Merge GH #549 MP and SIMD header refactoring

27 files changed, 1554 insertions, 2183 deletions

diff --git a/src/lib/math/mp/info.txt b/src/lib/math/mp/info.txt
index 6aa0142f3..b5db12648 100644
--- a/src/lib/math/mp/info.txt
+++ b/src/lib/math/mp/info.txt
@@ -1,12 +1,10 @@
 define BIGINT_MP 20151225
 
 <source>
-mp_asm.cpp
+mp_core.cpp
 mp_comba.cpp
 mp_karat.cpp
 mp_monty.cpp
-mp_misc.cpp
-mp_shift.cpp
 </source>
 
 <header:public>
@@ -15,8 +13,6 @@ mp_types.h
 
 <header:internal>
 mp_core.h
+mp_madd.h
+mp_asmi.h
 </header:internal>
-
-<requires>
-mp_x86_64|mp_x86_32|mp_x86_32_msvc|mp_generic
-</requires>
diff --git a/src/lib/math/mp/mp_asmi.h b/src/lib/math/mp/mp_asmi.h
new file mode 100644
index 000000000..afb4d1407
--- /dev/null
+++ b/src/lib/math/mp/mp_asmi.h
@@ -0,0 +1,820 @@
+/*
+* Lowest Level MPI Algorithms
+* (C) 1999-2010 Jack Lloyd
+*     2006 Luca Piccarreta
+*
+* Botan is released under the Simplified BSD License (see license.txt)
+*/
+
+#ifndef BOTAN_MP_ASM_INTERNAL_H__
+#define BOTAN_MP_ASM_INTERNAL_H__
+
+#include <botan/internal/mp_madd.h>
+
+namespace Botan {
+
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+
+#define ADDSUB2_OP(OPERATION, INDEX)                     \
+        ASM("movl 4*" #INDEX "(%[y]), %[carry]")         \
+        ASM(OPERATION " %[carry], 4*" #INDEX "(%[x])")   \
+
+#define ADDSUB3_OP(OPERATION, INDEX)                     \
+        ASM("movl 4*" #INDEX "(%[x]), %[carry]")         \
+        ASM(OPERATION " 4*" #INDEX "(%[y]), %[carry]")   \
+        ASM("movl %[carry], 4*" #INDEX "(%[z])")         \
+
+#define LINMUL_OP(WRITE_TO, INDEX)                       \
+        ASM("movl 4*" #INDEX "(%[x]),%%eax")             \
+        ASM("mull %[y]")                                 \
+        ASM("addl %[carry],%%eax")                       \
+        ASM("adcl $0,%%edx")                             \
+        ASM("movl %%edx,%[carry]")                       \
+        ASM("movl %%eax, 4*" #INDEX "(%[" WRITE_TO "])")
+
+#define MULADD_OP(IGNORED, INDEX)                        \
+        ASM("movl 4*" #INDEX "(%[x]),%%eax")             \
+        ASM("mull %[y]")                                 \
+        ASM("addl %[carry],%%eax")                       \
+        ASM("adcl $0,%%edx")                             \
+        ASM("addl 4*" #INDEX "(%[z]),%%eax")             \
+        ASM("adcl $0,%%edx")                             \
+        ASM("movl %%edx,%[carry]")                       \
+        ASM("movl %%eax, 4*" #INDEX " (%[z])")
+
+#define ADD_OR_SUBTRACT(CORE_CODE)     \
+        ASM("rorl %[carry]")           \
+        CORE_CODE                      \
+        ASM("sbbl %[carry],%[carry]")  \
+        ASM("negl %[carry]")
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+#define ADDSUB2_OP(OPERATION, INDEX)                     \
+        ASM("movq 8*" #INDEX "(%[y]), %[carry]")         \
+        ASM(OPERATION " %[carry], 8*" #INDEX "(%[x])")   \
+
+#define ADDSUB3_OP(OPERATION, INDEX)                     \
+        ASM("movq 8*" #INDEX "(%[x]), %[carry]")         \
+        ASM(OPERATION " 8*" #INDEX "(%[y]), %[carry]")   \
+        ASM("movq %[carry], 8*" #INDEX "(%[z])")         \
+
+#define LINMUL_OP(WRITE_TO, INDEX)                       \
+        ASM("movq 8*" #INDEX "(%[x]),%%rax")             \
+        ASM("mulq %[y]")                                 \
+        ASM("addq %[carry],%%rax")                       \
+        ASM("adcq $0,%%rdx")                             \
+        ASM("movq %%rdx,%[carry]")                       \
+        ASM("movq %%rax, 8*" #INDEX "(%[" WRITE_TO "])")
+
+#define MULADD_OP(IGNORED, INDEX)                        \
+        ASM("movq 8*" #INDEX "(%[x]),%%rax")             \
+        ASM("mulq %[y]")                                 \
+        ASM("addq %[carry],%%rax")                       \
+        ASM("adcq $0,%%rdx")                             \
+        ASM("addq 8*" #INDEX "(%[z]),%%rax")             \
+        ASM("adcq $0,%%rdx")                             \
+        ASM("movq %%rdx,%[carry]")                       \
+        ASM("movq %%rax, 8*" #INDEX " (%[z])")
+
+#define ADD_OR_SUBTRACT(CORE_CODE)     \
+        ASM("rorq %[carry]")           \
+        CORE_CODE                      \
+        ASM("sbbq %[carry],%[carry]")  \
+        ASM("negq %[carry]")
+
+#endif
+
+#if defined(ADD_OR_SUBTRACT)
+
+#define ASM(x) x "\n\t"
+
+#define DO_8_TIMES(MACRO, ARG) \
+        MACRO(ARG, 0) \
+        MACRO(ARG, 1) \
+        MACRO(ARG, 2) \
+        MACRO(ARG, 3) \
+        MACRO(ARG, 4) \
+        MACRO(ARG, 5) \
+        MACRO(ARG, 6) \
+        MACRO(ARG, 7)
+
+#endif
+
+/*
+* Word Addition
+*/
+inline word word_add(word x, word y, word* carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(ASM("adcl %[y],%[x]"))
+      : [x]"=r"(x), [carry]"=r"(*carry)
+      : "0"(x), [y]"rm"(y), "1"(*carry)
+      : "cc");
+   return x;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(ASM("adcq %[y],%[x]"))
+      : [x]"=r"(x), [carry]"=r"(*carry)
+      : "0"(x), [y]"rm"(y), "1"(*carry)
+      : "cc");
+   return x;
+
+#else
+   word z = x + y;
+   word c1 = (z < x);
+   z += *carry;
+   *carry = c1 | (z < *carry);
+   return z;
+#endif
+   }
+
+/*
+* Eight Word Block Addition, Two Argument
+*/
+inline word word8_add2(word x[8], const word y[8], word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "adcl"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "adcq"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_32_MSVC_ASM)
+
+   __asm {
+      mov edx,[x]
+      mov esi,[y]
+      xor eax,eax
+      sub eax,[carry] //force CF=1 iff *carry==1
+      mov eax,[esi]
+      adc [edx],eax
+      mov eax,[esi+4]
+      adc [edx+4],eax
+      mov eax,[esi+8]
+      adc [edx+8],eax
+      mov eax,[esi+12]
+      adc [edx+12],eax
+      mov eax,[esi+16]
+      adc [edx+16],eax
+      mov eax,[esi+20]
+      adc [edx+20],eax
+      mov eax,[esi+24]
+      adc [edx+24],eax
+      mov eax,[esi+28]
+      adc [edx+28],eax
+      sbb eax,eax
+      neg eax
+      }
+
+#else
+   x[0] = word_add(x[0], y[0], &carry);
+   x[1] = word_add(x[1], y[1], &carry);
+   x[2] = word_add(x[2], y[2], &carry);
+   x[3] = word_add(x[3], y[3], &carry);
+   x[4] = word_add(x[4], y[4], &carry);
+   x[5] = word_add(x[5], y[5], &carry);
+   x[6] = word_add(x[6], y[6], &carry);
+   x[7] = word_add(x[7], y[7], &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Eight Word Block Addition, Three Argument
+*/
+inline word word8_add3(word z[8], const word x[8],
+                       const word y[8], word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "adcl"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "adcq"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_32_MSVC_ASM)
+
+    __asm {
+      mov edi,[x]
+      mov esi,[y]
+      mov ebx,[z]
+      xor eax,eax
+      sub eax,[carry] //force CF=1 iff *carry==1
+      mov eax,[edi]
+      adc eax,[esi]
+      mov [ebx],eax
+
+      mov eax,[edi+4]
+      adc eax,[esi+4]
+      mov [ebx+4],eax
+
+      mov eax,[edi+8]
+      adc eax,[esi+8]
+      mov [ebx+8],eax
+
+      mov eax,[edi+12]
+      adc eax,[esi+12]
+      mov [ebx+12],eax
+
+      mov eax,[edi+16]
+      adc eax,[esi+16]
+      mov [ebx+16],eax
+
+      mov eax,[edi+20]
+      adc eax,[esi+20]
+      mov [ebx+20],eax
+
+      mov eax,[edi+24]
+      adc eax,[esi+24]
+      mov [ebx+24],eax
+
+      mov eax,[edi+28]
+      adc eax,[esi+28]
+      mov [ebx+28],eax
+
+      sbb eax,eax
+      neg eax
+      }
+
+#else
+   z[0] = word_add(x[0], y[0], &carry);
+   z[1] = word_add(x[1], y[1], &carry);
+   z[2] = word_add(x[2], y[2], &carry);
+   z[3] = word_add(x[3], y[3], &carry);
+   z[4] = word_add(x[4], y[4], &carry);
+   z[5] = word_add(x[5], y[5], &carry);
+   z[6] = word_add(x[6], y[6], &carry);
+   z[7] = word_add(x[7], y[7], &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Word Subtraction
+*/
+inline word word_sub(word x, word y, word* carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(ASM("sbbl %[y],%[x]"))
+      : [x]"=r"(x), [carry]"=r"(*carry)
+      : "0"(x), [y]"rm"(y), "1"(*carry)
+      : "cc");
+   return x;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(ASM("sbbq %[y],%[x]"))
+      : [x]"=r"(x), [carry]"=r"(*carry)
+      : "0"(x), [y]"rm"(y), "1"(*carry)
+      : "cc");
+   return x;
+
+#else
+   word t0 = x - y;
+   word c1 = (t0 > x);
+   word z = t0 - *carry;
+   *carry = c1 | (z > t0);
+   return z;
+#endif
+   }
+
+/*
+* Eight Word Block Subtraction, Two Argument
+*/
+inline word word8_sub2(word x[8], const word y[8], word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "sbbl"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "sbbq"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_32_MSVC_ASM)
+
+    __asm {
+      mov edi,[x]
+      mov esi,[y]
+      xor eax,eax
+      sub eax,[carry] //force CF=1 iff *carry==1
+      mov eax,[edi]
+      sbb eax,[esi]
+      mov [edi],eax
+      mov eax,[edi+4]
+      sbb eax,[esi+4]
+      mov [edi+4],eax
+      mov eax,[edi+8]
+      sbb eax,[esi+8]
+      mov [edi+8],eax
+      mov eax,[edi+12]
+      sbb eax,[esi+12]
+      mov [edi+12],eax
+      mov eax,[edi+16]
+      sbb eax,[esi+16]
+      mov [edi+16],eax
+      mov eax,[edi+20]
+      sbb eax,[esi+20]
+      mov [edi+20],eax
+      mov eax,[edi+24]
+      sbb eax,[esi+24]
+      mov [edi+24],eax
+      mov eax,[edi+28]
+      sbb eax,[esi+28]
+      mov [edi+28],eax
+      sbb eax,eax
+      neg eax
+      }
+
+#else
+   x[0] = word_sub(x[0], y[0], &carry);
+   x[1] = word_sub(x[1], y[1], &carry);
+   x[2] = word_sub(x[2], y[2], &carry);
+   x[3] = word_sub(x[3], y[3], &carry);
+   x[4] = word_sub(x[4], y[4], &carry);
+   x[5] = word_sub(x[5], y[5], &carry);
+   x[6] = word_sub(x[6], y[6], &carry);
+   x[7] = word_sub(x[7], y[7], &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Eight Word Block Subtraction, Two Argument
+*/
+inline word word8_sub2_rev(word x[8], const word y[8], word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbl"))
+      : [carry]"=r"(carry)
+      : [x]"r"(y), [y]"r"(x), [z]"r"(x), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbq"))
+      : [carry]"=r"(carry)
+      : [x]"r"(y), [y]"r"(x), [z]"r"(x), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#else
+   x[0] = word_sub(y[0], x[0], &carry);
+   x[1] = word_sub(y[1], x[1], &carry);
+   x[2] = word_sub(y[2], x[2], &carry);
+   x[3] = word_sub(y[3], x[3], &carry);
+   x[4] = word_sub(y[4], x[4], &carry);
+   x[5] = word_sub(y[5], x[5], &carry);
+   x[6] = word_sub(y[6], x[6], &carry);
+   x[7] = word_sub(y[7], x[7], &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Eight Word Block Subtraction, Three Argument
+*/
+inline word word8_sub3(word z[8], const word x[8],
+                       const word y[8], word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbl"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbq"))
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
+      : "cc", "memory");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_32_MSVC_ASM)
+
+   __asm {
+      mov edi,[x]
+      mov esi,[y]
+      xor eax,eax
+      sub eax,[carry] //force CF=1 iff *carry==1
+      mov ebx,[z]
+      mov eax,[edi]
+      sbb eax,[esi]
+      mov [ebx],eax
+      mov eax,[edi+4]
+      sbb eax,[esi+4]
+      mov [ebx+4],eax
+      mov eax,[edi+8]
+      sbb eax,[esi+8]
+      mov [ebx+8],eax
+      mov eax,[edi+12]
+      sbb eax,[esi+12]
+      mov [ebx+12],eax
+      mov eax,[edi+16]
+      sbb eax,[esi+16]
+      mov [ebx+16],eax
+      mov eax,[edi+20]
+      sbb eax,[esi+20]
+      mov [ebx+20],eax
+      mov eax,[edi+24]
+      sbb eax,[esi+24]
+      mov [ebx+24],eax
+      mov eax,[edi+28]
+      sbb eax,[esi+28]
+      mov [ebx+28],eax
+      sbb eax,eax
+      neg eax
+      }
+
+#else
+   z[0] = word_sub(x[0], y[0], &carry);
+   z[1] = word_sub(x[1], y[1], &carry);
+   z[2] = word_sub(x[2], y[2], &carry);
+   z[3] = word_sub(x[3], y[3], &carry);
+   z[4] = word_sub(x[4], y[4], &carry);
+   z[5] = word_sub(x[5], y[5], &carry);
+   z[6] = word_sub(x[6], y[6], &carry);
+   z[7] = word_sub(x[7], y[7], &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Eight Word Block Linear Multiplication
+*/
+inline word word8_linmul2(word x[8], word y, word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      DO_8_TIMES(LINMUL_OP, "x")
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"rm"(y), "0"(carry)
+      : "cc", "%eax", "%edx");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      DO_8_TIMES(LINMUL_OP, "x")
+      : [carry]"=r"(carry)
+      : [x]"r"(x), [y]"rm"(y), "0"(carry)
+      : "cc", "%rax", "%rdx");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_32_MSVC_ASM)
+
+   __asm {
+      mov esi,[x]
+      mov eax,[esi]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,[carry]      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi],eax        //load a
+
+      mov eax,[esi+4]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi+4],eax        //load a
+
+      mov eax,[esi+8]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi+8],eax        //load a
+
+      mov eax,[esi+12]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi+12],eax        //load a
+
+      mov eax,[esi+16]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi+16],eax        //load a
+
+      mov eax,[esi+20]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi+20],eax        //load a
+
+      mov eax,[esi+24]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [esi+24],eax        //load a
+
+      mov eax,[esi+28]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov [esi+28],eax        //load a
+
+      mov eax,edx      //store carry
+      }
+
+#else
+   x[0] = word_madd2(x[0], y, &carry);
+   x[1] = word_madd2(x[1], y, &carry);
+   x[2] = word_madd2(x[2], y, &carry);
+   x[3] = word_madd2(x[3], y, &carry);
+   x[4] = word_madd2(x[4], y, &carry);
+   x[5] = word_madd2(x[5], y, &carry);
+   x[6] = word_madd2(x[6], y, &carry);
+   x[7] = word_madd2(x[7], y, &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Eight Word Block Linear Multiplication
+*/
+inline word word8_linmul3(word z[8], const word x[8], word y, word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      DO_8_TIMES(LINMUL_OP, "z")
+      : [carry]"=r"(carry)
+      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
+      : "cc", "%eax", "%edx");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+   asm(
+      DO_8_TIMES(LINMUL_OP, "z")
+      : [carry]"=r"(carry)
+      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
+      : "cc", "%rax", "%rdx");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_32_MSVC_ASM)
+
+   __asm {
+      mov edi,[z]
+      mov esi,[x]
+      mov eax,[esi]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,[carry]    //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi],eax        //load a
+
+      mov eax,[esi+4]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi+4],eax        //load a
+
+      mov eax,[esi+8]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi+8],eax        //load a
+
+      mov eax,[esi+12]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi+12],eax        //load a
+
+      mov eax,[esi+16]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi+16],eax        //load a
+
+      mov eax,[esi+20]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi+20],eax        //load a
+
+      mov eax,[esi+24]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov ecx,edx      //store carry
+      mov [edi+24],eax        //load a
+
+      mov eax,[esi+28]        //load a
+      mul [y]           //edx(hi):eax(lo)=a*b
+      add eax,ecx      //sum lo carry
+      adc edx,0          //sum hi carry
+      mov [edi+28],eax        //load a
+      mov eax,edx      //store carry
+      }
+
+#else
+   z[0] = word_madd2(x[0], y, &carry);
+   z[1] = word_madd2(x[1], y, &carry);
+   z[2] = word_madd2(x[2], y, &carry);
+   z[3] = word_madd2(x[3], y, &carry);
+   z[4] = word_madd2(x[4], y, &carry);
+   z[5] = word_madd2(x[5], y, &carry);
+   z[6] = word_madd2(x[6], y, &carry);
+   z[7] = word_madd2(x[7], y, &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Eight Word Block Multiply/Add
+*/
+inline word word8_madd3(word z[8], const word x[8], word y, word carry)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      DO_8_TIMES(MULADD_OP, "")
+      : [carry]"=r"(carry)
+      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
+      : "cc", "%eax", "%edx");
+   return carry;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      DO_8_TIMES(MULADD_OP, "")
+      : [carry]"=r"(carry)
+      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
+      : "cc", "%rax", "%rdx");
+   return carry;
+
+#else
+   z[0] = word_madd3(x[0], y, z[0], &carry);
+   z[1] = word_madd3(x[1], y, z[1], &carry);
+   z[2] = word_madd3(x[2], y, z[2], &carry);
+   z[3] = word_madd3(x[3], y, z[3], &carry);
+   z[4] = word_madd3(x[4], y, z[4], &carry);
+   z[5] = word_madd3(x[5], y, z[5], &carry);
+   z[6] = word_madd3(x[6], y, z[6], &carry);
+   z[7] = word_madd3(x[7], y, z[7], &carry);
+   return carry;
+#endif
+   }
+
+/*
+* Multiply-Add Accumulator
+* (w2,w1,w0) += x * y
+*/
+inline void word3_muladd(word* w2, word* w1, word* w0, word x, word y)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ASM("mull %[y]")
+
+      ASM("addl %[x],%[w0]")
+      ASM("adcl %[y],%[w1]")
+      ASM("adcl $0,%[w2]")
+
+      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
+      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
+      : "cc");
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ASM("mulq %[y]")
+
+      ASM("addq %[x],%[w0]")
+      ASM("adcq %[y],%[w1]")
+      ASM("adcq $0,%[w2]")
+
+      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
+      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
+      : "cc");
+
+#else
+   word carry = *w0;
+   *w0 = word_madd2(x, y, &carry);
+   *w1 += carry;
+   *w2 += (*w1 < carry) ? 1 : 0;
+#endif
+   }
+
+/*
+* Multiply-Add Accumulator
+* (w2,w1,w0) += 2 * x * y
+*/
+inline void word3_muladd_2(word* w2, word* w1, word* w0, word x, word y)
+   {
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ASM("mull %[y]")
+
+      ASM("addl %[x],%[w0]")
+      ASM("adcl %[y],%[w1]")
+      ASM("adcl $0,%[w2]")
+
+      ASM("addl %[x],%[w0]")
+      ASM("adcl %[y],%[w1]")
+      ASM("adcl $0,%[w2]")
+
+      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
+      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
+      : "cc");
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+
+   asm(
+      ASM("mulq %[y]")
+
+      ASM("addq %[x],%[w0]")
+      ASM("adcq %[y],%[w1]")
+      ASM("adcq $0,%[w2]")
+
+      ASM("addq %[x],%[w0]")
+      ASM("adcq %[y],%[w1]")
+      ASM("adcq $0,%[w2]")
+
+      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
+      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
+      : "cc");
+
+#else
+   word carry = 0;
+   x = word_madd2(x, y, &carry);
+   y = carry;
+
+   word top = (y >> (BOTAN_MP_WORD_BITS-1));
+   y <<= 1;
+   y |= (x >> (BOTAN_MP_WORD_BITS-1));
+   x <<= 1;
+
+   carry = 0;
+   *w0 = word_add(*w0, x, &carry);
+   *w1 = word_add(*w1, y, &carry);
+   *w2 = word_add(*w2, top, &carry);
+#endif
+   }
+
+#if defined(ASM)
+  #undef ASM
+  #undef DO_8_TIMES
+  #undef ADD_OR_SUBTRACT
+  #undef ADDSUB2_OP
+  #undef ADDSUB3_OP
+  #undef LINMUL_OP
+  #undef MULADD_OP
+#endif
+
+}
+
+#endif
diff --git a/src/lib/math/mp/mp_asm.cpp b/src/lib/math/mp/mp_core.cpp
index cfbb027d7..2a0b08f67 100644
--- a/src/lib/math/mp/mp_asm.cpp
+++ b/src/lib/math/mp/mp_core.cpp
@@ -8,7 +8,6 @@
 
 #include <botan/internal/mp_core.h>
 #include <botan/internal/mp_asmi.h>
-#include <botan/internal/mp_core.h>
 #include <botan/internal/ct_utils.h>
 #include <botan/exceptn.h>
 #include <botan/mem_ops.h>
@@ -253,4 +252,189 @@ void bigint_linmul3(word z[], const word x[], size_t x_size, word y)
    z[x_size] = carry;
    }
 
+/*
+* Single Operand Left Shift
+*/
+void bigint_shl1(word x[], size_t x_size, size_t word_shift, size_t bit_shift)
+   {
+   if(word_shift)
+      {
+      copy_mem(x + word_shift, x, x_size);
+      clear_mem(x, word_shift);
+      }
+
+   if(bit_shift)
+      {
+      word carry = 0;
+      for(size_t j = word_shift; j != x_size + word_shift + 1; ++j)
+         {
+         word temp = x[j];
+         x[j] = (temp << bit_shift) | carry;
+         carry = (temp >> (MP_WORD_BITS - bit_shift));
+         }
+      }
+   }
+
+/*
+* Single Operand Right Shift
+*/
+void bigint_shr1(word x[], size_t x_size, size_t word_shift, size_t bit_shift)
+   {
+   if(x_size < word_shift)
+      {
+      clear_mem(x, x_size);
+      return;
+      }
+
+   if(word_shift)
+      {
+      copy_mem(x, x + word_shift, x_size - word_shift);
+      clear_mem(x + x_size - word_shift, word_shift);
+      }
+
+   if(bit_shift)
+      {
+      word carry = 0;
+
+      size_t top = x_size - word_shift;
+
+      while(top >= 4)
+         {
+         word w = x[top-1];
+         x[top-1] = (w >> bit_shift) | carry;
+         carry = (w << (MP_WORD_BITS - bit_shift));
+
+         w = x[top-2];
+         x[top-2] = (w >> bit_shift) | carry;
+         carry = (w << (MP_WORD_BITS - bit_shift));
+
+         w = x[top-3];
+         x[top-3] = (w >> bit_shift) | carry;
+         carry = (w << (MP_WORD_BITS - bit_shift));
+
+         w = x[top-4];
+         x[top-4] = (w >> bit_shift) | carry;
+         carry = (w << (MP_WORD_BITS - bit_shift));
+
+         top -= 4;
+         }
+
+      while(top)
+         {
+         word w = x[top-1];
+         x[top-1] = (w >> bit_shift) | carry;
+         carry = (w << (MP_WORD_BITS - bit_shift));
+
+         top--;
+         }
+      }
+   }
+
+/*
+* Two Operand Left Shift
+*/
+void bigint_shl2(word y[], const word x[], size_t x_size,
+                 size_t word_shift, size_t bit_shift)
+   {
+   for(size_t j = 0; j != x_size; ++j)
+      y[j + word_shift] = x[j];
+   if(bit_shift)
+      {
+      word carry = 0;
+      for(size_t j = word_shift; j != x_size + word_shift + 1; ++j)
+         {
+         word w = y[j];
+         y[j] = (w << bit_shift) | carry;
+         carry = (w >> (MP_WORD_BITS - bit_shift));
+         }
+      }
+   }
+
+/*
+* Two Operand Right Shift
+*/
+void bigint_shr2(word y[], const word x[], size_t x_size,
+                 size_t word_shift, size_t bit_shift)
+   {
+   if(x_size < word_shift) return;
+
+   for(size_t j = 0; j != x_size - word_shift; ++j)
+      y[j] = x[j + word_shift];
+   if(bit_shift)
+      {
+      word carry = 0;
+      for(size_t j = x_size - word_shift; j > 0; --j)
+         {
+         word w = y[j-1];
+         y[j-1] = (w >> bit_shift) | carry;
+         carry = (w << (MP_WORD_BITS - bit_shift));
+         }
+      }
+   }
+
+/*
+* Compare two MP integers
+*/
+s32bit bigint_cmp(const word x[], size_t x_size,
+                  const word y[], size_t y_size)
+   {
+   if(x_size < y_size) { return (-bigint_cmp(y, y_size, x, x_size)); }
+
+   while(x_size > y_size)
+      {
+      if(x[x_size-1])
+         return 1;
+      x_size--;
+      }
+
+   for(size_t i = x_size; i > 0; --i)
+      {
+      if(x[i-1] > y[i-1])
+         return 1;
+      if(x[i-1] < y[i-1])
+         return -1;
+      }
+
+   return 0;
+   }
+
+/*
+* Do a 2-word/1-word Division
+*/
+word bigint_divop(word n1, word n0, word d)
+   {
+   if(d == 0)
+      throw Invalid_Argument("bigint_divop divide by zero");
+
+   word high = n1 % d, quotient = 0;
+
+   for(size_t i = 0; i != MP_WORD_BITS; ++i)
+      {
+      word high_top_bit = (high & MP_WORD_TOP_BIT);
+
+      high <<= 1;
+      high |= (n0 >> (MP_WORD_BITS-1-i)) & 1;
+      quotient <<= 1;
+
+      if(high_top_bit || high >= d)
+         {
+         high -= d;
+         quotient |= 1;
+         }
+      }
+
+   return quotient;
+   }
+
+/*
+* Do a 2-word/1-word Modulo
+*/
+word bigint_modop(word n1, word n0, word d)
+   {
+   word z = bigint_divop(n1, n0, d);
+   word dummy = 0;
+   z = word_madd2(z, d, &dummy);
+   return (n0-z);
+   }
+
 }
diff --git a/src/lib/math/mp/mp_generic/info.txt b/src/lib/math/mp/mp_generic/info.txt
deleted file mode 100644
index c87dd00ca..000000000
--- a/src/lib/math/mp/mp_generic/info.txt
+++ /dev/null
@@ -1,6 +0,0 @@
-load_on dep
-
-<header:internal>
-mp_madd.h
-mp_asmi.h
-</header:internal>
diff --git a/src/lib/math/mp/mp_generic/mp_asmi.h b/src/lib/math/mp/mp_generic/mp_asmi.h
deleted file mode 100644
index 708afdfa0..000000000
--- a/src/lib/math/mp/mp_generic/mp_asmi.h
+++ /dev/null
@@ -1,203 +0,0 @@
-/*
-* Lowest Level MPI Algorithms
-* (C) 1999-2010 Jack Lloyd
-*     2006 Luca Piccarreta
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_MP_ASM_INTERNAL_H__
-#define BOTAN_MP_ASM_INTERNAL_H__
-
-#include <botan/internal/mp_madd.h>
-
-namespace Botan {
-
-/*
-* Word Addition
-*/
-inline word word_add(word x, word y, word* carry)
-   {
-   word z = x + y;
-   word c1 = (z < x);
-   z += *carry;
-   *carry = c1 | (z < *carry);
-   return z;
-   }
-
-/*
-* Eight Word Block Addition, Two Argument
-*/
-inline word word8_add2(word x[8], const word y[8], word carry)
-   {
-   x[0] = word_add(x[0], y[0], &carry);
-   x[1] = word_add(x[1], y[1], &carry);
-   x[2] = word_add(x[2], y[2], &carry);
-   x[3] = word_add(x[3], y[3], &carry);
-   x[4] = word_add(x[4], y[4], &carry);
-   x[5] = word_add(x[5], y[5], &carry);
-   x[6] = word_add(x[6], y[6], &carry);
-   x[7] = word_add(x[7], y[7], &carry);
-   return carry;
-   }
-
-/*
-* Eight Word Block Addition, Three Argument
-*/
-inline word word8_add3(word z[8], const word x[8],
-                       const word y[8], word carry)
-   {
-   z[0] = word_add(x[0], y[0], &carry);
-   z[1] = word_add(x[1], y[1], &carry);
-   z[2] = word_add(x[2], y[2], &carry);
-   z[3] = word_add(x[3], y[3], &carry);
-   z[4] = word_add(x[4], y[4], &carry);
-   z[5] = word_add(x[5], y[5], &carry);
-   z[6] = word_add(x[6], y[6], &carry);
-   z[7] = word_add(x[7], y[7], &carry);
-   return carry;
-   }
-
-/*
-* Word Subtraction
-*/
-inline word word_sub(word x, word y, word* carry)
-   {
-   word t0 = x - y;
-   word c1 = (t0 > x);
-   word z = t0 - *carry;
-   *carry = c1 | (z > t0);
-   return z;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2(word x[8], const word y[8], word carry)
-   {
-   x[0] = word_sub(x[0], y[0], &carry);
-   x[1] = word_sub(x[1], y[1], &carry);
-   x[2] = word_sub(x[2], y[2], &carry);
-   x[3] = word_sub(x[3], y[3], &carry);
-   x[4] = word_sub(x[4], y[4], &carry);
-   x[5] = word_sub(x[5], y[5], &carry);
-   x[6] = word_sub(x[6], y[6], &carry);
-   x[7] = word_sub(x[7], y[7], &carry);
-   return carry;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2_rev(word x[8], const word y[8], word carry)
-   {
-   x[0] = word_sub(y[0], x[0], &carry);
-   x[1] = word_sub(y[1], x[1], &carry);
-   x[2] = word_sub(y[2], x[2], &carry);
-   x[3] = word_sub(y[3], x[3], &carry);
-   x[4] = word_sub(y[4], x[4], &carry);
-   x[5] = word_sub(y[5], x[5], &carry);
-   x[6] = word_sub(y[6], x[6], &carry);
-   x[7] = word_sub(y[7], x[7], &carry);
-   return carry;
-   }
-
-/*
-* Eight Word Block Subtraction, Three Argument
-*/
-inline word word8_sub3(word z[8], const word x[8],
-                       const word y[8], word carry)
-   {
-   z[0] = word_sub(x[0], y[0], &carry);
-   z[1] = word_sub(x[1], y[1], &carry);
-   z[2] = word_sub(x[2], y[2], &carry);
-   z[3] = word_sub(x[3], y[3], &carry);
-   z[4] = word_sub(x[4], y[4], &carry);
-   z[5] = word_sub(x[5], y[5], &carry);
-   z[6] = word_sub(x[6], y[6], &carry);
-   z[7] = word_sub(x[7], y[7], &carry);
-   return carry;
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul2(word x[8], word y, word carry)
-   {
-   x[0] = word_madd2(x[0], y, &carry);
-   x[1] = word_madd2(x[1], y, &carry);
-   x[2] = word_madd2(x[2], y, &carry);
-   x[3] = word_madd2(x[3], y, &carry);
-   x[4] = word_madd2(x[4], y, &carry);
-   x[5] = word_madd2(x[5], y, &carry);
-   x[6] = word_madd2(x[6], y, &carry);
-   x[7] = word_madd2(x[7], y, &carry);
-   return carry;
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul3(word z[8], const word x[8], word y, word carry)
-   {
-   z[0] = word_madd2(x[0], y, &carry);
-   z[1] = word_madd2(x[1], y, &carry);
-   z[2] = word_madd2(x[2], y, &carry);
-   z[3] = word_madd2(x[3], y, &carry);
-   z[4] = word_madd2(x[4], y, &carry);
-   z[5] = word_madd2(x[5], y, &carry);
-   z[6] = word_madd2(x[6], y, &carry);
-   z[7] = word_madd2(x[7], y, &carry);
-   return carry;
-   }
-
-/*
-* Eight Word Block Multiply/Add
-*/
-inline word word8_madd3(word z[8], const word x[8], word y, word carry)
-   {
-   z[0] = word_madd3(x[0], y, z[0], &carry);
-   z[1] = word_madd3(x[1], y, z[1], &carry);
-   z[2] = word_madd3(x[2], y, z[2], &carry);
-   z[3] = word_madd3(x[3], y, z[3], &carry);
-   z[4] = word_madd3(x[4], y, z[4], &carry);
-   z[5] = word_madd3(x[5], y, z[5], &carry);
-   z[6] = word_madd3(x[6], y, z[6], &carry);
-   z[7] = word_madd3(x[7], y, z[7], &carry);
-   return carry;
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd(word* w2, word* w1, word* w0, word a, word b)
-   {
-   word carry = *w0;
-   *w0 = word_madd2(a, b, &carry);
-   *w1 += carry;
-   *w2 += (*w1 < carry) ? 1 : 0;
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd_2(word* w2, word* w1, word* w0, word a, word b)
-   {
-   word carry = 0;
-   a = word_madd2(a, b, &carry);
-   b = carry;
-
-   word top = (b >> (BOTAN_MP_WORD_BITS-1));
-   b <<= 1;
-   b |= (a >> (BOTAN_MP_WORD_BITS-1));
-   a <<= 1;
-
-   carry = 0;
-   *w0 = word_add(*w0, a, &carry);
-   *w1 = word_add(*w1, b, &carry);
-   *w2 = word_add(*w2, top, &carry);
-   }
-
-}
-
-#endif
diff --git a/src/lib/math/mp/mp_generic/mp_madd.h b/src/lib/math/mp/mp_madd.h
index 95a1069a4..0567622d9 100644
--- a/src/lib/math/mp/mp_generic/mp_madd.h
+++ b/src/lib/math/mp/mp_madd.h
@@ -35,12 +35,52 @@ namespace Botan {
   #error BOTAN_MP_WORD_BITS must be 8, 16, 32, or 64
 #endif
 
+#if defined(BOTAN_TARGET_ARCH_IS_X86_32) && (BOTAN_MP_WORD_BITS == 32)
+
+  #if defined(BOTAN_USE_GCC_INLINE_ASM)
+    #define BOTAN_MP_USE_X86_32_ASM
+    #define ASM(x) x "\n\t"
+  #elif defined(BOTAN_TARGET_COMPILER_IS_MSVC)
+    #define BOTAN_MP_USE_X86_32_MSVC_ASM
+  #endif
+
+#elif defined(BOTAN_TARGET_ARCH_IS_X86_64) && (BOTAN_MP_WORD_BITS == 64) && (BOTAN_USE_GCC_INLINE_ASM)
+  #define BOTAN_MP_USE_X86_64_ASM
+  #define ASM(x) x "\n\t"
+#endif
+
+#if defined(BOTAN_MP_USE_X86_32_ASM) || defined(BOTAN_MP_USE_X86_64_ASM)
+  #define ASM(x) x "\n\t"
+#endif
+
 /*
 * Word Multiply/Add
 */
 inline word word_madd2(word a, word b, word* c)
    {
-#if defined(BOTAN_HAS_MP_DWORD)
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ASM("mull %[b]")
+      ASM("addl %[c],%[a]")
+      ASM("adcl $0,%[carry]")
+
+      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*c)
+      : "0"(a), "1"(b), [c]"g"(*c) : "cc");
+
+   return a;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+      asm(
+      ASM("mulq %[b]")
+      ASM("addq %[c],%[a]")
+      ASM("adcq $0,%[carry]")
+
+      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*c)
+      : "0"(a), "1"(b), [c]"g"(*c) : "cc");
+
+   return a;
+
+#elif defined(BOTAN_HAS_MP_DWORD)
    const dword s = static_cast<dword>(a) * b + *c;
    *c = static_cast<word>(s >> BOTAN_MP_WORD_BITS);
    return static_cast<word>(s);
@@ -64,7 +104,37 @@ inline word word_madd2(word a, word b, word* c)
 */
 inline word word_madd3(word a, word b, word c, word* d)
    {
-#if defined(BOTAN_HAS_MP_DWORD)
+#if defined(BOTAN_MP_USE_X86_32_ASM)
+   asm(
+      ASM("mull %[b]")
+
+      ASM("addl %[c],%[a]")
+      ASM("adcl $0,%[carry]")
+
+      ASM("addl %[d],%[a]")
+      ASM("adcl $0,%[carry]")
+
+      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*d)
+      : "0"(a), "1"(b), [c]"g"(c), [d]"g"(*d) : "cc");
+
+   return a;
+
+#elif defined(BOTAN_MP_USE_X86_64_ASM)
+   asm(
+      ASM("mulq %[b]")
+
+      ASM("addq %[c],%[a]")
+      ASM("adcq $0,%[carry]")
+
+      ASM("addq %[d],%[a]")
+      ASM("adcq $0,%[carry]")
+
+      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*d)
+      : "0"(a), "1"(b), [c]"g"(c), [d]"g"(*d) : "cc");
+
+   return a;
+
+#elif defined(BOTAN_HAS_MP_DWORD)
    const dword s = static_cast<dword>(a) * b + c + *d;
    *d = static_cast<word>(s >> BOTAN_MP_WORD_BITS);
    return static_cast<word>(s);
@@ -86,6 +156,10 @@ inline word word_madd3(word a, word b, word c, word* d)
 #endif
    }
 
+#if defined(ASM)
+  #undef ASM
+#endif
+
 }
 
 #endif
diff --git a/src/lib/math/mp/mp_misc.cpp b/src/lib/math/mp/mp_misc.cpp
deleted file mode 100644
index 768543a64..000000000
--- a/src/lib/math/mp/mp_misc.cpp
+++ /dev/null
@@ -1,79 +0,0 @@
-/*
-* MP Misc Functions
-* (C) 1999-2008 Jack Lloyd
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#include <botan/internal/mp_core.h>
-#include <botan/internal/mp_madd.h>
-#include <botan/exceptn.h>
-
-namespace Botan {
-
-/*
-* Compare two MP integers
-*/
-s32bit bigint_cmp(const word x[], size_t x_size,
-                  const word y[], size_t y_size)
-   {
-   if(x_size < y_size) { return (-bigint_cmp(y, y_size, x, x_size)); }
-
-   while(x_size > y_size)
-      {
-      if(x[x_size-1])
-         return 1;
-      x_size--;
-      }
-
-   for(size_t i = x_size; i > 0; --i)
-      {
-      if(x[i-1] > y[i-1])
-         return 1;
-      if(x[i-1] < y[i-1])
-         return -1;
-      }
-
-   return 0;
-   }
-
-/*
-* Do a 2-word/1-word Division
-*/
-word bigint_divop(word n1, word n0, word d)
-   {
-   if(d == 0)
-      throw Invalid_Argument("bigint_divop divide by zero");
-
-   word high = n1 % d, quotient = 0;
-
-   for(size_t i = 0; i != MP_WORD_BITS; ++i)
-      {
-      word high_top_bit = (high & MP_WORD_TOP_BIT);
-
-      high <<= 1;
-      high |= (n0 >> (MP_WORD_BITS-1-i)) & 1;
-      quotient <<= 1;
-
-      if(high_top_bit || high >= d)
-         {
-         high -= d;
-         quotient |= 1;
-         }
-      }
-
-   return quotient;
-   }
-
-/*
-* Do a 2-word/1-word Modulo
-*/
-word bigint_modop(word n1, word n0, word d)
-   {
-   word z = bigint_divop(n1, n0, d);
-   word dummy = 0;
-   z = word_madd2(z, d, &dummy);
-   return (n0-z);
-   }
-
-}
diff --git a/src/lib/math/mp/mp_shift.cpp b/src/lib/math/mp/mp_shift.cpp
deleted file mode 100644
index 1850888a0..000000000
--- a/src/lib/math/mp/mp_shift.cpp
+++ /dev/null
@@ -1,133 +0,0 @@
-/*
-* MP Shift Algorithms
-* (C) 1999-2007,2014 Jack Lloyd
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#include <botan/internal/mp_core.h>
-#include <botan/mem_ops.h>
-
-namespace Botan {
-
-/*
-* Single Operand Left Shift
-*/
-void bigint_shl1(word x[], size_t x_size, size_t word_shift, size_t bit_shift)
-   {
-   if(word_shift)
-      {
-      copy_mem(x + word_shift, x, x_size);
-      clear_mem(x, word_shift);
-      }
-
-   if(bit_shift)
-      {
-      word carry = 0;
-      for(size_t j = word_shift; j != x_size + word_shift + 1; ++j)
-         {
-         word temp = x[j];
-         x[j] = (temp << bit_shift) | carry;
-         carry = (temp >> (MP_WORD_BITS - bit_shift));
-         }
-      }
-   }
-
-/*
-* Single Operand Right Shift
-*/
-void bigint_shr1(word x[], size_t x_size, size_t word_shift, size_t bit_shift)
-   {
-   if(x_size < word_shift)
-      {
-      clear_mem(x, x_size);
-      return;
-      }
-
-   if(word_shift)
-      {
-      copy_mem(x, x + word_shift, x_size - word_shift);
-      clear_mem(x + x_size - word_shift, word_shift);
-      }
-
-   if(bit_shift)
-      {
-      word carry = 0;
-
-      size_t top = x_size - word_shift;
-
-      while(top >= 4)
-         {
-         word w = x[top-1];
-         x[top-1] = (w >> bit_shift) | carry;
-         carry = (w << (MP_WORD_BITS - bit_shift));
-
-         w = x[top-2];
-         x[top-2] = (w >> bit_shift) | carry;
-         carry = (w << (MP_WORD_BITS - bit_shift));
-
-         w = x[top-3];
-         x[top-3] = (w >> bit_shift) | carry;
-         carry = (w << (MP_WORD_BITS - bit_shift));
-
-         w = x[top-4];
-         x[top-4] = (w >> bit_shift) | carry;
-         carry = (w << (MP_WORD_BITS - bit_shift));
-
-         top -= 4;
-         }
-
-      while(top)
-         {
-         word w = x[top-1];
-         x[top-1] = (w >> bit_shift) | carry;
-         carry = (w << (MP_WORD_BITS - bit_shift));
-
-         top--;
-         }
-      }
-   }
-
-/*
-* Two Operand Left Shift
-*/
-void bigint_shl2(word y[], const word x[], size_t x_size,
-                 size_t word_shift, size_t bit_shift)
-   {
-   for(size_t j = 0; j != x_size; ++j)
-      y[j + word_shift] = x[j];
-   if(bit_shift)
-      {
-      word carry = 0;
-      for(size_t j = word_shift; j != x_size + word_shift + 1; ++j)
-         {
-         word w = y[j];
-         y[j] = (w << bit_shift) | carry;
-         carry = (w >> (MP_WORD_BITS - bit_shift));
-         }
-      }
-   }
-
-/*
-* Two Operand Right Shift
-*/
-void bigint_shr2(word y[], const word x[], size_t x_size,
-                 size_t word_shift, size_t bit_shift)
-   {
-   if(x_size < word_shift) return;
-
-   for(size_t j = 0; j != x_size - word_shift; ++j)
-      y[j] = x[j + word_shift];
-   if(bit_shift)
-      {
-      word carry = 0;
-      for(size_t j = x_size - word_shift; j > 0; --j)
-         {
-         word w = y[j-1];
-         y[j-1] = (w >> bit_shift) | carry;
-         carry = (w << (MP_WORD_BITS - bit_shift));
-         }
-      }
-   }
-
-}
diff --git a/src/lib/math/mp/mp_x86_32/info.txt b/src/lib/math/mp/mp_x86_32/info.txt
deleted file mode 100644
index f36abaf62..000000000
--- a/src/lib/math/mp/mp_x86_32/info.txt
+++ /dev/null
@@ -1,18 +0,0 @@
-load_on dep
-
-mp_bits 32
-
-<header:internal>
-mp_madd.h
-mp_asmi.h
-</header:internal>
-
-<arch>
-x86_32
-</arch>
-
-<cc>
-clang
-gcc
-icc
-</cc>
diff --git a/src/lib/math/mp/mp_x86_32/mp_asmi.h b/src/lib/math/mp/mp_x86_32/mp_asmi.h
deleted file mode 100644
index 95af89fc0..000000000
--- a/src/lib/math/mp/mp_x86_32/mp_asmi.h
+++ /dev/null
@@ -1,236 +0,0 @@
-/*
-* Lowest Level MPI Algorithms
-* (C) 1999-2010 Jack Lloyd
-*     2006 Luca Piccarreta
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_MP_ASM_INTERNAL_H__
-#define BOTAN_MP_ASM_INTERNAL_H__
-
-#include <botan/internal/mp_madd.h>
-
-namespace Botan {
-
-/*
-* Helper Macros for x86 Assembly
-*/
-#ifndef ASM
-  #define ASM(x) x "\n\t"
-#endif
-
-#define ADDSUB2_OP(OPERATION, INDEX)                     \
-        ASM("movl 4*" #INDEX "(%[y]), %[carry]")         \
-        ASM(OPERATION " %[carry], 4*" #INDEX "(%[x])")   \
-
-#define ADDSUB3_OP(OPERATION, INDEX)                     \
-        ASM("movl 4*" #INDEX "(%[x]), %[carry]")         \
-        ASM(OPERATION " 4*" #INDEX "(%[y]), %[carry]")   \
-        ASM("movl %[carry], 4*" #INDEX "(%[z])")         \
-
-#define LINMUL_OP(WRITE_TO, INDEX)                       \
-        ASM("movl 4*" #INDEX "(%[x]),%%eax")             \
-        ASM("mull %[y]")                                 \
-        ASM("addl %[carry],%%eax")                       \
-        ASM("adcl $0,%%edx")                             \
-        ASM("movl %%edx,%[carry]")                       \
-        ASM("movl %%eax, 4*" #INDEX "(%[" WRITE_TO "])")
-
-#define MULADD_OP(IGNORED, INDEX)                        \
-        ASM("movl 4*" #INDEX "(%[x]),%%eax")             \
-        ASM("mull %[y]")                                 \
-        ASM("addl %[carry],%%eax")                       \
-        ASM("adcl $0,%%edx")                             \
-        ASM("addl 4*" #INDEX "(%[z]),%%eax")             \
-        ASM("adcl $0,%%edx")                             \
-        ASM("movl %%edx,%[carry]")                       \
-        ASM("movl %%eax, 4*" #INDEX " (%[z])")
-
-#define DO_8_TIMES(MACRO, ARG) \
-        MACRO(ARG, 0) \
-        MACRO(ARG, 1) \
-        MACRO(ARG, 2) \
-        MACRO(ARG, 3) \
-        MACRO(ARG, 4) \
-        MACRO(ARG, 5) \
-        MACRO(ARG, 6) \
-        MACRO(ARG, 7)
-
-#define ADD_OR_SUBTRACT(CORE_CODE)     \
-        ASM("rorl %[carry]")           \
-        CORE_CODE                      \
-        ASM("sbbl %[carry],%[carry]")  \
-        ASM("negl %[carry]")
-
-/*
-* Word Addition
-*/
-inline word word_add(word x, word y, word* carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(ASM("adcl %[y],%[x]"))
-      : [x]"=r"(x), [carry]"=r"(*carry)
-      : "0"(x), [y]"rm"(y), "1"(*carry)
-      : "cc");
-   return x;
-   }
-
-/*
-* Eight Word Block Addition, Two Argument
-*/
-inline word word8_add2(word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "adcl"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Addition, Three Argument
-*/
-inline word word8_add3(word z[8], const word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "adcl"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Word Subtraction
-*/
-inline word word_sub(word x, word y, word* carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(ASM("sbbl %[y],%[x]"))
-      : [x]"=r"(x), [carry]"=r"(*carry)
-      : "0"(x), [y]"rm"(y), "1"(*carry)
-      : "cc");
-   return x;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2(word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "sbbl"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2_rev(word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbl"))
-      : [carry]"=r"(carry)
-      : [x]"r"(y), [y]"r"(x), [z]"r"(x), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Subtraction, Three Argument
-*/
-inline word word8_sub3(word z[8], const word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbl"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul2(word x[8], word y, word carry)
-   {
-   asm(
-      DO_8_TIMES(LINMUL_OP, "x")
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"rm"(y), "0"(carry)
-      : "cc", "%eax", "%edx");
-   return carry;
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul3(word z[8], const word x[8], word y, word carry)
-   {
-   asm(
-      DO_8_TIMES(LINMUL_OP, "z")
-      : [carry]"=r"(carry)
-      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
-      : "cc", "%eax", "%edx");
-   return carry;
-   }
-
-/*
-* Eight Word Block Multiply/Add
-*/
-inline word word8_madd3(word z[8], const word x[8], word y, word carry)
-   {
-   asm(
-      DO_8_TIMES(MULADD_OP, "")
-      : [carry]"=r"(carry)
-      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
-      : "cc", "%eax", "%edx");
-   return carry;
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd(word* w2, word* w1, word* w0, word x, word y)
-   {
-   asm(
-      ASM("mull %[y]")
-
-      ASM("addl %[x],%[w0]")
-      ASM("adcl %[y],%[w1]")
-      ASM("adcl $0,%[w2]")
-
-      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
-      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
-      : "cc");
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd_2(word* w2, word* w1, word* w0, word x, word y)
-   {
-   asm(
-      ASM("mull %[y]")
-
-      ASM("addl %[x],%[w0]")
-      ASM("adcl %[y],%[w1]")
-      ASM("adcl $0,%[w2]")
-
-      ASM("addl %[x],%[w0]")
-      ASM("adcl %[y],%[w1]")
-      ASM("adcl $0,%[w2]")
-
-      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
-      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
-      : "cc");
-   }
-
-}
-
-#endif
diff --git a/src/lib/math/mp/mp_x86_32/mp_madd.h b/src/lib/math/mp/mp_x86_32/mp_madd.h
deleted file mode 100644
index 9c0990398..000000000
--- a/src/lib/math/mp/mp_x86_32/mp_madd.h
+++ /dev/null
@@ -1,63 +0,0 @@
-/*
-* Lowest Level MPI Algorithms
-* (C) 1999-2008 Jack Lloyd
-*     2006 Luca Piccarreta
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_MP_WORD_MULADD_H__
-#define BOTAN_MP_WORD_MULADD_H__
-
-#include <botan/mp_types.h>
-
-#if (BOTAN_MP_WORD_BITS != 32)
-   #error The mp_x86_32 module requires that BOTAN_MP_WORD_BITS == 32
-#endif
-
-namespace Botan {
-
-/*
-* Helper Macros for x86 Assembly
-*/
-#define ASM(x) x "\n\t"
-
-/*
-* Word Multiply
-*/
-inline word word_madd2(word a, word b, word* c)
-   {
-   asm(
-      ASM("mull %[b]")
-      ASM("addl %[c],%[a]")
-      ASM("adcl $0,%[carry]")
-
-      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*c)
-      : "0"(a), "1"(b), [c]"g"(*c) : "cc");
-
-   return a;
-   }
-
-/*
-* Word Multiply/Add
-*/
-inline word word_madd3(word a, word b, word c, word* d)
-   {
-   asm(
-      ASM("mull %[b]")
-
-      ASM("addl %[c],%[a]")
-      ASM("adcl $0,%[carry]")
-
-      ASM("addl %[d],%[a]")
-      ASM("adcl $0,%[carry]")
-
-      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*d)
-      : "0"(a), "1"(b), [c]"g"(c), [d]"g"(*d) : "cc");
-
-   return a;
-   }
-
-}
-
-#endif
diff --git a/src/lib/math/mp/mp_x86_32_msvc/info.txt b/src/lib/math/mp/mp_x86_32_msvc/info.txt
deleted file mode 100644
index 3029d6a61..000000000
--- a/src/lib/math/mp/mp_x86_32_msvc/info.txt
+++ /dev/null
@@ -1,16 +0,0 @@
-mp_bits 32
-
-load_on dep
-
-<header:internal>
-mp_generic:mp_madd.h
-mp_asmi.h
-</header:internal>
-
-<arch>
-x86_32
-</arch>
-
-<cc>
-msvc
-</cc>
diff --git a/src/lib/math/mp/mp_x86_32_msvc/mp_asmi.h b/src/lib/math/mp/mp_x86_32_msvc/mp_asmi.h
deleted file mode 100644
index 92bf7980d..000000000
--- a/src/lib/math/mp/mp_x86_32_msvc/mp_asmi.h
+++ /dev/null
@@ -1,454 +0,0 @@
-/*
-* Lowest Level MPI Algorithms
-* (C) 1999-2010 Jack Lloyd
-*     2006 Luca Piccarreta
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_MP_ASM_INTERNAL_H__
-#define BOTAN_MP_ASM_INTERNAL_H__
-
-#include <botan/internal/mp_madd.h>
-
-namespace Botan {
-
-/*
-* Word Addition
-*/
-inline word word_add(word x, word y, word* carry)
-   {
-   word z = x + y;
-   word c1 = (z < x);
-   z += *carry;
-   *carry = c1 | (z < *carry);
-   return z;
-   }
-
-/*
-* Eight Word Block Addition, Two Argument
-*/
-inline word word8_add2(word x[8], const word y[8], word carry)
-   {
-   __asm {
-      mov edx,[x]
-      mov esi,[y]
-      xor eax,eax
-      sub eax,[carry] //force CF=1 iff *carry==1
-      mov eax,[esi]
-      adc [edx],eax
-      mov eax,[esi+4]
-      adc [edx+4],eax
-      mov eax,[esi+8]
-      adc [edx+8],eax
-      mov eax,[esi+12]
-      adc [edx+12],eax
-      mov eax,[esi+16]
-      adc [edx+16],eax
-      mov eax,[esi+20]
-      adc [edx+20],eax
-      mov eax,[esi+24]
-      adc [edx+24],eax
-      mov eax,[esi+28]
-      adc [edx+28],eax
-      sbb eax,eax
-      neg eax
-      }
-   }
-
-/*
-* Eight Word Block Addition, Three Argument
-*/
-inline word word8_add3(word z[8], const word x[8], const word y[8], word carry)
-   {
-    __asm {
-      mov edi,[x]
-      mov esi,[y]
-      mov ebx,[z]
-      xor eax,eax
-      sub eax,[carry] //force CF=1 iff *carry==1
-      mov eax,[edi]
-      adc eax,[esi]
-      mov [ebx],eax
-
-      mov eax,[edi+4]
-      adc eax,[esi+4]
-      mov [ebx+4],eax
-
-      mov eax,[edi+8]
-      adc eax,[esi+8]
-      mov [ebx+8],eax
-
-      mov eax,[edi+12]
-      adc eax,[esi+12]
-      mov [ebx+12],eax
-
-      mov eax,[edi+16]
-      adc eax,[esi+16]
-      mov [ebx+16],eax
-
-      mov eax,[edi+20]
-      adc eax,[esi+20]
-      mov [ebx+20],eax
-
-      mov eax,[edi+24]
-      adc eax,[esi+24]
-      mov [ebx+24],eax
-
-      mov eax,[edi+28]
-      adc eax,[esi+28]
-      mov [ebx+28],eax
-
-      sbb eax,eax
-      neg eax
-      }
-   }
-
-/*
-* Word Subtraction
-*/
-inline word word_sub(word x, word y, word* carry)
-   {
-   word t0 = x - y;
-   word c1 = (t0 > x);
-   word z = t0 - *carry;
-   *carry = c1 | (z > t0);
-   return z;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2(word x[8], const word y[8], word carry)
-   {
-    __asm {
-      mov edi,[x]
-      mov esi,[y]
-      xor eax,eax
-      sub eax,[carry] //force CF=1 iff *carry==1
-      mov eax,[edi]
-      sbb eax,[esi]
-      mov [edi],eax
-      mov eax,[edi+4]
-      sbb eax,[esi+4]
-      mov [edi+4],eax
-      mov eax,[edi+8]
-      sbb eax,[esi+8]
-      mov [edi+8],eax
-      mov eax,[edi+12]
-      sbb eax,[esi+12]
-      mov [edi+12],eax
-      mov eax,[edi+16]
-      sbb eax,[esi+16]
-      mov [edi+16],eax
-      mov eax,[edi+20]
-      sbb eax,[esi+20]
-      mov [edi+20],eax
-      mov eax,[edi+24]
-      sbb eax,[esi+24]
-      mov [edi+24],eax
-      mov eax,[edi+28]
-      sbb eax,[esi+28]
-      mov [edi+28],eax
-      sbb eax,eax
-      neg eax
-      }
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2_rev(word x[8], const word y[8], word carry)
-   {
-   x[0] = word_sub(y[0], x[0], &carry);
-   x[1] = word_sub(y[1], x[1], &carry);
-   x[2] = word_sub(y[2], x[2], &carry);
-   x[3] = word_sub(y[3], x[3], &carry);
-   x[4] = word_sub(y[4], x[4], &carry);
-   x[5] = word_sub(y[5], x[5], &carry);
-   x[6] = word_sub(y[6], x[6], &carry);
-   x[7] = word_sub(y[7], x[7], &carry);
-   return carry;
-   }
-
-
-/*
-* Eight Word Block Subtraction, Three Argument
-*/
-inline word word8_sub3(word z[8], const word x[8],
-                       const word y[8], word carry)
-   {
-    __asm {
-      mov edi,[x]
-      mov esi,[y]
-      xor eax,eax
-      sub eax,[carry] //force CF=1 iff *carry==1
-      mov ebx,[z]
-      mov eax,[edi]
-      sbb eax,[esi]
-      mov [ebx],eax
-      mov eax,[edi+4]
-      sbb eax,[esi+4]
-      mov [ebx+4],eax
-      mov eax,[edi+8]
-      sbb eax,[esi+8]
-      mov [ebx+8],eax
-      mov eax,[edi+12]
-      sbb eax,[esi+12]
-      mov [ebx+12],eax
-      mov eax,[edi+16]
-      sbb eax,[esi+16]
-      mov [ebx+16],eax
-      mov eax,[edi+20]
-      sbb eax,[esi+20]
-      mov [ebx+20],eax
-      mov eax,[edi+24]
-      sbb eax,[esi+24]
-      mov [ebx+24],eax
-      mov eax,[edi+28]
-      sbb eax,[esi+28]
-      mov [ebx+28],eax
-      sbb eax,eax
-      neg eax
-      }
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul2(word x[8], word y, word carry)
-   {
-   __asm {
-      mov esi,[x]
-      mov eax,[esi]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,[carry]      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi],eax        //load a
-
-      mov eax,[esi+4]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi+4],eax        //load a
-
-      mov eax,[esi+8]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi+8],eax        //load a
-
-      mov eax,[esi+12]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi+12],eax        //load a
-
-      mov eax,[esi+16]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi+16],eax        //load a
-
-      mov eax,[esi+20]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi+20],eax        //load a
-
-      mov eax,[esi+24]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [esi+24],eax        //load a
-
-      mov eax,[esi+28]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov [esi+28],eax        //load a
-
-      mov eax,edx      //store carry
-      }
-   }
-
-inline word word8_linmul3(word z[4], const word x[4], word y, word carry)
-   {
-   __asm {
-#if 0
-      //it's slower!!!
-      mov edx,[z]
-      mov eax,[x]
-      movd mm7,[y]
-
-      movd mm0,[eax]
-      movd mm1,[eax+4]
-      movd mm2,[eax+8]
-      pmuludq mm0,mm7
-      pmuludq mm1,mm7
-      pmuludq mm2,mm7
-
-      movd mm6,[carry]
-      paddq mm0,mm6
-      movd [edx],mm0
-
-      psrlq mm0,32
-      paddq mm1,mm0
-      movd [edx+4],mm1
-
-      movd mm3,[eax+12]
-      psrlq mm1,32
-      paddq mm2,mm1
-      movd [edx+8],mm2
-
-      pmuludq mm3,mm7
-      movd mm4,[eax+16]
-      psrlq mm2,32
-      paddq mm3,mm2
-      movd [edx+12],mm3
-
-      pmuludq mm4,mm7
-      movd mm5,[eax+20]
-      psrlq mm3,32
-      paddq mm4,mm3
-      movd [edx+16],mm4
-
-      pmuludq mm5,mm7
-      movd mm0,[eax+24]
-      psrlq mm4,32
-      paddq mm5,mm4
-      movd [edx+20],mm5
-
-      pmuludq mm0,mm7
-      movd mm1,[eax+28]
-      psrlq mm5,32
-      paddq mm0,mm5
-      movd [edx+24],mm0
-
-      pmuludq mm1,mm7
-      psrlq mm0,32
-      paddq mm1,mm0
-      movd [edx+28],mm1
-      psrlq mm1,32
-
-      movd eax,mm1
-      emms
-#else
-      mov edi,[z]
-      mov esi,[x]
-      mov eax,[esi]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,[carry]    //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi],eax        //load a
-
-      mov eax,[esi+4]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi+4],eax        //load a
-
-      mov eax,[esi+8]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi+8],eax        //load a
-
-      mov eax,[esi+12]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi+12],eax        //load a
-
-      mov eax,[esi+16]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi+16],eax        //load a
-
-      mov eax,[esi+20]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi+20],eax        //load a
-
-      mov eax,[esi+24]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov ecx,edx      //store carry
-      mov [edi+24],eax        //load a
-
-      mov eax,[esi+28]        //load a
-      mul [y]           //edx(hi):eax(lo)=a*b
-      add eax,ecx      //sum lo carry
-      adc edx,0          //sum hi carry
-      mov [edi+28],eax        //load a
-      mov eax,edx      //store carry
-#endif
-      }
-   }
-
-/*
-* Eight Word Block Multiply/Add
-*/
-inline word word8_madd3(word z[8], const word x[8], word y, word carry)
-   {
-   z[0] = word_madd3(x[0], y, z[0], &carry);
-   z[1] = word_madd3(x[1], y, z[1], &carry);
-   z[2] = word_madd3(x[2], y, z[2], &carry);
-   z[3] = word_madd3(x[3], y, z[3], &carry);
-   z[4] = word_madd3(x[4], y, z[4], &carry);
-   z[5] = word_madd3(x[5], y, z[5], &carry);
-   z[6] = word_madd3(x[6], y, z[6], &carry);
-   z[7] = word_madd3(x[7], y, z[7], &carry);
-   return carry;
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd(word* w2, word* w1, word* w0, word a, word b)
-   {
-   word carry = *w0;
-   *w0 = word_madd2(a, b, &carry);
-   *w1 += carry;
-   *w2 += (*w1 < carry) ? 1 : 0;
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd_2(word* w2, word* w1, word* w0, word a, word b)
-   {
-   word carry = 0;
-   a = word_madd2(a, b, &carry);
-   b = carry;
-
-   word top = (b >> (BOTAN_MP_WORD_BITS-1));
-   b <<= 1;
-   b |= (a >> (BOTAN_MP_WORD_BITS-1));
-   a <<= 1;
-
-   carry = 0;
-   *w0 = word_add(*w0, a, &carry);
-   *w1 = word_add(*w1, b, &carry);
-   *w2 = word_add(*w2, top, &carry);
-   }
-
-}
-
-#endif
diff --git a/src/lib/math/mp/mp_x86_64/info.txt b/src/lib/math/mp/mp_x86_64/info.txt
deleted file mode 100644
index 75c42ddc1..000000000
--- a/src/lib/math/mp/mp_x86_64/info.txt
+++ /dev/null
@@ -1,18 +0,0 @@
-load_on dep
-
-mp_bits 64
-
-<header:internal>
-mp_madd.h
-mp_asmi.h
-</header:internal>
-
-<arch>
-x86_64
-</arch>
-
-<cc>
-clang
-gcc
-icc
-</cc>
diff --git a/src/lib/math/mp/mp_x86_64/mp_asmi.h b/src/lib/math/mp/mp_x86_64/mp_asmi.h
deleted file mode 100644
index cd5884867..000000000
--- a/src/lib/math/mp/mp_x86_64/mp_asmi.h
+++ /dev/null
@@ -1,244 +0,0 @@
-/*
-* Lowest Level MPI Algorithms
-* (C) 1999-2010 Jack Lloyd
-*     2006 Luca Piccarreta
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_MP_ASM_INTERNAL_H__
-#define BOTAN_MP_ASM_INTERNAL_H__
-
-#include <botan/internal/mp_madd.h>
-
-namespace Botan {
-
-/*
-* Helper Macros for x86-64 Assembly
-*/
-#ifndef ASM
-  #define ASM(x) x "\n\t"
-#endif
-
-#define ADDSUB2_OP(OPERATION, INDEX)                     \
-        ASM("movq 8*" #INDEX "(%[y]), %[carry]")         \
-        ASM(OPERATION " %[carry], 8*" #INDEX "(%[x])")   \
-
-#define ADDSUB3_OP(OPERATION, INDEX)                     \
-        ASM("movq 8*" #INDEX "(%[x]), %[carry]")         \
-        ASM(OPERATION " 8*" #INDEX "(%[y]), %[carry]")   \
-        ASM("movq %[carry], 8*" #INDEX "(%[z])")         \
-
-#define LINMUL_OP(WRITE_TO, INDEX)                       \
-        ASM("movq 8*" #INDEX "(%[x]),%%rax")             \
-        ASM("mulq %[y]")                                 \
-        ASM("addq %[carry],%%rax")                       \
-        ASM("adcq $0,%%rdx")                             \
-        ASM("movq %%rdx,%[carry]")                       \
-        ASM("movq %%rax, 8*" #INDEX "(%[" WRITE_TO "])")
-
-#define MULADD_OP(IGNORED, INDEX)                        \
-        ASM("movq 8*" #INDEX "(%[x]),%%rax")             \
-        ASM("mulq %[y]")                                 \
-        ASM("addq %[carry],%%rax")                       \
-        ASM("adcq $0,%%rdx")                             \
-        ASM("addq 8*" #INDEX "(%[z]),%%rax")             \
-        ASM("adcq $0,%%rdx")                             \
-        ASM("movq %%rdx,%[carry]")                       \
-        ASM("movq %%rax, 8*" #INDEX " (%[z])")
-
-#define DO_8_TIMES(MACRO, ARG) \
-        MACRO(ARG, 0) \
-        MACRO(ARG, 1) \
-        MACRO(ARG, 2) \
-        MACRO(ARG, 3) \
-        MACRO(ARG, 4) \
-        MACRO(ARG, 5) \
-        MACRO(ARG, 6) \
-        MACRO(ARG, 7)
-
-#define ADD_OR_SUBTRACT(CORE_CODE)     \
-        ASM("rorq %[carry]")           \
-        CORE_CODE                      \
-        ASM("sbbq %[carry],%[carry]")  \
-        ASM("negq %[carry]")
-
-/*
-* Word Addition
-*/
-inline word word_add(word x, word y, word* carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(ASM("adcq %[y],%[x]"))
-      : [x]"=r"(x), [carry]"=r"(*carry)
-      : "0"(x), [y]"rm"(y), "1"(*carry)
-      : "cc");
-   return x;
-   }
-
-/*
-* Eight Word Block Addition, Two Argument
-*/
-inline word word8_add2(word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "adcq"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Addition, Three Argument
-*/
-inline word word8_add3(word z[8], const word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "adcq"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Word Subtraction
-*/
-inline word word_sub(word x, word y, word* carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(ASM("sbbq %[y],%[x]"))
-      : [x]"=r"(x), [carry]"=r"(*carry)
-      : "0"(x), [y]"rm"(y), "1"(*carry)
-      : "cc");
-   return x;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2(word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB2_OP, "sbbq"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Subtraction, Two Argument
-*/
-inline word word8_sub2_rev(word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbq"))
-      : [carry]"=r"(carry)
-      : [x]"r"(y), [y]"r"(x), [z]"r"(x), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Subtraction, Three Argument
-*/
-inline word word8_sub3(word z[8], const word x[8], const word y[8], word carry)
-   {
-   asm(
-      ADD_OR_SUBTRACT(DO_8_TIMES(ADDSUB3_OP, "sbbq"))
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"r"(y), [z]"r"(z), "0"(carry)
-      : "cc", "memory");
-   return carry;
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul2(word x[8], word y, word carry)
-   {
-   asm(
-      DO_8_TIMES(LINMUL_OP, "x")
-      : [carry]"=r"(carry)
-      : [x]"r"(x), [y]"rm"(y), "0"(carry)
-      : "cc", "%rax", "%rdx");
-   return carry;
-   }
-
-/*
-* Eight Word Block Linear Multiplication
-*/
-inline word word8_linmul3(word z[8], const word x[8], word y, word carry)
-   {
-   asm(
-      DO_8_TIMES(LINMUL_OP, "z")
-      : [carry]"=r"(carry)
-      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
-      : "cc", "%rax", "%rdx");
-   return carry;
-   }
-
-/*
-* Eight Word Block Multiply/Add
-*/
-inline word word8_madd3(word z[8], const word x[8], word y, word carry)
-   {
-   asm(
-      DO_8_TIMES(MULADD_OP, "")
-      : [carry]"=r"(carry)
-      : [z]"r"(z), [x]"r"(x), [y]"rm"(y), "0"(carry)
-      : "cc", "%rax", "%rdx");
-   return carry;
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd(word* w2, word* w1, word* w0, word x, word y)
-   {
-   asm(
-      ASM("mulq %[y]")
-
-      ASM("addq %[x],%[w0]")
-      ASM("adcq %[y],%[w1]")
-      ASM("adcq $0,%[w2]")
-
-      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
-      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
-      : "cc");
-   }
-
-/*
-* Multiply-Add Accumulator
-*/
-inline void word3_muladd_2(word* w2, word* w1, word* w0, word x, word y)
-   {
-   asm(
-      ASM("mulq %[y]")
-
-      ASM("addq %[x],%[w0]")
-      ASM("adcq %[y],%[w1]")
-      ASM("adcq $0,%[w2]")
-
-      ASM("addq %[x],%[w0]")
-      ASM("adcq %[y],%[w1]")
-      ASM("adcq $0,%[w2]")
-
-      : [w0]"=r"(*w0), [w1]"=r"(*w1), [w2]"=r"(*w2)
-      : [x]"a"(x), [y]"d"(y), "0"(*w0), "1"(*w1), "2"(*w2)
-      : "cc");
-   }
-
-#undef ASM
-#undef DO_8_TIMES
-#undef ADD_OR_SUBTRACT
-#undef ADDSUB2_OP
-#undef ADDSUB3_OP
-#undef LINMUL_OP
-#undef MULADD_OP
-
-}
-
-#endif
diff --git a/src/lib/math/mp/mp_x86_64/mp_madd.h b/src/lib/math/mp/mp_x86_64/mp_madd.h
deleted file mode 100644
index 6f9185dc0..000000000
--- a/src/lib/math/mp/mp_x86_64/mp_madd.h
+++ /dev/null
@@ -1,65 +0,0 @@
-/*
-* Lowest Level MPI Algorithms
-* (C) 1999-2008 Jack Lloyd
-*     2006 Luca Piccarreta
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_MP_WORD_MULADD_H__
-#define BOTAN_MP_WORD_MULADD_H__
-
-#include <botan/mp_types.h>
-
-#if (BOTAN_MP_WORD_BITS != 64)
-   #error The mp_x86_64 module requires that BOTAN_MP_WORD_BITS == 64
-#endif
-
-namespace Botan {
-
-/*
-* Helper Macros for x86-64 Assembly
-*/
-#define ASM(x) x "\n\t"
-
-/*
-* Word Multiply
-*/
-inline word word_madd2(word a, word b, word* c)
-   {
-   asm(
-      ASM("mulq %[b]")
-      ASM("addq %[c],%[a]")
-      ASM("adcq $0,%[carry]")
-
-      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*c)
-      : "0"(a), "1"(b), [c]"g"(*c) : "cc");
-
-   return a;
-   }
-
-/*
-* Word Multiply/Add
-*/
-inline word word_madd3(word a, word b, word c, word* d)
-   {
-   asm(
-      ASM("mulq %[b]")
-
-      ASM("addq %[c],%[a]")
-      ASM("adcq $0,%[carry]")
-
-      ASM("addq %[d],%[a]")
-      ASM("adcq $0,%[carry]")
-
-      : [a]"=a"(a), [b]"=rm"(b), [carry]"=&d"(*d)
-      : "0"(a), "1"(b), [c]"g"(c), [d]"g"(*d) : "cc");
-
-   return a;
-   }
-
-#undef ASM
-
-}
-
-#endif
diff --git a/src/lib/utils/cpuid.cpp b/src/lib/utils/cpuid.cpp
index 695a28550..d3def91ed 100644
--- a/src/lib/utils/cpuid.cpp
+++ b/src/lib/utils/cpuid.cpp
@@ -159,14 +159,12 @@ bool altivec_check_pvr_emul()
 
 bool CPUID::has_simd_32()
    {
-#if defined(BOTAN_HAS_SIMD_SSE2)
+#if defined(BOTAN_TARGET_SUPPORTS_SSE2)
    return CPUID::has_sse2();
-#elif defined(BOTAN_HAS_SIMD_ALTIVEC)
+#elif defined(BOTAN_TARGET_SUPPORTS_ALTIVEC)
    return CPUID::has_altivec();
-#elif defined(BOTAN_HAS_SIMD_SCALAR)
-   return true;
 #else
-   return false;
+   return true;
 #endif
    }
 
diff --git a/src/lib/utils/simd/info.txt b/src/lib/utils/simd/info.txt
index 35620c940..6b9e381fa 100644
--- a/src/lib/utils/simd/info.txt
+++ b/src/lib/utils/simd/info.txt
@@ -3,7 +3,3 @@ define SIMD_32 20131128
 <header:internal>
 simd_32.h
 </header:internal>
-
-<requires>
-simd_sse2|simd_altivec|simd_scalar
-</requires>
diff --git a/src/lib/utils/simd/simd_32.h b/src/lib/utils/simd/simd_32.h
index 265e347a9..351146f22 100644
--- a/src/lib/utils/simd/simd_32.h
+++ b/src/lib/utils/simd/simd_32.h
@@ -1,6 +1,6 @@
 /*
 * Lightweight wrappers for SIMD operations
-* (C) 2009,2011 Jack Lloyd
+* (C) 2009,2011,2016 Jack Lloyd
 *
 * Botan is released under the Simplified BSD License (see license.txt)
 */
@@ -9,22 +9,470 @@
 #define BOTAN_SIMD_32_H__
 
 #include <botan/types.h>
+#include <botan/loadstor.h>
+#include <botan/bswap.h>
 
-#if defined(BOTAN_HAS_SIMD_SSE2)
-  #include <botan/internal/simd_sse2.h>
-  namespace Botan { typedef SIMD_SSE2 SIMD_32; }
+#if defined(BOTAN_TARGET_SUPPORTS_SSE2) && 0
+  #include <emmintrin.h>
+  #define BOTAN_SIMD_USE_SSE2
 
-#elif defined(BOTAN_HAS_SIMD_ALTIVEC)
-  #include <botan/internal/simd_altivec.h>
-  namespace Botan { typedef SIMD_Altivec SIMD_32; }
+#elif defined(BOTAN_TARGET_SUPPORTS_ALTIVEC)
+  #include <altivec.h>
+  #undef vector
+  #undef bool
+  #define BOTAN_SIMD_USE_ALTIVEC
+#endif
+
+// TODO: NEON support
+
+namespace Botan {
+
+/**
+* This class is not a general purpose SIMD type, and only offers
+* instructions needed for evaluation of specific crypto primitives.
+* For example it does not currently have equality operators of any
+* kind.
+*/
+class SIMD_4x32
+   {
+   public:
 
-#elif defined(BOTAN_HAS_SIMD_SCALAR)
-  #include <botan/internal/simd_scalar.h>
-  namespace Botan { typedef SIMD_Scalar<u32bit,4> SIMD_32; }
+      SIMD_4x32() // zero initialized
+         {
+#if defined(BOTAN_SIMD_USE_SSE2) || defined(BOTAN_SIMD_USE_ALTIVEC)
+         ::memset(&m_reg, 0, sizeof(m_reg));
+#else
+         ::memset(m_reg, 0, sizeof(m_reg));
+#endif
+         }
 
+      explicit SIMD_4x32(const u32bit B[4])
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_loadu_si128(reinterpret_cast<const __m128i*>(B));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = (__vector unsigned int){B[0], B[1], B[2], B[3]};
 #else
-  #error "No SIMD module defined"
+         m_reg[0] = B[0];
+         m_reg[1] = B[1];
+         m_reg[2] = B[2];
+         m_reg[3] = B[3];
+#endif
+         }
+
+      SIMD_4x32(u32bit B0, u32bit B1, u32bit B2, u32bit B3)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_set_epi32(B0, B1, B2, B3);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = (__vector unsigned int){B0, B1, B2, B3};
+#else
+         m_reg[0] = B0;
+         m_reg[1] = B1;
+         m_reg[2] = B2;
+         m_reg[3] = B3;
+#endif
+         }
+
+      explicit SIMD_4x32(u32bit B)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_set1_epi32(B);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = (__vector unsigned int){B, B, B, B};
+#else
+         m_reg[0] = B;
+         m_reg[1] = B;
+         m_reg[2] = B;
+         m_reg[3] = B;
+#endif
+         }
+
+      static SIMD_4x32 load_le(const void* in)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_loadu_si128(reinterpret_cast<const __m128i*>(in)));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         const u32bit* in_32 = static_cast<const u32bit*>(in);
 
+         __vector unsigned int R0 = vec_ld(0, in_32);
+         __vector unsigned int R1 = vec_ld(12, in_32);
+
+         __vector unsigned char perm = vec_lvsl(0, in_32);
+
+         perm = vec_xor(perm, vec_splat_u8(3));
+
+         R0 = vec_perm(R0, R1, perm);
+
+         return SIMD_4x32(R0);
+#else
+         SIMD_4x32 out;
+         Botan::load_le(out.m_reg, static_cast<const uint8_t*>(in), 4);
+         return out;
+#endif
+         }
+
+      static SIMD_4x32 load_be(const void* in)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return load_le(in).bswap();
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         const u32bit* in_32 = static_cast<const u32bit*>(in);
+
+         __vector unsigned int R0 = vec_ld(0, in_32);
+         __vector unsigned int R1 = vec_ld(12, in_32);
+
+         __vector unsigned char perm = vec_lvsl(0, in_32);
+
+         R0 = vec_perm(R0, R1, perm);
+
+         return SIMD_4x32(R0);
+
+#else
+         SIMD_4x32 out;
+         Botan::load_be(out.m_reg, static_cast<const uint8_t*>(in), 4);
+         return out;
+#endif
+         }
+
+      void store_le(uint8_t out[]) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         _mm_storeu_si128(reinterpret_cast<__m128i*>(out), m_reg);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         __vector unsigned char perm = vec_lvsl(0, static_cast<u32bit*>(nullptr));
+
+         perm = vec_xor(perm, vec_splat_u8(3)); // bswap vector
+
+         union {
+            __vector unsigned int V;
+            u32bit R[4];
+            } vec;
+
+         vec.V = vec_perm(m_reg, m_reg, perm);
+
+         Botan::store_be(out, vec.R[0], vec.R[1], vec.R[2], vec.R[3]);
+#else
+         Botan::store_le(out, m_reg[0], m_reg[1], m_reg[2], m_reg[3]);
+#endif
+         }
+
+      void store_be(uint8_t out[]) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         bswap().store_le(out);
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+
+         union {
+            __vector unsigned int V;
+            u32bit R[4];
+            } vec;
+
+         vec.V = m_reg;
+
+         Botan::store_be(out, vec.R[0], vec.R[1], vec.R[2], vec.R[3]);
+#else
+         Botan::store_be(out, m_reg[0], m_reg[1], m_reg[2], m_reg[3]);
+#endif
+         }
+
+      void rotate_left(size_t rot)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_or_si128(_mm_slli_epi32(m_reg, static_cast<int>(rot)),
+                              _mm_srli_epi32(m_reg, static_cast<int>(32-rot)));
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         const unsigned int r = static_cast<unsigned int>(rot);
+         m_reg = vec_rl(m_reg, (__vector unsigned int){r, r, r, r});
+
+#else
+         m_reg[0] = Botan::rotate_left(m_reg[0], rot);
+         m_reg[1] = Botan::rotate_left(m_reg[1], rot);
+         m_reg[2] = Botan::rotate_left(m_reg[2], rot);
+         m_reg[3] = Botan::rotate_left(m_reg[3], rot);
+#endif
+         }
+
+      void rotate_right(size_t rot)
+         {
+         rotate_left(32 - rot);
+         }
+
+      void operator+=(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_add_epi32(m_reg, other.m_reg);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = vec_add(m_reg, other.m_reg);
+#else
+         m_reg[0] += other.m_reg[0];
+         m_reg[1] += other.m_reg[1];
+         m_reg[2] += other.m_reg[2];
+         m_reg[3] += other.m_reg[3];
+#endif
+         }
+
+      SIMD_4x32 operator+(const SIMD_4x32& other) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_add_epi32(m_reg, other.m_reg));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         return SIMD_4x32(vec_add(m_reg, other.m_reg));
+#else
+         return SIMD_4x32(m_reg[0] + other.m_reg[0],
+                          m_reg[1] + other.m_reg[1],
+                          m_reg[2] + other.m_reg[2],
+                          m_reg[3] + other.m_reg[3]);
+#endif
+         }
+
+      void operator-=(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_sub_epi32(m_reg, other.m_reg);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = vec_sub(m_reg, other.m_reg);
+#else
+         m_reg[0] -= other.m_reg[0];
+         m_reg[1] -= other.m_reg[1];
+         m_reg[2] -= other.m_reg[2];
+         m_reg[3] -= other.m_reg[3];
 #endif
+         }
+
+      SIMD_4x32 operator-(const SIMD_4x32& other) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_sub_epi32(m_reg, other.m_reg));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         return SIMD_4x32(vec_sub(m_reg, other.m_reg));
+#else
+         return SIMD_4x32(m_reg[0] - other.m_reg[0],
+                          m_reg[1] - other.m_reg[1],
+                          m_reg[2] - other.m_reg[2],
+                          m_reg[3] - other.m_reg[3]);
+#endif
+         }
+
+      void operator^=(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_xor_si128(m_reg, other.m_reg);
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = vec_xor(m_reg, other.m_reg);
+#else
+         m_reg[0] ^= other.m_reg[0];
+         m_reg[1] ^= other.m_reg[1];
+         m_reg[2] ^= other.m_reg[2];
+         m_reg[3] ^= other.m_reg[3];
+#endif
+         }
+
+      SIMD_4x32 operator^(const SIMD_4x32& other) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_xor_si128(m_reg, other.m_reg));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         return SIMD_4x32(vec_xor(m_reg, other.m_reg));
+#else
+         return SIMD_4x32(m_reg[0] ^ other.m_reg[0],
+                          m_reg[1] ^ other.m_reg[1],
+                          m_reg[2] ^ other.m_reg[2],
+                          m_reg[3] ^ other.m_reg[3]);
+#endif
+         }
+
+      void operator|=(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_or_si128(m_reg, other.m_reg);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = vec_or(m_reg, other.m_reg);
+#else
+         m_reg[0] |= other.m_reg[0];
+         m_reg[1] |= other.m_reg[1];
+         m_reg[2] |= other.m_reg[2];
+         m_reg[3] |= other.m_reg[3];
+#endif
+         }
+
+      SIMD_4x32 operator&(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_and_si128(m_reg, other.m_reg));
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         return SIMD_4x32(vec_and(m_reg, other.m_reg));
+#else
+         return SIMD_4x32(m_reg[0] & other.m_reg[0],
+                          m_reg[1] & other.m_reg[1],
+                          m_reg[2] & other.m_reg[2],
+                          m_reg[3] & other.m_reg[3]);
+#endif
+         }
+
+      void operator&=(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         m_reg = _mm_and_si128(m_reg, other.m_reg);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         m_reg = vec_and(m_reg, other.m_reg);
+#else
+         m_reg[0] &= other.m_reg[0];
+         m_reg[1] &= other.m_reg[1];
+         m_reg[2] &= other.m_reg[2];
+         m_reg[3] &= other.m_reg[3];
+#endif
+         }
+
+      SIMD_4x32 operator<<(size_t shift) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_slli_epi32(m_reg, static_cast<int>(shift)));
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         const unsigned int s = static_cast<unsigned int>(shift);
+         return SIMD_4x32(vec_sl(m_reg, (__vector unsigned int){s, s, s, s}));
+#else
+         return SIMD_4x32(m_reg[0] << shift,
+                          m_reg[1] << shift,
+                          m_reg[2] << shift,
+                          m_reg[3] << shift);
+#endif
+         }
+
+      SIMD_4x32 operator>>(size_t shift) const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_srli_epi32(m_reg, static_cast<int>(shift)));
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         const unsigned int s = static_cast<unsigned int>(shift);
+         return SIMD_4x32(vec_sr(m_reg, (__vector unsigned int){s, s, s, s}));
+#else
+         return SIMD_4x32(m_reg[0] >> shift,
+                          m_reg[1] >> shift,
+                          m_reg[2] >> shift,
+                          m_reg[3] >> shift);
+
+#endif
+         }
+
+      SIMD_4x32 operator~() const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_xor_si128(m_reg, _mm_set1_epi32(0xFFFFFFFF)));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         return SIMD_4x32(vec_nor(m_reg, m_reg));
+#else
+         return SIMD_4x32(~m_reg[0],
+                          ~m_reg[1],
+                          ~m_reg[2],
+                          ~m_reg[3]);
+#endif
+         }
+
+      // (~reg) & other
+      SIMD_4x32 andc(const SIMD_4x32& other)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         return SIMD_4x32(_mm_andnot_si128(m_reg, other.m_reg));
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         /*
+         AltiVec does arg1 & ~arg2 rather than SSE's ~arg1 & arg2
+         so swap the arguments
+         */
+         return SIMD_4x32(vec_andc(other.m_reg, m_reg));
+#else
+         return SIMD_4x32((~m_reg[0]) & other.m_reg[0],
+                          (~m_reg[1]) & other.m_reg[1],
+                          (~m_reg[2]) & other.m_reg[2],
+                          (~m_reg[3]) & other.m_reg[3]);
+#endif
+         }
+
+      SIMD_4x32 bswap() const
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         __m128i T = m_reg;
+
+         T = _mm_shufflehi_epi16(T, _MM_SHUFFLE(2, 3, 0, 1));
+         T = _mm_shufflelo_epi16(T, _MM_SHUFFLE(2, 3, 0, 1));
+
+         return SIMD_4x32(_mm_or_si128(_mm_srli_epi16(T, 8),
+                                       _mm_slli_epi16(T, 8)));
+
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+
+         __vector unsigned char perm = vec_lvsl(0, static_cast<u32bit*>(nullptr));
+
+         perm = vec_xor(perm, vec_splat_u8(3));
+
+         return SIMD_4x32(vec_perm(m_reg, m_reg, perm));
+#else
+         return SIMD_4x32(reverse_bytes(m_reg[0]),
+                          reverse_bytes(m_reg[1]),
+                          reverse_bytes(m_reg[2]),
+                          reverse_bytes(m_reg[3]));
+#endif
+         }
+
+      static void transpose(SIMD_4x32& B0, SIMD_4x32& B1,
+                            SIMD_4x32& B2, SIMD_4x32& B3)
+         {
+#if defined(BOTAN_SIMD_USE_SSE2)
+         __m128i T0 = _mm_unpacklo_epi32(B0.m_reg, B1.m_reg);
+         __m128i T1 = _mm_unpacklo_epi32(B2.m_reg, B3.m_reg);
+         __m128i T2 = _mm_unpackhi_epi32(B0.m_reg, B1.m_reg);
+         __m128i T3 = _mm_unpackhi_epi32(B2.m_reg, B3.m_reg);
+         B0.m_reg = _mm_unpacklo_epi64(T0, T1);
+         B1.m_reg = _mm_unpackhi_epi64(T0, T1);
+         B2.m_reg = _mm_unpacklo_epi64(T2, T3);
+         B3.m_reg = _mm_unpackhi_epi64(T2, T3);
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+         __vector unsigned int T0 = vec_mergeh(B0.m_reg, B2.m_reg);
+         __vector unsigned int T1 = vec_mergel(B0.m_reg, B2.m_reg);
+         __vector unsigned int T2 = vec_mergeh(B1.m_reg, B3.m_reg);
+         __vector unsigned int T3 = vec_mergel(B1.m_reg, B3.m_reg);
+
+         B0.m_reg = vec_mergeh(T0, T2);
+         B1.m_reg = vec_mergel(T0, T2);
+         B2.m_reg = vec_mergeh(T1, T3);
+         B3.m_reg = vec_mergel(T1, T3);
+#else
+         SIMD_4x32 T0(B0.m_reg[0], B1.m_reg[0], B2.m_reg[0], B3.m_reg[0]);
+         SIMD_4x32 T1(B0.m_reg[1], B1.m_reg[1], B2.m_reg[1], B3.m_reg[1]);
+         SIMD_4x32 T2(B0.m_reg[2], B1.m_reg[2], B2.m_reg[2], B3.m_reg[2]);
+         SIMD_4x32 T3(B0.m_reg[3], B1.m_reg[3], B2.m_reg[3], B3.m_reg[3]);
+
+         B0 = T0;
+         B1 = T1;
+         B2 = T2;
+         B3 = T3;
+#endif
+         }
+
+   private:
+#if defined(BOTAN_SIMD_USE_SSE2)
+      explicit SIMD_4x32(__m128i in) { m_reg = in; }
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+      explicit SIMD_4x32(__vector unsigned int input) { m_reg = input; }
+#endif
+
+#if defined(BOTAN_SIMD_USE_SSE2)
+      __m128i m_reg;
+#elif defined(BOTAN_SIMD_USE_ALTIVEC)
+      __vector unsigned int m_reg;
+#else
+      uint32_t m_reg[4];
+#endif
+   };
+
+typedef SIMD_4x32 SIMD_32;
+
+}
 
 #endif
diff --git a/src/lib/utils/simd/simd_altivec/info.txt b/src/lib/utils/simd/simd_altivec/info.txt
deleted file mode 100644
index 19168a928..000000000
--- a/src/lib/utils/simd/simd_altivec/info.txt
+++ /dev/null
@@ -1,9 +0,0 @@
-define SIMD_ALTIVEC 20131128
-
-need_isa altivec
-
-load_on dep
-
-<header:internal>
-simd_altivec.h
-</header:internal>
diff --git a/src/lib/utils/simd/simd_altivec/simd_altivec.h b/src/lib/utils/simd/simd_altivec/simd_altivec.h
deleted file mode 100644
index 3963f2817..000000000
--- a/src/lib/utils/simd/simd_altivec/simd_altivec.h
+++ /dev/null
@@ -1,213 +0,0 @@
-/*
-* Lightweight wrappers around AltiVec for 32-bit operations
-* (C) 2009 Jack Lloyd
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_SIMD_ALTIVEC_H__
-#define BOTAN_SIMD_ALTIVEC_H__
-
-#if defined(BOTAN_TARGET_SUPPORTS_ALTIVEC)
-
-#include <botan/loadstor.h>
-#include <botan/cpuid.h>
-
-#include <altivec.h>
-#undef vector
-#undef bool
-
-namespace Botan {
-
-class SIMD_Altivec
-   {
-   public:
-      SIMD_Altivec(const u32bit B[4])
-         {
-         m_reg = (__vector unsigned int){B[0], B[1], B[2], B[3]};
-         }
-
-      SIMD_Altivec(u32bit B0, u32bit B1, u32bit B2, u32bit B3)
-         {
-         m_reg = (__vector unsigned int){B0, B1, B2, B3};
-         }
-
-      SIMD_Altivec(u32bit B)
-         {
-         m_reg = (__vector unsigned int){B, B, B, B};
-         }
-
-      static SIMD_Altivec load_le(const void* in)
-         {
-         const u32bit* in_32 = static_cast<const u32bit*>(in);
-
-         __vector unsigned int R0 = vec_ld(0, in_32);
-         __vector unsigned int R1 = vec_ld(12, in_32);
-
-         __vector unsigned char perm = vec_lvsl(0, in_32);
-
-         perm = vec_xor(perm, vec_splat_u8(3));
-
-         R0 = vec_perm(R0, R1, perm);
-
-         return SIMD_Altivec(R0);
-         }
-
-      static SIMD_Altivec load_be(const void* in)
-         {
-         const u32bit* in_32 = static_cast<const u32bit*>(in);
-
-         __vector unsigned int R0 = vec_ld(0, in_32);
-         __vector unsigned int R1 = vec_ld(12, in_32);
-
-         __vector unsigned char perm = vec_lvsl(0, in_32);
-
-         R0 = vec_perm(R0, R1, perm);
-
-         return SIMD_Altivec(R0);
-         }
-
-      void store_le(byte out[]) const
-         {
-         __vector unsigned char perm = vec_lvsl(0, static_cast<u32bit*>(nullptr));
-
-         perm = vec_xor(perm, vec_splat_u8(3));
-
-         union {
-            __vector unsigned int V;
-            u32bit R[4];
-            } vec;
-
-         vec.V = vec_perm(m_reg, m_reg, perm);
-
-         Botan::store_be(out, vec.R[0], vec.R[1], vec.R[2], vec.R[3]);
-         }
-
-      void store_be(byte out[]) const
-         {
-         union {
-            __vector unsigned int V;
-            u32bit R[4];
-            } vec;
-
-         vec.V = m_reg;
-
-         Botan::store_be(out, vec.R[0], vec.R[1], vec.R[2], vec.R[3]);
-         }
-
-      void rotate_left(size_t rot)
-         {
-         const unsigned int r = static_cast<unsigned int>(rot);
-         m_reg = vec_rl(m_reg, (__vector unsigned int){r, r, r, r});
-         }
-
-      void rotate_right(size_t rot)
-         {
-         rotate_left(32 - rot);
-         }
-
-      void operator+=(const SIMD_Altivec& other)
-         {
-         m_reg = vec_add(m_reg, other.m_reg);
-         }
-
-      SIMD_Altivec operator+(const SIMD_Altivec& other) const
-         {
-         return vec_add(m_reg, other.m_reg);
-         }
-
-      void operator-=(const SIMD_Altivec& other)
-         {
-         m_reg = vec_sub(m_reg, other.m_reg);
-         }
-
-      SIMD_Altivec operator-(const SIMD_Altivec& other) const
-         {
-         return vec_sub(m_reg, other.m_reg);
-         }
-
-      void operator^=(const SIMD_Altivec& other)
-         {
-         m_reg = vec_xor(m_reg, other.m_reg);
-         }
-
-      SIMD_Altivec operator^(const SIMD_Altivec& other) const
-         {
-         return vec_xor(m_reg, other.m_reg);
-         }
-
-      void operator|=(const SIMD_Altivec& other)
-         {
-         m_reg = vec_or(m_reg, other.m_reg);
-         }
-
-      SIMD_Altivec operator&(const SIMD_Altivec& other)
-         {
-         return vec_and(m_reg, other.m_reg);
-         }
-
-      void operator&=(const SIMD_Altivec& other)
-         {
-         m_reg = vec_and(m_reg, other.m_reg);
-         }
-
-      SIMD_Altivec operator<<(size_t shift) const
-         {
-         const unsigned int s = static_cast<unsigned int>(shift);
-         return vec_sl(m_reg, (__vector unsigned int){s, s, s, s});
-         }
-
-      SIMD_Altivec operator>>(size_t shift) const
-         {
-         const unsigned int s = static_cast<unsigned int>(shift);
-         return vec_sr(m_reg, (__vector unsigned int){s, s, s, s});
-         }
-
-      SIMD_Altivec operator~() const
-         {
-         return vec_nor(m_reg, m_reg);
-         }
-
-      SIMD_Altivec andc(const SIMD_Altivec& other)
-         {
-         /*
-         AltiVec does arg1 & ~arg2 rather than SSE's ~arg1 & arg2
-         so swap the arguments
-         */
-         return vec_andc(other.m_reg, m_reg);
-         }
-
-      SIMD_Altivec bswap() const
-         {
-         __vector unsigned char perm = vec_lvsl(0, static_cast<u32bit*>(nullptr));
-
-         perm = vec_xor(perm, vec_splat_u8(3));
-
-         return SIMD_Altivec(vec_perm(m_reg, m_reg, perm));
-         }
-
-      static void transpose(SIMD_Altivec& B0, SIMD_Altivec& B1,
-                            SIMD_Altivec& B2, SIMD_Altivec& B3)
-         {
-         __vector unsigned int T0 = vec_mergeh(B0.m_reg, B2.m_reg);
-         __vector unsigned int T1 = vec_mergel(B0.m_reg, B2.m_reg);
-         __vector unsigned int T2 = vec_mergeh(B1.m_reg, B3.m_reg);
-         __vector unsigned int T3 = vec_mergel(B1.m_reg, B3.m_reg);
-
-         B0.m_reg = vec_mergeh(T0, T2);
-         B1.m_reg = vec_mergel(T0, T2);
-         B2.m_reg = vec_mergeh(T1, T3);
-         B3.m_reg = vec_mergel(T1, T3);
-         }
-
-   private:
-      SIMD_Altivec(__vector unsigned int input) { m_reg = input; }
-
-      __vector unsigned int m_reg;
-   };
-
-}
-
-#endif
-
-#endif
diff --git a/src/lib/utils/simd/simd_scalar/info.txt b/src/lib/utils/simd/simd_scalar/info.txt
deleted file mode 100644
index 26a9fbfee..000000000
--- a/src/lib/utils/simd/simd_scalar/info.txt
+++ /dev/null
@@ -1,7 +0,0 @@
-define SIMD_SCALAR 20131128
-
-load_on dep
-
-<header:internal>
-simd_scalar.h
-</header:internal>
diff --git a/src/lib/utils/simd/simd_scalar/simd_scalar.h b/src/lib/utils/simd/simd_scalar/simd_scalar.h
deleted file mode 100644
index 28d72c615..000000000
--- a/src/lib/utils/simd/simd_scalar/simd_scalar.h
+++ /dev/null
@@ -1,213 +0,0 @@
-/*
-* Scalar emulation of SIMD
-* (C) 2009,2013 Jack Lloyd
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_SIMD_SCALAR_H__
-#define BOTAN_SIMD_SCALAR_H__
-
-#include <botan/loadstor.h>
-#include <botan/bswap.h>
-
-namespace Botan {
-
-/**
-* Fake SIMD, using plain scalar operations
-* Often still faster than iterative on superscalar machines
-*/
-template<typename T, size_t N>
-class SIMD_Scalar
-   {
-   public:
-      static size_t size() { return N; }
-
-      SIMD_Scalar() { /* uninitialized */ }
-
-      SIMD_Scalar(const T B[N])
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] = B[i];
-         }
-
-      SIMD_Scalar(T B)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] = B;
-         }
-
-      static SIMD_Scalar<T,N> load_le(const void* in)
-         {
-         SIMD_Scalar<T,N> out;
-         const byte* in_b = static_cast<const byte*>(in);
-
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] = Botan::load_le<T>(in_b, i);
-
-         return out;
-         }
-
-      static SIMD_Scalar<T,N> load_be(const void* in)
-         {
-         SIMD_Scalar<T,N> out;
-         const byte* in_b = static_cast<const byte*>(in);
-
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] = Botan::load_be<T>(in_b, i);
-
-         return out;
-         }
-
-      void store_le(byte out[]) const
-         {
-         for(size_t i = 0; i != size(); ++i)
-            Botan::store_le(m_v[i], out + i*sizeof(T));
-         }
-
-      void store_be(byte out[]) const
-         {
-         for(size_t i = 0; i != size(); ++i)
-            Botan::store_be(m_v[i], out + i*sizeof(T));
-         }
-
-      void rotate_left(size_t rot)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] = Botan::rotate_left(m_v[i], rot);
-         }
-
-      void rotate_right(size_t rot)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] = Botan::rotate_right(m_v[i], rot);
-         }
-
-      void operator+=(const SIMD_Scalar<T,N>& other)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] += other.m_v[i];
-         }
-
-      void operator-=(const SIMD_Scalar<T,N>& other)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] -= other.m_v[i];
-         }
-
-      SIMD_Scalar<T,N> operator+(const SIMD_Scalar<T,N>& other) const
-         {
-         SIMD_Scalar<T,N> out = *this;
-         out += other;
-         return out;
-         }
-
-      SIMD_Scalar<T,N> operator-(const SIMD_Scalar<T,N>& other) const
-         {
-         SIMD_Scalar<T,N> out = *this;
-         out -= other;
-         return out;
-         }
-
-      void operator^=(const SIMD_Scalar<T,N>& other)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] ^= other.m_v[i];
-         }
-
-      SIMD_Scalar<T,N> operator^(const SIMD_Scalar<T,N>& other) const
-         {
-         SIMD_Scalar<T,N> out = *this;
-         out ^= other;
-         return out;
-         }
-
-      void operator|=(const SIMD_Scalar<T,N>& other)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] |= other.m_v[i];
-         }
-
-      void operator&=(const SIMD_Scalar<T,N>& other)
-         {
-         for(size_t i = 0; i != size(); ++i)
-            m_v[i] &= other.m_v[i];
-         }
-
-      SIMD_Scalar<T,N> operator&(const SIMD_Scalar<T,N>& other)
-         {
-         SIMD_Scalar<T,N> out = *this;
-         out &= other;
-         return out;
-         }
-
-      SIMD_Scalar<T,N> operator<<(size_t shift) const
-         {
-         SIMD_Scalar<T,N> out = *this;
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] <<= shift;
-         return out;
-         }
-
-      SIMD_Scalar<T,N> operator>>(size_t shift) const
-         {
-         SIMD_Scalar<T,N> out = *this;
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] >>= shift;
-         return out;
-         }
-
-      SIMD_Scalar<T,N> operator~() const
-         {
-         SIMD_Scalar<T,N> out = *this;
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] = ~out.m_v[i];
-         return out;
-         }
-
-      // (~reg) & other
-      SIMD_Scalar<T,N> andc(const SIMD_Scalar<T,N>& other)
-         {
-         SIMD_Scalar<T,N> out;
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] = (~m_v[i]) & other.m_v[i];
-         return out;
-         }
-
-      SIMD_Scalar<T,N> bswap() const
-         {
-         SIMD_Scalar<T,N> out;
-         for(size_t i = 0; i != size(); ++i)
-            out.m_v[i] = reverse_bytes(m_v[i]);
-         return out;
-         }
-
-      static void transpose(SIMD_Scalar<T,N>& B0, SIMD_Scalar<T,N>& B1,
-                            SIMD_Scalar<T,N>& B2, SIMD_Scalar<T,N>& B3)
-         {
-         static_assert(N == 4, "4x4 transpose");
-         SIMD_Scalar<T,N> T0({B0.m_v[0], B1.m_v[0], B2.m_v[0], B3.m_v[0]});
-         SIMD_Scalar<T,N> T1({B0.m_v[1], B1.m_v[1], B2.m_v[1], B3.m_v[1]});
-         SIMD_Scalar<T,N> T2({B0.m_v[2], B1.m_v[2], B2.m_v[2], B3.m_v[2]});
-         SIMD_Scalar<T,N> T3({B0.m_v[3], B1.m_v[3], B2.m_v[3], B3.m_v[3]});
-
-         B0 = T0;
-         B1 = T1;
-         B2 = T2;
-         B3 = T3;
-         }
-
-   private:
-      SIMD_Scalar(std::initializer_list<T> B)
-         {
-         size_t i = 0;
-         for(auto v = B.begin(); v != B.end(); ++v)
-            m_v[i++] = *v;
-         }
-
-      T m_v[N];
-   };
-
-}
-
-#endif
diff --git a/src/lib/utils/simd/simd_sse2/info.txt b/src/lib/utils/simd/simd_sse2/info.txt
deleted file mode 100644
index bd9e430cb..000000000
--- a/src/lib/utils/simd/simd_sse2/info.txt
+++ /dev/null
@@ -1,9 +0,0 @@
-define SIMD_SSE2 20131128
-
-need_isa sse2
-
-load_on dep
-
-<header:internal>
-simd_sse2.h
-</header:internal>
diff --git a/src/lib/utils/simd/simd_sse2/simd_sse2.h b/src/lib/utils/simd/simd_sse2/simd_sse2.h
deleted file mode 100644
index 551e9189c..000000000
--- a/src/lib/utils/simd/simd_sse2/simd_sse2.h
+++ /dev/null
@@ -1,167 +0,0 @@
-/*
-* Lightweight wrappers for SSE2 intrinsics for 32-bit operations
-* (C) 2009 Jack Lloyd
-*
-* Botan is released under the Simplified BSD License (see license.txt)
-*/
-
-#ifndef BOTAN_SIMD_SSE_H__
-#define BOTAN_SIMD_SSE_H__
-
-#if defined(BOTAN_TARGET_SUPPORTS_SSE2)
-
-#include <botan/cpuid.h>
-#include <emmintrin.h>
-
-namespace Botan {
-
-class SIMD_SSE2
-   {
-   public:
-      explicit SIMD_SSE2(const u32bit B[4])
-         {
-         m_reg = _mm_loadu_si128(reinterpret_cast<const __m128i*>(B));
-         }
-
-      SIMD_SSE2(u32bit B0, u32bit B1, u32bit B2, u32bit B3)
-         {
-         m_reg = _mm_set_epi32(B0, B1, B2, B3);
-         }
-
-      explicit SIMD_SSE2(u32bit B)
-         {
-         m_reg = _mm_set1_epi32(B);
-         }
-
-      static SIMD_SSE2 load_le(const void* in)
-         {
-         return SIMD_SSE2(_mm_loadu_si128(reinterpret_cast<const __m128i*>(in)));
-         }
-
-      static SIMD_SSE2 load_be(const void* in)
-         {
-         return load_le(in).bswap();
-         }
-
-      void store_le(byte out[]) const
-         {
-         _mm_storeu_si128(reinterpret_cast<__m128i*>(out), m_reg);
-         }
-
-      void store_be(byte out[]) const
-         {
-         bswap().store_le(out);
-         }
-
-      void rotate_left(size_t rot)
-         {
-         m_reg = _mm_or_si128(_mm_slli_epi32(m_reg, static_cast<int>(rot)),
-                            _mm_srli_epi32(m_reg, static_cast<int>(32-rot)));
-         }
-
-      void rotate_right(size_t rot)
-         {
-         rotate_left(32 - rot);
-         }
-
-      void operator+=(const SIMD_SSE2& other)
-         {
-         m_reg = _mm_add_epi32(m_reg, other.m_reg);
-         }
-
-      SIMD_SSE2 operator+(const SIMD_SSE2& other) const
-         {
-         return SIMD_SSE2(_mm_add_epi32(m_reg, other.m_reg));
-         }
-
-      void operator-=(const SIMD_SSE2& other)
-         {
-         m_reg = _mm_sub_epi32(m_reg, other.m_reg);
-         }
-
-      SIMD_SSE2 operator-(const SIMD_SSE2& other) const
-         {
-         return SIMD_SSE2(_mm_sub_epi32(m_reg, other.m_reg));
-         }
-
-      void operator^=(const SIMD_SSE2& other)
-         {
-         m_reg = _mm_xor_si128(m_reg, other.m_reg);
-         }
-
-      SIMD_SSE2 operator^(const SIMD_SSE2& other) const
-         {
-         return SIMD_SSE2(_mm_xor_si128(m_reg, other.m_reg));
-         }
-
-      void operator|=(const SIMD_SSE2& other)
-         {
-         m_reg = _mm_or_si128(m_reg, other.m_reg);
-         }
-
-      SIMD_SSE2 operator&(const SIMD_SSE2& other)
-         {
-         return SIMD_SSE2(_mm_and_si128(m_reg, other.m_reg));
-         }
-
-      void operator&=(const SIMD_SSE2& other)
-         {
-         m_reg = _mm_and_si128(m_reg, other.m_reg);
-         }
-
-      SIMD_SSE2 operator<<(size_t shift) const
-         {
-         return SIMD_SSE2(_mm_slli_epi32(m_reg, static_cast<int>(shift)));
-         }
-
-      SIMD_SSE2 operator>>(size_t shift) const
-         {
-         return SIMD_SSE2(_mm_srli_epi32(m_reg, static_cast<int>(shift)));
-         }
-
-      SIMD_SSE2 operator~() const
-         {
-         return SIMD_SSE2(_mm_xor_si128(m_reg, _mm_set1_epi32(0xFFFFFFFF)));
-         }
-
-      // (~reg) & other
-      SIMD_SSE2 andc(const SIMD_SSE2& other)
-         {
-         return SIMD_SSE2(_mm_andnot_si128(m_reg, other.m_reg));
-         }
-
-      SIMD_SSE2 bswap() const
-         {
-         __m128i T = m_reg;
-
-         T = _mm_shufflehi_epi16(T, _MM_SHUFFLE(2, 3, 0, 1));
-         T = _mm_shufflelo_epi16(T, _MM_SHUFFLE(2, 3, 0, 1));
-
-         return SIMD_SSE2(_mm_or_si128(_mm_srli_epi16(T, 8),
-                             _mm_slli_epi16(T, 8)));
-         }
-
-      static void transpose(SIMD_SSE2& B0, SIMD_SSE2& B1,
-                            SIMD_SSE2& B2, SIMD_SSE2& B3)
-         {
-         __m128i T0 = _mm_unpacklo_epi32(B0.m_reg, B1.m_reg);
-         __m128i T1 = _mm_unpacklo_epi32(B2.m_reg, B3.m_reg);
-         __m128i T2 = _mm_unpackhi_epi32(B0.m_reg, B1.m_reg);
-         __m128i T3 = _mm_unpackhi_epi32(B2.m_reg, B3.m_reg);
-         B0.m_reg = _mm_unpacklo_epi64(T0, T1);
-         B1.m_reg = _mm_unpackhi_epi64(T0, T1);
-         B2.m_reg = _mm_unpacklo_epi64(T2, T3);
-         B3.m_reg = _mm_unpackhi_epi64(T2, T3);
-         }
-
-   private:
-      explicit SIMD_SSE2(__m128i in) { m_reg = in; }
-
-      __m128i m_reg;
-   };
-
-}
-
-#endif
-
-#endif
diff --git a/src/tests/data/block/serpent.vec b/src/tests/data/block/serpent.vec
index 75511f62d..623ee6be8 100644
--- a/src/tests/data/block/serpent.vec
+++ b/src/tests/data/block/serpent.vec
@@ -3,6 +3,10 @@ Key = 00000000000000000000000000000000
 In = D29D576FCEA3A3A7ED9099F29273D78E2D62A890CEA3A3A7ED9099F29273D78ED29D576F315C5C58ED9099F29273D78E2D62A890315C5C58ED9099F29273D78ED29D576FCEA3A3A7126F660D9273D78E2D62A890CEA3A3A7126F660D9273D78ED29D576F315C5C58126F660D9273D78E2D62A890315C5C58126F660D9273D78ED29D576FCEA3A3A7ED9099F26D8C28712D62A890CEA3A3A7ED9099F26D8C2871D29D576F315C5C58ED9099F26D8C28712D62A890315C5C58ED9099F26D8C2871D29D576FCEA3A3A7126F660D6D8C28712D62A890CEA3A3A7126F660D6D8C2871D29D576F315C5C58126F660D6D8C2871
 Out = B2288B968AE8B08648D1CE9606FD992D717EB02EB81A2E939D54ACA91087112D0D809C5EE82F477EBA7B956DBB23463B0F0190D616F5294112FFB7884E8B37F941BA1B505386B7428B88338188F7E718A3348230BF5CFA552F88D22463D9703A115351622E016BCA26918D17E13225F67EE4E3F2C46FE52ECBDA044C585717DC563A8403FF5309D62370B1DCF5A11EDD2F7D73602B70CD2553E44C1D3F170126155BBD9BE3A965B345E834718F651CEF6CC65E8C5C566E894817350F497816F1EEFA51FC91FEBB6E9F8CB141CC0EB6AF3C6F8380CBD3C996167F2F0E90E71B756C87EB62A4975356B28DCBF6A64A0BD1
 
+Key = 0000000000000000000000000000000000000000000000000000000000000000
+In = D095576FCEA3E3A7ED98D9F29073D78E2F6AA890CEA3E3A7ED98D9F29073D78ED095576F315C1C58ED98D9F29073D78E2F6AA890315C1C58ED98D9F29073D78ED095576FCEA3E3A71267260D9073D78E2F6AA890CEA3E3A71267260D9073D78ED095576F315C1C581267260D9073D78E2F6AA890315C1C581267260D9073D78E5718F3B4ED59D02DE5C40E9A826FFF30AC56479B43617B1C7C804629E36E93F520027AB27AB4CCF48F9973CC052E0A35C8DD678E5D5E8F67FAA0CB599C7886BD
+Out = B90EE5862DE69168F2BDD5125B45472BA3E17E2DF4EA6F41B2017E37023F202AFD2951E466110BC6F4816EBF968CDD01C36EE750C51192B0C85D5BB2584AE06286069E27F65F454C6E3F7E9097DCEFAA00FA885A8854844C970338841CA2E24DA18A399A0AA2FD20EAA740A0CB6145B53D79E39A79CC14D97147F3D172F66BB19263BE9CA3CB9661695C1CCD1041F2E8C35E27B56BAA1AF03542B3DA703BFEF722AAA7F89676ADAA2923903B2A3DB526734448F52EDB9316CAFA7F095C24B0E7
+
 Key = 000000000000000000000000000000000000000000000000
 In = D29D576FCEABA3A7ED9899F2927BD78E
 Out = 130E353E1037C22405E8FAEFB2C3C3E9
diff --git a/src/tests/data/block/xtea.vec b/src/tests/data/block/xtea.vec
index 94b25ded2..e2377c887 100644
--- a/src/tests/data/block/xtea.vec
+++ b/src/tests/data/block/xtea.vec
@@ -7,6 +7,10 @@ Key = 00112233445566778899AABBCCDDEEFF
 In = 0123456789ABCDEF
 Out = B8BF2821622B5B30
 
+Key = 00112233445566778899AABBCCDDEEFF
+In = FEFEFEFEFEFEFEFE0123456789ABCDEFFEFEFEFEFEFEFEFE0123456789ABCDEFFEFEFEFEFEFEFEFE0123456789ABCDEFFEFEFEFEFEFEFEFE0123456789ABCDEF
+Out = 36552E889D6A41CFB8BF2821622B5B3036552E889D6A41CFB8BF2821622B5B3036552E889D6A41CFB8BF2821622B5B3036552E889D6A41CFB8BF2821622B5B30
+
 Key = D94576CD1A5F99E4155DC7BEC5D33DF2
 In = D51399607F7AA9D7
 Out = 782E4E131C5BA746