Fixed the problem with PCLMUL and gf_complete.h. Removed

ARCH_64 from everything but 128/GROUP/SSE.  Fortunately, no
one ever uses that.

7 files changed, 45 insertions, 48 deletions

diff --git a/include/gf_complete.h b/include/gf_complete.h
index ef685f9..57b439e 100644
--- a/include/gf_complete.h
+++ b/include/gf_complete.h
@@ -24,14 +24,8 @@
   #include <emmintrin.h>
 #endif
 
-#ifdef INTEL_PCLMUL
+#ifdef INTEL_SSE4_PCLMUL
   #include <wmmintrin.h>
-  #ifdef INTEL_SSE4
-    #define INTEL_SSE4_PCLMUL
-  #endif
-  #ifdef INTEL_SSSE3
-    #define INTEL_SSSE3_PCLMUL
-  #endif
 #endif
 
 
diff --git a/src/gf_w128.c b/src/gf_w128.c
index fc08a3f..c888f44 100644
--- a/src/gf_w128.c
+++ b/src/gf_w128.c
@@ -91,7 +91,7 @@ int xor)
     gf_val_128_t d128;
     uint64_t c128[2];
     gf_region_data rd;
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
     __m128i     a,b;
     __m128i     result0,result1;
     __m128i     prim_poly;
@@ -296,7 +296,7 @@ gf_w128_shift_multiply(gf_t *gf, gf_val_128_t a128, gf_val_128_t b128, gf_val_12
 void
 gf_w128_clm_multiply(gf_t *gf, gf_val_128_t a128, gf_val_128_t b128, gf_val_128_t c128)
 {
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
     __m128i     a,b;
     __m128i     result0,result1;
@@ -382,7 +382,7 @@ gf_w128_bytwo_p_multiply(gf_t *gf, gf_val_128_t a128, gf_val_128_t b128, gf_val_
 void
 gf_w128_sse_bytwo_p_multiply(gf_t *gf, gf_val_128_t a128, gf_val_128_t b128, gf_val_128_t c128)
 {
-#if defined(INTEL_SSE4) && defined(ARCH_64)
+#if defined(INTEL_SSE4)
   int i;
   __m128i a, b, pp, one, prod, amask, l_middle_one, u_middle_one; 
   /*John: pmask is always the highest bit set, and the rest zeros. amask changes, it's a countdown.*/
@@ -440,7 +440,7 @@ gf_w128_sse_bytwo_p_multiply(gf_t *gf, gf_val_128_t a128, gf_val_128_t b128, gf_
 void
 gf_w128_sse_bytwo_b_multiply(gf_t *gf, gf_val_128_t a128, gf_val_128_t b128, gf_val_128_t c128)
 {
-#if defined(INTEL_SSE4) && defined(ARCH_64)
+#if defined(INTEL_SSE4)
   __m128i a, b, lmask, hmask, pp, c, middle_one;
   gf_internal_t *h;
   uint64_t topbit, middlebit;
@@ -987,7 +987,7 @@ void gf_w128_group_m_init(gf_t *gf, gf_val_128_t b128)
 static
 void gf_w128_group_m_sse_init(gf_t *gf, gf_val_128_t b128)
 {
-#if defined(INTEL_SSE4) && defined(ARCH_64)
+#if defined(INTEL_SSE4)
   int i, j;
   int g_m;
   uint64_t lbit, middlebit;
@@ -1277,7 +1277,7 @@ static
 void
 gf_w128_group_sse_multiply_region(gf_t *gf, void *src, void *dest, gf_val_128_t val, int bytes, int xor)
 {
-#if defined(INTEL_SSE4) && defined(ARCH_64)
+#if defined(INTEL_SSE4)
   int i;
   int i_r, i_m, t_m;
   int mask_m, mask_r, mask_s;
@@ -1706,7 +1706,7 @@ int gf_w128_composite_init(gf_t *gf)
 static
 int gf_w128_cfm_init(gf_t *gf)
 {
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   gf->inverse.w128 = gf_w128_euclid;
   gf->multiply.w128 = gf_w128_clm_multiply;
   gf->multiply_region.w128 = gf_w128_clm_multiply_region_from_single;
@@ -1779,7 +1779,7 @@ void gf_w128_group_r_init(gf_t *gf)
   static
 void gf_w128_group_r_sse_init(gf_t *gf)
 {
-#if defined(INTEL_SSE4) && defined(ARCH_64)
+#if defined(INTEL_SSE4)
   int i, j;
   int g_r;
   uint64_t pp;
@@ -1814,7 +1814,7 @@ int gf_w128_split_init(gf_t *gf)
   h = (gf_internal_t *) gf->scratch;
 
   gf->multiply.w128 = gf_w128_bytwo_p_multiply;
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   if (!(h->region_type & GF_REGION_NOSSE)){
     gf->multiply.w128 = gf_w128_clm_multiply;
   }
@@ -1880,6 +1880,9 @@ int gf_w128_group_init(gf_t *gf)
   gf->inverse.w128 = gf_w128_euclid;
   gf->multiply_region.w128 = gf_w128_group_multiply_region;
 
+  /* JSP: I've got a problem compiling here -- something about "vmovq", and
+     I don't have the time to chase it down right now. */
+
   #if defined(INTEL_SSE4) && defined(ARCH_64)
     if(!(scratch->region_type & GF_REGION_NOSSE))
     {
diff --git a/src/gf_w16.c b/src/gf_w16.c
index 2d73034..454c6cc 100644
--- a/src/gf_w16.c
+++ b/src/gf_w16.c
@@ -133,7 +133,7 @@ gf_w16_clm_multiply_region_from_single_2(gf_t *gf, void *src, void *dest, gf_val
   uint16_t *s16;
   uint16_t *d16;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   __m128i         a, b;
   __m128i         result;
   __m128i         prim_poly;
@@ -197,7 +197,7 @@ gf_w16_clm_multiply_region_from_single_3(gf_t *gf, void *src, void *dest, gf_val
   uint16_t *s16;
   uint16_t *d16;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -266,7 +266,7 @@ gf_w16_clm_multiply_region_from_single_4(gf_t *gf, void *src, void *dest, gf_val
   uint16_t *s16;
   uint16_t *d16;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -448,7 +448,7 @@ gf_w16_clm_multiply_2 (gf_t *gf, gf_val_32_t a16, gf_val_32_t b16)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -495,7 +495,7 @@ gf_w16_clm_multiply_3 (gf_t *gf, gf_val_32_t a16, gf_val_32_t b16)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -535,7 +535,7 @@ gf_w16_clm_multiply_4 (gf_t *gf, gf_val_32_t a16, gf_val_32_t b16)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -611,7 +611,7 @@ int gf_w16_cfm_init(gf_t *gf)
   
   /*Ben: Determining how many reductions to do */
   
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   if ((0xfe00 & h->prim_poly) == 0) {
     gf->multiply.w32 = gf_w16_clm_multiply_2;
     gf->multiply_region.w32 = gf_w16_clm_multiply_region_from_single_2;
@@ -739,7 +739,7 @@ int gf_w16_log_init(gf_t *gf)
   if (check) {
     if (h->mult_type != GF_MULT_LOG_TABLE) {
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
       return gf_w16_cfm_init(gf);
 #endif
       return gf_w16_shift_init(gf);
diff --git a/src/gf_w32.c b/src/gf_w32.c
index e2fb0f9..03f285f 100644
--- a/src/gf_w32.c
+++ b/src/gf_w32.c
@@ -125,7 +125,7 @@ void
 gf_w32_clm_multiply_region_from_single_2(gf_t *gf, void *src, void *dest, uint32_t val, int bytes, int xor)
 {
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   int i;
   uint32_t *s32;
@@ -175,7 +175,7 @@ void
 gf_w32_clm_multiply_region_from_single_3(gf_t *gf, void *src, void *dest, uint32_t val, int bytes, int xor)
 {
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
 
   int i;
   uint32_t *s32;
@@ -229,7 +229,7 @@ static
 void
 gf_w32_clm_multiply_region_from_single_4(gf_t *gf, void *src, void *dest, uint32_t val, int bytes, int xor)
 {
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   int i;
   uint32_t *s32;
   uint32_t *d32;
@@ -409,7 +409,7 @@ gf_w32_clm_multiply_2 (gf_t *gf, gf_val_32_t a32, gf_val_32_t b32)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -453,7 +453,7 @@ gf_w32_clm_multiply_3 (gf_t *gf, gf_val_32_t a32, gf_val_32_t b32)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -492,7 +492,7 @@ gf_w32_clm_multiply_4 (gf_t *gf, gf_val_32_t a32, gf_val_32_t b32)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -565,7 +565,7 @@ int gf_w32_cfm_init(gf_t *gf)
   /*Ben: We also check to see if the prim poly will work for pclmul */
   /*Ben: Check to see how many reduction steps it will take*/
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   if ((0xfffe0000 & h->prim_poly) == 0){ 
     gf->multiply.w32 = gf_w32_clm_multiply_2;
     gf->multiply_region.w32 = gf_w32_clm_multiply_region_from_single_2;
@@ -2176,7 +2176,7 @@ int gf_w32_split_init(gf_t *gf)
   int i, j, exp, ispclmul, issse3;
 
   ispclmul = 0;
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   ispclmul = 1;
 #endif
 
diff --git a/src/gf_w4.c b/src/gf_w4.c
index 3e00cd2..2504ec6 100644
--- a/src/gf_w4.c
+++ b/src/gf_w4.c
@@ -182,7 +182,7 @@ gf_w4_clm_multiply (gf_t *gf, gf_val_32_t a4, gf_val_32_t b4)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -1967,7 +1967,7 @@ int gf_w4_cfm_init(gf_t *gf)
 
   h = (gf_internal_t *) gf->scratch;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   gf->multiply.w32 = gf_w4_clm_multiply;
   return 1;
 #endif
diff --git a/src/gf_w64.c b/src/gf_w64.c
index b8baa8f..73bf164 100644
--- a/src/gf_w64.c
+++ b/src/gf_w64.c
@@ -96,7 +96,7 @@ xor)
   gf_val_64_t *s64, *d64, *top;
   gf_region_data rd;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
   __m128i         a, b;
   __m128i         result, r1;
   __m128i         prim_poly;
@@ -187,7 +187,7 @@ xor)
   gf_val_64_t *s64, *d64, *top;
   gf_region_data rd;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
   __m128i         a, b;
   __m128i         result, r1;
   __m128i         prim_poly;
@@ -385,7 +385,7 @@ gf_w64_clm_multiply_2 (gf_t *gf, gf_val_64_t a64, gf_val_64_t b64)
 {
        gf_val_64_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
 
         __m128i         a, b;
         __m128i         result;
@@ -427,7 +427,7 @@ gf_w64_clm_multiply_4 (gf_t *gf, gf_val_64_t a64, gf_val_64_t b64)
 {
   gf_val_64_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
 
   __m128i         a, b;
   __m128i         result;
@@ -466,7 +466,7 @@ gf_w64_clm_multiply_4 (gf_t *gf, gf_val_64_t a64, gf_val_64_t b64)
   void
 gf_w64_clm_multiply_region(gf_t *gf, void *src, void *dest, uint64_t val, int bytes, int xor)
 {
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
   gf_internal_t *h;
   int i, j, k;
   uint8_t *s8, *d8, *dtop;
@@ -759,7 +759,7 @@ int gf_w64_cfm_init(gf_t *gf)
   gf->inverse.w64 = gf_w64_euclid;
   gf->multiply_region.w64 = gf_w64_multiply_region_from_single;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
   if ((0xfffffffe00000000ULL & h->prim_poly) == 0){ 
     gf->multiply.w64 = gf_w64_clm_multiply_2;
     gf->multiply_region.w64 = gf_w64_clm_multiply_region_from_single_2; 
@@ -2030,7 +2030,7 @@ int gf_w64_split_init(gf_t *gf)
 
   gf->multiply.w64 = gf_w64_bytwo_p_multiply; 
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL) 
   if ((!(h->region_type & GF_REGION_NOSSE) &&
      (h->arg1 == 64 || h->arg2 == 64)) ||
      h->mult_type == GF_MULT_DEFAULT){
diff --git a/src/gf_w8.c b/src/gf_w8.c
index da34968..7661aad 100644
--- a/src/gf_w8.c
+++ b/src/gf_w8.c
@@ -211,7 +211,7 @@ gf_w8_clm_multiply_2 (gf_t *gf, gf_val_32_t a8, gf_val_32_t b8)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -257,7 +257,7 @@ gf_w8_clm_multiply_3 (gf_t *gf, gf_val_32_t a8, gf_val_32_t b8)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -296,7 +296,7 @@ gf_w8_clm_multiply_4 (gf_t *gf, gf_val_32_t a8, gf_val_32_t b8)
 {
   gf_val_32_t rv = 0;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -373,7 +373,7 @@ gf_w8_clm_multiply_region_from_single_2(gf_t *gf, void *src, void *dest, gf_val_
   uint8_t *s8;
   uint8_t *d8;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -432,7 +432,7 @@ gf_w8_clm_multiply_region_from_single_3(gf_t *gf, void *src, void *dest, gf_val_
   uint8_t *s8;
   uint8_t *d8;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -495,7 +495,7 @@ gf_w8_clm_multiply_region_from_single_4(gf_t *gf, void *src, void *dest, gf_val_
   uint8_t *s8;
   uint8_t *d8;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
 
   __m128i         a, b;
   __m128i         result;
@@ -592,7 +592,7 @@ int gf_w8_cfm_init(gf_t *gf)
 
   h = (gf_internal_t *) gf->scratch;
 
-#if defined(INTEL_SSE4_PCLMUL) && defined(ARCH_64)
+#if defined(INTEL_SSE4_PCLMUL)
     if ((0xe0 & h->prim_poly) == 0){
       gf->multiply.w32 = gf_w8_clm_multiply_2;
       gf->multiply_region.w32 = gf_w8_clm_multiply_region_from_single_2;