Remove the 32-bit x86 asm for MSVC

The numbers in #256 suggest that it does nothing at all for performance.

2 files changed, 4 insertions, 272 deletions

diff --git a/src/lib/math/mp/mp_asmi.h b/src/lib/math/mp/mp_asmi.h
index c99d8b7e5..e1518d51c 100644
--- a/src/lib/math/mp/mp_asmi.h
+++ b/src/lib/math/mp/mp_asmi.h
@@ -154,33 +154,6 @@ inline word word8_add2(word x[8], const word y[8], word 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);
@@ -217,50 +190,6 @@ inline word word8_add3(word z[8], const word x[8],
       : "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);
@@ -327,41 +256,6 @@ inline word word8_sub2(word x[8], const word y[8], word 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);
@@ -433,42 +327,6 @@ inline word word8_sub3(word z[8], const word x[8],
       : "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);
@@ -504,68 +362,6 @@ inline word word8_linmul2(word x[8], word y, word 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);
@@ -600,68 +396,6 @@ inline word word8_linmul3(word z[8], const word x[8], word y, word 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);
diff --git a/src/lib/math/mp/mp_madd.h b/src/lib/math/mp/mp_madd.h
index 25659172b..531d6e663 100644
--- a/src/lib/math/mp/mp_madd.h
+++ b/src/lib/math/mp/mp_madd.h
@@ -30,16 +30,14 @@ namespace Botan {
   #error BOTAN_MP_WORD_BITS must be 32 or 64
 #endif
 
-#if defined(BOTAN_TARGET_ARCH_IS_X86_32) && (BOTAN_MP_WORD_BITS == 32)
+#if defined(BOTAN_USE_GCC_INLINE_ASM)
 
-  #if defined(BOTAN_USE_GCC_INLINE_ASM)
+  #if defined(BOTAN_TARGET_ARCH_IS_X86_32) && (BOTAN_MP_WORD_BITS == 32)
     #define BOTAN_MP_USE_X86_32_ASM
-  #elif defined(BOTAN_BUILD_COMPILER_IS_MSVC)
-    #define BOTAN_MP_USE_X86_32_MSVC_ASM
+  #elif defined(BOTAN_TARGET_ARCH_IS_X86_64) && (BOTAN_MP_WORD_BITS == 64)
+    #define BOTAN_MP_USE_X86_64_ASM
   #endif
 
-#elif defined(BOTAN_TARGET_ARCH_IS_X86_64) && (BOTAN_MP_WORD_BITS == 64) && defined(BOTAN_USE_GCC_INLINE_ASM)
-  #define BOTAN_MP_USE_X86_64_ASM
 #endif
 
 /*