function, unsigned stackTotal, void* memoryTable,
@@ -150,13 +156,22 @@ dynamicCall(void* function, uintptr_t* arguments, uint8_t* argumentTypes,
             unsigned argumentCount, unsigned argumentsSize,
             unsigned returnType)
 {
+#ifdef __APPLE__
+#  define SKIP(var, count) var += count;
+#  define ALIGN(var)
   const unsigned LinkageArea = 24;
+  const unsigned FprCount = 13;
+#else
+#  define SKIP(var, count)
+#  define ALIGN(var) if (var & 1) ++var;
+  const unsigned LinkageArea = 8;
+  const unsigned FprCount = 8;
+#endif
 
   const unsigned GprCount = 8;
   uintptr_t gprTable[GprCount];
   unsigned gprIndex = 0;
 
-  const unsigned FprCount = 13;
   uint64_t fprTable[FprCount];
   unsigned fprIndex = 0;
 
@@ -172,8 +187,8 @@ dynamicCall(void* function, uintptr_t* arguments, uint8_t* argumentTypes,
         double d = bitsToFloat(arguments[ai]);
         memcpy(fprTable + fprIndex, &d, 8);
         ++ fprIndex;
-        ++ gprIndex;
-        ++ stackSkip;
+        SKIP(gprIndex, 1);
+        SKIP(stackSkip, 1);
       } else {
         stack[stackIndex++] = arguments[ai];
       }
@@ -184,9 +199,10 @@ dynamicCall(void* function, uintptr_t* arguments, uint8_t* argumentTypes,
       if (fprIndex + (8 / BytesPerWord) <= FprCount) {
         memcpy(fprTable + fprIndex, arguments + ai, 8);
         ++ fprIndex;
-        gprIndex += 8 / BytesPerWord;
-        stackSkip += 8 / BytesPerWord;
+        SKIP(gprIndex, 8 / BytesPerWord);
+        SKIP(stackSkip, 8 / BytesPerWord);
       } else {
+        ALIGN(stackIndex);
         memcpy(stack + stackIndex, arguments + ai, 8);
         stackIndex += 8 / BytesPerWord;
       }
@@ -195,10 +211,12 @@ dynamicCall(void* function, uintptr_t* arguments, uint8_t* argumentTypes,
 
     case INT64_TYPE: {
       if (gprIndex + (8 / BytesPerWord) <= GprCount) {
+        ALIGN(gprIndex);
         memcpy(gprTable + gprIndex, arguments + ai, 8);
         gprIndex += 8 / BytesPerWord;
-        stackSkip += 8 / BytesPerWord;
+        SKIP(stackSkip, 8 / BytesPerWord);
       } else {
+        ALIGN(stackIndex);
         memcpy(stack + stackIndex, arguments + ai, 8);
         stackIndex += 8 / BytesPerWord;
       }
@@ -208,7 +226,7 @@ dynamicCall(void* function, uintptr_t* arguments, uint8_t* argumentTypes,
     default: {
       if (gprIndex < GprCount) {
         gprTable[gprIndex++] = arguments[ai];
-        ++ stackSkip;
+        SKIP(stackSkip, 1);
       } else {
         stack[stackIndex++] = arguments[ai];
       }
@@ -219,8 +237,7 @@ dynamicCall(void* function, uintptr_t* arguments, uint8_t* argumentTypes,
 
   return vmNativeCall
     (function,
-     - ((((1 + stackSkip + stackIndex) * BytesPerWord) + LinkageArea + 15)
-        & -16),
+     (((1 + stackSkip + stackIndex) * BytesPerWord) + LinkageArea + 15) & -16,
      stack, stackIndex * BytesPerWord,
      (gprIndex ? gprTable : 0),
      (fprIndex ? fprTable : 0), returnType);