Use X64 ASM ed25519 signatures on Linux/x64, which are about 10X faster. Will matter a lot for network controllers, not so much for other things.

ext/ed25519-amd64-asm/batch.c

@@ -0,0 +1,94 @@
+#include "crypto_sign.h"
+
+#include "crypto_verify_32.h"
+#include "crypto_hash_sha512.h"
+#include "randombytes.h"
+
+#include "ge25519.h"
+#include "hram.h"
+
+#define MAXBATCH 64
+
+int crypto_sign_open_batch(
+    unsigned char* const m[],unsigned long long mlen[],
+    unsigned char* const sm[],const unsigned long long smlen[],
+    unsigned char* const pk[], 
+    unsigned long long num
+    )
+{
+  int ret = 0;
+  unsigned long long i, j;
+  shortsc25519 r[MAXBATCH];
+  sc25519 scalars[2*MAXBATCH+1];
+  ge25519 points[2*MAXBATCH+1];
+  unsigned char hram[crypto_hash_sha512_BYTES];
+  unsigned long long batchsize;
+
+  for (i = 0;i < num;++i) mlen[i] = -1;
+
+  while (num >= 3) {
+    batchsize = num;
+    if (batchsize > MAXBATCH) batchsize = MAXBATCH;
+
+    for (i = 0;i < batchsize;++i)
+      if (smlen[i] < 64) goto fallback;
+
+    randombytes((unsigned char*)r,sizeof(shortsc25519) * batchsize);
+
+    /* Computing scalars[0] = ((r1s1 + r2s2 + ...)) */
+    for(i=0;i<batchsize;i++)
+    {
+      sc25519_from32bytes(&scalars[i], sm[i]+32); 
+      sc25519_mul_shortsc(&scalars[i], &scalars[i], &r[i]);
+    }
+    for(i=1;i<batchsize;i++)
+      sc25519_add(&scalars[0], &scalars[0], &scalars[i]);
+    
+    /* Computing scalars[1] ... scalars[batchsize] as r[i]*H(R[i],A[i],m[i]) */
+    for(i=0;i<batchsize;i++)
+    {
+      get_hram(hram, sm[i], pk[i], m[i], smlen[i]);
+      sc25519_from64bytes(&scalars[i+1],hram);
+      sc25519_mul_shortsc(&scalars[i+1],&scalars[i+1],&r[i]);
+    }
+    /* Setting scalars[batchsize+1] ... scalars[2*batchsize] to r[i] */
+    for(i=0;i<batchsize;i++)
+      sc25519_from_shortsc(&scalars[batchsize+i+1],&r[i]);
+  
+    /* Computing points */
+    points[0] = ge25519_base;
+  
+    for(i=0;i<batchsize;i++)
+      if (ge25519_unpackneg_vartime(&points[i+1], pk[i])) goto fallback;
+    for(i=0;i<batchsize;i++)
+      if (ge25519_unpackneg_vartime(&points[batchsize+i+1], sm[i])) goto fallback;
+  
+    ge25519_multi_scalarmult_vartime(points, points, scalars, 2*batchsize+1);
+  
+    if (ge25519_isneutral_vartime(points)) {
+      for(i=0;i<batchsize;i++)
+      {
+        for(j=0;j<smlen[i]-64;j++)
+          m[i][j] = sm[i][j + 64];
+        mlen[i] = smlen[i]-64;
+      }
+    } else {
+      fallback:
+
+      for (i = 0;i < batchsize;++i)
+        ret |= crypto_sign_open(m[i], &mlen[i], sm[i], smlen[i], pk[i]);
+    }
+
+    m += batchsize;
+    mlen += batchsize;
+    sm += batchsize;
+    smlen += batchsize;
+    pk += batchsize;
+    num -= batchsize;
+  }
+
+  for (i = 0;i < num;++i)
+    ret |= crypto_sign_open(m[i], &mlen[i], sm[i], smlen[i], pk[i]);
+
+  return ret;
+}

ext/ed25519-amd64-asm/consts.s

@@ -0,0 +1,39 @@
+.data
+
+.globl crypto_sign_ed25519_amd64_64_121666
+.globl crypto_sign_ed25519_amd64_64_MU0
+.globl crypto_sign_ed25519_amd64_64_MU1
+.globl crypto_sign_ed25519_amd64_64_MU2
+.globl crypto_sign_ed25519_amd64_64_MU3
+.globl crypto_sign_ed25519_amd64_64_MU4
+.globl crypto_sign_ed25519_amd64_64_ORDER0
+.globl crypto_sign_ed25519_amd64_64_ORDER1
+.globl crypto_sign_ed25519_amd64_64_ORDER2
+.globl crypto_sign_ed25519_amd64_64_ORDER3
+.globl crypto_sign_ed25519_amd64_64_EC2D0
+.globl crypto_sign_ed25519_amd64_64_EC2D1
+.globl crypto_sign_ed25519_amd64_64_EC2D2
+.globl crypto_sign_ed25519_amd64_64_EC2D3
+.globl crypto_sign_ed25519_amd64_64_38
+
+.p2align 4
+
+crypto_sign_ed25519_amd64_64_121666:     .quad 121666
+
+crypto_sign_ed25519_amd64_64_MU0:         .quad 0xED9CE5A30A2C131B
+crypto_sign_ed25519_amd64_64_MU1:         .quad 0x2106215D086329A7
+crypto_sign_ed25519_amd64_64_MU2:         .quad 0xFFFFFFFFFFFFFFEB
+crypto_sign_ed25519_amd64_64_MU3:         .quad 0xFFFFFFFFFFFFFFFF
+crypto_sign_ed25519_amd64_64_MU4:         .quad 0x000000000000000F
+
+crypto_sign_ed25519_amd64_64_ORDER0:      .quad 0x5812631A5CF5D3ED
+crypto_sign_ed25519_amd64_64_ORDER1:      .quad 0x14DEF9DEA2F79CD6
+crypto_sign_ed25519_amd64_64_ORDER2:      .quad 0x0000000000000000
+crypto_sign_ed25519_amd64_64_ORDER3:      .quad 0x1000000000000000
+
+crypto_sign_ed25519_amd64_64_EC2D0:       .quad 0xEBD69B9426B2F146
+crypto_sign_ed25519_amd64_64_EC2D1:       .quad 0x00E0149A8283B156
+crypto_sign_ed25519_amd64_64_EC2D2:       .quad 0x198E80F2EEF3D130
+crypto_sign_ed25519_amd64_64_EC2D3:       .quad 0xA406D9DC56DFFCE7
+
+crypto_sign_ed25519_amd64_64_38:         .quad 38

ext/ed25519-amd64-asm/fe25519.h

@@ -0,0 +1,64 @@
+#ifndef FE25519_H
+#define FE25519_H
+
+#define fe25519                crypto_sign_ed25519_amd64_64_fe25519
+#define fe25519_freeze         crypto_sign_ed25519_amd64_64_fe25519_freeze
+#define fe25519_unpack         crypto_sign_ed25519_amd64_64_fe25519_unpack
+#define fe25519_pack           crypto_sign_ed25519_amd64_64_fe25519_pack
+#define fe25519_iszero_vartime crypto_sign_ed25519_amd64_64_fe25519_iszero_vartime
+#define fe25519_iseq_vartime   crypto_sign_ed25519_amd64_64_fe25519_iseq_vartime
+#define fe25519_cmov           crypto_sign_ed25519_amd64_64_fe25519_cmov
+#define fe25519_setint         crypto_sign_ed25519_amd64_64_fe25519_setint
+#define fe25519_neg            crypto_sign_ed25519_amd64_64_fe25519_neg
+#define fe25519_getparity      crypto_sign_ed25519_amd64_64_fe25519_getparity
+#define fe25519_add            crypto_sign_ed25519_amd64_64_fe25519_add
+#define fe25519_sub            crypto_sign_ed25519_amd64_64_fe25519_sub
+#define fe25519_mul            crypto_sign_ed25519_amd64_64_fe25519_mul
+#define fe25519_mul121666      crypto_sign_ed25519_amd64_64_fe25519_mul121666
+#define fe25519_square         crypto_sign_ed25519_amd64_64_fe25519_square
+#define fe25519_invert         crypto_sign_ed25519_amd64_64_fe25519_invert
+#define fe25519_pow2523        crypto_sign_ed25519_amd64_64_fe25519_pow2523
+
+typedef struct 
+{
+  unsigned long long v[4]; 
+}
+fe25519;
+
+void fe25519_freeze(fe25519 *r);
+
+void fe25519_unpack(fe25519 *r, const unsigned char x[32]);
+
+void fe25519_pack(unsigned char r[32], const fe25519 *x);
+
+void fe25519_cmov(fe25519 *r, const fe25519 *x, unsigned char b);
+
+void fe25519_cswap(fe25519 *r, fe25519 *x, unsigned char b);
+
+void fe25519_setint(fe25519 *r, unsigned int v);
+
+void fe25519_neg(fe25519 *r, const fe25519 *x);
+
+unsigned char fe25519_getparity(const fe25519 *x);
+
+int fe25519_iszero_vartime(const fe25519 *x);
+
+int fe25519_iseq_vartime(const fe25519 *x, const fe25519 *y);
+
+void fe25519_add(fe25519 *r, const fe25519 *x, const fe25519 *y);
+
+void fe25519_sub(fe25519 *r, const fe25519 *x, const fe25519 *y);
+
+void fe25519_mul(fe25519 *r, const fe25519 *x, const fe25519 *y);
+
+void fe25519_mul121666(fe25519 *r, const fe25519 *x);
+
+void fe25519_square(fe25519 *r, const fe25519 *x);
+
+void fe25519_pow(fe25519 *r, const fe25519 *x, const unsigned char *e);
+
+void fe25519_invert(fe25519 *r, const fe25519 *x);
+
+void fe25519_pow2523(fe25519 *r, const fe25519 *x);
+
+#endif

ext/ed25519-amd64-asm/fe25519_add.s

@@ -0,0 +1,189 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm: int64 addt0
+
+# qhasm: int64 addt1
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_fe25519_add
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_fe25519_add
+.globl crypto_sign_ed25519_amd64_64_fe25519_add
+_crypto_sign_ed25519_amd64_64_fe25519_add:
+crypto_sign_ed25519_amd64_64_fe25519_add:
+mov %rsp,%r11
+and $31,%r11
+add $0,%r11
+sub %r11,%rsp
+
+# qhasm: r0 = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>r0=int64#4
+# asm 2: movq   0(<xp=%rsi),>r0=%rcx
+movq   0(%rsi),%rcx
+
+# qhasm: r1 = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>r1=int64#5
+# asm 2: movq   8(<xp=%rsi),>r1=%r8
+movq   8(%rsi),%r8
+
+# qhasm: r2 = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>r2=int64#6
+# asm 2: movq   16(<xp=%rsi),>r2=%r9
+movq   16(%rsi),%r9
+
+# qhasm: r3 = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>r3=int64#2
+# asm 2: movq   24(<xp=%rsi),>r3=%rsi
+movq   24(%rsi),%rsi
+
+# qhasm:   carry? r0 += *(uint64 *)(yp + 0) 
+# asm 1: addq 0(<yp=int64#3),<r0=int64#4
+# asm 2: addq 0(<yp=%rdx),<r0=%rcx
+addq 0(%rdx),%rcx
+
+# qhasm:   carry? r1 += *(uint64 *)(yp + 8) + carry
+# asm 1: adcq 8(<yp=int64#3),<r1=int64#5
+# asm 2: adcq 8(<yp=%rdx),<r1=%r8
+adcq 8(%rdx),%r8
+
+# qhasm:   carry? r2 += *(uint64 *)(yp + 16) + carry
+# asm 1: adcq 16(<yp=int64#3),<r2=int64#6
+# asm 2: adcq 16(<yp=%rdx),<r2=%r9
+adcq 16(%rdx),%r9
+
+# qhasm:   carry? r3 += *(uint64 *)(yp + 24) + carry
+# asm 1: adcq 24(<yp=int64#3),<r3=int64#2
+# asm 2: adcq 24(<yp=%rdx),<r3=%rsi
+adcq 24(%rdx),%rsi
+
+# qhasm:   addt0 = 0
+# asm 1: mov  $0,>addt0=int64#3
+# asm 2: mov  $0,>addt0=%rdx
+mov  $0,%rdx
+
+# qhasm:   addt1 = 38
+# asm 1: mov  $38,>addt1=int64#7
+# asm 2: mov  $38,>addt1=%rax
+mov  $38,%rax
+
+# qhasm:   addt1 = addt0 if !carry
+# asm 1: cmovae <addt0=int64#3,<addt1=int64#7
+# asm 2: cmovae <addt0=%rdx,<addt1=%rax
+cmovae %rdx,%rax
+
+# qhasm:   carry? r0 += addt1
+# asm 1: add  <addt1=int64#7,<r0=int64#4
+# asm 2: add  <addt1=%rax,<r0=%rcx
+add  %rax,%rcx
+
+# qhasm:   carry? r1 += addt0 + carry
+# asm 1: adc <addt0=int64#3,<r1=int64#5
+# asm 2: adc <addt0=%rdx,<r1=%r8
+adc %rdx,%r8
+
+# qhasm:   carry? r2 += addt0 + carry
+# asm 1: adc <addt0=int64#3,<r2=int64#6
+# asm 2: adc <addt0=%rdx,<r2=%r9
+adc %rdx,%r9
+
+# qhasm:   carry? r3 += addt0 + carry
+# asm 1: adc <addt0=int64#3,<r3=int64#2
+# asm 2: adc <addt0=%rdx,<r3=%rsi
+adc %rdx,%rsi
+
+# qhasm:   addt0 = addt1 if carry
+# asm 1: cmovc <addt1=int64#7,<addt0=int64#3
+# asm 2: cmovc <addt1=%rax,<addt0=%rdx
+cmovc %rax,%rdx
+
+# qhasm:   r0 += addt0
+# asm 1: add  <addt0=int64#3,<r0=int64#4
+# asm 2: add  <addt0=%rdx,<r0=%rcx
+add  %rdx,%rcx
+
+# qhasm: *(uint64 *)(rp + 0) = r0
+# asm 1: movq   <r0=int64#4,0(<rp=int64#1)
+# asm 2: movq   <r0=%rcx,0(<rp=%rdi)
+movq   %rcx,0(%rdi)
+
+# qhasm: *(uint64 *)(rp + 8) = r1
+# asm 1: movq   <r1=int64#5,8(<rp=int64#1)
+# asm 2: movq   <r1=%r8,8(<rp=%rdi)
+movq   %r8,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#6,16(<rp=int64#1)
+# asm 2: movq   <r2=%r9,16(<rp=%rdi)
+movq   %r9,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#2,24(<rp=int64#1)
+# asm 2: movq   <r3=%rsi,24(<rp=%rdi)
+movq   %rsi,24(%rdi)
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/fe25519_freeze.s

@@ -0,0 +1,322 @@
+
+# qhasm: int64 rp
+
+# qhasm: input rp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm: int64 two63
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_fe25519_freeze
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_fe25519_freeze
+.globl crypto_sign_ed25519_amd64_64_fe25519_freeze
+_crypto_sign_ed25519_amd64_64_fe25519_freeze:
+crypto_sign_ed25519_amd64_64_fe25519_freeze:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm:   caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm:   caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm:   caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm:   caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm:   caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm:   caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm:   caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: r0 = *(uint64 *) (rp + 0)
+# asm 1: movq   0(<rp=int64#1),>r0=int64#2
+# asm 2: movq   0(<rp=%rdi),>r0=%rsi
+movq   0(%rdi),%rsi
+
+# qhasm: r1 = *(uint64 *) (rp + 8)
+# asm 1: movq   8(<rp=int64#1),>r1=int64#3
+# asm 2: movq   8(<rp=%rdi),>r1=%rdx
+movq   8(%rdi),%rdx
+
+# qhasm: r2 = *(uint64 *) (rp + 16)
+# asm 1: movq   16(<rp=int64#1),>r2=int64#4
+# asm 2: movq   16(<rp=%rdi),>r2=%rcx
+movq   16(%rdi),%rcx
+
+# qhasm: r3 = *(uint64 *) (rp + 24)
+# asm 1: movq   24(<rp=int64#1),>r3=int64#5
+# asm 2: movq   24(<rp=%rdi),>r3=%r8
+movq   24(%rdi),%r8
+
+# qhasm: t0 = r0
+# asm 1: mov  <r0=int64#2,>t0=int64#6
+# asm 2: mov  <r0=%rsi,>t0=%r9
+mov  %rsi,%r9
+
+# qhasm: t1 = r1
+# asm 1: mov  <r1=int64#3,>t1=int64#7
+# asm 2: mov  <r1=%rdx,>t1=%rax
+mov  %rdx,%rax
+
+# qhasm: t2 = r2
+# asm 1: mov  <r2=int64#4,>t2=int64#8
+# asm 2: mov  <r2=%rcx,>t2=%r10
+mov  %rcx,%r10
+
+# qhasm: t3 = r3
+# asm 1: mov  <r3=int64#5,>t3=int64#9
+# asm 2: mov  <r3=%r8,>t3=%r11
+mov  %r8,%r11
+
+# qhasm: two63 = 1
+# asm 1: mov  $1,>two63=int64#10
+# asm 2: mov  $1,>two63=%r12
+mov  $1,%r12
+
+# qhasm: two63 <<= 63
+# asm 1: shl  $63,<two63=int64#10
+# asm 2: shl  $63,<two63=%r12
+shl  $63,%r12
+
+# qhasm: carry? t0 += 19
+# asm 1: add  $19,<t0=int64#6
+# asm 2: add  $19,<t0=%r9
+add  $19,%r9
+
+# qhasm: carry? t1 += 0 + carry
+# asm 1: adc $0,<t1=int64#7
+# asm 2: adc $0,<t1=%rax
+adc $0,%rax
+
+# qhasm: carry? t2 += 0 + carry
+# asm 1: adc $0,<t2=int64#8
+# asm 2: adc $0,<t2=%r10
+adc $0,%r10
+
+# qhasm: carry? t3 += two63 + carry
+# asm 1: adc <two63=int64#10,<t3=int64#9
+# asm 2: adc <two63=%r12,<t3=%r11
+adc %r12,%r11
+
+# qhasm: r0 = t0 if carry
+# asm 1: cmovc <t0=int64#6,<r0=int64#2
+# asm 2: cmovc <t0=%r9,<r0=%rsi
+cmovc %r9,%rsi
+
+# qhasm: r1 = t1 if carry
+# asm 1: cmovc <t1=int64#7,<r1=int64#3
+# asm 2: cmovc <t1=%rax,<r1=%rdx
+cmovc %rax,%rdx
+
+# qhasm: r2 = t2 if carry
+# asm 1: cmovc <t2=int64#8,<r2=int64#4
+# asm 2: cmovc <t2=%r10,<r2=%rcx
+cmovc %r10,%rcx
+
+# qhasm: r3 = t3 if carry
+# asm 1: cmovc <t3=int64#9,<r3=int64#5
+# asm 2: cmovc <t3=%r11,<r3=%r8
+cmovc %r11,%r8
+
+# qhasm: t0 = r0
+# asm 1: mov  <r0=int64#2,>t0=int64#6
+# asm 2: mov  <r0=%rsi,>t0=%r9
+mov  %rsi,%r9
+
+# qhasm: t1 = r1
+# asm 1: mov  <r1=int64#3,>t1=int64#7
+# asm 2: mov  <r1=%rdx,>t1=%rax
+mov  %rdx,%rax
+
+# qhasm: t2 = r2
+# asm 1: mov  <r2=int64#4,>t2=int64#8
+# asm 2: mov  <r2=%rcx,>t2=%r10
+mov  %rcx,%r10
+
+# qhasm: t3 = r3
+# asm 1: mov  <r3=int64#5,>t3=int64#9
+# asm 2: mov  <r3=%r8,>t3=%r11
+mov  %r8,%r11
+
+# qhasm: carry? t0 += 19
+# asm 1: add  $19,<t0=int64#6
+# asm 2: add  $19,<t0=%r9
+add  $19,%r9
+
+# qhasm: carry? t1 += 0 + carry
+# asm 1: adc $0,<t1=int64#7
+# asm 2: adc $0,<t1=%rax
+adc $0,%rax
+
+# qhasm: carry? t2 += 0 + carry
+# asm 1: adc $0,<t2=int64#8
+# asm 2: adc $0,<t2=%r10
+adc $0,%r10
+
+# qhasm: carry? t3 += two63 + carry
+# asm 1: adc <two63=int64#10,<t3=int64#9
+# asm 2: adc <two63=%r12,<t3=%r11
+adc %r12,%r11
+
+# qhasm: r0 = t0 if carry
+# asm 1: cmovc <t0=int64#6,<r0=int64#2
+# asm 2: cmovc <t0=%r9,<r0=%rsi
+cmovc %r9,%rsi
+
+# qhasm: r1 = t1 if carry
+# asm 1: cmovc <t1=int64#7,<r1=int64#3
+# asm 2: cmovc <t1=%rax,<r1=%rdx
+cmovc %rax,%rdx
+
+# qhasm: r2 = t2 if carry
+# asm 1: cmovc <t2=int64#8,<r2=int64#4
+# asm 2: cmovc <t2=%r10,<r2=%rcx
+cmovc %r10,%rcx
+
+# qhasm: r3 = t3 if carry
+# asm 1: cmovc <t3=int64#9,<r3=int64#5
+# asm 2: cmovc <t3=%r11,<r3=%r8
+cmovc %r11,%r8
+
+# qhasm: *(uint64 *)(rp + 0) = r0
+# asm 1: movq   <r0=int64#2,0(<rp=int64#1)
+# asm 2: movq   <r0=%rsi,0(<rp=%rdi)
+movq   %rsi,0(%rdi)
+
+# qhasm: *(uint64 *)(rp + 8) = r1
+# asm 1: movq   <r1=int64#3,8(<rp=int64#1)
+# asm 2: movq   <r1=%rdx,8(<rp=%rdi)
+movq   %rdx,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#4,16(<rp=int64#1)
+# asm 2: movq   <r2=%rcx,16(<rp=%rdi)
+movq   %rcx,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#5,24(<rp=int64#1)
+# asm 2: movq   <r3=%r8,24(<rp=%rdi)
+movq   %r8,24(%rdi)
+
+# qhasm:   caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm:   caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm:   caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm:   caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm:   caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm:   caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm:   caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/fe25519_getparity.c

@@ -0,0 +1,8 @@
+#include "fe25519.h"
+
+unsigned char fe25519_getparity(const fe25519 *x)
+{
+  fe25519 t = *x;
+  fe25519_freeze(&t);
+  return (unsigned char)t.v[0] & 1;
+}

ext/ed25519-amd64-asm/fe25519_invert.c

@@ -0,0 +1,60 @@
+#include "fe25519.h"
+
+void fe25519_invert(fe25519 *r, const fe25519 *x)
+{
+	fe25519 z2;
+	fe25519 z9;
+	fe25519 z11;
+	fe25519 z2_5_0;
+	fe25519 z2_10_0;
+	fe25519 z2_20_0;
+	fe25519 z2_50_0;
+	fe25519 z2_100_0;
+	fe25519 t;
+	int i;
+	
+	/* 2 */ fe25519_square(&z2,x);
+	/* 4 */ fe25519_square(&t,&z2);
+	/* 8 */ fe25519_square(&t,&t);
+	/* 9 */ fe25519_mul(&z9,&t,x);
+	/* 11 */ fe25519_mul(&z11,&z9,&z2);
+	/* 22 */ fe25519_square(&t,&z11);
+	/* 2^5 - 2^0 = 31 */ fe25519_mul(&z2_5_0,&t,&z9);
+
+	/* 2^6 - 2^1 */ fe25519_square(&t,&z2_5_0);
+	/* 2^20 - 2^10 */ for (i = 1;i < 5;i++) { fe25519_square(&t,&t); }
+	/* 2^10 - 2^0 */ fe25519_mul(&z2_10_0,&t,&z2_5_0);
+
+	/* 2^11 - 2^1 */ fe25519_square(&t,&z2_10_0);
+	/* 2^20 - 2^10 */ for (i = 1;i < 10;i++) { fe25519_square(&t,&t); }
+	/* 2^20 - 2^0 */ fe25519_mul(&z2_20_0,&t,&z2_10_0);
+
+	/* 2^21 - 2^1 */ fe25519_square(&t,&z2_20_0);
+	/* 2^40 - 2^20 */ for (i = 1;i < 20;i++) { fe25519_square(&t,&t); }
+	/* 2^40 - 2^0 */ fe25519_mul(&t,&t,&z2_20_0);
+
+	/* 2^41 - 2^1 */ fe25519_square(&t,&t);
+	/* 2^50 - 2^10 */ for (i = 1;i < 10;i++) { fe25519_square(&t,&t); }
+	/* 2^50 - 2^0 */ fe25519_mul(&z2_50_0,&t,&z2_10_0);
+
+	/* 2^51 - 2^1 */ fe25519_square(&t,&z2_50_0);
+	/* 2^100 - 2^50 */ for (i = 1;i < 50;i++) { fe25519_square(&t,&t); }
+	/* 2^100 - 2^0 */ fe25519_mul(&z2_100_0,&t,&z2_50_0);
+
+	/* 2^101 - 2^1 */ fe25519_square(&t,&z2_100_0);
+	/* 2^200 - 2^100 */ for (i = 1;i < 100;i++) { fe25519_square(&t,&t); }
+	/* 2^200 - 2^0 */ fe25519_mul(&t,&t,&z2_100_0);
+
+	/* 2^201 - 2^1 */ fe25519_square(&t,&t);
+	/* 2^250 - 2^50 */ for (i = 1;i < 50;i++) { fe25519_square(&t,&t); }
+	/* 2^250 - 2^0 */ fe25519_mul(&t,&t,&z2_50_0);
+
+	/* 2^251 - 2^1 */ fe25519_square(&t,&t);
+	/* 2^252 - 2^2 */ fe25519_square(&t,&t);
+	/* 2^253 - 2^3 */ fe25519_square(&t,&t);
+
+	/* 2^254 - 2^4 */ fe25519_square(&t,&t);
+
+	/* 2^255 - 2^5 */ fe25519_square(&t,&t);
+	/* 2^255 - 21 */ fe25519_mul(r,&t,&z11);
+}

ext/ed25519-amd64-asm/fe25519_iseq.c

@@ -0,0 +1,14 @@
+#include "fe25519.h"
+
+int fe25519_iseq_vartime(const fe25519 *x, const fe25519 *y)
+{
+  fe25519 t1 = *x;
+  fe25519 t2 = *y;
+  fe25519_freeze(&t1);
+  fe25519_freeze(&t2);
+  if(t1.v[0] != t2.v[0]) return 0;
+  if(t1.v[1] != t2.v[1]) return 0;
+  if(t1.v[2] != t2.v[2]) return 0;
+  if(t1.v[3] != t2.v[3]) return 0;
+  return 1;
+}

ext/ed25519-amd64-asm/fe25519_iszero.c

@@ -0,0 +1,12 @@
+#include "fe25519.h"
+
+int fe25519_iszero_vartime(const fe25519 *x)
+{
+  fe25519 t = *x;
+  fe25519_freeze(&t);
+  if (t.v[0]) return 0;
+  if (t.v[1]) return 0;
+  if (t.v[2]) return 0;
+  if (t.v[3]) return 0;
+  return 1;
+}

ext/ed25519-amd64-asm/fe25519_mul.s

@@ -0,0 +1,865 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: int64 mulr4
+
+# qhasm: int64 mulr5
+
+# qhasm: int64 mulr6
+
+# qhasm: int64 mulr7
+
+# qhasm: int64 mulr8
+
+# qhasm: int64 mulrax
+
+# qhasm: int64 mulrdx
+
+# qhasm: int64 mulx0
+
+# qhasm: int64 mulx1
+
+# qhasm: int64 mulx2
+
+# qhasm: int64 mulx3
+
+# qhasm: int64 mulc
+
+# qhasm: int64 mulzero
+
+# qhasm: int64 muli38
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_fe25519_mul
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_fe25519_mul
+.globl crypto_sign_ed25519_amd64_64_fe25519_mul
+_crypto_sign_ed25519_amd64_64_fe25519_mul:
+crypto_sign_ed25519_amd64_64_fe25519_mul:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm:   caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm:   caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm:   caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm:   caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm:   caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm:   caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm:   caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: yp = yp
+# asm 1: mov  <yp=int64#3,>yp=int64#4
+# asm 2: mov  <yp=%rdx,>yp=%rcx
+mov  %rdx,%rcx
+
+# qhasm:   mulr4 = 0
+# asm 1: mov  $0,>mulr4=int64#5
+# asm 2: mov  $0,>mulr4=%r8
+mov  $0,%r8
+
+# qhasm:   mulr5 = 0
+# asm 1: mov  $0,>mulr5=int64#6
+# asm 2: mov  $0,>mulr5=%r9
+mov  $0,%r9
+
+# qhasm:   mulr6 = 0
+# asm 1: mov  $0,>mulr6=int64#8
+# asm 2: mov  $0,>mulr6=%r10
+mov  $0,%r10
+
+# qhasm:   mulr7 = 0
+# asm 1: mov  $0,>mulr7=int64#9
+# asm 2: mov  $0,>mulr7=%r11
+mov  $0,%r11
+
+# qhasm:   mulx0 = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>mulx0=int64#10
+# asm 2: movq   0(<xp=%rsi),>mulx0=%r12
+movq   0(%rsi),%r12
+
+# qhasm:   mulrax = *(uint64 *)(yp + 0)
+# asm 1: movq   0(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   0(<yp=%rcx),>mulrax=%rax
+movq   0(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx0
+# asm 1: mul  <mulx0=int64#10
+# asm 2: mul  <mulx0=%r12
+mul  %r12
+
+# qhasm:   r0 = mulrax
+# asm 1: mov  <mulrax=int64#7,>r0=int64#11
+# asm 2: mov  <mulrax=%rax,>r0=%r13
+mov  %rax,%r13
+
+# qhasm:   r1 = mulrdx
+# asm 1: mov  <mulrdx=int64#3,>r1=int64#12
+# asm 2: mov  <mulrdx=%rdx,>r1=%r14
+mov  %rdx,%r14
+
+# qhasm:   mulrax = *(uint64 *)(yp + 8)
+# asm 1: movq   8(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   8(<yp=%rcx),>mulrax=%rax
+movq   8(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx0
+# asm 1: mul  <mulx0=int64#10
+# asm 2: mul  <mulx0=%r12
+mul  %r12
+
+# qhasm:   carry? r1 += mulrax
+# asm 1: add  <mulrax=int64#7,<r1=int64#12
+# asm 2: add  <mulrax=%rax,<r1=%r14
+add  %rax,%r14
+
+# qhasm:   r2 = 0
+# asm 1: mov  $0,>r2=int64#13
+# asm 2: mov  $0,>r2=%r15
+mov  $0,%r15
+
+# qhasm:   r2 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<r2=int64#13
+# asm 2: adc <mulrdx=%rdx,<r2=%r15
+adc %rdx,%r15
+
+# qhasm:   mulrax = *(uint64 *)(yp + 16)
+# asm 1: movq   16(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   16(<yp=%rcx),>mulrax=%rax
+movq   16(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx0
+# asm 1: mul  <mulx0=int64#10
+# asm 2: mul  <mulx0=%r12
+mul  %r12
+
+# qhasm:   carry? r2 += mulrax
+# asm 1: add  <mulrax=int64#7,<r2=int64#13
+# asm 2: add  <mulrax=%rax,<r2=%r15
+add  %rax,%r15
+
+# qhasm:   r3 = 0
+# asm 1: mov  $0,>r3=int64#14
+# asm 2: mov  $0,>r3=%rbx
+mov  $0,%rbx
+
+# qhasm:   r3 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<r3=int64#14
+# asm 2: adc <mulrdx=%rdx,<r3=%rbx
+adc %rdx,%rbx
+
+# qhasm:   mulrax = *(uint64 *)(yp + 24)
+# asm 1: movq   24(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   24(<yp=%rcx),>mulrax=%rax
+movq   24(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx0
+# asm 1: mul  <mulx0=int64#10
+# asm 2: mul  <mulx0=%r12
+mul  %r12
+
+# qhasm:   carry? r3 += mulrax
+# asm 1: add  <mulrax=int64#7,<r3=int64#14
+# asm 2: add  <mulrax=%rax,<r3=%rbx
+add  %rax,%rbx
+
+# qhasm:   mulr4 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr4=int64#5
+# asm 2: adc <mulrdx=%rdx,<mulr4=%r8
+adc %rdx,%r8
+
+# qhasm:   mulx1 = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>mulx1=int64#10
+# asm 2: movq   8(<xp=%rsi),>mulx1=%r12
+movq   8(%rsi),%r12
+
+# qhasm:   mulrax = *(uint64 *)(yp + 0)
+# asm 1: movq   0(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   0(<yp=%rcx),>mulrax=%rax
+movq   0(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx1
+# asm 1: mul  <mulx1=int64#10
+# asm 2: mul  <mulx1=%r12
+mul  %r12
+
+# qhasm:   carry? r1 += mulrax
+# asm 1: add  <mulrax=int64#7,<r1=int64#12
+# asm 2: add  <mulrax=%rax,<r1=%r14
+add  %rax,%r14
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#15
+# asm 2: mov  $0,>mulc=%rbp
+mov  $0,%rbp
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#15
+# asm 2: adc <mulrdx=%rdx,<mulc=%rbp
+adc %rdx,%rbp
+
+# qhasm:   mulrax = *(uint64 *)(yp + 8)
+# asm 1: movq   8(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   8(<yp=%rcx),>mulrax=%rax
+movq   8(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx1
+# asm 1: mul  <mulx1=int64#10
+# asm 2: mul  <mulx1=%r12
+mul  %r12
+
+# qhasm:   carry? r2 += mulrax
+# asm 1: add  <mulrax=int64#7,<r2=int64#13
+# asm 2: add  <mulrax=%rax,<r2=%r15
+add  %rax,%r15
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? r2 += mulc
+# asm 1: add  <mulc=int64#15,<r2=int64#13
+# asm 2: add  <mulc=%rbp,<r2=%r15
+add  %rbp,%r15
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#15
+# asm 2: mov  $0,>mulc=%rbp
+mov  $0,%rbp
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#15
+# asm 2: adc <mulrdx=%rdx,<mulc=%rbp
+adc %rdx,%rbp
+
+# qhasm:   mulrax = *(uint64 *)(yp + 16)
+# asm 1: movq   16(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   16(<yp=%rcx),>mulrax=%rax
+movq   16(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx1
+# asm 1: mul  <mulx1=int64#10
+# asm 2: mul  <mulx1=%r12
+mul  %r12
+
+# qhasm:   carry? r3 += mulrax 
+# asm 1: add  <mulrax=int64#7,<r3=int64#14
+# asm 2: add  <mulrax=%rax,<r3=%rbx
+add  %rax,%rbx
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? r3 += mulc
+# asm 1: add  <mulc=int64#15,<r3=int64#14
+# asm 2: add  <mulc=%rbp,<r3=%rbx
+add  %rbp,%rbx
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#15
+# asm 2: mov  $0,>mulc=%rbp
+mov  $0,%rbp
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#15
+# asm 2: adc <mulrdx=%rdx,<mulc=%rbp
+adc %rdx,%rbp
+
+# qhasm:   mulrax = *(uint64 *)(yp + 24)
+# asm 1: movq   24(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   24(<yp=%rcx),>mulrax=%rax
+movq   24(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx1
+# asm 1: mul  <mulx1=int64#10
+# asm 2: mul  <mulx1=%r12
+mul  %r12
+
+# qhasm:   carry? mulr4 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr4=int64#5
+# asm 2: add  <mulrax=%rax,<mulr4=%r8
+add  %rax,%r8
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? mulr4 += mulc
+# asm 1: add  <mulc=int64#15,<mulr4=int64#5
+# asm 2: add  <mulc=%rbp,<mulr4=%r8
+add  %rbp,%r8
+
+# qhasm:   mulr5 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr5=int64#6
+# asm 2: adc <mulrdx=%rdx,<mulr5=%r9
+adc %rdx,%r9
+
+# qhasm:   mulx2 = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>mulx2=int64#10
+# asm 2: movq   16(<xp=%rsi),>mulx2=%r12
+movq   16(%rsi),%r12
+
+# qhasm:   mulrax = *(uint64 *)(yp + 0)
+# asm 1: movq   0(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   0(<yp=%rcx),>mulrax=%rax
+movq   0(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx2
+# asm 1: mul  <mulx2=int64#10
+# asm 2: mul  <mulx2=%r12
+mul  %r12
+
+# qhasm:   carry? r2 += mulrax
+# asm 1: add  <mulrax=int64#7,<r2=int64#13
+# asm 2: add  <mulrax=%rax,<r2=%r15
+add  %rax,%r15
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#15
+# asm 2: mov  $0,>mulc=%rbp
+mov  $0,%rbp
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#15
+# asm 2: adc <mulrdx=%rdx,<mulc=%rbp
+adc %rdx,%rbp
+
+# qhasm:   mulrax = *(uint64 *)(yp + 8)
+# asm 1: movq   8(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   8(<yp=%rcx),>mulrax=%rax
+movq   8(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx2
+# asm 1: mul  <mulx2=int64#10
+# asm 2: mul  <mulx2=%r12
+mul  %r12
+
+# qhasm:   carry? r3 += mulrax
+# asm 1: add  <mulrax=int64#7,<r3=int64#14
+# asm 2: add  <mulrax=%rax,<r3=%rbx
+add  %rax,%rbx
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? r3 += mulc
+# asm 1: add  <mulc=int64#15,<r3=int64#14
+# asm 2: add  <mulc=%rbp,<r3=%rbx
+add  %rbp,%rbx
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#15
+# asm 2: mov  $0,>mulc=%rbp
+mov  $0,%rbp
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#15
+# asm 2: adc <mulrdx=%rdx,<mulc=%rbp
+adc %rdx,%rbp
+
+# qhasm:   mulrax = *(uint64 *)(yp + 16)
+# asm 1: movq   16(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   16(<yp=%rcx),>mulrax=%rax
+movq   16(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx2
+# asm 1: mul  <mulx2=int64#10
+# asm 2: mul  <mulx2=%r12
+mul  %r12
+
+# qhasm:   carry? mulr4 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr4=int64#5
+# asm 2: add  <mulrax=%rax,<mulr4=%r8
+add  %rax,%r8
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? mulr4 += mulc
+# asm 1: add  <mulc=int64#15,<mulr4=int64#5
+# asm 2: add  <mulc=%rbp,<mulr4=%r8
+add  %rbp,%r8
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#15
+# asm 2: mov  $0,>mulc=%rbp
+mov  $0,%rbp
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#15
+# asm 2: adc <mulrdx=%rdx,<mulc=%rbp
+adc %rdx,%rbp
+
+# qhasm:   mulrax = *(uint64 *)(yp + 24)
+# asm 1: movq   24(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   24(<yp=%rcx),>mulrax=%rax
+movq   24(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx2
+# asm 1: mul  <mulx2=int64#10
+# asm 2: mul  <mulx2=%r12
+mul  %r12
+
+# qhasm:   carry? mulr5 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr5=int64#6
+# asm 2: add  <mulrax=%rax,<mulr5=%r9
+add  %rax,%r9
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? mulr5 += mulc
+# asm 1: add  <mulc=int64#15,<mulr5=int64#6
+# asm 2: add  <mulc=%rbp,<mulr5=%r9
+add  %rbp,%r9
+
+# qhasm:   mulr6 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr6=int64#8
+# asm 2: adc <mulrdx=%rdx,<mulr6=%r10
+adc %rdx,%r10
+
+# qhasm:   mulx3 = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>mulx3=int64#2
+# asm 2: movq   24(<xp=%rsi),>mulx3=%rsi
+movq   24(%rsi),%rsi
+
+# qhasm:   mulrax = *(uint64 *)(yp + 0)
+# asm 1: movq   0(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   0(<yp=%rcx),>mulrax=%rax
+movq   0(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx3
+# asm 1: mul  <mulx3=int64#2
+# asm 2: mul  <mulx3=%rsi
+mul  %rsi
+
+# qhasm:   carry? r3 += mulrax
+# asm 1: add  <mulrax=int64#7,<r3=int64#14
+# asm 2: add  <mulrax=%rax,<r3=%rbx
+add  %rax,%rbx
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#10
+# asm 2: mov  $0,>mulc=%r12
+mov  $0,%r12
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#10
+# asm 2: adc <mulrdx=%rdx,<mulc=%r12
+adc %rdx,%r12
+
+# qhasm:   mulrax = *(uint64 *)(yp + 8)
+# asm 1: movq   8(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   8(<yp=%rcx),>mulrax=%rax
+movq   8(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx3
+# asm 1: mul  <mulx3=int64#2
+# asm 2: mul  <mulx3=%rsi
+mul  %rsi
+
+# qhasm:   carry? mulr4 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr4=int64#5
+# asm 2: add  <mulrax=%rax,<mulr4=%r8
+add  %rax,%r8
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? mulr4 += mulc
+# asm 1: add  <mulc=int64#10,<mulr4=int64#5
+# asm 2: add  <mulc=%r12,<mulr4=%r8
+add  %r12,%r8
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#10
+# asm 2: mov  $0,>mulc=%r12
+mov  $0,%r12
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#10
+# asm 2: adc <mulrdx=%rdx,<mulc=%r12
+adc %rdx,%r12
+
+# qhasm:   mulrax = *(uint64 *)(yp + 16)
+# asm 1: movq   16(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   16(<yp=%rcx),>mulrax=%rax
+movq   16(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx3
+# asm 1: mul  <mulx3=int64#2
+# asm 2: mul  <mulx3=%rsi
+mul  %rsi
+
+# qhasm:   carry? mulr5 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr5=int64#6
+# asm 2: add  <mulrax=%rax,<mulr5=%r9
+add  %rax,%r9
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? mulr5 += mulc
+# asm 1: add  <mulc=int64#10,<mulr5=int64#6
+# asm 2: add  <mulc=%r12,<mulr5=%r9
+add  %r12,%r9
+
+# qhasm:   mulc = 0
+# asm 1: mov  $0,>mulc=int64#10
+# asm 2: mov  $0,>mulc=%r12
+mov  $0,%r12
+
+# qhasm:   mulc += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulc=int64#10
+# asm 2: adc <mulrdx=%rdx,<mulc=%r12
+adc %rdx,%r12
+
+# qhasm:   mulrax = *(uint64 *)(yp + 24)
+# asm 1: movq   24(<yp=int64#4),>mulrax=int64#7
+# asm 2: movq   24(<yp=%rcx),>mulrax=%rax
+movq   24(%rcx),%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * mulx3
+# asm 1: mul  <mulx3=int64#2
+# asm 2: mul  <mulx3=%rsi
+mul  %rsi
+
+# qhasm:   carry? mulr6 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr6=int64#8
+# asm 2: add  <mulrax=%rax,<mulr6=%r10
+add  %rax,%r10
+
+# qhasm:   mulrdx += 0 + carry
+# asm 1: adc $0,<mulrdx=int64#3
+# asm 2: adc $0,<mulrdx=%rdx
+adc $0,%rdx
+
+# qhasm:   carry? mulr6 += mulc
+# asm 1: add  <mulc=int64#10,<mulr6=int64#8
+# asm 2: add  <mulc=%r12,<mulr6=%r10
+add  %r12,%r10
+
+# qhasm:   mulr7 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr7=int64#9
+# asm 2: adc <mulrdx=%rdx,<mulr7=%r11
+adc %rdx,%r11
+
+# qhasm:   mulrax = mulr4
+# asm 1: mov  <mulr4=int64#5,>mulrax=int64#7
+# asm 2: mov  <mulr4=%r8,>mulrax=%rax
+mov  %r8,%rax
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   mulr4 = mulrax
+# asm 1: mov  <mulrax=int64#7,>mulr4=int64#2
+# asm 2: mov  <mulrax=%rax,>mulr4=%rsi
+mov  %rax,%rsi
+
+# qhasm:   mulrax = mulr5
+# asm 1: mov  <mulr5=int64#6,>mulrax=int64#7
+# asm 2: mov  <mulr5=%r9,>mulrax=%rax
+mov  %r9,%rax
+
+# qhasm:   mulr5 = mulrdx
+# asm 1: mov  <mulrdx=int64#3,>mulr5=int64#4
+# asm 2: mov  <mulrdx=%rdx,>mulr5=%rcx
+mov  %rdx,%rcx
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   carry? mulr5 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr5=int64#4
+# asm 2: add  <mulrax=%rax,<mulr5=%rcx
+add  %rax,%rcx
+
+# qhasm:   mulrax = mulr6
+# asm 1: mov  <mulr6=int64#8,>mulrax=int64#7
+# asm 2: mov  <mulr6=%r10,>mulrax=%rax
+mov  %r10,%rax
+
+# qhasm:   mulr6 = 0
+# asm 1: mov  $0,>mulr6=int64#5
+# asm 2: mov  $0,>mulr6=%r8
+mov  $0,%r8
+
+# qhasm:   mulr6 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr6=int64#5
+# asm 2: adc <mulrdx=%rdx,<mulr6=%r8
+adc %rdx,%r8
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   carry? mulr6 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr6=int64#5
+# asm 2: add  <mulrax=%rax,<mulr6=%r8
+add  %rax,%r8
+
+# qhasm:   mulrax = mulr7
+# asm 1: mov  <mulr7=int64#9,>mulrax=int64#7
+# asm 2: mov  <mulr7=%r11,>mulrax=%rax
+mov  %r11,%rax
+
+# qhasm:   mulr7 = 0
+# asm 1: mov  $0,>mulr7=int64#6
+# asm 2: mov  $0,>mulr7=%r9
+mov  $0,%r9
+
+# qhasm:   mulr7 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr7=int64#6
+# asm 2: adc <mulrdx=%rdx,<mulr7=%r9
+adc %rdx,%r9
+
+# qhasm:   (uint128) mulrdx mulrax = mulrax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   carry? mulr7 += mulrax
+# asm 1: add  <mulrax=int64#7,<mulr7=int64#6
+# asm 2: add  <mulrax=%rax,<mulr7=%r9
+add  %rax,%r9
+
+# qhasm:   mulr8 = 0
+# asm 1: mov  $0,>mulr8=int64#7
+# asm 2: mov  $0,>mulr8=%rax
+mov  $0,%rax
+
+# qhasm:   mulr8 += mulrdx + carry
+# asm 1: adc <mulrdx=int64#3,<mulr8=int64#7
+# asm 2: adc <mulrdx=%rdx,<mulr8=%rax
+adc %rdx,%rax
+
+# qhasm:   carry? r0 += mulr4
+# asm 1: add  <mulr4=int64#2,<r0=int64#11
+# asm 2: add  <mulr4=%rsi,<r0=%r13
+add  %rsi,%r13
+
+# qhasm:   carry? r1 += mulr5 + carry
+# asm 1: adc <mulr5=int64#4,<r1=int64#12
+# asm 2: adc <mulr5=%rcx,<r1=%r14
+adc %rcx,%r14
+
+# qhasm:   carry? r2 += mulr6 + carry
+# asm 1: adc <mulr6=int64#5,<r2=int64#13
+# asm 2: adc <mulr6=%r8,<r2=%r15
+adc %r8,%r15
+
+# qhasm:   carry? r3 += mulr7 + carry
+# asm 1: adc <mulr7=int64#6,<r3=int64#14
+# asm 2: adc <mulr7=%r9,<r3=%rbx
+adc %r9,%rbx
+
+# qhasm:   mulzero = 0
+# asm 1: mov  $0,>mulzero=int64#2
+# asm 2: mov  $0,>mulzero=%rsi
+mov  $0,%rsi
+
+# qhasm:   mulr8 += mulzero + carry
+# asm 1: adc <mulzero=int64#2,<mulr8=int64#7
+# asm 2: adc <mulzero=%rsi,<mulr8=%rax
+adc %rsi,%rax
+
+# qhasm:   mulr8 *= 38
+# asm 1: imulq  $38,<mulr8=int64#7,>mulr8=int64#3
+# asm 2: imulq  $38,<mulr8=%rax,>mulr8=%rdx
+imulq  $38,%rax,%rdx
+
+# qhasm:   carry? r0 += mulr8
+# asm 1: add  <mulr8=int64#3,<r0=int64#11
+# asm 2: add  <mulr8=%rdx,<r0=%r13
+add  %rdx,%r13
+
+# qhasm:   carry? r1 += mulzero + carry
+# asm 1: adc <mulzero=int64#2,<r1=int64#12
+# asm 2: adc <mulzero=%rsi,<r1=%r14
+adc %rsi,%r14
+
+# qhasm:   carry? r2 += mulzero + carry
+# asm 1: adc <mulzero=int64#2,<r2=int64#13
+# asm 2: adc <mulzero=%rsi,<r2=%r15
+adc %rsi,%r15
+
+# qhasm:   carry? r3 += mulzero + carry
+# asm 1: adc <mulzero=int64#2,<r3=int64#14
+# asm 2: adc <mulzero=%rsi,<r3=%rbx
+adc %rsi,%rbx
+
+# qhasm:   mulzero += mulzero + carry
+# asm 1: adc <mulzero=int64#2,<mulzero=int64#2
+# asm 2: adc <mulzero=%rsi,<mulzero=%rsi
+adc %rsi,%rsi
+
+# qhasm:   mulzero *= 38
+# asm 1: imulq  $38,<mulzero=int64#2,>mulzero=int64#2
+# asm 2: imulq  $38,<mulzero=%rsi,>mulzero=%rsi
+imulq  $38,%rsi,%rsi
+
+# qhasm:   r0 += mulzero
+# asm 1: add  <mulzero=int64#2,<r0=int64#11
+# asm 2: add  <mulzero=%rsi,<r0=%r13
+add  %rsi,%r13
+
+# qhasm: *(uint64 *)(rp + 8) = r1
+# asm 1: movq   <r1=int64#12,8(<rp=int64#1)
+# asm 2: movq   <r1=%r14,8(<rp=%rdi)
+movq   %r14,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#13,16(<rp=int64#1)
+# asm 2: movq   <r2=%r15,16(<rp=%rdi)
+movq   %r15,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#14,24(<rp=int64#1)
+# asm 2: movq   <r3=%rbx,24(<rp=%rdi)
+movq   %rbx,24(%rdi)
+
+# qhasm: *(uint64 *)(rp + 0) = r0
+# asm 1: movq   <r0=int64#11,0(<rp=int64#1)
+# asm 2: movq   <r0=%r13,0(<rp=%rdi)
+movq   %r13,0(%rdi)
+
+# qhasm:   caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm:   caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm:   caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm:   caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm:   caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm:   caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm:   caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/fe25519_neg.c

@@ -0,0 +1,8 @@
+#include "fe25519.h"
+
+void fe25519_neg(fe25519 *r, const fe25519 *x)
+{
+  fe25519 t;
+  fe25519_setint(&t,0);
+  fe25519_sub(r,&t,x);
+}

ext/ed25519-amd64-asm/fe25519_pack.c

@@ -0,0 +1,13 @@
+#include "fe25519.h"
+
+/* Assumes input x being reduced below 2^255 */
+void fe25519_pack(unsigned char r[32], const fe25519 *x)
+{
+  int i;
+  fe25519 t;
+  t = *x;
+  fe25519_freeze(&t);
+  /* assuming little-endian */
+  for(i=0;i<32;i++) r[i] = i[(unsigned char *)&t.v]; 
+}
+

ext/ed25519-amd64-asm/fe25519_pow2523.c

@@ -0,0 +1,55 @@
+#include "fe25519.h"
+
+void fe25519_pow2523(fe25519 *r, const fe25519 *x)
+{
+	fe25519 z2;
+	fe25519 z9;
+	fe25519 z11;
+	fe25519 z2_5_0;
+	fe25519 z2_10_0;
+	fe25519 z2_20_0;
+	fe25519 z2_50_0;
+	fe25519 z2_100_0;
+	fe25519 t;
+	int i;
+		
+	/* 2 */ fe25519_square(&z2,x);
+	/* 4 */ fe25519_square(&t,&z2);
+	/* 8 */ fe25519_square(&t,&t);
+	/* 9 */ fe25519_mul(&z9,&t,x);
+	/* 11 */ fe25519_mul(&z11,&z9,&z2);
+	/* 22 */ fe25519_square(&t,&z11);
+	/* 2^5 - 2^0 = 31 */ fe25519_mul(&z2_5_0,&t,&z9);
+
+	/* 2^6 - 2^1 */ fe25519_square(&t,&z2_5_0);
+	/* 2^10 - 2^5 */ for (i = 1;i < 5;i++) { fe25519_square(&t,&t); }
+	/* 2^10 - 2^0 */ fe25519_mul(&z2_10_0,&t,&z2_5_0);
+
+	/* 2^11 - 2^1 */ fe25519_square(&t,&z2_10_0);
+	/* 2^20 - 2^10 */ for (i = 1;i < 10;i++) { fe25519_square(&t,&t); }
+	/* 2^20 - 2^0 */ fe25519_mul(&z2_20_0,&t,&z2_10_0);
+
+	/* 2^21 - 2^1 */ fe25519_square(&t,&z2_20_0);
+	/* 2^40 - 2^20 */ for (i = 1;i < 20;i++) { fe25519_square(&t,&t); }
+	/* 2^40 - 2^0 */ fe25519_mul(&t,&t,&z2_20_0);
+
+	/* 2^41 - 2^1 */ fe25519_square(&t,&t);
+	/* 2^50 - 2^10 */ for (i = 1;i < 10;i++) { fe25519_square(&t,&t); }
+	/* 2^50 - 2^0 */ fe25519_mul(&z2_50_0,&t,&z2_10_0);
+
+	/* 2^51 - 2^1 */ fe25519_square(&t,&z2_50_0);
+	/* 2^100 - 2^50 */ for (i = 1;i < 50;i++) { fe25519_square(&t,&t); }
+	/* 2^100 - 2^0 */ fe25519_mul(&z2_100_0,&t,&z2_50_0);
+
+	/* 2^101 - 2^1 */ fe25519_square(&t,&z2_100_0);
+	/* 2^200 - 2^100 */ for (i = 1;i < 100;i++) { fe25519_square(&t,&t); }
+	/* 2^200 - 2^0 */ fe25519_mul(&t,&t,&z2_100_0);
+
+	/* 2^201 - 2^1 */ fe25519_square(&t,&t);
+	/* 2^250 - 2^50 */ for (i = 1;i < 50;i++) { fe25519_square(&t,&t); }
+	/* 2^250 - 2^0 */ fe25519_mul(&t,&t,&z2_50_0);
+
+	/* 2^251 - 2^1 */ fe25519_square(&t,&t);
+	/* 2^252 - 2^2 */ fe25519_square(&t,&t);
+	/* 2^252 - 3 */ fe25519_mul(r,&t,x);
+}

ext/ed25519-amd64-asm/fe25519_setint.c

@@ -0,0 +1,9 @@
+#include "fe25519.h"
+
+void fe25519_setint(fe25519 *r, unsigned int v)
+{
+  r->v[0] = v;
+  r->v[1] = 0;
+  r->v[2] = 0;
+  r->v[3] = 0;
+}

ext/ed25519-amd64-asm/fe25519_square.s

@@ -0,0 +1,639 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: int64 squarer4
+
+# qhasm: int64 squarer5
+
+# qhasm: int64 squarer6
+
+# qhasm: int64 squarer7
+
+# qhasm: int64 squarer8
+
+# qhasm: int64 squarerax
+
+# qhasm: int64 squarerdx
+
+# qhasm: int64 squaret1
+
+# qhasm: int64 squaret2
+
+# qhasm: int64 squaret3
+
+# qhasm: int64 squarec
+
+# qhasm: int64 squarezero
+
+# qhasm: int64 squarei38
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_fe25519_square
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_fe25519_square
+.globl crypto_sign_ed25519_amd64_64_fe25519_square
+_crypto_sign_ed25519_amd64_64_fe25519_square:
+crypto_sign_ed25519_amd64_64_fe25519_square:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm:   caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm:   caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm:   caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm:   caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm:   caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm:   caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm:   caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm:   squarer7 = 0
+# asm 1: mov  $0,>squarer7=int64#4
+# asm 2: mov  $0,>squarer7=%rcx
+mov  $0,%rcx
+
+# qhasm:   squarerax = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   8(<xp=%rsi),>squarerax=%rax
+movq   8(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 0)
+# asm 1: mulq  0(<xp=int64#2)
+# asm 2: mulq  0(<xp=%rsi)
+mulq  0(%rsi)
+
+# qhasm:   r1 = squarerax
+# asm 1: mov  <squarerax=int64#7,>r1=int64#5
+# asm 2: mov  <squarerax=%rax,>r1=%r8
+mov  %rax,%r8
+
+# qhasm:   r2 = squarerdx
+# asm 1: mov  <squarerdx=int64#3,>r2=int64#6
+# asm 2: mov  <squarerdx=%rdx,>r2=%r9
+mov  %rdx,%r9
+
+# qhasm:   squarerax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   16(<xp=%rsi),>squarerax=%rax
+movq   16(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 8)
+# asm 1: mulq  8(<xp=int64#2)
+# asm 2: mulq  8(<xp=%rsi)
+mulq  8(%rsi)
+
+# qhasm:   r3 = squarerax 
+# asm 1: mov  <squarerax=int64#7,>r3=int64#8
+# asm 2: mov  <squarerax=%rax,>r3=%r10
+mov  %rax,%r10
+
+# qhasm:   squarer4 = squarerdx
+# asm 1: mov  <squarerdx=int64#3,>squarer4=int64#9
+# asm 2: mov  <squarerdx=%rdx,>squarer4=%r11
+mov  %rdx,%r11
+
+# qhasm:   squarerax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   24(<xp=%rsi),>squarerax=%rax
+movq   24(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 16)
+# asm 1: mulq  16(<xp=int64#2)
+# asm 2: mulq  16(<xp=%rsi)
+mulq  16(%rsi)
+
+# qhasm:   squarer5 = squarerax
+# asm 1: mov  <squarerax=int64#7,>squarer5=int64#10
+# asm 2: mov  <squarerax=%rax,>squarer5=%r12
+mov  %rax,%r12
+
+# qhasm:   squarer6 = squarerdx
+# asm 1: mov  <squarerdx=int64#3,>squarer6=int64#11
+# asm 2: mov  <squarerdx=%rdx,>squarer6=%r13
+mov  %rdx,%r13
+
+# qhasm:   squarerax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   16(<xp=%rsi),>squarerax=%rax
+movq   16(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 0)
+# asm 1: mulq  0(<xp=int64#2)
+# asm 2: mulq  0(<xp=%rsi)
+mulq  0(%rsi)
+
+# qhasm:   carry? r2 += squarerax
+# asm 1: add  <squarerax=int64#7,<r2=int64#6
+# asm 2: add  <squarerax=%rax,<r2=%r9
+add  %rax,%r9
+
+# qhasm:   carry? r3 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<r3=int64#8
+# asm 2: adc <squarerdx=%rdx,<r3=%r10
+adc %rdx,%r10
+
+# qhasm:   squarer4 += 0 + carry
+# asm 1: adc $0,<squarer4=int64#9
+# asm 2: adc $0,<squarer4=%r11
+adc $0,%r11
+
+# qhasm:   squarerax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   24(<xp=%rsi),>squarerax=%rax
+movq   24(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 8)
+# asm 1: mulq  8(<xp=int64#2)
+# asm 2: mulq  8(<xp=%rsi)
+mulq  8(%rsi)
+
+# qhasm:   carry? squarer4 += squarerax
+# asm 1: add  <squarerax=int64#7,<squarer4=int64#9
+# asm 2: add  <squarerax=%rax,<squarer4=%r11
+add  %rax,%r11
+
+# qhasm:   carry? squarer5 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer5=int64#10
+# asm 2: adc <squarerdx=%rdx,<squarer5=%r12
+adc %rdx,%r12
+
+# qhasm:   squarer6 += 0 + carry
+# asm 1: adc $0,<squarer6=int64#11
+# asm 2: adc $0,<squarer6=%r13
+adc $0,%r13
+
+# qhasm:   squarerax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   24(<xp=%rsi),>squarerax=%rax
+movq   24(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 0)
+# asm 1: mulq  0(<xp=int64#2)
+# asm 2: mulq  0(<xp=%rsi)
+mulq  0(%rsi)
+
+# qhasm:   carry? r3 += squarerax
+# asm 1: add  <squarerax=int64#7,<r3=int64#8
+# asm 2: add  <squarerax=%rax,<r3=%r10
+add  %rax,%r10
+
+# qhasm:   carry? squarer4 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer4=int64#9
+# asm 2: adc <squarerdx=%rdx,<squarer4=%r11
+adc %rdx,%r11
+
+# qhasm:   carry? squarer5 += 0 + carry
+# asm 1: adc $0,<squarer5=int64#10
+# asm 2: adc $0,<squarer5=%r12
+adc $0,%r12
+
+# qhasm:   carry? squarer6 += 0 + carry
+# asm 1: adc $0,<squarer6=int64#11
+# asm 2: adc $0,<squarer6=%r13
+adc $0,%r13
+
+# qhasm:   squarer7 += 0 + carry
+# asm 1: adc $0,<squarer7=int64#4
+# asm 2: adc $0,<squarer7=%rcx
+adc $0,%rcx
+
+# qhasm:   carry? r1 += r1
+# asm 1: add  <r1=int64#5,<r1=int64#5
+# asm 2: add  <r1=%r8,<r1=%r8
+add  %r8,%r8
+
+# qhasm:   carry? r2 += r2 + carry
+# asm 1: adc <r2=int64#6,<r2=int64#6
+# asm 2: adc <r2=%r9,<r2=%r9
+adc %r9,%r9
+
+# qhasm:   carry? r3 += r3 + carry
+# asm 1: adc <r3=int64#8,<r3=int64#8
+# asm 2: adc <r3=%r10,<r3=%r10
+adc %r10,%r10
+
+# qhasm:   carry? squarer4 += squarer4 + carry
+# asm 1: adc <squarer4=int64#9,<squarer4=int64#9
+# asm 2: adc <squarer4=%r11,<squarer4=%r11
+adc %r11,%r11
+
+# qhasm:   carry? squarer5 += squarer5 + carry
+# asm 1: adc <squarer5=int64#10,<squarer5=int64#10
+# asm 2: adc <squarer5=%r12,<squarer5=%r12
+adc %r12,%r12
+
+# qhasm:   carry? squarer6 += squarer6 + carry
+# asm 1: adc <squarer6=int64#11,<squarer6=int64#11
+# asm 2: adc <squarer6=%r13,<squarer6=%r13
+adc %r13,%r13
+
+# qhasm:   squarer7 += squarer7 + carry
+# asm 1: adc <squarer7=int64#4,<squarer7=int64#4
+# asm 2: adc <squarer7=%rcx,<squarer7=%rcx
+adc %rcx,%rcx
+
+# qhasm:   squarerax = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   0(<xp=%rsi),>squarerax=%rax
+movq   0(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 0)
+# asm 1: mulq  0(<xp=int64#2)
+# asm 2: mulq  0(<xp=%rsi)
+mulq  0(%rsi)
+
+# qhasm:   r0 = squarerax
+# asm 1: mov  <squarerax=int64#7,>r0=int64#12
+# asm 2: mov  <squarerax=%rax,>r0=%r14
+mov  %rax,%r14
+
+# qhasm:   squaret1 = squarerdx
+# asm 1: mov  <squarerdx=int64#3,>squaret1=int64#13
+# asm 2: mov  <squarerdx=%rdx,>squaret1=%r15
+mov  %rdx,%r15
+
+# qhasm:   squarerax = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   8(<xp=%rsi),>squarerax=%rax
+movq   8(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 8)
+# asm 1: mulq  8(<xp=int64#2)
+# asm 2: mulq  8(<xp=%rsi)
+mulq  8(%rsi)
+
+# qhasm:   squaret2 = squarerax
+# asm 1: mov  <squarerax=int64#7,>squaret2=int64#14
+# asm 2: mov  <squarerax=%rax,>squaret2=%rbx
+mov  %rax,%rbx
+
+# qhasm:   squaret3 = squarerdx
+# asm 1: mov  <squarerdx=int64#3,>squaret3=int64#15
+# asm 2: mov  <squarerdx=%rdx,>squaret3=%rbp
+mov  %rdx,%rbp
+
+# qhasm:   squarerax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   16(<xp=%rsi),>squarerax=%rax
+movq   16(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 16)
+# asm 1: mulq  16(<xp=int64#2)
+# asm 2: mulq  16(<xp=%rsi)
+mulq  16(%rsi)
+
+# qhasm:   carry? r1 += squaret1
+# asm 1: add  <squaret1=int64#13,<r1=int64#5
+# asm 2: add  <squaret1=%r15,<r1=%r8
+add  %r15,%r8
+
+# qhasm:   carry? r2 += squaret2 + carry
+# asm 1: adc <squaret2=int64#14,<r2=int64#6
+# asm 2: adc <squaret2=%rbx,<r2=%r9
+adc %rbx,%r9
+
+# qhasm:   carry? r3 += squaret3 + carry
+# asm 1: adc <squaret3=int64#15,<r3=int64#8
+# asm 2: adc <squaret3=%rbp,<r3=%r10
+adc %rbp,%r10
+
+# qhasm:   carry? squarer4 += squarerax + carry
+# asm 1: adc <squarerax=int64#7,<squarer4=int64#9
+# asm 2: adc <squarerax=%rax,<squarer4=%r11
+adc %rax,%r11
+
+# qhasm:   carry? squarer5 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer5=int64#10
+# asm 2: adc <squarerdx=%rdx,<squarer5=%r12
+adc %rdx,%r12
+
+# qhasm:   carry? squarer6 += 0 + carry
+# asm 1: adc $0,<squarer6=int64#11
+# asm 2: adc $0,<squarer6=%r13
+adc $0,%r13
+
+# qhasm:   squarer7 += 0 + carry 
+# asm 1: adc $0,<squarer7=int64#4
+# asm 2: adc $0,<squarer7=%rcx
+adc $0,%rcx
+
+# qhasm:   squarerax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>squarerax=int64#7
+# asm 2: movq   24(<xp=%rsi),>squarerax=%rax
+movq   24(%rsi),%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)(xp + 24)
+# asm 1: mulq  24(<xp=int64#2)
+# asm 2: mulq  24(<xp=%rsi)
+mulq  24(%rsi)
+
+# qhasm:   carry? squarer6 += squarerax
+# asm 1: add  <squarerax=int64#7,<squarer6=int64#11
+# asm 2: add  <squarerax=%rax,<squarer6=%r13
+add  %rax,%r13
+
+# qhasm:   squarer7 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer7=int64#4
+# asm 2: adc <squarerdx=%rdx,<squarer7=%rcx
+adc %rdx,%rcx
+
+# qhasm:   squarerax = squarer4
+# asm 1: mov  <squarer4=int64#9,>squarerax=int64#7
+# asm 2: mov  <squarer4=%r11,>squarerax=%rax
+mov  %r11,%rax
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   squarer4 = squarerax
+# asm 1: mov  <squarerax=int64#7,>squarer4=int64#2
+# asm 2: mov  <squarerax=%rax,>squarer4=%rsi
+mov  %rax,%rsi
+
+# qhasm:   squarerax = squarer5
+# asm 1: mov  <squarer5=int64#10,>squarerax=int64#7
+# asm 2: mov  <squarer5=%r12,>squarerax=%rax
+mov  %r12,%rax
+
+# qhasm:   squarer5 = squarerdx
+# asm 1: mov  <squarerdx=int64#3,>squarer5=int64#9
+# asm 2: mov  <squarerdx=%rdx,>squarer5=%r11
+mov  %rdx,%r11
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   carry? squarer5 += squarerax
+# asm 1: add  <squarerax=int64#7,<squarer5=int64#9
+# asm 2: add  <squarerax=%rax,<squarer5=%r11
+add  %rax,%r11
+
+# qhasm:   squarerax = squarer6
+# asm 1: mov  <squarer6=int64#11,>squarerax=int64#7
+# asm 2: mov  <squarer6=%r13,>squarerax=%rax
+mov  %r13,%rax
+
+# qhasm:   squarer6 = 0
+# asm 1: mov  $0,>squarer6=int64#10
+# asm 2: mov  $0,>squarer6=%r12
+mov  $0,%r12
+
+# qhasm:   squarer6 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer6=int64#10
+# asm 2: adc <squarerdx=%rdx,<squarer6=%r12
+adc %rdx,%r12
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   carry? squarer6 += squarerax
+# asm 1: add  <squarerax=int64#7,<squarer6=int64#10
+# asm 2: add  <squarerax=%rax,<squarer6=%r12
+add  %rax,%r12
+
+# qhasm:   squarerax = squarer7
+# asm 1: mov  <squarer7=int64#4,>squarerax=int64#7
+# asm 2: mov  <squarer7=%rcx,>squarerax=%rax
+mov  %rcx,%rax
+
+# qhasm:   squarer7 = 0
+# asm 1: mov  $0,>squarer7=int64#4
+# asm 2: mov  $0,>squarer7=%rcx
+mov  $0,%rcx
+
+# qhasm:   squarer7 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer7=int64#4
+# asm 2: adc <squarerdx=%rdx,<squarer7=%rcx
+adc %rdx,%rcx
+
+# qhasm:   (uint128) squarerdx squarerax = squarerax * *(uint64 *)&crypto_sign_ed25519_amd64_64_38
+mulq  crypto_sign_ed25519_amd64_64_38
+
+# qhasm:   carry? squarer7 += squarerax
+# asm 1: add  <squarerax=int64#7,<squarer7=int64#4
+# asm 2: add  <squarerax=%rax,<squarer7=%rcx
+add  %rax,%rcx
+
+# qhasm:   squarer8 = 0
+# asm 1: mov  $0,>squarer8=int64#7
+# asm 2: mov  $0,>squarer8=%rax
+mov  $0,%rax
+
+# qhasm:   squarer8 += squarerdx + carry
+# asm 1: adc <squarerdx=int64#3,<squarer8=int64#7
+# asm 2: adc <squarerdx=%rdx,<squarer8=%rax
+adc %rdx,%rax
+
+# qhasm:   carry? r0 += squarer4
+# asm 1: add  <squarer4=int64#2,<r0=int64#12
+# asm 2: add  <squarer4=%rsi,<r0=%r14
+add  %rsi,%r14
+
+# qhasm:   carry? r1 += squarer5 + carry
+# asm 1: adc <squarer5=int64#9,<r1=int64#5
+# asm 2: adc <squarer5=%r11,<r1=%r8
+adc %r11,%r8
+
+# qhasm:   carry? r2 += squarer6 + carry
+# asm 1: adc <squarer6=int64#10,<r2=int64#6
+# asm 2: adc <squarer6=%r12,<r2=%r9
+adc %r12,%r9
+
+# qhasm:   carry? r3 += squarer7 + carry
+# asm 1: adc <squarer7=int64#4,<r3=int64#8
+# asm 2: adc <squarer7=%rcx,<r3=%r10
+adc %rcx,%r10
+
+# qhasm:   squarezero = 0
+# asm 1: mov  $0,>squarezero=int64#2
+# asm 2: mov  $0,>squarezero=%rsi
+mov  $0,%rsi
+
+# qhasm:   squarer8 += squarezero + carry
+# asm 1: adc <squarezero=int64#2,<squarer8=int64#7
+# asm 2: adc <squarezero=%rsi,<squarer8=%rax
+adc %rsi,%rax
+
+# qhasm:   squarer8 *= 38
+# asm 1: imulq  $38,<squarer8=int64#7,>squarer8=int64#3
+# asm 2: imulq  $38,<squarer8=%rax,>squarer8=%rdx
+imulq  $38,%rax,%rdx
+
+# qhasm:   carry? r0 += squarer8
+# asm 1: add  <squarer8=int64#3,<r0=int64#12
+# asm 2: add  <squarer8=%rdx,<r0=%r14
+add  %rdx,%r14
+
+# qhasm:   carry? r1 += squarezero + carry
+# asm 1: adc <squarezero=int64#2,<r1=int64#5
+# asm 2: adc <squarezero=%rsi,<r1=%r8
+adc %rsi,%r8
+
+# qhasm:   carry? r2 += squarezero + carry
+# asm 1: adc <squarezero=int64#2,<r2=int64#6
+# asm 2: adc <squarezero=%rsi,<r2=%r9
+adc %rsi,%r9
+
+# qhasm:   carry? r3 += squarezero + carry
+# asm 1: adc <squarezero=int64#2,<r3=int64#8
+# asm 2: adc <squarezero=%rsi,<r3=%r10
+adc %rsi,%r10
+
+# qhasm:   squarezero += squarezero + carry
+# asm 1: adc <squarezero=int64#2,<squarezero=int64#2
+# asm 2: adc <squarezero=%rsi,<squarezero=%rsi
+adc %rsi,%rsi
+
+# qhasm:   squarezero *= 38
+# asm 1: imulq  $38,<squarezero=int64#2,>squarezero=int64#2
+# asm 2: imulq  $38,<squarezero=%rsi,>squarezero=%rsi
+imulq  $38,%rsi,%rsi
+
+# qhasm:   r0 += squarezero
+# asm 1: add  <squarezero=int64#2,<r0=int64#12
+# asm 2: add  <squarezero=%rsi,<r0=%r14
+add  %rsi,%r14
+
+# qhasm: *(uint64 *)(rp + 8) = r1
+# asm 1: movq   <r1=int64#5,8(<rp=int64#1)
+# asm 2: movq   <r1=%r8,8(<rp=%rdi)
+movq   %r8,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#6,16(<rp=int64#1)
+# asm 2: movq   <r2=%r9,16(<rp=%rdi)
+movq   %r9,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#8,24(<rp=int64#1)
+# asm 2: movq   <r3=%r10,24(<rp=%rdi)
+movq   %r10,24(%rdi)
+
+# qhasm: *(uint64 *)(rp + 0) = r0
+# asm 1: movq   <r0=int64#12,0(<rp=int64#1)
+# asm 2: movq   <r0=%r14,0(<rp=%rdi)
+movq   %r14,0(%rdi)
+
+# qhasm:   caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm:   caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm:   caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm:   caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm:   caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm:   caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm:   caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/fe25519_sub.s

@@ -0,0 +1,189 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm: int64 subt0
+
+# qhasm: int64 subt1
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_fe25519_sub
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_fe25519_sub
+.globl crypto_sign_ed25519_amd64_64_fe25519_sub
+_crypto_sign_ed25519_amd64_64_fe25519_sub:
+crypto_sign_ed25519_amd64_64_fe25519_sub:
+mov %rsp,%r11
+and $31,%r11
+add $0,%r11
+sub %r11,%rsp
+
+# qhasm: r0 = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>r0=int64#4
+# asm 2: movq   0(<xp=%rsi),>r0=%rcx
+movq   0(%rsi),%rcx
+
+# qhasm: r1 = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>r1=int64#5
+# asm 2: movq   8(<xp=%rsi),>r1=%r8
+movq   8(%rsi),%r8
+
+# qhasm: r2 = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>r2=int64#6
+# asm 2: movq   16(<xp=%rsi),>r2=%r9
+movq   16(%rsi),%r9
+
+# qhasm: r3 = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>r3=int64#2
+# asm 2: movq   24(<xp=%rsi),>r3=%rsi
+movq   24(%rsi),%rsi
+
+# qhasm:   carry? r0 -= *(uint64 *)(yp + 0) 
+# asm 1: subq 0(<yp=int64#3),<r0=int64#4
+# asm 2: subq 0(<yp=%rdx),<r0=%rcx
+subq 0(%rdx),%rcx
+
+# qhasm:   carry? r1 -= *(uint64 *)(yp + 8) - carry
+# asm 1: sbbq 8(<yp=int64#3),<r1=int64#5
+# asm 2: sbbq 8(<yp=%rdx),<r1=%r8
+sbbq 8(%rdx),%r8
+
+# qhasm:   carry? r2 -= *(uint64 *)(yp + 16) - carry
+# asm 1: sbbq 16(<yp=int64#3),<r2=int64#6
+# asm 2: sbbq 16(<yp=%rdx),<r2=%r9
+sbbq 16(%rdx),%r9
+
+# qhasm:   carry? r3 -= *(uint64 *)(yp + 24) - carry
+# asm 1: sbbq 24(<yp=int64#3),<r3=int64#2
+# asm 2: sbbq 24(<yp=%rdx),<r3=%rsi
+sbbq 24(%rdx),%rsi
+
+# qhasm:   subt0 = 0
+# asm 1: mov  $0,>subt0=int64#3
+# asm 2: mov  $0,>subt0=%rdx
+mov  $0,%rdx
+
+# qhasm:   subt1 = 38
+# asm 1: mov  $38,>subt1=int64#7
+# asm 2: mov  $38,>subt1=%rax
+mov  $38,%rax
+
+# qhasm:   subt1 = subt0 if !carry
+# asm 1: cmovae <subt0=int64#3,<subt1=int64#7
+# asm 2: cmovae <subt0=%rdx,<subt1=%rax
+cmovae %rdx,%rax
+
+# qhasm:   carry? r0 -= subt1
+# asm 1: sub  <subt1=int64#7,<r0=int64#4
+# asm 2: sub  <subt1=%rax,<r0=%rcx
+sub  %rax,%rcx
+
+# qhasm:   carry? r1 -= subt0 - carry
+# asm 1: sbb  <subt0=int64#3,<r1=int64#5
+# asm 2: sbb  <subt0=%rdx,<r1=%r8
+sbb  %rdx,%r8
+
+# qhasm:   carry? r2 -= subt0 - carry
+# asm 1: sbb  <subt0=int64#3,<r2=int64#6
+# asm 2: sbb  <subt0=%rdx,<r2=%r9
+sbb  %rdx,%r9
+
+# qhasm:   carry? r3 -= subt0 - carry
+# asm 1: sbb  <subt0=int64#3,<r3=int64#2
+# asm 2: sbb  <subt0=%rdx,<r3=%rsi
+sbb  %rdx,%rsi
+
+# qhasm:   subt0 = subt1 if carry
+# asm 1: cmovc <subt1=int64#7,<subt0=int64#3
+# asm 2: cmovc <subt1=%rax,<subt0=%rdx
+cmovc %rax,%rdx
+
+# qhasm:   r0 -= subt0
+# asm 1: sub  <subt0=int64#3,<r0=int64#4
+# asm 2: sub  <subt0=%rdx,<r0=%rcx
+sub  %rdx,%rcx
+
+# qhasm: *(uint64 *)(rp + 0) = r0
+# asm 1: movq   <r0=int64#4,0(<rp=int64#1)
+# asm 2: movq   <r0=%rcx,0(<rp=%rdi)
+movq   %rcx,0(%rdi)
+
+# qhasm: *(uint64 *)(rp + 8) = r1
+# asm 1: movq   <r1=int64#5,8(<rp=int64#1)
+# asm 2: movq   <r1=%r8,8(<rp=%rdi)
+movq   %r8,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#6,16(<rp=int64#1)
+# asm 2: movq   <r2=%r9,16(<rp=%rdi)
+movq   %r9,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#2,24(<rp=int64#1)
+# asm 2: movq   <r3=%rsi,24(<rp=%rdi)
+movq   %rsi,24(%rdi)
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/fe25519_unpack.c

@@ -0,0 +1,11 @@
+#include "fe25519.h"
+
+void fe25519_unpack(fe25519 *r, const unsigned char x[32])
+{
+  /* assuming little-endian */
+  r->v[0] = *(unsigned long long *)x;
+  r->v[1] = *(((unsigned long long *)x)+1);
+  r->v[2] = *(((unsigned long long *)x)+2);
+  r->v[3] = *(((unsigned long long *)x)+3);
+  r->v[3] &= 0x7fffffffffffffffULL;
+}

ext/ed25519-amd64-asm/ge25519.h

@@ -0,0 +1,95 @@
+#ifndef GE25519_H
+#define GE25519_H
+
+#include "fe25519.h"
+#include "sc25519.h"
+
+#define ge25519                           crypto_sign_ed25519_amd64_64_ge25519
+#define ge25519_base                      crypto_sign_ed25519_amd64_64_ge25519_base
+#define ge25519_unpackneg_vartime         crypto_sign_ed25519_amd64_64_unpackneg_vartime
+#define ge25519_pack                      crypto_sign_ed25519_amd64_64_pack
+#define ge25519_isneutral_vartime         crypto_sign_ed25519_amd64_64_isneutral_vartime
+#define ge25519_add                       crypto_sign_ed25519_amd64_64_ge25519_add
+#define ge25519_double                    crypto_sign_ed25519_amd64_64_ge25519_double
+#define ge25519_double_scalarmult_vartime crypto_sign_ed25519_amd64_64_double_scalarmult_vartime
+#define ge25519_multi_scalarmult_vartime  crypto_sign_ed25519_amd64_64_ge25519_multi_scalarmult_vartime
+#define ge25519_scalarmult_base           crypto_sign_ed25519_amd64_64_scalarmult_base
+#define ge25519_p1p1_to_p2                crypto_sign_ed25519_amd64_64_ge25519_p1p1_to_p2
+#define ge25519_p1p1_to_p3                crypto_sign_ed25519_amd64_64_ge25519_p1p1_to_p3
+#define ge25519_add_p1p1                  crypto_sign_ed25519_amd64_64_ge25519_add_p1p1
+#define ge25519_dbl_p1p1                  crypto_sign_ed25519_amd64_64_ge25519_dbl_p1p1
+#define choose_t                          crypto_sign_ed25519_amd64_64_choose_t
+#define ge25519_nielsadd2                 crypto_sign_ed25519_amd64_64_ge25519_nielsadd2
+#define ge25519_nielsadd_p1p1             crypto_sign_ed25519_amd64_64_ge25519_nielsadd_p1p1
+#define ge25519_pnielsadd_p1p1            crypto_sign_ed25519_amd64_64_ge25519_pnielsadd_p1p1
+
+
+#define ge25519_p3 ge25519
+
+typedef struct
+{
+  fe25519 x;
+  fe25519 y;
+  fe25519 z;
+  fe25519 t;
+} ge25519;
+
+typedef struct
+{
+  fe25519 x;
+  fe25519 z;
+  fe25519 y;
+  fe25519 t;
+} ge25519_p1p1;
+
+typedef struct
+{
+  fe25519 x;
+  fe25519 y;
+  fe25519 z;
+} ge25519_p2;
+
+typedef struct
+{
+  fe25519 ysubx;
+  fe25519 xaddy;
+  fe25519 t2d;
+} ge25519_niels;
+
+typedef struct
+{
+  fe25519 ysubx;
+  fe25519 xaddy;
+  fe25519 z;
+  fe25519 t2d;
+} ge25519_pniels;
+
+extern void ge25519_p1p1_to_p2(ge25519_p2 *r, const ge25519_p1p1 *p);
+extern void ge25519_p1p1_to_p3(ge25519_p3 *r, const ge25519_p1p1 *p);
+extern void ge25519_add_p1p1(ge25519_p1p1 *r, const ge25519_p3 *p, const ge25519_p3 *q);
+extern void ge25519_dbl_p1p1(ge25519_p1p1 *r, const ge25519_p2 *p);
+extern void choose_t(ge25519_niels *t, unsigned long long pos, signed long long b, const ge25519_niels *base_multiples);
+extern void ge25519_nielsadd2(ge25519_p3 *r, const ge25519_niels *q);
+extern void ge25519_nielsadd_p1p1(ge25519_p1p1 *r, const ge25519_p3 *p, const ge25519_niels *q);
+extern void ge25519_pnielsadd_p1p1(ge25519_p1p1 *r, const ge25519_p3 *p, const ge25519_pniels *q);
+
+extern const ge25519 ge25519_base;
+
+extern int ge25519_unpackneg_vartime(ge25519 *r, const unsigned char p[32]);
+
+extern void ge25519_pack(unsigned char r[32], const ge25519 *p);
+
+extern int ge25519_isneutral_vartime(const ge25519 *p);
+
+extern void ge25519_add(ge25519 *r, const ge25519 *p, const ge25519 *q);
+
+extern void ge25519_double(ge25519 *r, const ge25519 *p);
+
+/* computes [s1]p1 + [s2]ge25519_base */
+extern void ge25519_double_scalarmult_vartime(ge25519 *r, const ge25519 *p1, const sc25519 *s1, const sc25519 *s2);
+
+extern void ge25519_multi_scalarmult_vartime(ge25519 *r, ge25519 *p, sc25519 *s, const unsigned long long npoints);
+
+extern void ge25519_scalarmult_base(ge25519 *r, const sc25519 *s);
+
+#endif

ext/ed25519-amd64-asm/ge25519_add.c

@@ -0,0 +1,8 @@
+#include "ge25519.h"
+
+void ge25519_add(ge25519_p3 *r, const ge25519_p3 *p, const ge25519_p3 *q)
+{
+  ge25519_p1p1 grp1p1;
+  ge25519_add_p1p1(&grp1p1, p, q);
+  ge25519_p1p1_to_p3(r, &grp1p1);
+}

ext/ed25519-amd64-asm/ge25519_base.c

@@ -0,0 +1,7 @@
+#include "ge25519.h"
+
+const ge25519 ge25519_base = {{{0xC9562D608F25D51A, 0x692CC7609525A7B2, 0xC0A4E231FDD6DC5C, 0x216936D3CD6E53FE}},
+                              {{0x6666666666666658, 0x6666666666666666, 0x6666666666666666, 0x6666666666666666}},
+                              {{0x0000000000000001, 0x0000000000000000, 0x0000000000000000, 000000000000000000}},
+                              {{0x6DDE8AB3A5B7DDA3, 0x20F09F80775152F5, 0x66EA4E8E64ABE37D, 0x67875F0FD78B7665}}};
+

ext/ed25519-amd64-asm/ge25519_base_niels_smalltables.data

@@ -0,0 +1,768 @@
+{{{0x9d103905d740913e, 0xfd399f05d140beb3, 0xa5c18434688f8a09, 0x44fd2f9298f81267}},
+ {{0x2fbc93c6f58c3b85, 0xcf932dc6fb8c0e19, 0x270b4898643d42c2, 0x07cf9d3a33d4ba65}},
+ {{0xdbbd15674b6fbb59, 0x41e13f00eea2a5ea, 0xcdd49d1cc957c6fa, 0x4f0ebe1faf16ecca}}},
+{{{0x8a99a56042b4d5a8, 0x8f2b810c4e60acf6, 0xe09e236bb16e37aa, 0x6bb595a669c92555}},
+ {{0x9224e7fc933c71d7, 0x9f469d967a0ff5b5, 0x5aa69a65e1d60702, 0x590c063fa87d2e2e}},
+ {{0x6e347eaadad36802, 0xbaf3599383ee4805, 0x3bcabe10e6076826, 0x49314f0a165ed1b8}}},
+{{{0x56611fe8a4fcd265, 0x3bd353fde5c1ba7d, 0x8131f31a214bd6bd, 0x2ab91587555bda62}},
+ {{0xaf25b0a84cee9730, 0x025a8430e8864b8a, 0xc11b50029f016732, 0x7a164e1b9a80f8f4}},
+ {{0x9bf211f4f1674834, 0xb84e6b17f62df895, 0xd7de6f075b722a4e, 0x549a04b963bb2a21}}},
+{{{0x95fe050a056818bf, 0x327e89715660faa9, 0xc3e8e3cd06a05073, 0x27933f4c7445a49a}},
+ {{0x287351b98efc099f, 0x6765c6f47dfd2538, 0xca348d3dfb0a9265, 0x680e910321e58727}},
+ {{0xbf1e45ece51426b0, 0xe32bc63d6dba0f94, 0xe42974d58cf852c0, 0x44f079b1b0e64c18}}},
+{{{0x7f9182c3a447d6ba, 0xd50014d14b2729b7, 0xe33cf11cb864a087, 0x154a7e73eb1b55f3}},
+ {{0xa212bc4408a5bb33, 0x8d5048c3c75eed02, 0xdd1beb0c5abfec44, 0x2945ccf146e206eb}},
+ {{0xc832a179e7d003b3, 0x5f729d0a00124d7e, 0x62c1d4a10e6d8ff3, 0x68b8ac5938b27a98}}},
+{{{0x499806b67b7d8ca4, 0x575be28427d22739, 0xbb085ce7204553b9, 0x38b64c41ae417884}},
+ {{0x3a0ceeeb77157131, 0x9b27158900c8af88, 0x8065b668da59a736, 0x51e57bb6a2cc38bd}},
+ {{0x8f9dad91689de3a4, 0x175f2428f8fb9137, 0x050ab5329fcfb988, 0x7865dfa21354c09f}}},
+{{{0xba6f2c9aaa3221b1, 0x6ca021533bba23a7, 0x9dea764f92192c3a, 0x1d6edd5d2e5317e0}},
+ {{0x6b1a5cd0944ea3bf, 0x7470353ab39dc0d2, 0x71b2528228542e49, 0x461bea69283c927e}},
+ {{0x217a8aacab0fda36, 0xa528c6543d3549c8, 0x37d05b8b13ab7568, 0x233cef623a2cbc37}}},
+{{{0xe2a75dedf39234d9, 0x963d7680e1b558f9, 0x2c2741ac6e3c23fb, 0x3a9024a1320e01c3}},
+ {{0x59b7596604dd3e8f, 0x6cb30377e288702c, 0xb1339c665ed9c323, 0x0915e76061bce52f}},
+ {{0xdf7de835a834a37e, 0x8be19cda689857ea, 0x2c1185367167b326, 0x589eb3d9dbefd5c2}}},
+{{{0xed5b635449aa515e, 0xa865c49f0bc6823a, 0x850c1fe95b42d1c4, 0x30d76d6f03d315b9}},
+ {{0x2eccdd0e632f9c1d, 0x51d0b69676893115, 0x52dfb76ba8637a58, 0x6dd37d49a00eef39}},
+ {{0x6c4444172106e4c7, 0xfb53d680928d7f69, 0xb4739ea4694d3f26, 0x10c697112e864bb0}}},
+{{{0x6493c4277dbe5fde, 0x265d4fad19ad7ea2, 0x0e00dfc846304590, 0x25e61cabed66fe09}},
+ {{0x0ca62aa08358c805, 0x6a3d4ae37a204247, 0x7464d3a63b11eddc, 0x03bf9baf550806ef}},
+ {{0x3f13e128cc586604, 0x6f5873ecb459747e, 0xa0b63dedcc1268f5, 0x566d78634586e22c}}},
+{{{0x1637a49f9cc10834, 0xbc8e56d5a89bc451, 0x1cb5ec0f7f7fd2db, 0x33975bca5ecc35d9}},
+ {{0xa1054285c65a2fd0, 0x6c64112af31667c3, 0x680ae240731aee58, 0x14fba5f34793b22a}},
+ {{0x3cd746166985f7d4, 0x593e5e84c9c80057, 0x2fc3f2b67b61131e, 0x14829cea83fc526c}}},
+{{{0xff437b8497dd95c2, 0x6c744e30aa4eb5a7, 0x9e0c5d613c85e88b, 0x2fd9c71e5f758173}},
+ {{0x21e70b2f4e71ecb8, 0xe656ddb940a477e3, 0xbf6556cece1d4f80, 0x05fc3bc4535d7b7e}},
+ {{0x24b8b3ae52afdedd, 0x3495638ced3b30cf, 0x33a4bc83a9be8195, 0x373767475c651f04}}},
+{{{0x2fba99fd40d1add9, 0xb307166f96f4d027, 0x4363f05215f03bae, 0x1fbea56c3b18f999}},
+ {{0x634095cb14246590, 0xef12144016c15535, 0x9e38140c8910bc60, 0x6bf5905730907c8c}},
+ {{0x0fa778f1e1415b8a, 0x06409ff7bac3a77e, 0x6f52d7b89aa29a50, 0x02521cf67a635a56}}},
+{{{0x513fee0b0a9d5294, 0x8f98e75c0fdf5a66, 0xd4618688bfe107ce, 0x3fa00a7e71382ced}},
+ {{0xb1146720772f5ee4, 0xe8f894b196079ace, 0x4af8224d00ac824a, 0x001753d9f7cd6cc4}},
+ {{0x3c69232d963ddb34, 0x1dde87dab4973858, 0xaad7d1f9a091f285, 0x12b5fe2fa048edb6}}},
+{{{0x71f0fbc496fce34d, 0x73b9826badf35bed, 0xd2047261ff28c561, 0x749b76f96fb1206f}},
+ {{0xdf2b7c26ad6f1e92, 0x4b66d323504b8913, 0x8c409dc0751c8bc3, 0x6f7e93c20796c7b8}},
+ {{0x1f5af604aea6ae05, 0xc12351f1bee49c99, 0x61a808b5eeff6b66, 0x0fcec10f01e02151}}},
+{{{0x644d58a649fe1e44, 0x21fcaea231ad777e, 0x02441c5a887fd0d2, 0x4901aa7183c511f3}},
+ {{0x3df2d29dc4244e45, 0x2b020e7493d8de0a, 0x6cc8067e820c214d, 0x413779166feab90a}},
+ {{0x08b1b7548c1af8f0, 0xce0f7a7c246299b4, 0xf760b0f91e06d939, 0x41bb887b726d1213}}},
+{{{0x40e87d44744346be, 0x1d48dad415b52b25, 0x7c3a8a18a13b603e, 0x4eb728c12fcdbdf7}},
+ {{0x7e234c597c6691ae, 0x64889d3d0a85b4c8, 0xdae2c90c354afae7, 0x0a871e070c6a9e1d}},
+ {{0x3301b5994bbc8989, 0x736bae3a5bdd4260, 0x0d61ade219d59e3c, 0x3ee7300f2685d464}}},
+{{{0xf5d255e49e7dd6b7, 0x8016115c610b1eac, 0x3c99975d92e187ca, 0x13815762979125c2}},
+ {{0x43fa7947841e7518, 0xe5c6fa59639c46d7, 0xa1065e1de3052b74, 0x7d47c6a2cfb89030}},
+ {{0x3fdad0148ef0d6e0, 0x9d3e749a91546f3c, 0x71ec621026bb8157, 0x148cf58d34c9ec80}}},
+{{{0x46a492f67934f027, 0x469984bef6840aa9, 0x5ca1bc2a89611854, 0x3ff2fa1ebd5dbbd4}},
+ {{0xe2572f7d9ae4756d, 0x56c345bb88f3487f, 0x9fd10b6d6960a88d, 0x278febad4eaea1b9}},
+ {{0xb1aa681f8c933966, 0x8c21949c20290c98, 0x39115291219d3c52, 0x4104dd02fe9c677b}}},
+{{{0x72b2bf5e1124422a, 0xa1fa0c3398a33ab5, 0x94cb6101fa52b666, 0x2c863b00afaf53d5}},
+ {{0x81214e06db096ab8, 0x21a8b6c90ce44f35, 0x6524c12a409e2af5, 0x0165b5a48efca481}},
+ {{0xf190a474a0846a76, 0x12eff984cd2f7cc0, 0x695e290658aa2b8f, 0x591b67d9bffec8b8}}},
+{{{0x312f0d1c80b49bfa, 0x5979515eabf3ec8a, 0x727033c09ef01c88, 0x3de02ec7ca8f7bcb}},
+ {{0x99b9b3719f18b55d, 0xe465e5faa18c641e, 0x61081136c29f05ed, 0x489b4f867030128b}},
+ {{0xd232102d3aeb92ef, 0xe16253b46116a861, 0x3d7eabe7190baa24, 0x49f5fbba496cbebf}}},
+{{{0x30949a108a5bcfd4, 0xdc40dd70bc6473eb, 0x92c294c1307c0d1c, 0x5604a86dcbfa6e74}},
+ {{0x155d628c1e9c572e, 0x8a4d86acc5884741, 0x91a352f6515763eb, 0x06a1a6c28867515b}},
+ {{0x7288d1d47c1764b6, 0x72541140e0418b51, 0x9f031a6018acf6d1, 0x20989e89fe2742c6}}},
+{{{0x499777fd3a2dcc7f, 0x32857c2ca54fd892, 0xa279d864d207e3a0, 0x0403ed1d0ca67e29}},
+ {{0x1674278b85eaec2e, 0x5621dc077acb2bdf, 0x640a4c1661cbf45a, 0x730b9950f70595d3}},
+ {{0xc94b2d35874ec552, 0xc5e6c8cf98246f8d, 0xf7cb46fa16c035ce, 0x5bd7454308303dcc}}},
+{{{0x7f9ad19528b24cc2, 0x7f6b54656335c181, 0x66b8b66e4fc07236, 0x133a78007380ad83}},
+ {{0x85c4932115e7792a, 0xc64c89a2bdcdddc9, 0x9d1e3da8ada3d762, 0x5bb7db123067f82c}},
+ {{0x0961f467c6ca62be, 0x04ec21d6211952ee, 0x182360779bd54770, 0x740dca6d58f0e0d2}}},
+{{{0xdf48ee0752cfce4e, 0xc3fffaf306ec08b7, 0x05710b2ab95459c4, 0x161d25fa963ea38d}},
+ {{0x231a8c570478433c, 0xb7b5270ec281439d, 0xdbaa99eae3d9079f, 0x2c03f5256c2b03d9}},
+ {{0x790f18757b53a47d, 0x307b0130cf0c5879, 0x31903d77257ef7f9, 0x699468bdbd96bbaf}}},
+{{{0xbd1f2f46f4dafecf, 0x7cef0114a47fd6f7, 0xd31ffdda4a47b37f, 0x525219a473905785}},
+ {{0xd8dd3de66aa91948, 0x485064c22fc0d2cc, 0x9b48246634fdea2f, 0x293e1c4e6c4a2e3a}},
+ {{0x376e134b925112e1, 0x703778b5dca15da0, 0xb04589af461c3111, 0x5b605c447f032823}}},
+{{{0xb965805920c47c89, 0xe7f0100c923b8fcc, 0x0001256502e2ef77, 0x24a76dcea8aeb3ee}},
+ {{0x3be9fec6f0e7f04c, 0x866a579e75e34962, 0x5542ef161e1de61a, 0x2f12fef4cc5abdd5}},
+ {{0x0a4522b2dfc0c740, 0x10d06e7f40c9a407, 0xc6cf144178cff668, 0x5e607b2518a43790}}},
+{{{0x58b31d8f6cdf1818, 0x35cfa74fc36258a2, 0xe1b3ff4f66e61d6e, 0x5067acab6ccdd5f7}},
+ {{0xa02c431ca596cf14, 0xe3c42d40aed3e400, 0xd24526802e0f26db, 0x201f33139e457068}},
+ {{0xfd527f6b08039d51, 0x18b14964017c0006, 0xd5220eb02e25a4a8, 0x397cba8862460375}}},
+{{{0x30c13093f05959b2, 0xe23aa18de9a97976, 0x222fd491721d5e26, 0x2339d320766e6c3a}},
+ {{0x7815c3fbc81379e7, 0xa6619420dde12af1, 0xffa9c0f885a8fdd5, 0x771b4022c1e1c252}},
+ {{0xd87dd986513a2fa7, 0xf5ac9b71f9d4cf08, 0xd06bc31b1ea283b3, 0x331a189219971a76}}},
+{{{0xf5166f45fb4f80c6, 0x9c36c7de61c775cf, 0xe3d4e81b9041d91c, 0x31167c6b83bdfe21}},
+ {{0x26512f3a9d7572af, 0x5bcbe28868074a9e, 0x84edc1c11180f7c4, 0x1ac9619ff649a67b}},
+ {{0xf22b3842524b1068, 0x5068343bee9ce987, 0xfc9d71844a6250c8, 0x612436341f08b111}}},
+{{{0xd99d41db874e898d, 0x09fea5f16c07dc20, 0x793d2c67d00f9bbc, 0x46ebe2309e5eff40}},
+ {{0x8b6349e31a2d2638, 0x9ddfb7009bd3fd35, 0x7f8bf1b8a3a06ba4, 0x1522aa3178d90445}},
+ {{0x2c382f5369614938, 0xdafe409ab72d6d10, 0xe8c83391b646f227, 0x45fe70f50524306c}}},
+{{{0xda4875a6960c0b8c, 0x5b68d076ef0e2f20, 0x07fb51cf3d0b8fd4, 0x428d1623a0e392d4}},
+ {{0x62f24920c8951491, 0x05f007c83f630ca2, 0x6fbb45d2f5c9d4b8, 0x16619f6db57a2245}},
+ {{0x084f4a4401a308fd, 0xa82219c376a5caac, 0xdeb8de4643d1bc7d, 0x1d81592d60bd38c6}}},
+{{{0x61368756a60dac5f, 0x17e02f6aebabdc57, 0x7f193f2d4cce0f7d, 0x20234a7789ecdcf0}},
+ {{0x8765b69f7b85c5e8, 0x6ff0678bd168bab2, 0x3a70e77c1d330f9b, 0x3a5f6d51b0af8e7c}},
+ {{0x76d20db67178b252, 0x071c34f9d51ed160, 0xf62a4a20b3e41170, 0x7cd682353cffe366}}},
+{{{0x0be1a45bd887fab6, 0x2a846a32ba403b6e, 0xd9921012e96e6000, 0x2838c8863bdc0943}},
+ {{0xa665cd6068acf4f3, 0x42d92d183cd7e3d3, 0x5759389d336025d9, 0x3ef0253b2b2cd8ff}},
+ {{0xd16bb0cf4a465030, 0xfa496b4115c577ab, 0x82cfae8af4ab419d, 0x21dcb8a606a82812}}},
+{{{0x5c6004468c9d9fc8, 0x2540096ed42aa3cb, 0x125b4d4c12ee2f9c, 0x0bc3d08194a31dab}},
+ {{0x9a8d00fabe7731ba, 0x8203607e629e1889, 0xb2cc023743f3d97f, 0x5d840dbf6c6f678b}},
+ {{0x706e380d309fe18b, 0x6eb02da6b9e165c7, 0x57bbba997dae20ab, 0x3a4276232ac196dd}}},
+{{{0x4b42432c8a7084fa, 0x898a19e3dfb9e545, 0xbe9f00219c58e45d, 0x1ff177cea16debd1}},
+ {{0x3bf8c172db447ecb, 0x5fcfc41fc6282dbd, 0x80acffc075aa15fe, 0x0770c9e824e1a9f9}},
+ {{0xcf61d99a45b5b5fd, 0x860984e91b3a7924, 0xe7300919303e3e89, 0x39f264fd41500b1e}}},
+{{{0xa7ad3417dbe7e29c, 0xbd94376a2b9c139c, 0xa0e91b8e93597ba9, 0x1712d73468889840}},
+ {{0xd19b4aabfe097be1, 0xa46dfce1dfe01929, 0xc3c908942ca6f1ff, 0x65c621272c35f14e}},
+ {{0xe72b89f8ce3193dd, 0x4d103356a125c0bb, 0x0419a93d2e1cfe83, 0x22f9800ab19ce272}}},
+{{{0x605a368a3e9ef8cb, 0xe3e9c022a5504715, 0x553d48b05f24248f, 0x13f416cd647626e5}},
+ {{0x42029fdd9a6efdac, 0xb912cebe34a54941, 0x640f64b987bdf37b, 0x4171a4d38598cab4}},
+ {{0xfa2758aa99c94c8c, 0x23006f6fb000b807, 0xfbd291ddadda5392, 0x508214fa574bd1ab}}},
+{{{0xc20269153ed6fe4b, 0xa65a6739511d77c4, 0xcbde26462c14af94, 0x22f960ec6faba74b}},
+ {{0x461a15bb53d003d6, 0xb2102888bcf3c965, 0x27c576756c683a5a, 0x3a7758a4c86cb447}},
+ {{0x548111f693ae5076, 0x1dae21df1dfd54a6, 0x12248c90f3115e65, 0x5d9fd15f8de7f494}}},
+{{{0x031408d36d63727f, 0x6a379aefd7c7b533, 0xa9e18fc5ccaee24b, 0x332f35914f8fbed3}},
+ {{0x3f244d2aeed7521e, 0x8e3a9028432e9615, 0xe164ba772e9c16d4, 0x3bc187fa47eb98d8}},
+ {{0x6d470115ea86c20c, 0x998ab7cb6c46d125, 0xd77832b53a660188, 0x450d81ce906fba03}}},
+{{{0x6e7bb6a1a6205275, 0xaa4f21d7413c8e83, 0x6f56d155e88f5cb2, 0x2de25d4ba6345be1}},
+ {{0xd074d8961cae743f, 0xf86d18f5ee1c63ed, 0x97bdc55be7f4ed29, 0x4cbad279663ab108}},
+ {{0x80d19024a0d71fcd, 0xc525c20afb288af8, 0xb1a3974b5f3a6419, 0x7d7fbcefe2007233}}},
+{{{0xfaef1e6a266b2801, 0x866c68c4d5739f16, 0xf68a2fbc1b03762c, 0x5975435e87b75a8d}},
+ {{0xcd7c5dc5f3c29094, 0xc781a29a2a9105ab, 0x80c61d36421c3058, 0x4f9cd196dcd8d4d7}},
+ {{0x199297d86a7b3768, 0xd0d058241ad17a63, 0xba029cad5c1c0c17, 0x7ccdd084387a0307}}},
+{{{0xdca6422c6d260417, 0xae153d50948240bd, 0xa9c0c1b4fb68c677, 0x428bd0ed61d0cf53}},
+ {{0x9b0c84186760cc93, 0xcdae007a1ab32a99, 0xa88dec86620bda18, 0x3593ca848190ca44}},
+ {{0x9213189a5e849aa7, 0xd4d8c33565d8facd, 0x8c52545b53fdbbd1, 0x27398308da2d63e6}}},
+{{{0x42c38d28435ed413, 0xbd50f3603278ccc9, 0xbb07ab1a79da03ef, 0x269597aebe8c3355}},
+ {{0xb9a10e4c0a702453, 0x0fa25866d57d1bde, 0xffb9d9b5cd27daf7, 0x572c2945492c33fd}},
+ {{0xc77fc745d6cd30be, 0xe4dfe8d3e3baaefb, 0xa22c8830aa5dda0c, 0x7f985498c05bca80}}},
+{{{0x3849ce889f0be117, 0x8005ad1b7b54a288, 0x3da3c39f23fc921c, 0x76c2ec470a31f304}},
+ {{0xd35615520fbf6363, 0x08045a45cf4dfba6, 0xeec24fbc873fa0c2, 0x30f2653cd69b12e7}},
+ {{0x8a08c938aac10c85, 0x46179b60db276bcb, 0xa920c01e0e6fac70, 0x2f1273f1596473da}}},
+{{{0x4739fc7c8ae01e11, 0xfd5274904a6aab9f, 0x41d98a8287728f2e, 0x5d9e572ad85b69f2}},
+ {{0x30488bd755a70bc0, 0x06d6b5a4f1d442e7, 0xead1a69ebc596162, 0x38ac1997edc5f784}},
+ {{0x0666b517a751b13b, 0x747d06867e9b858c, 0xacacc011454dde49, 0x22dfcd9cbfe9e69c}}},
+{{{0x8ddbd2e0c30d0cd9, 0xad8e665facbb4333, 0x8f6b258c322a961f, 0x6b2916c05448c1c7}},
+ {{0x56ec59b4103be0a1, 0x2ee3baecd259f969, 0x797cb29413f5cd32, 0x0fe9877824cde472}},
+ {{0x7edb34d10aba913b, 0x4ea3cd822e6dac0e, 0x66083dff6578f815, 0x4c303f307ff00a17}}},
+{{{0xd30a3bd617b28c85, 0xc5d377b739773bea, 0xc6c6e78c1e6a5cbf, 0x0d61b8f78b2ab7c4}},
+ {{0x29fc03580dd94500, 0xecd27aa46fbbec93, 0x130a155fc2e2a7f8, 0x416b151ab706a1d5}},
+ {{0x56a8d7efe9c136b0, 0xbd07e5cd58e44b20, 0xafe62fda1b57e0ab, 0x191a2af74277e8d2}}},
+{{{0xce16f74bc53c1431, 0x2b9725ce2072edde, 0xb8b9c36fb5b23ee7, 0x7e2e0e450b5cc908}},
+ {{0x9fe62b434f460efb, 0xded303d4a63607d6, 0xf052210eb7a0da24, 0x237e7dbe00545b93}},
+ {{0x013575ed6701b430, 0x231094e69f0bfd10, 0x75320f1583e47f22, 0x71afa699b11155e3}}},
+{{{0x65ce6f9b3953b61d, 0xc65839eaafa141e6, 0x0f435ffda9f759fe, 0x021142e9c2b1c28e}},
+ {{0xea423c1c473b50d6, 0x51e87a1f3b38ef10, 0x9b84bf5fb2c9be95, 0x00731fbc78f89a1c}},
+ {{0xe430c71848f81880, 0xbf960c225ecec119, 0xb6dae0836bba15e3, 0x4c4d6f3347e15808}}},
+{{{0x18f7eccfc17d1fc9, 0x6c75f5a651403c14, 0xdbde712bf7ee0cdf, 0x193fddaaa7e47a22}},
+ {{0x2f0cddfc988f1970, 0x6b916227b0b9f51b, 0x6ec7b6c4779176be, 0x38bf9500a88f9fa8}},
+ {{0x1fd2c93c37e8876f, 0xa2f61e5a18d1462c, 0x5080f58239241276, 0x6a6fb99ebf0d4969}}},
+{{{0x6a46c1bb560855eb, 0x2416bb38f893f09d, 0xd71d11378f71acc1, 0x75f76914a31896ea}},
+ {{0xeeb122b5b6e423c6, 0x939d7010f286ff8e, 0x90a92a831dcf5d8c, 0x136fda9f42c5eb10}},
+ {{0xf94cdfb1a305bdd1, 0x0f364b9d9ff82c08, 0x2a87d8a5c3bb588a, 0x022183510be8dcba}}},
+{{{0x4af766385ead2d14, 0xa08ed880ca7c5830, 0x0d13a6e610211e3d, 0x6a071ce17b806c03}},
+ {{0x9d5a710143307a7f, 0xb063de9ec47da45f, 0x22bbfe52be927ad3, 0x1387c441fd40426c}},
+ {{0xb5d3c3d187978af8, 0x722b5a3d7f0e4413, 0x0d7b4848bb477ca0, 0x3171b26aaf1edc92}}},
+{{{0xa92f319097564ca8, 0xff7bb84c2275e119, 0x4f55fe37a4875150, 0x221fd4873cf0835a}},
+ {{0xa60db7d8b28a47d1, 0xa6bf14d61770a4f1, 0xd4a1f89353ddbd58, 0x6c514a63344243e9}},
+ {{0x2322204f3a156341, 0xfb73e0e9ba0a032d, 0xfce0dd4c410f030e, 0x48daa596fb924aaa}}},
+{{{0x6eca8e665ca59cc7, 0xa847254b2e38aca0, 0x31afc708d21e17ce, 0x676dd6fccad84af7}},
+ {{0x14f61d5dc84c9793, 0x9941f9e3ef418206, 0xcdf5b88f346277ac, 0x58c837fa0e8a79a9}},
+ {{0x0cf9688596fc9058, 0x1ddcbbf37b56a01b, 0xdcc2e77d4935d66a, 0x1c4f73f2c6a57f0a}}},
+{{{0x0e7a4fbd305fa0bb, 0x829d4ce054c663ad, 0xf421c3832fe33848, 0x795ac80d1bf64c42}},
+ {{0xb36e706efc7c3484, 0x73dfc9b4c3c1cf61, 0xeb1d79c9781cc7e5, 0x70459adb7daf675c}},
+ {{0x1b91db4991b42bb3, 0x572696234b02dcca, 0x9fdf9ee51f8c78dc, 0x5fe162848ce21fd3}}},
+{{{0x4e59214fe194961a, 0x49be7dc70d71cd4f, 0x9300cfd23b50f22d, 0x4789d446fc917232}},
+ {{0x2879852d5d7cb208, 0xb8dedd70687df2e7, 0xdc0bffab21687891, 0x2b44c043677daa35}},
+ {{0x1a1c87ab074eb78e, 0xfac6d18e99daf467, 0x3eacbbcd484f9067, 0x60c52eef2bb9a4e4}}},
+{{{0x0b5d89bc3bfd8bf1, 0xb06b9237c9f3551a, 0x0e4c16b0d53028f5, 0x10bc9c312ccfcaab}},
+ {{0x702bc5c27cae6d11, 0x44c7699b54a48cab, 0xefbc4056ba492eb2, 0x70d77248d9b6676d}},
+ {{0xaa8ae84b3ec2a05b, 0x98699ef4ed1781e0, 0x794513e4708e85d1, 0x63755bd3a976f413}}},
+{{{0xb55fa03e2ad10853, 0x356f75909ee63569, 0x9ff9f1fdbe69b890, 0x0d8cc1c48bc16f84}},
+ {{0x3dc7101897f1acb7, 0x5dda7d5ec165bbd8, 0x508e5b9c0fa1020f, 0x2763751737c52a56}},
+ {{0x029402d36eb419a9, 0xf0b44e7e77b460a5, 0xcfa86230d43c4956, 0x70c2dd8a7ad166e7}}},
+{{{0x656194509f6fec0e, 0xee2e7ea946c6518d, 0x9733c1f367e09b5c, 0x2e0fac6363948495}},
+ {{0x91d4967db8ed7e13, 0x74252f0ad776817a, 0xe40982e00d852564, 0x32b8613816a53ce5}},
+ {{0x79e7f7bee448cd64, 0x6ac83a67087886d0, 0xf89fd4d9a0e4db2e, 0x4179215c735a4f41}}},
+{{{0x8c7094e7d7dced2a, 0x97fb8ac347d39c70, 0xe13be033a906d902, 0x700344a30cd99d76}},
+ {{0xe4ae33b9286bcd34, 0xb7ef7eb6559dd6dc, 0x278b141fb3d38e1f, 0x31fa85662241c286}},
+ {{0xaf826c422e3622f4, 0xc12029879833502d, 0x9bc1b7e12b389123, 0x24bb2312a9952489}}},
+{{{0xb1a8ed1732de67c3, 0x3cb49418461b4948, 0x8ebd434376cfbcd2, 0x0fee3e871e188008}},
+ {{0x41f80c2af5f85c6b, 0x687284c304fa6794, 0x8945df99a3ba1bad, 0x0d1d2af9ffeb5d16}},
+ {{0xa9da8aa132621edf, 0x30b822a159226579, 0x4004197ba79ac193, 0x16acd79718531d76}}},
+{{{0x72df72af2d9b1d3d, 0x63462a36a432245a, 0x3ecea07916b39637, 0x123e0ef6b9302309}},
+ {{0xc959c6c57887b6ad, 0x94e19ead5f90feba, 0x16e24e62a342f504, 0x164ed34b18161700}},
+ {{0x487ed94c192fe69a, 0x61ae2cea3a911513, 0x877bf6d3b9a4de27, 0x78da0fc61073f3eb}}},
+{{{0x5bf15d28e52bc66a, 0x2c47e31870f01a8e, 0x2419afbc06c28bdd, 0x2d25deeb256b173a}},
+ {{0xa29f80f1680c3a94, 0x71f77e151ae9e7e6, 0x1100f15848017973, 0x054aa4b316b38ddd}},
+ {{0xdfc8468d19267cb8, 0x0b28789c66e54daf, 0x2aeb1d2a666eec17, 0x134610a6ab7da760}}},
+{{{0x51138ec78df6b0fe, 0x5397da89e575f51b, 0x09207a1d717af1b9, 0x2102fdba2b20d650}},
+ {{0xcd2a65e777d1f515, 0x548991878faa60f1, 0xb1b73bbcdabc06e5, 0x654878cba97cc9fb}},
+ {{0x969ee405055ce6a1, 0x36bca7681251ad29, 0x3a1af517aa7da415, 0x0ad725db29ecb2ba}}},
+{{{0xdc4267b1834e2457, 0xb67544b570ce1bc5, 0x1af07a0bf7d15ed7, 0x4aefcffb71a03650}},
+ {{0xfec7bc0c9b056f85, 0x537d5268e7f5ffd7, 0x77afc6624312aefa, 0x4f675f5302399fd9}},
+ {{0xc32d36360415171e, 0xcd2bef118998483b, 0x870a6eadd0945110, 0x0bccbb72a2a86561}}},
+{{{0x185e962feab1a9c8, 0x86e7e63565147dcd, 0xb092e031bb5b6df2, 0x4024f0ab59d6b73e}},
+ {{0x186d5e4c50fe1296, 0xe0397b82fee89f7e, 0x3bc7f6c5507031b0, 0x6678fd69108f37c2}},
+ {{0x1586fa31636863c2, 0x07f68c48572d33f2, 0x4f73cc9f789eaefc, 0x2d42e2108ead4701}}},
+{{{0x97f5131594dfd29b, 0x6155985d313f4c6a, 0xeba13f0708455010, 0x676b2608b8d2d322}},
+ {{0x21717b0d0f537593, 0x914e690b131e064c, 0x1bb687ae752ae09f, 0x420bf3a79b423c6e}},
+ {{0x8138ba651c5b2b47, 0x8671b6ec311b1b80, 0x7bff0cb1bc3135b0, 0x745d2ffa9c0cf1e0}}},
+{{{0xbf525a1e2bc9c8bd, 0xea5b260826479d81, 0xd511c70edf0155db, 0x1ae23ceb960cf5d0}},
+ {{0x6036df5721d34e6a, 0xb1db8827997bb3d0, 0xd3c209c3c8756afa, 0x06e15be54c1dc839}},
+ {{0x5b725d871932994a, 0x32351cb5ceb1dab0, 0x7dc41549dab7ca05, 0x58ded861278ec1f7}}},
+{{{0xd8173793f266c55c, 0xc8c976c5cc454e49, 0x5ce382f8bc26c3a8, 0x2ff39de85485f6f9}},
+ {{0x2dfb5ba8b6c2c9a8, 0x48eeef8ef52c598c, 0x33809107f12d1573, 0x08ba696b531d5bd8}},
+ {{0x77ed3eeec3efc57a, 0x04e05517d4ff4811, 0xea3d7a3ff1a671cb, 0x120633b4947cfe54}}},
+{{{0x0b94987891610042, 0x4ee7b13cecebfae8, 0x70be739594f0a4c0, 0x35d30a99b4d59185}},
+ {{0x82bd31474912100a, 0xde237b6d7e6fbe06, 0xe11e761911ea79c6, 0x07433be3cb393bde}},
+ {{0xff7944c05ce997f4, 0x575d3de4b05c51a3, 0x583381fd5a76847c, 0x2d873ede7af6da9f}}},
+{{{0x157a316443373409, 0xfab8b7eef4aa81d9, 0xb093fee6f5a64806, 0x2e773654707fa7b6}},
+ {{0xaa6202e14e5df981, 0xa20d59175015e1f5, 0x18a275d3bae21d6c, 0x0543618a01600253}},
+ {{0x0deabdf4974c23c1, 0xaa6f0a259dce4693, 0x04202cb8a29aba2c, 0x4b1443362d07960d}}},
+{{{0xccc4b7c7b66e1f7a, 0x44157e25f50c2f7e, 0x3ef06dfc713eaf1c, 0x582f446752da63f7}},
+ {{0x967c54e91c529ccb, 0x30f6269264c635fb, 0x2747aff478121965, 0x17038418eaf66f5c}},
+ {{0xc6317bd320324ce4, 0xa81042e8a4488bc4, 0xb21ef18b4e5a1364, 0x0c2a1c4bcda28dc9}}},
+{{{0xd24dc7d06f1f0447, 0xb2269e3edb87c059, 0xd15b0272fbb2d28f, 0x7c558bd1c6f64877}},
+ {{0xedc4814869bd6945, 0x0d6d907dbe1c8d22, 0xc63bd212d55cc5ab, 0x5a6a9b30a314dc83}},
+ {{0xd0ec1524d396463d, 0x12bb628ac35a24f0, 0xa50c3a791cbc5fa4, 0x0404a5ca0afbafc3}}},
+{{{0x8c1f40070aa743d6, 0xccbad0cb5b265ee8, 0x574b046b668fd2de, 0x46395bfdcadd9633}},
+ {{0x62bc9e1b2a416fd1, 0xb5c6f728e350598b, 0x04343fd83d5d6967, 0x39527516e7f8ee98}},
+ {{0x117fdb2d1a5d9a9c, 0x9c7745bcd1005c2a, 0xefd4bef154d56fea, 0x76579a29e822d016}}},
+{{{0x45b68e7e49c02a17, 0x23cd51a2bca9a37f, 0x3ed65f11ec224c1b, 0x43a384dc9e05bdb1}},
+ {{0x333cb51352b434f2, 0xd832284993de80e1, 0xb5512887750d35ce, 0x02c514bb2a2777c1}},
+ {{0x684bd5da8bf1b645, 0xfb8bd37ef6b54b53, 0x313916d7a9b0d253, 0x1160920961548059}}},
+{{{0xb44d166929dacfaa, 0xda529f4c8413598f, 0xe9ef63ca453d5559, 0x351e125bc5698e0b}},
+ {{0x7a385616369b4dcd, 0x75c02ca7655c3563, 0x7dc21bf9d4f18021, 0x2f637d7491e6e042}},
+ {{0xd4b49b461af67bbe, 0xd603037ac8ab8961, 0x71dee19ff9a699fb, 0x7f182d06e7ce2a9a}}},
+{{{0x7a7c8e64ab0168ec, 0xcb5a4a5515edc543, 0x095519d347cd0eda, 0x67d4ac8c343e93b0}},
+ {{0x09454b728e217522, 0xaa58e8f4d484b8d8, 0xd358254d7f46903c, 0x44acc043241c5217}},
+ {{0x1c7d6bbb4f7a5777, 0x8b35fed4918313e1, 0x4adca1c6c96b4684, 0x556d1c8312ad71bd}}},
+{{{0x17ef40e30c8d3982, 0x31f7073e15a3fa34, 0x4f21f3cb0773646e, 0x746c6c6d1d824eff}},
+ {{0x81f06756b11be821, 0x0faff82310a3f3dd, 0xf8b2d0556a99465d, 0x097abe38cc8c7f05}},
+ {{0x0c49c9877ea52da4, 0x4c4369559bdc1d43, 0x022c3809f7ccebd2, 0x577e14a34bee84bd}}},
+{{{0xf0e268ac61a73b0a, 0xf2fafa103791a5f5, 0xc1e13e826b6d00e9, 0x60fa7ee96fd78f42}},
+ {{0x94fecebebd4dd72b, 0xf46a4fda060f2211, 0x124a5977c0c8d1ff, 0x705304b8fb009295}},
+ {{0xb63d1d354d296ec6, 0xf3c3053e5fad31d8, 0x670b958cb4bd42ec, 0x21398e0ca16353fd}}},
+{{{0x89f5058a382b33f3, 0x5ae2ba0bad48c0b4, 0x8f93b503a53db36e, 0x5aa3ed9d95a232e6}},
+ {{0x2798aaf9b4b75601, 0x5eac72135c8dad72, 0xd2ceaa6161b7a023, 0x1bbfb284e98f7d4e}},
+ {{0x656777e9c7d96561, 0xcb2b125472c78036, 0x65053299d9506eee, 0x4a07e14e5e8957cc}}},
+{{{0x4ee412cb980df999, 0xa315d76f3c6ec771, 0xbba5edde925c77fd, 0x3f0bac391d313402}},
+ {{0x240b58cdc477a49b, 0xfd38dade6447f017, 0x19928d32a7c86aad, 0x50af7aed84afa081}},
+ {{0x6e4fde0115f65be5, 0x29982621216109b2, 0x780205810badd6d9, 0x1921a316baebd006}}},
+{{{0x89422f7edfb870fc, 0x2c296beb4f76b3bd, 0x0738f1d436c24df7, 0x6458df41e273aeb0}},
+ {{0xd75aad9ad9f3c18b, 0x566a0eef60b1c19c, 0x3e9a0bac255c0ed9, 0x7b049deca062c7f5}},
+ {{0xdccbe37a35444483, 0x758879330fedbe93, 0x786004c312c5dd87, 0x6093dccbc2950e64}}},
+{{{0x1ff39a8585e0706d, 0x36d0a5d8b3e73933, 0x43b9f2e1718f453b, 0x57d1ea084827a97c}},
+ {{0x6bdeeebe6084034b, 0x3199c2b6780fb854, 0x973376abb62d0695, 0x6e3180c98b647d90}},
+ {{0xee7ab6e7a128b071, 0xa4c1596d93a88baa, 0xf7b4de82b2216130, 0x363e999ddd97bd18}}},
+{{{0x96a843c135ee1fc4, 0x976eb35508e4c8cf, 0xb42f6801b58cd330, 0x48ee9b78693a052b}},
+ {{0x2f1848dce24baec6, 0x769b7255babcaf60, 0x90cb3c6e3cefe931, 0x231f979bc6f9b355}},
+ {{0x5c31de4bcc2af3c6, 0xb04bb030fe208d1f, 0xb78d7009c14fb466, 0x079bfa9b08792413}}},
+{{{0xe3903a51da300df4, 0x843964233da95ab0, 0xed3cf12d0b356480, 0x038c77f684817194}},
+ {{0xf3c9ed80a2d54245, 0x0aa08b7877f63952, 0xd76dac63d1085475, 0x1ef4fb159470636b}},
+ {{0x854e5ee65b167bec, 0x59590a4296d0cdc2, 0x72b2df3498102199, 0x575ee92a4a0bff56}}},
+{{{0xd4c080908a182fcf, 0x30e170c299489dbd, 0x05babd5752f733de, 0x43d4e7112cd3fd00}},
+ {{0x5d46bc450aa4d801, 0xc3af1227a533b9d8, 0x389e3b262b8906c2, 0x200a1e7e382f581b}},
+ {{0x518db967eaf93ac5, 0x71bc989b056652c0, 0xfe2b85d9567197f5, 0x050eca52651e4e38}}},
+{{{0xc3431ade453f0c9c, 0xe9f5045eff703b9b, 0xfcd97ac9ed847b3d, 0x4b0ee6c21c58f4c6}},
+ {{0x97ac397660e668ea, 0x9b19bbfe153ab497, 0x4cb179b534eca79f, 0x6151c09fa131ae57}},
+ {{0x3af55c0dfdf05d96, 0xdd262ee02ab4ee7a, 0x11b2bb8712171709, 0x1fef24fa800f030b}}},
+{{{0x37d653fb1aa73196, 0x0f9495303fd76418, 0xad200b09fb3a17b2, 0x544d49292fc8613e}},
+ {{0x22d2aff530976b86, 0x8d90b806c2d24604, 0xdca1896c4de5bae5, 0x28005fe6c8340c17}},
+ {{0x6aefba9f34528688, 0x5c1bff9425107da1, 0xf75bbbcd66d94b36, 0x72e472930f316dfa}}},
+{{{0x2695208c9781084f, 0xb1502a0b23450ee1, 0xfd9daea603efde02, 0x5a9d2e8c2733a34c}},
+ {{0x07f3f635d32a7627, 0x7aaa4d865f6566f0, 0x3c85e79728d04450, 0x1fee7f000fe06438}},
+ {{0x765305da03dbf7e5, 0xa4daf2491434cdbd, 0x7b4ad5cdd24a88ec, 0x00f94051ee040543}}},
+{{{0x8d356b23c3d330b2, 0xf21c8b9bb0471b06, 0xb36c316c6e42b83c, 0x07d79c7e8beab10d}},
+ {{0xd7ef93bb07af9753, 0x583ed0cf3db766a7, 0xce6998bf6e0b1ec5, 0x47b7ffd25dd40452}},
+ {{0x87fbfb9cbc08dd12, 0x8a066b3ae1eec29b, 0x0d57242bdb1fc1bf, 0x1c3520a35ea64bb6}}},
+{{{0x80d253a6bccba34a, 0x3e61c3a13838219b, 0x90c3b6019882e396, 0x1c3d05775d0ee66f}},
+ {{0xcda86f40216bc059, 0x1fbb231d12bcd87e, 0xb4956a9e17c70990, 0x38750c3b66d12e55}},
+ {{0x692ef1409422e51a, 0xcbc0c73c2b5df671, 0x21014fe7744ce029, 0x0621e2c7d330487c}}},
+{{{0xaf9860cc8259838d, 0x90ea48c1c69f9adc, 0x6526483765581e30, 0x0007d6097bd3a5bc}},
+ {{0xb7ae1796b0dbf0f3, 0x54dfafb9e17ce196, 0x25923071e9aaa3b4, 0x5d8e589ca1002e9d}},
+ {{0xc0bf1d950842a94b, 0xb2d3c363588f2e3e, 0x0a961438bb51e2ef, 0x1583d7783c1cbf86}}},
+{{{0xeceea2ef5da27ae1, 0x597c3a1455670174, 0xc9a62a126609167a, 0x252a5f2e81ed8f70}},
+ {{0x90034704cc9d28c7, 0x1d1b679ef72cc58f, 0x16e12b5fbe5b8726, 0x4958064e83c5580a}},
+ {{0x0d2894265066e80d, 0xfcc3f785307c8c6b, 0x1b53da780c1112fd, 0x079c170bd843b388}}},
+{{{0x0506ece464fa6fff, 0xbee3431e6205e523, 0x3579422451b8ea42, 0x6dec05e34ac9fb00}},
+ {{0xcdd6cd50c0d5d056, 0x9af7686dbb03573b, 0x3ca6723ff3c3ef48, 0x6768c0d7317b8acc}},
+ {{0x94b625e5f155c1b3, 0x417bf3a7997b7b91, 0xc22cbddc6d6b2600, 0x51445e14ddcd52f4}}},
+{{{0x57502b4b3b144951, 0x8e67ff6b444bbcb3, 0xb8bd6927166385db, 0x13186f31e39295c8}},
+ {{0x893147ab2bbea455, 0x8c53a24f92079129, 0x4b49f948be30f7a7, 0x12e990086e4fd43d}},
+ {{0xf10c96b37fdfbb2e, 0x9f9a935e121ceaf9, 0xdf1136c43a5b983f, 0x77b2e3f05d3e99af}}},
+{{{0x296fa9c59c2ec4de, 0xbc8b61bf4f84f3cb, 0x1c7706d917a8f908, 0x63b795fc7ad3255d}},
+ {{0xd598639c12ddb0a4, 0xa5d19f30c024866b, 0xd17c2f0358fce460, 0x07a195152e095e8a}},
+ {{0xa8368f02389e5fc8, 0x90433b02cf8de43b, 0xafa1fd5dc5412643, 0x3e8fe83d032f0137}}},
+{{{0x2f8b15b90570a294, 0x94f2427067084549, 0xde1c5ae161bbfd84, 0x75ba3b797fac4007}},
+ {{0x08704c8de8efd13c, 0xdfc51a8e33e03731, 0xa59d5da51260cde3, 0x22d60899a6258c86}},
+ {{0x6239dbc070cdd196, 0x60fe8a8b6c7d8a9a, 0xb38847bceb401260, 0x0904d07b87779e5e}}},
+{{{0xb4ce1fd4ddba919c, 0xcf31db3ec74c8daa, 0x2c63cc63ad86cc51, 0x43e2143fbc1dde07}},
+ {{0xf4322d6648f940b9, 0x06952f0cbd2d0c39, 0x167697ada081f931, 0x6240aacebaf72a6c}},
+ {{0xf834749c5ba295a0, 0xd6947c5bca37d25a, 0x66f13ba7e7c9316a, 0x56bdaf238db40cac}}},
+{{{0x362ab9e3f53533eb, 0x338568d56eb93d40, 0x9e0e14521d5a5572, 0x1d24a86d83741318}},
+ {{0x1310d36cc19d3bb2, 0x062a6bb7622386b9, 0x7c9b8591d7a14f5c, 0x03aa31507e1e5754}},
+ {{0xf4ec7648ffd4ce1f, 0xe045eaf054ac8c1c, 0x88d225821d09357c, 0x43b261dc9aeb4859}}},
+{{{0xe55b1e1988bb79bb, 0xa09ed07dc17a359d, 0xb02c2ee2603dea33, 0x326055cf5b276bc2}},
+ {{0x19513d8b6c951364, 0x94fe7126000bf47b, 0x028d10ddd54f9567, 0x02b4d5e242940964}},
+ {{0xb4a155cb28d18df2, 0xeacc4646186ce508, 0xc49cf4936c824389, 0x27a6c809ae5d3410}}},
+{{{0x8ba6ebcd1f0db188, 0x37d3d73a675a5be8, 0xf22edfa315f5585a, 0x2cb67174ff60a17e}},
+ {{0xcd2c270ac43d6954, 0xdd4a3e576a66cab2, 0x79fa592469d7036c, 0x221503603d8c2599}},
+ {{0x59eecdf9390be1d0, 0xa9422044728ce3f1, 0x82891c667a94f0f4, 0x7b1df4b73890f436}}},
+{{{0xe492f2e0b3b2a224, 0x7c6c9e062b551160, 0x15eb8fe20d7f7b0e, 0x61fcef2658fc5992}},
+ {{0x5f2e221807f8f58c, 0xe3555c9fd49409d4, 0xb2aaa88d1fb6a630, 0x68698245d352e03d}},
+ {{0xdbb15d852a18187a, 0xf3e4aad386ddacd7, 0x44bae2810ff6c482, 0x46cf4c473daf01cf}}},
+{{{0x426525ed9ec4e5f9, 0x0e5eda0116903303, 0x72b1a7f2cbe5cadc, 0x29387bcd14eb5f40}},
+ {{0x213c6ea7f1498140, 0x7c1e7ef8392b4854, 0x2488c38c5629ceba, 0x1065aae50d8cc5bb}},
+ {{0x1c2c4525df200d57, 0x5c3b2dd6bfca674a, 0x0a07e7b1e1834030, 0x69a198e64f1ce716}}},
+{{{0x9062b2e0d91a78bc, 0x47c9889cc8509667, 0x9df54a66405070b8, 0x7369e6a92493a1bf}},
+ {{0xe1014434dcc5caed, 0x47ed5d963c84fb33, 0x70019576ed86a0e7, 0x25b2697bd267f9e4}},
+ {{0x9d673ffb13986864, 0x3ca5fbd9415dc7b8, 0xe04ecc3bdf273b5e, 0x1420683db54e4cd2}}},
+{{{0xb478bd1e249dd197, 0x620c35005e58c102, 0xfb02d32fccbaac5c, 0x60b63bebf508a72d}},
+ {{0x34eebb6fc1cc5ad0, 0x6a1b0ce99646ac8b, 0xd3b0da49a66bde53, 0x31e83b4161d081c1}},
+ {{0x97e8c7129e062b4f, 0x49e48f4f29320ad8, 0x5bece14b6f18683f, 0x55cf1eb62d550317}}},
+{{{0x5879101065c23d58, 0x8b9d086d5094819c, 0xe2402fa912c55fa7, 0x669a6564570891d4}},
+ {{0x3076b5e37df58c52, 0xd73ab9dde799cc36, 0xbd831ce34913ee20, 0x1a56fbaa62ba0133}},
+ {{0x943e6b505c9dc9ec, 0x302557bba77c371a, 0x9873ae5641347651, 0x13c4836799c58a5c}}},
+{{{0x423a5d465ab3e1b9, 0xfc13c187c7f13f61, 0x19f83664ecb5b9b6, 0x66f80c93a637b607}},
+ {{0xc4dcfb6a5d8bd080, 0xdeebc4ec571a4842, 0xd4b2e883b8e55365, 0x50bdc87dc8e5b827}},
+ {{0x606d37836edfe111, 0x32353e15f011abd9, 0x64b03ac325b73b96, 0x1dd56444725fd5ae}}},
+{{{0x8fa47ff83362127d, 0xbc9f6ac471cd7c15, 0x6e71454349220c8b, 0x0e645912219f732e}},
+ {{0xc297e60008bac89a, 0x7d4cea11eae1c3e0, 0xf3e38be19fe7977c, 0x3a3a450f63a305cd}},
+ {{0x078f2f31d8394627, 0x389d3183de94a510, 0xd1e36c6d17996f80, 0x318c8d9393a9a87b}}},
+{{{0xf2745d032afffe19, 0x0c9f3c497f24db66, 0xbc98d3e3ba8598ef, 0x224c7c679a1d5314}},
+ {{0x5d669e29ab1dd398, 0xfc921658342d9e3b, 0x55851dfdf35973cd, 0x509a41c325950af6}},
+ {{0xbdc06edca6f925e9, 0x793ef3f4641b1f33, 0x82ec12809d833e89, 0x05bff02328a11389}}},
+{{{0x3632137023cae00b, 0x544acf0ad1accf59, 0x96741049d21a1c88, 0x780b8cc3fa2a44a7}},
+ {{0x6881a0dd0dc512e4, 0x4fe70dc844a5fafe, 0x1f748e6b8f4a5240, 0x576277cdee01a3ea}},
+ {{0x1ef38abc234f305f, 0x9a577fbd1405de08, 0x5e82a51434e62a0d, 0x5ff418726271b7a1}}},
+{{{0x398e080c1789db9d, 0xa7602025f3e778f5, 0xfa98894c06bd035d, 0x106a03dc25a966be}},
+ {{0xe5db47e813b69540, 0xf35d2a3b432610e1, 0xac1f26e938781276, 0x29d4db8ca0a0cb69}},
+ {{0xd9ad0aaf333353d0, 0x38669da5acd309e5, 0x3c57658ac888f7f0, 0x4ab38a51052cbefa}}},
+{{{0xda7c2b256768d593, 0x98c1c0574422ca13, 0xf1a80bd5ca0ace1d, 0x29cdd1adc088a690}},
+ {{0xd6cfd1ef5fddc09c, 0xe82b3efdf7575dce, 0x25d56b5d201634c2, 0x3041c6bb04ed2b9b}},
+ {{0x0ff2f2f9d956e148, 0xade797759f356b2e, 0x1a4698bb5f6c025c, 0x104bbd6814049a7b}}},
+{{{0x51f0fd3168f1ed67, 0x2c811dcdd86f3bc2, 0x44dc5c4304d2f2de, 0x5be8cc57092a7149}},
+ {{0xa95d9a5fd67ff163, 0xe92be69d4cc75681, 0xb7f8024cde20f257, 0x204f2a20fb072df5}},
+ {{0xc8143b3d30ebb079, 0x7589155abd652e30, 0x653c3c318f6d5c31, 0x2570fb17c279161f}}},
+{{{0x3efa367f2cb61575, 0xf5f96f761cd6026c, 0xe8c7142a65b52562, 0x3dcb65ea53030acd}},
+ {{0x192ea9550bb8245a, 0xc8e6fba88f9050d1, 0x7986ea2d88a4c935, 0x241c5f91de018668}},
+ {{0x28d8172940de6caa, 0x8fbf2cf022d9733a, 0x16d7fcdd235b01d1, 0x08420edd5fcdf0e5}}},
+{{{0xcdff20ab8362fa4a, 0x57e118d4e21a3e6e, 0xe3179617fc39e62b, 0x0d9a53efbc1769fd}},
+ {{0x0358c34e04f410ce, 0xb6135b5a276e0685, 0x5d9670c7ebb91521, 0x04d654f321db889c}},
+ {{0x5e7dc116ddbdb5d5, 0x2954deb68da5dd2d, 0x1cb608173334a292, 0x4a7a4f2618991ad7}}},
+{{{0xf4a718025fb15f95, 0x3df65f346b5c1b8f, 0xcdfcf08500e01112, 0x11b50c4cddd31848}},
+ {{0x24c3b291af372a4b, 0x93da8270718147f2, 0xdd84856486899ef2, 0x4a96314223e0ee33}},
+ {{0xa6e8274408a4ffd6, 0x738e177e9c1576d9, 0x773348b63d02b3f2, 0x4f4bce4dce6bcc51}}},
+{{{0xa71fce5ae2242584, 0x26ea725692f58a9e, 0xd21a09d71cea3cf4, 0x73fcdd14b71c01e6}},
+ {{0x30e2616ec49d0b6f, 0xe456718fcaec2317, 0x48eb409bf26b4fa6, 0x3042cee561595f37}},
+ {{0x427e7079449bac41, 0x855ae36dbce2310a, 0x4cae76215f841a7c, 0x389e740c9a9ce1d6}}},
+{{{0x64fcb3ae34dcb9ce, 0x97500323e348d0ad, 0x45b3f07d62c6381b, 0x61545379465a6788}},
+ {{0xc9bd78f6570eac28, 0xe55b0b3227919ce1, 0x65fc3eaba19b91ed, 0x25c425e5d6263690}},
+ {{0x3f3e06a6f1d7de6e, 0x3ef976278e062308, 0x8c14f6264e8a6c77, 0x6539a08915484759}}},
+{{{0xe9d21f74c3d2f773, 0xc150544125c46845, 0x624e5ce8f9b99e33, 0x11c5e4aac5cd186c}},
+ {{0xddc4dbd414bb4a19, 0x19b2bc3c98424f8e, 0x48a89fd736ca7169, 0x0f65320ef019bd90}},
+ {{0xd486d1b1cafde0c6, 0x4f3fe6e3163b5181, 0x59a8af0dfaf2939a, 0x4cabc7bdec33072a}}},
+{{{0x239e9624089c0a2e, 0xc748c4c03afe4738, 0x17dbed2a764fa12a, 0x639b93f0321c8582}},
+ {{0xc08f788f3f78d289, 0xfe30a72ca1404d9f, 0xf2778bfccf65cc9d, 0x7ee498165acb2021}},
+ {{0x7bd508e39111a1c3, 0x2b2b90d480907489, 0xe7d2aec2ae72fd19, 0x0edf493c85b602a6}}},
+{{{0xaecc8158599b5a68, 0xea574f0febade20e, 0x4fe41d7422b67f07, 0x403b92e3019d4fb4}},
+ {{0x6767c4d284764113, 0xa090403ff7f5f835, 0x1c8fcffacae6bede, 0x04c00c54d1dfa369}},
+ {{0x4dc22f818b465cf8, 0x71a0f35a1480eff8, 0xaee8bfad04c7d657, 0x355bb12ab26176f4}}},
+{{{0xa71e64cc7493bbf4, 0xe5bd84d9eca3b0c3, 0x0a6bc50cfa05e785, 0x0f9b8132182ec312}},
+ {{0xa301dac75a8c7318, 0xed90039db3ceaa11, 0x6f077cbf3bae3f2d, 0x7518eaf8e052ad8e}},
+ {{0xa48859c41b7f6c32, 0x0f2d60bcf4383298, 0x1815a929c9b1d1d9, 0x47c3871bbb1755c4}}},
+{{{0x5144539771ec4f48, 0xf805b17dc98c5d6e, 0xf762c11a47c3c66b, 0x00b89b85764699dc}},
+ {{0xfbe65d50c85066b0, 0x62ecc4b0b3a299b0, 0xe53754ea441ae8e0, 0x08fea02ce8d48d5f}},
+ {{0x824ddd7668deead0, 0xc86445204b685d23, 0xb514cfcd5d89d665, 0x473829a74f75d537}}},
+{{{0x82d2da754679c418, 0xe63bd7d8b2618df0, 0x355eef24ac47eb0a, 0x2078684c4833c6b4}},
+ {{0x23d9533aad3902c9, 0x64c2ddceef03588f, 0x15257390cfe12fb4, 0x6c668b4d44e4d390}},
+ {{0x3b48cf217a78820c, 0xf76a0ab281273e97, 0xa96c65a78c8eed7b, 0x7411a6054f8a433f}}},
+{{{0x4d659d32b99dc86d, 0x044cdc75603af115, 0xb34c712cdcc2e488, 0x7c136574fb8134ff}},
+ {{0x579ae53d18b175b4, 0x68713159f392a102, 0x8455ecba1eef35f5, 0x1ec9a872458c398f}},
+ {{0xb8e6a4d400a2509b, 0x9b81d7020bc882b4, 0x57e7cc9bf1957561, 0x3add88a5c7cd6460}}},
+{{{0xab895770b635dcf2, 0x02dfef6cf66c1fbc, 0x85530268beb6d187, 0x249929fccc879e74}},
+ {{0x85c298d459393046, 0x8f7e35985ff659ec, 0x1d2ca22af2f66e3a, 0x61ba1131a406a720}},
+ {{0xa3d0a0f116959029, 0x023b6b6cba7ebd89, 0x7bf15a3e26783307, 0x5620310cbbd8ece7}}},
+{{{0x528993434934d643, 0xb9dbf806a51222f5, 0x8f6d878fc3f41c22, 0x37676a2a4d9d9730}},
+ {{0x6646b5f477e285d6, 0x40e8ff676c8f6193, 0xa6ec7311abb594dd, 0x7ec846f3658cec4d}},
+ {{0x9b5e8f3f1da22ec7, 0x130f1d776c01cd13, 0x214c8fcfa2989fb8, 0x6daaf723399b9dd5}}},
+{{{0x5f3a7562eb3dbe47, 0xf7ea38548ebda0b8, 0x00c3e53145747299, 0x1304e9e71627d551}},
+ {{0x583b04bfacad8ea2, 0x29b743e8148be884, 0x2b1e583b0810c5db, 0x2b5449e58eb3bbaa}},
+ {{0x789814d26adc9cfe, 0x3c1bab3f8b48dd0b, 0xda0fe1fff979c60a, 0x4468de2d7c2dd693}}},
+{{{0x51bb355e9419469e, 0x33e6dc4c23ddc754, 0x93a5b6d6447f9962, 0x6cce7c6ffb44bd63}},
+ {{0x4b9ad8c6f86307ce, 0x21113531435d0c28, 0xd4a866c5657a772c, 0x5da6427e63247352}},
+ {{0x1a94c688deac22ca, 0xb9066ef7bbae1ff8, 0x88ad8c388d59580f, 0x58f29abfe79f2ca8}}},
+{{{0xe90ecfab8de73e68, 0x54036f9f377e76a5, 0xf0495b0bbe015982, 0x577629c4a7f41e36}},
+ {{0x4b5a64bf710ecdf6, 0xb14ce538462c293c, 0x3643d056d50b3ab9, 0x6af93724185b4870}},
+ {{0x3220024509c6a888, 0xd2e036134b558973, 0x83e236233c33289f, 0x701f25bb0caec18f}}},
+{{{0xc3a8b0f8e4616ced, 0xf700660e9e25a87d, 0x61e3061ff4bca59c, 0x2e0c92bfbdc40be9}},
+ {{0x9d18f6d97cbec113, 0x844a06e674bfdbe4, 0x20f5b522ac4e60d6, 0x720a5bc050955e51}},
+ {{0x0c3f09439b805a35, 0xe84e8b376242abfc, 0x691417f35c229346, 0x0e9b9cbb144ef0ec}}},
+{{{0xfbbad48ffb5720ad, 0xee81916bdbf90d0e, 0xd4813152635543bf, 0x221104eb3f337bd8}},
+ {{0x8dee9bd55db1beee, 0xc9c3ab370a723fb9, 0x44a8f1bf1c68d791, 0x366d44191cfd3cde}},
+ {{0x9e3c1743f2bc8c14, 0x2eda26fcb5856c3b, 0xccb82f0e68a7fb97, 0x4167a4e6bc593244}}},
+{{{0x643b9d2876f62700, 0x5d1d9d400e7668eb, 0x1b4b430321fc0684, 0x7938bb7e2255246a}},
+ {{0xc2be2665f8ce8fee, 0xe967ff14e880d62c, 0xf12e6e7e2f364eee, 0x34b33370cb7ed2f6}},
+ {{0xcdc591ee8681d6cc, 0xce02109ced85a753, 0xed7485c158808883, 0x1176fc6e2dfe65e4}}},
+{{{0xb4af6cd05b9c619b, 0x2ddfc9f4b2a58480, 0x3d4fa502ebe94dc4, 0x08fc3a4c677d5f34}},
+ {{0xdb90e28949770eb8, 0x98fbcc2aacf440a3, 0x21354ffeded7879b, 0x1f6a3e54f26906b6}},
+ {{0x60a4c199d30734ea, 0x40c085b631165cd6, 0xe2333e23f7598295, 0x4f2fad0116b900d1}}},
+{{{0x44beb24194ae4e54, 0x5f541c511857ef6c, 0xa61e6b2d368d0498, 0x445484a4972ef7ab}},
+ {{0x962cd91db73bb638, 0xe60577aafc129c08, 0x6f619b39f3b61689, 0x3451995f2944ee81}},
+ {{0x9152fcd09fea7d7c, 0x4a816c94b0935cf6, 0x258e9aaa47285c40, 0x10b89ca6042893b7}}},
+{{{0x3d5947499718289c, 0x12ebf8c524533f26, 0x0262bfcb14c3ef15, 0x20b878d577b7518e}},
+ {{0x753941be5a45f06e, 0xd07caeed6d9c5f65, 0x11776b9c72ff51b6, 0x17d2d1d9ef0d4da9}},
+ {{0x27f2af18073f3e6a, 0xfd3fe519d7521069, 0x22e3b72c3ca60022, 0x72214f63cc65c6a7}}},
+{{{0xb4e37f405307a693, 0xaba714d72f336795, 0xd6fbd0a773761099, 0x5fdf48c58171cbc9}},
+ {{0x1d9db7b9f43b29c9, 0xd605824a4f518f75, 0xf2c072bd312f9dc4, 0x1f24ac855a1545b0}},
+ {{0x24d608328e9505aa, 0x4748c1d10c1420ee, 0xc7ffe45c06fb25a2, 0x00ba739e2ae395e6}}},
+{{{0x592e98de5c8790d6, 0xe5bfb7d345c2a2df, 0x115a3b60f9b49922, 0x03283a3e67ad78f3}},
+ {{0xae4426f5ea88bb26, 0x360679d984973bfb, 0x5c9f030c26694e50, 0x72297de7d518d226}},
+ {{0x48241dc7be0cb939, 0x32f19b4d8b633080, 0xd3dfc90d02289308, 0x05e1296846271945}}},
+{{{0xba82eeb32d9c495a, 0xceefc8fcf12bb97c, 0xb02dabae93b5d1e0, 0x39c00c9c13698d9b}},
+ {{0xadbfbbc8242c4550, 0xbcc80cecd03081d9, 0x843566a6f5c8df92, 0x78cf25d38258ce4c}},
+ {{0x15ae6b8e31489d68, 0xaa851cab9c2bf087, 0xc9a75a97f04efa05, 0x006b52076b3ff832}}},
+{{{0x29e0cfe19d95781c, 0xb681df18966310e2, 0x57df39d370516b39, 0x4d57e3443bc76122}},
+ {{0xf5cb7e16b9ce082d, 0x3407f14c417abc29, 0xd4b36bce2bf4a7ab, 0x7de2e9561a9f75ce}},
+ {{0xde70d4f4b6a55ecb, 0x4801527f5d85db99, 0xdbc9c440d3ee9a81, 0x6b2a90af1a6029ed}}},
+{{{0x6923f4fc9ae61e97, 0x5735281de03f5fd1, 0xa764ae43e6edd12d, 0x5fd8f4e9d12d3e4a}},
+ {{0x77ebf3245bb2d80a, 0xd8301b472fb9079b, 0xc647e6f24cee7333, 0x465812c8276c2109}},
+ {{0x4d43beb22a1062d9, 0x7065fb753831dc16, 0x180d4a7bde2968d7, 0x05b32c2b1cb16790}}},
+{{{0xc8c05eccd24da8fd, 0xa1cf1aac05dfef83, 0xdbbeeff27df9cd61, 0x3b5556a37b471e99}},
+ {{0xf7fca42c7ad58195, 0x3214286e4333f3cc, 0xb6c29d0d340b979d, 0x31771a48567307e1}},
+ {{0x32b0c524e14dd482, 0xedb351541a2ba4b6, 0xa3d16048282b5af3, 0x4fc079d27a7336eb}}},
+{{{0x51c938b089bf2f7f, 0x2497bd6502dfe9a7, 0xffffc09c7880e453, 0x124567cecaf98e92}},
+ {{0xdc348b440c86c50d, 0x1337cbc9cc94e651, 0x6422f74d643e3cb9, 0x241170c2bae3cd08}},
+ {{0x3ff9ab860ac473b4, 0xf0911dee0113e435, 0x4ae75060ebc6c4af, 0x3f8612966c87000d}}},
+{{{0x559a0cc9782a0dde, 0x551dcdb2ea718385, 0x7f62865b31ef238c, 0x504aa7767973613d}},
+ {{0x9c18fcfa36048d13, 0x29159db373899ddd, 0xdc9f350b9f92d0aa, 0x26f57eee878a19d4}},
+ {{0x0cab2cd55687efb1, 0x5180d162247af17b, 0x85c15a344f5a2467, 0x4041943d9dba3069}}},
+{{{0xc3c0eeba43ebcc96, 0x8d749c9c26ea9caf, 0xd9fa95ee1c77ccc6, 0x1420a1d97684340f}},
+ {{0x4b217743a26caadd, 0x47a6b424648ab7ce, 0xcb1d4f7a03fbc9e3, 0x12d931429800d019}},
+ {{0x00c67799d337594f, 0x5e3c5140b23aa47b, 0x44182854e35ff395, 0x1b4f92314359a012}}},
+{{{0x3e5c109d89150951, 0x39cefa912de9696a, 0x20eae43f975f3020, 0x239b572a7f132dae}},
+ {{0x33cf3030a49866b1, 0x251f73d2215f4859, 0xab82aa4051def4f6, 0x5ff191d56f9a23f6}},
+ {{0x819ed433ac2d9068, 0x2883ab795fc98523, 0xef4572805593eb3d, 0x020c526a758f36cb}}},
+{{{0x779834f89ed8dbbc, 0xc8f2aaf9dc7ca46c, 0xa9524cdca3e1b074, 0x02aacc4615313877}},
+ {{0xe931ef59f042cc89, 0x2c589c9d8e124bb6, 0xadc8e18aaec75997, 0x452cfe0a5602c50c}},
+ {{0x86a0f7a0647877df, 0xbbc464270e607c9f, 0xab17ea25f1fb11c9, 0x4cfb7d7b304b877b}}},
+{{{0x72b43d6cb89b75fe, 0x54c694d99c6adc80, 0xb8c3aa373ee34c9f, 0x14b4622b39075364}},
+ {{0xe28699c29789ef12, 0x2b6ecd71df57190d, 0xc343c857ecc970d0, 0x5b1d4cbc434d3ac5}},
+ {{0xb6fb2615cc0a9f26, 0x3a4f0e2bb88dcce5, 0x1301498b3369a705, 0x2f98f71258592dd1}}},
+{{{0x0c94a74cb50f9e56, 0x5b1ff4a98e8e1320, 0x9a2acc2182300f67, 0x3a6ae249d806aaf9}},
+ {{0x2e12ae444f54a701, 0xfcfe3ef0a9cbd7de, 0xcebf890d75835de0, 0x1d8062e9e7614554}},
+ {{0x657ada85a9907c5a, 0x1a0ea8b591b90f62, 0x8d0e1dfbdf34b4e9, 0x298b8ce8aef25ff3}}},
+{{{0x2a927953eff70cb2, 0x4b89c92a79157076, 0x9418457a30a7cf6a, 0x34b8a8404d5ce485}},
+ {{0x837a72ea0a2165de, 0x3fab07b40bcf79f6, 0x521636c77738ae70, 0x6ba6271803a7d7dc}},
+ {{0xc26eecb583693335, 0xd5a813df63b5fefd, 0xa293aa9aa4b22573, 0x71d62bdd465e1c6a}}},
+{{{0x6533cc28d378df80, 0xf6db43790a0fa4b4, 0xe3645ff9f701da5a, 0x74d5f317f3172ba4}},
+ {{0xcd2db5dab1f75ef5, 0xd77f95cf16b065f5, 0x14571fea3f49f085, 0x1c333621262b2b3d}},
+ {{0xa86fe55467d9ca81, 0x398b7c752b298c37, 0xda6d0892e3ac623b, 0x4aebcc4547e9d98c}}},
+{{{0x12f0071b276d01c9, 0xe7b8bac586c48c70, 0x5308129b71d6fba9, 0x5d88fbf95a3db792}},
+ {{0x0b408d9e7354b610, 0x806b32535ba85b6e, 0xdbe63a034a58a207, 0x173bd9ddc9a1df2c}},
+ {{0x2b500f1efe5872df, 0x58d6582ed43918c1, 0xe6ed278ec9673ae0, 0x06e1cd13b19ea319}}},
+{{{0x40d0ad516f166f23, 0x118e32931fab6abe, 0x3fe35e14a04d088e, 0x3080603526e16266}},
+ {{0x472baf629e5b0353, 0x3baa0b90278d0447, 0x0c785f469643bf27, 0x7f3a6a1a8d837b13}},
+ {{0xf7e644395d3d800b, 0x95a8d555c901edf6, 0x68cd7830592c6339, 0x30d0fded2e51307e}}},
+{{{0xe0594d1af21233b3, 0x1bdbe78ef0cc4d9c, 0x6965187f8f499a77, 0x0a9214202c099868}},
+ {{0x9cb4971e68b84750, 0xa09572296664bbcf, 0x5c8de72672fa412b, 0x4615084351c589d9}},
+ {{0xbc9019c0aeb9a02e, 0x55c7110d16034cae, 0x0e6df501659932ec, 0x3bca0d2895ca5dfe}}},
+{{{0x40f031bc3c5d62a4, 0x19fc8b3ecff07a60, 0x98183da2130fb545, 0x5631deddae8f13cd}},
+ {{0x9c688eb69ecc01bf, 0xf0bc83ada644896f, 0xca2d955f5f7a9fe2, 0x4ea8b4038df28241}},
+ {{0x2aed460af1cad202, 0x46305305a48cee83, 0x9121774549f11a5f, 0x24ce0930542ca463}}},
+{{{0x1fe890f5fd06c106, 0xb5c468355d8810f2, 0x827808fe6e8caf3e, 0x41d4e3c28a06d74b}},
+ {{0x3fcfa155fdf30b85, 0xd2f7168e36372ea4, 0xb2e064de6492f844, 0x549928a7324f4280}},
+ {{0xf26e32a763ee1a2e, 0xae91e4b7d25ffdea, 0xbc3bd33bd17f4d69, 0x491b66dec0dcff6a}}},
+{{{0x98f5b13dc7ea32a7, 0xe3d5f8cc7e16db98, 0xac0abf52cbf8d947, 0x08f338d0c85ee4ac}},
+ {{0x75f04a8ed0da64a1, 0xed222caf67e2284b, 0x8234a3791f7b7ba4, 0x4cf6b8b0b7018b67}},
+ {{0xc383a821991a73bd, 0xab27bc01df320c7a, 0xc13d331b84777063, 0x530d4a82eb078a99}}},
+{{{0x004c3630e1f94825, 0x7e2d78268cab535a, 0xc7482323cc84ff8b, 0x65ea753f101770b9}},
+ {{0x6d6973456c9abf9e, 0x257fb2fc4900a880, 0x2bacf412c8cfb850, 0x0db3e7e00cbfbd5b}},
+ {{0x3d66fc3ee2096363, 0x81d62c7f61b5cb6b, 0x0fbe044213443b1a, 0x02a4ec1921e1a1db}}},
+{{{0x5ce6259a3b24b8a2, 0xb8577acc45afa0b8, 0xcccbe6e88ba07037, 0x3d143c51127809bf}},
+ {{0xf5c86162f1cf795f, 0x118c861926ee57f2, 0x172124851c063578, 0x36d12b5dec067fcf}},
+ {{0x126d279179154557, 0xd5e48f5cfc783a0a, 0x36bdb6e8df179bac, 0x2ef517885ba82859}}},
+{{{0x4637974e8c58aedc, 0xb9ef22fbabf041a4, 0xe185d956e980718a, 0x2f1b78fab143a8a6}},
+ {{0x96eebffb305b2f51, 0xd3f938ad889596b8, 0xf0f52dc746d5dd25, 0x57968290bb3a0095}},
+ {{0xf71ab8430a20e101, 0xf393658d24f0ec47, 0xcf7509a86ee2eed1, 0x7dc43e35dc2aa3e1}}},
+{{{0x85966665887dd9c3, 0xc90f9b314bb05355, 0xc6e08df8ef2079b1, 0x7ef72016758cc12f}},
+ {{0x5a782a5c273e9718, 0x3576c6995e4efd94, 0x0f2ed8051f237d3e, 0x044fb81d82d50a99}},
+ {{0xc1df18c5a907e3d9, 0x57b3371dce4c6359, 0xca704534b201bb49, 0x7f79823f9c30dd2e}}},
+{{{0x8334d239a3b513e8, 0xc13670d4b91fa8d8, 0x12b54136f590bd33, 0x0a4e0373d784d9b4}},
+ {{0x6a9c1ff068f587ba, 0x0827894e0050c8de, 0x3cbf99557ded5be7, 0x64a9b0431c06d6f0}},
+ {{0x2eb3d6a15b7d2919, 0xb0b4f6a0d53a8235, 0x7156ce4389a45d47, 0x071a7d0ace18346c}}},
+{{{0xd3072daac887ba0b, 0x01262905bfa562ee, 0xcf543002c0ef768b, 0x2c3bcc7146ea7e9c}},
+ {{0xcc0c355220e14431, 0x0d65950709b15141, 0x9af5621b209d5f36, 0x7c69bcf7617755d3}},
+ {{0x07f0d7eb04e8295f, 0x10db18252f50f37d, 0xe951a9a3171798d7, 0x6f5a9a7322aca51d}}},
+{{{0x8ba1000c2f41c6c5, 0xc49f79c10cfefb9b, 0x4efa47703cc51c9f, 0x494e21a2e147afca}},
+ {{0xe729d4eba3d944be, 0x8d9e09408078af9e, 0x4525567a47869c03, 0x02ab9680ee8d3b24}},
+ {{0xefa48a85dde50d9a, 0x219a224e0fb9a249, 0xfa091f1dd91ef6d9, 0x6b5d76cbea46bb34}}},
+{{{0x8857556cec0cd994, 0x6472dc6f5cd01dba, 0xaf0169148f42b477, 0x0ae333f685277354}},
+ {{0xe0f941171e782522, 0xf1e6ae74036936d3, 0x408b3ea2d0fcc746, 0x16fb869c03dd313e}},
+ {{0x288e199733b60962, 0x24fc72b4d8abe133, 0x4811f7ed0991d03e, 0x3f81e38b8f70d075}}},
+{{{0x7f910fcc7ed9affe, 0x545cb8a12465874b, 0xa8397ed24b0c4704, 0x50510fc104f50993}},
+ {{0x0adb7f355f17c824, 0x74b923c3d74299a4, 0xd57c3e8bcbf8eaf7, 0x0ad3e2d34cdedc3d}},
+ {{0x6f0c0fc5336e249d, 0x745ede19c331cfd9, 0xf2d6fd0009eefe1c, 0x127c158bf0fa1ebe}}},
+{{{0xf6197c422e9879a2, 0xa44addd452ca3647, 0x9b413fc14b4eaccb, 0x354ef87d07ef4f68}},
+ {{0xdea28fc4ae51b974, 0x1d9973d3744dfe96, 0x6240680b873848a8, 0x4ed82479d167df95}},
+ {{0xfee3b52260c5d975, 0x50352efceb41b0b8, 0x8808ac30a9f6653c, 0x302d92d20539236d}}},
+{{{0x7813c1a2bca4283d, 0xed62f091a1863dd9, 0xaec7bcb8c268fa86, 0x10e5d3b76f1cae4c}},
+ {{0x2dbc6fb6e4e0f177, 0x04e1bf29a4bd6a93, 0x5e1966d4787af6e8, 0x0edc5f5eb426d060}},
+ {{0x5453bfd653da8e67, 0xe9dc1eec24a9f641, 0xbf87263b03578a23, 0x45b46c51361cba72}}},
+{{{0xa9402abf314f7fa1, 0xe257f1dc8e8cf450, 0x1dbbd54b23a8be84, 0x2177bfa36dcb713b}},
+ {{0xce9d4ddd8a7fe3e4, 0xab13645676620e30, 0x4b594f7bb30e9958, 0x5c1c0aef321229df}},
+ {{0x37081bbcfa79db8f, 0x6048811ec25f59b3, 0x087a76659c832487, 0x4ae619387d8ab5bb}}},
+{{{0x8ddbf6aa5344a32e, 0x7d88eab4b41b4078, 0x5eb0eb974a130d60, 0x1a00d91b17bf3e03}},
+ {{0x61117e44985bfb83, 0xfce0462a71963136, 0x83ac3448d425904b, 0x75685abe5ba43d64}},
+ {{0x6e960933eb61f2b2, 0x543d0fa8c9ff4952, 0xdf7275107af66569, 0x135529b623b0e6aa}}},
+{{{0x18f0dbd7add1d518, 0x979f7888cfc11f11, 0x8732e1f07114759b, 0x79b5b81a65ca3a01}},
+ {{0xf5c716bce22e83fe, 0xb42beb19e80985c1, 0xec9da63714254aae, 0x5972ea051590a613}},
+ {{0x0fd4ac20dc8f7811, 0x9a9ad294ac4d4fa8, 0xc01b2d64b3360434, 0x4f7e9c95905f3bdb}}},
+{{{0x62674bbc5781302e, 0xd8520f3989addc0f, 0x8c2999ae53fbd9c6, 0x31993ad92e638e4c}},
+ {{0x71c8443d355299fe, 0x8bcd3b1cdbebead7, 0x8092499ef1a49466, 0x1942eec4a144adc8}},
+ {{0x7dac5319ae234992, 0x2c1b3d910cea3e92, 0x553ce494253c1122, 0x2a0a65314ef9ca75}}},
+{{{0x2db7937ff7f927c2, 0xdb741f0617d0a635, 0x5982f3a21155af76, 0x4cf6e218647c2ded}},
+ {{0xcf361acd3c1c793a, 0x2f9ebcac5a35bc3b, 0x60e860e9a8cda6ab, 0x055dc39b6dea1a13}},
+ {{0xb119227cc28d5bb6, 0x07e24ebc774dffab, 0xa83c78cee4a32c89, 0x121a307710aa24b6}}},
+{{{0xe4db5d5e9f034a97, 0xe153fc093034bc2d, 0x460546919551d3b1, 0x333fc76c7a40e52d}},
+ {{0xd659713ec77483c9, 0x88bfe077b82b96af, 0x289e28231097bcd3, 0x527bb94a6ced3a9b}},
+ {{0x563d992a995b482e, 0x3405d07c6e383801, 0x485035de2f64d8e5, 0x6b89069b20a7a9f7}}},
+{{{0x812aa0416270220d, 0x995a89faf9245b4e, 0xffadc4ce5072ef05, 0x23bc2103aa73eb73}},
+ {{0x4082fa8cb5c7db77, 0x068686f8c734c155, 0x29e6c8d9f6e7a57e, 0x0473d308a7639bcf}},
+ {{0xcaee792603589e05, 0x2b4b421246dcc492, 0x02a1ef74e601a94f, 0x102f73bfde04341a}}},
+{{{0xeb18b9ab7f5745c6, 0x023a8aee5787c690, 0xb72712da2df7afa9, 0x36597d25ea5c013d}},
+ {{0xa2b4dae0b5511c9a, 0x7ac860292bffff06, 0x981f375df5504234, 0x3f6bd725da4ea12d}},
+ {{0x734d8d7b106058ac, 0xd940579e6fc6905f, 0x6466f8f99202932d, 0x7b7ecc19da60d6d0}}},
+{{{0x78c2373c695c690d, 0xdd252e660642906e, 0x951d44444ae12bd2, 0x4235ad7601743956}},
+ {{0x6dae4a51a77cfa9b, 0x82263654e7a38650, 0x09bbffcd8f2d82db, 0x03bedc661bf5caba}},
+ {{0x6258cb0d078975f5, 0x492942549189f298, 0xa0cab423e2e36ee4, 0x0e7ce2b0cdf066a1}}},
+{{{0xc494643ac48c85a3, 0xfd361df43c6139ad, 0x09db17dd3ae94d48, 0x666e0a5d8fb4674a}},
+ {{0xfea6fedfd94b70f9, 0xf130c051c1fcba2d, 0x4882d47e7f2fab89, 0x615256138aeceeb5}},
+ {{0x2abbf64e4870cb0d, 0xcd65bcf0aa458b6b, 0x9abe4eba75e8985d, 0x7f0bc810d514dee4}}},
+{{{0xb9006ba426f4136f, 0x8d67369e57e03035, 0xcbc8dfd94f463c28, 0x0d1f8dbcf8eedbf5}},
+ {{0x83ac9dad737213a0, 0x9ff6f8ba2ef72e98, 0x311e2edd43ec6957, 0x1d3a907ddec5ab75}},
+ {{0xba1693313ed081dc, 0x29329fad851b3480, 0x0128013c030321cb, 0x00011b44a31bfde3}}},
+{{{0x3fdfa06c3fc66c0c, 0x5d40e38e4dd60dd2, 0x7ae38b38268e4d71, 0x3ac48d916e8357e1}},
+ {{0x16561f696a0aa75c, 0xc1bf725c5852bd6a, 0x11a8dd7f9a7966ad, 0x63d988a2d2851026}},
+ {{0x00120753afbd232e, 0xe92bceb8fdd8f683, 0xf81669b384e72b91, 0x33fad52b2368a066}}},
+{{{0x540649c6c5e41e16, 0x0af86430333f7735, 0xb2acfcd2f305e746, 0x16c0f429a256dca7}},
+ {{0x8d2cc8d0c422cfe8, 0x072b4f7b05a13acb, 0xa3feb6e6ecf6a56f, 0x3cc355ccb90a71e2}},
+ {{0xe9b69443903e9131, 0xb8a494cb7a5637ce, 0xc87cd1a4baba9244, 0x631eaf426bae7568}}},
+{{{0xb3e90410da66fe9f, 0x85dd4b526c16e5a6, 0xbc3d97611ef9bf83, 0x5599648b1ea919b5}},
+ {{0x47d975b9a3700de8, 0x7280c5fbe2f80552, 0x53658f2732e45de1, 0x431f2c7f665f80b5}},
+ {{0xd6026344858f7b19, 0x14ab352fa1ea514a, 0x8900441a2090a9d7, 0x7b04715f91253b26}}},
+{{{0x83edbd28acf6ae43, 0x86357c8b7d5c7ab4, 0xc0404769b7eb2c44, 0x59b37bf5c2f6583f}},
+ {{0xb376c280c4e6bac6, 0x970ed3dd6d1d9b0b, 0xb09a9558450bf944, 0x48d0acfa57cde223}},
+ {{0xb60f26e47dabe671, 0xf1d1a197622f3a37, 0x4208ce7ee9960394, 0x16234191336d3bdb}}},
+{{{0xb9e499def6267ff6, 0x7772ca7b742c0843, 0x23a0153fe9a4f2b1, 0x2cdfdfecd5d05006}},
+ {{0xdd499cd61ff38640, 0x29cd9bc3063625a0, 0x51e2d8023dd73dc3, 0x4a25707a203b9231}},
+ {{0x2ab7668a53f6ed6a, 0x304242581dd170a1, 0x4000144c3ae20161, 0x5721896d248e49fc}}},
+{{{0x0b6e5517fd181bae, 0x9022629f2bb963b4, 0x5509bce932064625, 0x578edd74f63c13da}},
+ {{0x285d5091a1d0da4e, 0x4baa6fa7b5fe3e08, 0x63e5177ce19393b3, 0x03c935afc4b030fd}},
+ {{0x997276c6492b0c3d, 0x47ccc2c4dfe205fc, 0xdcd29b84dd623a3c, 0x3ec2ab590288c7a2}}},
+{{{0xa1a0d27be4d87bb9, 0xa98b4deb61391aed, 0x99a0ddd073cb9b83, 0x2dd5c25a200fcace}},
+ {{0xa7213a09ae32d1cb, 0x0f2b87df40f5c2d5, 0x0baea4c6e81eab29, 0x0e1bf66c6adbac5e}},
+ {{0xe2abd5e9792c887e, 0x1a020018cb926d5d, 0xbfba69cdbaae5f1e, 0x730548b35ae88f5f}}},
+{{{0xc43551a3cba8b8ee, 0x65a26f1db2115f16, 0x760f4f52ab8c3850, 0x3043443b411db8ca}},
+ {{0x805b094ba1d6e334, 0xbf3ef17709353f19, 0x423f06cb0622702b, 0x585a2277d87845dd}},
+ {{0xa18a5f8233d48962, 0x6698c4b5ec78257f, 0xa78e6fa5373e41ff, 0x7656278950ef981f}}},
+{{{0x38c3cf59d51fc8c0, 0x9bedd2fd0506b6f2, 0x26bf109fab570e8f, 0x3f4160a8c1b846a6}},
+ {{0xe17073a3ea86cf9d, 0x3a8cfbb707155fdc, 0x4853e7fc31838a8e, 0x28bbf484b613f616}},
+ {{0xf2612f5c6f136c7c, 0xafead107f6dd11be, 0x527e9ad213de6f33, 0x1e79cb358188f75d}}},
+{{{0x013436c3eef7e3f1, 0x828b6a7ffe9e10f8, 0x7ff908e5bcf9defc, 0x65d7951b3a3b3831}},
+ {{0x77e953d8f5e08181, 0x84a50c44299dded9, 0xdc6c2d0c864525e5, 0x478ab52d39d1f2f4}},
+ {{0x66a6a4d39252d159, 0xe5dde1bc871ac807, 0xb82c6b40a6c1c96f, 0x16d87a411a212214}}},
+{{{0xb3bd7e5a42066215, 0x879be3cd0c5a24c1, 0x57c05db1d6f994b7, 0x28f87c8165f38ca6}},
+ {{0xfba4d5e2d54e0583, 0xe21fafd72ebd99fa, 0x497ac2736ee9778f, 0x1f990b577a5a6dde}},
+ {{0xa3344ead1be8f7d6, 0x7d1e50ebacea798f, 0x77c6569e520de052, 0x45882fe1534d6d3e}}},
+{{{0x6669345d757983d6, 0x62b6ed1117aa11a6, 0x7ddd1857985e128f, 0x688fe5b8f626f6dd}},
+ {{0xd8ac9929943c6fe4, 0xb5f9f161a38392a2, 0x2699db13bec89af3, 0x7dcf843ce405f074}},
+ {{0x6c90d6484a4732c0, 0xd52143fdca563299, 0xb3be28c3915dc6e1, 0x6739687e7327191b}}},
+{{{0xef782014385675a6, 0xa2649f30aafda9e8, 0x4cd1eb505cdfa8cb, 0x46115aba1d4dc0b3}},
+ {{0xa66dcc9dc80c1ac0, 0x97a05cf41b38a436, 0xa7ebf3be95dbd7c6, 0x7da0b8f68d7e7dab}},
+ {{0xd40f1953c3b5da76, 0x1dac6f7321119e9b, 0x03cc6021feb25960, 0x5a5f887e83674b4b}}},
+{{{0x8f6301cf70a13d11, 0xcfceb815350dd0c4, 0xf70297d4a4bca47e, 0x3669b656e44d1434}},
+ {{0x9e9628d3a0a643b9, 0xb5c3cb00e6c32064, 0x9b5302897c2dec32, 0x43e37ae2d5d1c70c}},
+ {{0x387e3f06eda6e133, 0x67301d5199a13ac0, 0xbd5ad8f836263811, 0x6a21e6cd4fd5e9be}}},
+{{{0xf1c6170a3046e65f, 0x58712a2a00d23524, 0x69dbbd3c8c82b755, 0x586bf9f1a195ff57}},
+ {{0xef4129126699b2e3, 0x71d30847708d1301, 0x325432d01182b0bd, 0x45371b07001e8b36}},
+ {{0xa6db088d5ef8790b, 0x5278f0dc610937e5, 0xac0349d261a16eb8, 0x0eafb03790e52179}}},
+{{{0x960555c13748042f, 0x219a41e6820baa11, 0x1c81f73873486d0c, 0x309acc675a02c661}},
+ {{0x5140805e0f75ae1d, 0xec02fbe32662cc30, 0x2cebdf1eea92396d, 0x44ae3344c5435bb3}},
+ {{0x9cf289b9bba543ee, 0xf3760e9d5ac97142, 0x1d82e5c64f9360aa, 0x62d5221b7f94678f}}},
+{{{0x524c299c18d0936d, 0xc86bb56c8a0c1a0c, 0xa375052edb4a8631, 0x5c0efde4bc754562}},
+ {{0x7585d4263af77a3c, 0xdfae7b11fee9144d, 0xa506708059f7193d, 0x14f29a5383922037}},
+ {{0xdf717edc25b2d7f5, 0x21f970db99b53040, 0xda9234b7c3ed4c62, 0x5e72365c7bee093e}}},
+{{{0x575bfc074571217f, 0x3779675d0694d95b, 0x9a0a37bbf4191e33, 0x77f1104c47b4eabc}},
+ {{0x7d9339062f08b33e, 0x5b9659e5df9f32be, 0xacff3dad1f9ebdfd, 0x70b20555cb7349b7}},
+ {{0xbe5113c555112c4c, 0x6688423a9a881fcd, 0x446677855e503b47, 0x0e34398f4a06404a}}},
+{{{0xb67d22d93ecebde8, 0x09b3e84127822f07, 0x743fa61fb05b6d8d, 0x5e5405368a362372}},
+ {{0x18930b093e4b1928, 0x7de3e10e73f3f640, 0xf43217da73395d6f, 0x6f8aded6ca379c3e}},
+ {{0xe340123dfdb7b29a, 0x487b97e1a21ab291, 0xf9967d02fde6949e, 0x780de72ec8d3de97}}},
+{{{0x0ae28545089ae7bc, 0x388ddecf1c7f4d06, 0x38ac15510a4811b8, 0x0eb28bf671928ce4}},
+ {{0x671feaf300f42772, 0x8f72eb2a2a8c41aa, 0x29a17fd797373292, 0x1defc6ad32b587a6}},
+ {{0xaf5bbe1aef5195a7, 0x148c1277917b15ed, 0x2991f7fb7ae5da2e, 0x467d201bf8dd2867}}},
+{{{0x95fe919a74ef4fad, 0x3a827becf6a308a2, 0x964e01d309a47b01, 0x71c43c4f5ba3c797}},
+ {{0xbc1ef4bd567ae7a9, 0x3f624cb2d64498bd, 0xe41064d22c1f4ec8, 0x2ef9c5a5ba384001}},
+ {{0xb6fd6df6fa9e74cd, 0xf18278bce4af267a, 0x8255b3d0f1ef990e, 0x5a758ca390c5f293}}},
+{{{0xa2b72710d9462495, 0x3aa8c6d2d57d5003, 0xe3d400bfa0b487ca, 0x2dbae244b3eb72ec}},
+ {{0x8ce0918b1d61dc94, 0x8ded36469a813066, 0xd4e6a829afe8aad3, 0x0a738027f639d43f}},
+ {{0x980f4a2f57ffe1cc, 0x00670d0de1839843, 0x105c3f4a49fb15fd, 0x2698ca635126a69c}}},
+{{{0xe765318832b0ba78, 0x381831f7925cff8b, 0x08a81b91a0291fcc, 0x1fb43dcc49caeb07}},
+ {{0x2e3d702f5e3dd90e, 0x9e3f0918e4d25386, 0x5e773ef6024da96a, 0x3c004b0c4afa3332}},
+ {{0x9aa946ac06f4b82b, 0x1ca284a5a806c4f3, 0x3ed3265fc6cd4787, 0x6b43fd01cd1fd217}}},
+{{{0xc7a75d4b4697c544, 0x15fdf848df0fffbf, 0x2868b9ebaa46785a, 0x5a68d7105b52f714}},
+ {{0xb5c742583e760ef3, 0x75dc52b9ee0ab990, 0xbf1427c2072b923f, 0x73420b2d6ff0d9f0}},
+ {{0xaf2cf6cb9e851e06, 0x8f593913c62238c4, 0xda8ab89699fbf373, 0x3db5632fea34bc9e}}},
+{{{0xf46eee2bf75dd9d8, 0x0d17b1f6396759a5, 0x1bf2d131499e7273, 0x04321adf49d75f13}},
+ {{0x2e4990b1829825d5, 0xedeaeb873e9a8991, 0xeef03d394c704af8, 0x59197ea495df2b0e}},
+ {{0x04e16019e4e55aae, 0xe77b437a7e2f92e9, 0xc7ce2dc16f159aa4, 0x45eafdc1f4d70cc0}}},
+{{{0x698401858045d72b, 0x4c22faa2cf2f0651, 0x941a36656b222dc6, 0x5a5eebc80362dade}},
+ {{0xb60e4624cfccb1ed, 0x59dbc292bd5c0395, 0x31a09d1ddc0481c9, 0x3f73ceea5d56d940}},
+ {{0xb7a7bfd10a4e8dc6, 0xbe57007e44c9b339, 0x60c1207f1557aefa, 0x26058891266218db}}},
+{{{0x59f704a68360ff04, 0xc3d93fde7661e6f4, 0x831b2a7312873551, 0x54ad0c2e4e615d57}},
+ {{0x4c818e3cc676e542, 0x5e422c9303ceccad, 0xec07cccab4129f08, 0x0dedfa10b24443b8}},
+ {{0xee3b67d5b82b522a, 0x36f163469fa5c1eb, 0xa5b4d2f26ec19fd3, 0x62ecb2baa77a9408}}},
+{{{0xe5ed795261152b3d, 0x4962357d0eddd7d1, 0x7482c8d0b96b4c71, 0x2e59f919a966d8be}},
+ {{0x92072836afb62874, 0x5fcd5e8579e104a5, 0x5aad01adc630a14a, 0x61913d5075663f98}},
+ {{0x0dc62d361a3231da, 0xfa47583294200270, 0x02d801513f9594ce, 0x3ddbc2a131c05d5c}}},
+{{{0x9adc0ff9ce5ec54b, 0x039c2a6b8c2f130d, 0x028007c7f0f89515, 0x78968314ac04b36b}},
+ {{0xf3aa57a22796bb14, 0x883abab79b07da21, 0xe54be21831a0391c, 0x5ee7fb38d83205f9}},
+ {{0x538dfdcb41446a8e, 0xa5acfda9434937f9, 0x46af908d263c8c78, 0x61d0633c9bca0d09}}},
+{{{0x63744935ffdb2566, 0xc5bd6b89780b68bb, 0x6f1b3280553eec03, 0x6e965fd847aed7f5}},
+ {{0xada328bcf8fc73df, 0xee84695da6f037fc, 0x637fb4db38c2a909, 0x5b23ac2df8067bdc}},
+ {{0x9ad2b953ee80527b, 0xe88f19aafade6d8d, 0x0e711704150e82cf, 0x79b9bbb9dd95dedc}}},
+{{{0xebb355406a3126c2, 0xd26383a868c8c393, 0x6c0c6429e5b97a82, 0x5065f158c9fd2147}},
+ {{0xd1997dae8e9f7374, 0xa032a2f8cfbb0816, 0xcd6cba126d445f0a, 0x1ba811460accb834}},
+ {{0x708169fb0c429954, 0xe14600acd76ecf67, 0x2eaab98a70e645ba, 0x3981f39e58a4faf2}}},
+{{{0x18fb8a7559230a93, 0x1d168f6960e6f45d, 0x3a85a94514a93cb5, 0x38dc083705acd0fd}},
+ {{0xc845dfa56de66fde, 0xe152a5002c40483a, 0xe9d2e163c7b4f632, 0x30f4452edcbc1b65}},
+ {{0x856d2782c5759740, 0xfa134569f99cbecc, 0x8844fc73c0ea4e71, 0x632d9a1a593f2469}}},
+{{{0xf6bb6b15b807cba6, 0x1823c7dfbc54f0d7, 0xbb1d97036e29670b, 0x0b24f48847ed4a57}},
+ {{0xbf09fd11ed0c84a7, 0x63f071810d9f693a, 0x21908c2d57cf8779, 0x3a5a7df28af64ba2}},
+ {{0xdcdad4be511beac7, 0xa4538075ed26ccf2, 0xe19cff9f005f9a65, 0x34fcf74475481f63}}},
+{{{0xc197e04c789767ca, 0xb8714dcb38d9467d, 0x55de888283f95fa8, 0x3d3bdc164dfa63f7}},
+ {{0xa5bb1dab78cfaa98, 0x5ceda267190b72f2, 0x9309c9110a92608e, 0x0119a3042fb374b0}},
+ {{0x67a2d89ce8c2177d, 0x669da5f66895d0c1, 0xf56598e5b282a2b0, 0x56c088f1ede20a73}}},
+{{{0x336d3d1110a86e17, 0xd7f388320b75b2fa, 0xf915337625072988, 0x09674c6b99108b87}},
+ {{0x581b5fac24f38f02, 0xa90be9febae30cbd, 0x9a2169028acf92f0, 0x038b7ea48359038f}},
+ {{0x9f4ef82199316ff8, 0x2f49d282eaa78d4f, 0x0971a5ab5aef3174, 0x6e5e31025969eb65}}},
+{{{0xb16c62f587e593fb, 0x4999eddeca5d3e71, 0xb491c1e014cc3e6d, 0x08f5114789a8dba8}},
+ {{0x3304fb0e63066222, 0xfb35068987acba3f, 0xbd1924778c1061a3, 0x3058ad43d1838620}},
+ {{0x323c0ffde57663d0, 0x05c3df38a22ea610, 0xbdc78abdac994f9a, 0x26549fa4efe3dc99}}},
+{{{0x741d5a461e6bf9d6, 0x2305b3fc7777a581, 0xd45574a26474d3d9, 0x1926e1dc6401e0ff}},
+ {{0xdb468549af3f666e, 0xd77fcf04f14a0ea5, 0x3df23ff7a4ba0c47, 0x3a10dfe132ce3c85}},
+ {{0xe07f4e8aea17cea0, 0x2fd515463a1fc1fd, 0x175322fd31f2c0f1, 0x1fa1d01d861e5d15}}},
+{{{0xcc8055947d599832, 0x1e4656da37f15520, 0x99f6f7744e059320, 0x773563bc6a75cf33}},
+ {{0x38dcac00d1df94ab, 0x2e712bddd1080de9, 0x7f13e93efdd5e262, 0x73fced18ee9a01e5}},
+ {{0x06b1e90863139cb3, 0xa493da67c5a03ecd, 0x8d77cec8ad638932, 0x1f426b701b864f44}}},
+{{{0xefc9264c41911c01, 0xf1a3b7b817a22c25, 0x5875da6bf30f1447, 0x4e1af5271d31b090}},
+ {{0xf17e35c891a12552, 0xb76b8153575e9c76, 0xfa83406f0d9b723e, 0x0b76bb1b3fa7e438}},
+ {{0x08b8c1f97f92939b, 0xbe6771cbd444ab6e, 0x22e5646399bb8017, 0x7b6dd61eb772a955}}},
+{{{0xb7adc1e850f33d92, 0x7998fa4f608cd5cf, 0xad962dbd8dfc5bdb, 0x703e9bceaf1d2f4f}},
+ {{0x5730abf9ab01d2c7, 0x16fb76dc40143b18, 0x866cbe65a0cbb281, 0x53fa9b659bff6afe}},
+ {{0x6c14c8e994885455, 0x843a5d6665aed4e5, 0x181bb73ebcd65af1, 0x398d93e5c4c61f50}}},
+{{{0x1c4bd16733e248f3, 0xbd9e128715bf0a5f, 0xd43f8cf0a10b0376, 0x53b09b5ddf191b13}},
+ {{0xc3877c60d2e7e3f2, 0x3b34aaa030828bb1, 0x283e26e7739ef138, 0x699c9c9002c30577}},
+ {{0xf306a7235946f1cc, 0x921718b5cce5d97d, 0x28cdd24781b4e975, 0x51caf30c6fcdd907}}},
+{{{0xa60ba7427674e00a, 0x630e8570a17a7bf3, 0x3758563dcf3324cc, 0x5504aa292383fdaa}},
+ {{0x737af99a18ac54c7, 0x903378dcc51cb30f, 0x2b89bc334ce10cc7, 0x12ae29c189f8e99a}},
+ {{0xa99ec0cb1f0d01cf, 0x0dd1efcc3a34f7ae, 0x55ca7521d09c4e22, 0x5fd14fe958eba5ea}}},
+{{{0xb5dc2ddf2845ab2c, 0x069491b10a7fe993, 0x4daaf3d64002e346, 0x093ff26e586474d1}},
+ {{0x3c42fe5ebf93cb8e, 0xbedfa85136d4565f, 0xe0f0859e884220e8, 0x7dd73f960725d128}},
+ {{0xb10d24fe68059829, 0x75730672dbaf23e5, 0x1367253ab457ac29, 0x2f59bcbc86b470a4}}},
+{{{0x83847d429917135f, 0xad1b911f567d03d7, 0x7e7748d9be77aad1, 0x5458b42e2e51af4a}},
+ {{0x7041d560b691c301, 0x85201b3fadd7e71e, 0x16c2e16311335585, 0x2aa55e3d010828b1}},
+ {{0xed5192e60c07444f, 0x42c54e2d74421d10, 0x352b4c82fdb5c864, 0x13e9004a8a768664}}},
+{{{0x739d8845832fcedb, 0xfa38d6c9ae6bf863, 0x32bc0dcab74ffef7, 0x73937e8814bce45e}},
+ {{0xbb2e00c9193b877f, 0xece3a890e0dc506b, 0xecf3b7c036de649f, 0x5f46040898de9e1a}},
+ {{0xb9037116297bf48d, 0xa9d13b22d4f06834, 0xe19715574696bdc6, 0x2cf8a4e891d5e835}}},
+{{{0x6d93fd8707110f67, 0xdd4c09d37c38b549, 0x7cb16a4cc2736a86, 0x2049bd6e58252a09}},
+ {{0x2cb5487e17d06ba2, 0x24d2381c3950196b, 0xd7659c8185978a30, 0x7a6f7f2891d6a4f6}},
+ {{0x7d09fd8d6a9aef49, 0xf0ee60be5b3db90b, 0x4c21b52c519ebfd4, 0x6011aadfc545941d}}},
+{{{0x5f67926dcf95f83c, 0x7c7e856171289071, 0xd6a1e7f3998f7a5b, 0x6fc5cc1b0b62f9e0}},
+ {{0x63ded0c802cbf890, 0xfbd098ca0dff6aaa, 0x624d0afdb9b6ed99, 0x69ce18b779340b1e}},
+ {{0xd1ef5528b29879cb, 0xdd1aae3cd47e9092, 0x127e0442189f2352, 0x15596b3ae57101f1}}},
+{{{0x462739d23f9179a2, 0xff83123197d6ddcf, 0x1307deb553f2148a, 0x0d2237687b5f4dda}},
+ {{0x09ff31167e5124ca, 0x0be4158bd9c745df, 0x292b7d227ef556e5, 0x3aa4e241afb6d138}},
+ {{0x2cc138bf2a3305f5, 0x48583f8fa2e926c3, 0x083ab1a25549d2eb, 0x32fcaa6e4687a36c}}},
+{{{0x7bc56e8dc57d9af5, 0x3e0bd2ed9df0bdf2, 0xaac014de22efe4a3, 0x4627e9cefebd6a5c}},
+ {{0x3207a4732787ccdf, 0x17e31908f213e3f8, 0xd5b2ecd7f60d964e, 0x746f6336c2600be9}},
+ {{0x3f4af345ab6c971c, 0xe288eb729943731f, 0x33596a8a0344186d, 0x7b4917007ed66293}}},
+{{{0x2d85fb5cab84b064, 0x497810d289f3bc14, 0x476adc447b15ce0c, 0x122ba376f844fd7b}},
+ {{0x54341b28dd53a2dd, 0xaa17905bdf42fc3f, 0x0ff592d94dd2f8f4, 0x1d03620fe08cd37d}},
+ {{0xc20232cda2b4e554, 0x9ed0fd42115d187f, 0x2eabb4be7dd479d9, 0x02c70bf52b68ec4c}}},
+{{{0xa287ec4b5d0b2fbb, 0x415c5790074882ca, 0xe044a61ec1d0815c, 0x26334f0a409ef5e0}},
+ {{0xace532bf458d72e1, 0x5be768e07cb73cb5, 0x56cf7d94ee8bbde7, 0x6b0697e3feb43a03}},
+ {{0xb6c8f04adf62a3c0, 0x3ef000ef076da45d, 0x9c9cb95849f0d2a9, 0x1cc37f43441b2fae}}},
+{{{0x508f565a5cc7324f, 0xd061c4c0e506a922, 0xfb18abdb5c45ac19, 0x6c6809c10380314a}},
+ {{0xd76656f1c9ceaeb9, 0x1c5b15f818e5656a, 0x26e72832844c2334, 0x3a346f772f196838}},
+ {{0xd2d55112e2da6ac8, 0xe9bd0331b1e851ed, 0x960746dd8ec67262, 0x05911b9f6ef7c5d0}}},
+{{{0xc1339983f5df0ebb, 0xc0f3758f512c4cac, 0x2cf1130a0bb398e1, 0x6b3cecf9aa270c62}},
+ {{0x5349acf3512eeaef, 0x20c141d31cc1cb49, 0x24180c07a99a688d, 0x555ef9d1c64b2d17}},
+ {{0x36a770ba3b73bd08, 0x624aef08a3afbf0c, 0x5737ff98b40946f2, 0x675f4de13381749d}}},
+{{{0x0e2c52036b1782fc, 0x64816c816cad83b4, 0xd0dcbdd96964073e, 0x13d99df70164c520}},
+ {{0xa12ff6d93bdab31d, 0x0725d80f9d652dfe, 0x019c4ff39abe9487, 0x60f450b882cd3c43}},
+ {{0x014b5ec321e5c0ca, 0x4fcb69c9d719bfa2, 0x4e5f1c18750023a0, 0x1c06de9e55edac80}}},
+{{{0x990f7ad6a33ec4e2, 0x6608f938be2ee08e, 0x9ca143c563284515, 0x4cf38a1fec2db60d}},
+ {{0xffd52b40ff6d69aa, 0x34530b18dc4049bb, 0x5e4a5c2fa34d9897, 0x78096f8e7d32ba2d}},
+ {{0xa0aaaa650dfa5ce7, 0xf9c49e2a48b5478c, 0x4f09cc7d7003725b, 0x373cad3a26091abe}}},
+{{{0xb294634d82c9f57c, 0x1fcbfde124934536, 0x9e9c4db3418cdb5a, 0x0040f3d9454419fc}},
+ {{0xf1bea8fb89ddbbad, 0x3bcb2cbc61aeaecb, 0x8f58a7bb1f9b8d9d, 0x21547eda5112a686}},
+ {{0xdefde939fd5986d3, 0xf4272c89510a380c, 0xb72ba407bb3119b9, 0x63550a334a254df4}}},
+{{{0x6507d6edb569cf37, 0x178429b00ca52ee1, 0xea7c0090eb6bd65d, 0x3eea62c7daf78f51}},
+ {{0x9bba584572547b49, 0xf305c6fae2c408e0, 0x60e8fa69c734f18d, 0x39a92bafaa7d767a}},
+ {{0x9d24c713e693274e, 0x5f63857768dbd375, 0x70525560eb8ab39a, 0x68436a0665c9c4cd}}},
+{{{0xbc0235e8202f3f27, 0xc75c00e264f975b0, 0x91a4e9d5a38c2416, 0x17b6e7f68ab789f9}},
+ {{0x1e56d317e820107c, 0xc5266844840ae965, 0xc1e0a1c6320ffc7a, 0x5373669c91611472}},
+ {{0x5d2814ab9a0e5257, 0x908f2084c9cab3fc, 0xafcaf5885b2d1eca, 0x1cb4b5a678f87d11}}},
+{{{0xb664c06b394afc6c, 0x0c88de2498da5fb1, 0x4f8d03164bcad834, 0x330bca78de7434a2}},
+ {{0x6b74aa62a2a007e7, 0xf311e0b0f071c7b1, 0x5707e438000be223, 0x2dc0fd2d82ef6eac}},
+ {{0x982eff841119744e, 0xf9695e962b074724, 0xc58ac14fbfc953fb, 0x3c31be1b369f1cf5}}},
+{{{0xb0f4864d08948aee, 0x07dc19ee91ba1c6f, 0x7975cdaea6aca158, 0x330b61134262d4bb}},
+ {{0xc168bc93f9cb4272, 0xaeb8711fc7cedb98, 0x7f0e52aa34ac8d7a, 0x41cec1097e7d55bb}},
+ {{0xf79619d7a26d808a, 0xbb1fd49e1d9e156d, 0x73d7c36cdba1df27, 0x26b44cd91f28777d}}},
+{{{0x51f048478f387475, 0xb25dbcf49cbecb3c, 0x9aab1244d99f2055, 0x2c709e6c1c10a5d6}},
+ {{0xe1b7f29362730383, 0x4b5279ffebca8a2c, 0xdafc778abfd41314, 0x7deb10149c72610f}},
+ {{0xcb62af6a8766ee7a, 0x66cbec045553cd0e, 0x588001380f0be4b5, 0x08e68e9ff62ce2ea}}},
+{{{0x34ad500a4bc130ad, 0x8d38db493d0bd49c, 0xa25c3d98500a89be, 0x2f1f3f87eeba3b09}},
+ {{0x2f2d09d50ab8f2f9, 0xacb9218dc55923df, 0x4a8f342673766cb9, 0x4cb13bd738f719f5}},
+ {{0xf7848c75e515b64a, 0xa59501badb4a9038, 0xc20d313f3f751b50, 0x19a1e353c0ae2ee8}}},
+{{{0x7d1c7560bafa05c3, 0xb3e1a0a0c6e55e61, 0xe3529718c0d66473, 0x41546b11c20c3486}},
+ {{0xb42172cdd596bdbd, 0x93e0454398eefc40, 0x9fb15347b44109b5, 0x736bd3990266ae34}},
+ {{0x85532d509334b3b4, 0x46fd114b60816573, 0xcc5f5f30425c8375, 0x412295a2b87fab5c}}},
+{{{0x19c99b88f57ed6e9, 0x5393cb266df8c825, 0x5cee3213b30ad273, 0x14e153ebb52d2e34}},
+ {{0x2e655261e293eac6, 0x845a92032133acdb, 0x460975cb7900996b, 0x0760bb8d195add80}},
+ {{0x413e1a17cde6818a, 0x57156da9ed69a084, 0x2cbf268f46caccb1, 0x6b34be9bc33ac5f2}}},
+{{{0xf3df2f643a78c0b2, 0x4c3e971ef22e027c, 0xec7d1c5e49c1b5a3, 0x2012c18f0922dd2d}},
+ {{0x11fc69656571f2d3, 0xc6c9e845530e737a, 0xe33ae7a2d4fe5035, 0x01b9c7b62e6dd30b}},
+ {{0x880b55e55ac89d29, 0x1483241f45a0a763, 0x3d36efdfc2e76c1f, 0x08af5b784e4bade8}}},
+{{{0x283499dc881f2533, 0x9d0525da779323b6, 0x897addfb673441f4, 0x32b79d71163a168d}},
+ {{0xe27314d289cc2c4b, 0x4be4bd11a287178d, 0x18d528d6fa3364ce, 0x6423c1d5afd9826e}},
+ {{0xcc85f8d9edfcb36a, 0x22bcc28f3746e5f9, 0xe49de338f9e5d3cd, 0x480a5efbc13e2dcc}}},
+{{{0x0b51e70b01622071, 0x06b505cf8b1dafc5, 0x2c6bb061ef5aabcd, 0x47aa27600cb7bf31}},
+ {{0xb6614ce442ce221f, 0x6e199dcc4c053928, 0x663fb4a4dc1cbe03, 0x24b31d47691c8e06}},
+ {{0x2a541eedc015f8c3, 0x11a4fe7e7c693f7c, 0xf0af66134ea278d6, 0x545b585d14dda094}}},
+{{{0x67bf275ea0d43a0f, 0xade68e34089beebe, 0x4289134cd479e72e, 0x0f62f9c332ba5454}},
+ {{0x6204e4d0e3b321e1, 0x3baa637a28ff1e95, 0x0b0ccffd5b99bd9e, 0x4d22dc3e64c8d071}},
+ {{0xfcb46589d63b5f39, 0x5cae6a3f57cbcf61, 0xfebac2d2953afa05, 0x1c0fa01a36371436}}},
+{{{0xd2c604b622943dff, 0xbc8cbece44cfb3a0, 0x5d254ff397808678, 0x0fa3614f3b1ca6bf}},
+ {{0x69082b0e8c936a50, 0xf9c9a035c1dac5b6, 0x6fb73e54c4dfb634, 0x4005419b1d2bc140}},
+ {{0xa003febdb9be82f0, 0x2089c1af3a44ac90, 0xf8499f911954fa8e, 0x1fba218aef40ab42}}},
+{{{0xab549448fac8f53e, 0x81f6e89a7ba63741, 0x74fd6c7d6c2b5e01, 0x392e3acaa8c86e42}},
+ {{0x4f3e57043e7b0194, 0xa81d3eee08daaf7f, 0xc839c6ab99dcdef1, 0x6c535d13ff7761d5}},
+ {{0x4cbd34e93e8a35af, 0x2e0781445887e816, 0x19319c76f29ab0ab, 0x25e17fe4d50ac13b}}},
+{{{0x0a289bd71e04f676, 0x208e1c52d6420f95, 0x5186d8b034691fab, 0x255751442a9fb351}},
+ {{0x915f7ff576f121a7, 0xc34a32272fcd87e3, 0xccba2fde4d1be526, 0x6bba828f8969899b}},
+ {{0xe2d1bc6690fe3901, 0x4cb54a18a0997ad5, 0x971d6914af8460d4, 0x559d504f7f6b7be4}}},
+{{{0xa7738378b3eb54d5, 0x1d69d366a5553c7c, 0x0a26cf62f92800ba, 0x01ab12d5807e3217}},
+ {{0x9c4891e7f6d266fd, 0x0744a19b0307781b, 0x88388f1d6061e23b, 0x123ea6a3354bd50e}},
+ {{0x118d189041e32d96, 0xb9ede3c2d8315848, 0x1eab4271d83245d9, 0x4a3961e2c918a154}}},
+{{{0x71dc3be0f8e6bba0, 0xd6cef8347effe30a, 0xa992425fe13a476a, 0x2cd6bce3fb1db763}},
+ {{0x0327d644f3233f1e, 0x499a260e34fcf016, 0x83b5a716f2dab979, 0x68aceead9bd4111f}},
+ {{0x38b4c90ef3d7c210, 0x308e6e24b7ad040c, 0x3860d9f1b7e73e23, 0x595760d5b508f597}}},
+{{{0x6129bfe104aa6397, 0x8f960008a4a7fccb, 0x3f8bc0897d909458, 0x709fa43edcb291a9}},
+ {{0x882acbebfd022790, 0x89af3305c4115760, 0x65f492e37d3473f4, 0x2cb2c5df54515a2b}},
+ {{0xeb0a5d8c63fd2aca, 0xd22bc1662e694eff, 0x2723f36ef8cbb03a, 0x70f029ecf0c8131f}}},
+{{{0x461307b32eed3e33, 0xae042f33a45581e7, 0xc94449d3195f0366, 0x0b7d5d8a6c314858}},
+ {{0x2a6aafaa5e10b0b9, 0x78f0a370ef041aa9, 0x773efb77aa3ad61f, 0x44eca5a2a74bd9e1}},
+ {{0x25d448327b95d543, 0x70d38300a3340f1d, 0xde1c531c60e1c52b, 0x272224512c7de9e4}}},
+{{{0x1abc92af49c5342e, 0xffeed811b2e6fad0, 0xefa28c8dfcc84e29, 0x11b5df18a44cc543}},
+ {{0xbf7bbb8a42a975fc, 0x8c5c397796ada358, 0xe27fc76fcdedaa48, 0x19735fd7f6bc20a6}},
+ {{0xe3ab90d042c84266, 0xeb848e0f7f19547e, 0x2503a1d065a497b9, 0x0fef911191df895f}}}

ext/ed25519-amd64-asm/ge25519_base_slide_multiples.data

@@ -0,0 +1,96 @@
+{{{0x9d103905d740913e, 0xfd399f05d140beb3, 0xa5c18434688f8a09, 0x44fd2f9298f81267}},
+ {{0x2fbc93c6f58c3b85, 0xcf932dc6fb8c0e19, 0x270b4898643d42c2, 0x07cf9d3a33d4ba65}},
+ {{0xabc91205877aaa68, 0x26d9e823ccaac49e, 0x5a1b7dcbdd43598c, 0x6f117b689f0c65a8}}},
+{{{0x56611fe8a4fcd265, 0x3bd353fde5c1ba7d, 0x8131f31a214bd6bd, 0x2ab91587555bda62}},
+ {{0xaf25b0a84cee9730, 0x025a8430e8864b8a, 0xc11b50029f016732, 0x7a164e1b9a80f8f4}},
+ {{0x14ae933f0dd0d889, 0x589423221c35da62, 0xd170e5458cf2db4c, 0x5a2826af12b9b4c6}}},
+{{{0x7f9182c3a447d6ba, 0xd50014d14b2729b7, 0xe33cf11cb864a087, 0x154a7e73eb1b55f3}},
+ {{0xa212bc4408a5bb33, 0x8d5048c3c75eed02, 0xdd1beb0c5abfec44, 0x2945ccf146e206eb}},
+ {{0xbcbbdbf1812a8285, 0x270e0807d0bdd1fc, 0xb41b670b1bbda72d, 0x43aabe696b3bb69a}}},
+{{{0xba6f2c9aaa3221b1, 0x6ca021533bba23a7, 0x9dea764f92192c3a, 0x1d6edd5d2e5317e0}},
+ {{0x6b1a5cd0944ea3bf, 0x7470353ab39dc0d2, 0x71b2528228542e49, 0x461bea69283c927e}},
+ {{0xf1836dc801b8b3a2, 0xb3035f47053ea49a, 0x529c41ba5877adf3, 0x7a9fbb1c6a0f90a7}}},
+{{{0xf36e217e039d8064, 0x98a081b6f520419b, 0x96cbc608e75eb044, 0x49c05a51fadc9c8f}},
+ {{0x9b2e678aa6a8632f, 0xa6509e6f51bc46c5, 0xceb233c9c686f5b5, 0x34b9ed338add7f59}},
+ {{0x06b4e8bf9045af1b, 0xe2ff83e8a719d22f, 0xaaf6fc2993d4cf16, 0x73c172021b008b06}}},
+{{{0x315f5b0249864348, 0x3ed6b36977088381, 0xa3a075556a8deb95, 0x18ab598029d5c77f}},
+ {{0x2fbf00848a802ade, 0xe5d9fecf02302e27, 0x113e847117703406, 0x4275aae2546d8faf}},
+ {{0xd82b2cc5fd6089e9, 0x031eb4a13282e4a4, 0x44311199b51a8622, 0x3dc65522b53df948}}},
+{{{0x506f013b327fbf93, 0xaefcebc99b776f6b, 0x9d12b232aaad5968, 0x0267882d176024a7}},
+ {{0xbf70c222a2007f6d, 0xbf84b39ab5bcdedb, 0x537a0e12fb07ba07, 0x234fd7eec346f241}},
+ {{0x5360a119732ea378, 0x2437e6b1df8dd471, 0xa2ef37f891a7e533, 0x497ba6fdaa097863}}},
+{{{0x040bcd86468ccf0b, 0xd3829ba42a9910d6, 0x7508300807b25192, 0x43b5cd4218d05ebf}},
+ {{0x24cecc0313cfeaa0, 0x8648c28d189c246d, 0x2dbdbdfac1f2d4d0, 0x61e22917f12de72b}},
+ {{0x5d9a762f9bd0b516, 0xeb38af4e373fdeee, 0x032e5a7d93d64270, 0x511d61210ae4d842}}},
+{{{0x081386484420de87, 0x8a1cf016b592edb4, 0x39fa4e2729942d25, 0x71a7fe6fe2482810}},
+ {{0x92c676ef950e9d81, 0xa54620cdc0d7044f, 0xaa9b36646f8f1248, 0x6d325924ddb855e3}},
+ {{0x6c7182b8a5c8c854, 0x33fd1479fe5f2a03, 0x72cf591883778d0c, 0x4746c4b6559eeaa9}}},
+{{{0x348546c864741147, 0x7d35aedd0efcc849, 0xff939a760672a332, 0x219663497db5e6d6}},
+ {{0xd3777b3c6dc69a2b, 0xdefab2276f89f617, 0x45651cf7b53a16b5, 0x5c9a51de34fe9fb7}},
+ {{0xf510f1cf79f10e67, 0xffdddaa1e658515b, 0x09c3a71710142277, 0x4804503c608223bb}}},
+{{{0x3b6821d23a36d175, 0xbbb40aa7e99b9e32, 0x5d9e5ce420838a47, 0x771e098858de4c5e}},
+ {{0xc4249ed02ca37fc7, 0xa059a0e3a615acab, 0x88a96ed7c96e0e23, 0x553398a51650696d}},
+ {{0x9a12f5d278451edf, 0x3ada5d7985899ccb, 0x477f4a2d9fa59508, 0x5a5ed1d68ff5a611}}},
+{{{0xbae5e0c558527359, 0x392e5c19cadb9d7e, 0x28653c1eda1cabe9, 0x019b60135fefdc44}},
+ {{0x1195122afe150e83, 0xcf209a257e4b35d8, 0x7387f8291e711e20, 0x44acb897d8bf92f0}},
+ {{0x1e6068145e134b83, 0xc4f5e64f24304c16, 0x506e88a8fc1a3ed7, 0x150c49fde6ad2f92}}},
+{{{0xb849863c9cdca868, 0xc83f44dbb8714ad0, 0xfe3ee3560c36168d, 0x78a6d7791e05fbc1}},
+ {{0x8e7bf29509471138, 0x5d6fef394f75a651, 0x10af79c425a708ad, 0x6b2b5a075bb99922}},
+ {{0x58bf704b47a0b976, 0xa601b355741748d5, 0xaa2b1fb1d542f590, 0x725c7ffc4ad55d00}}},
+{{{0x91802bf71cd098c0, 0xfe416ca4ed5e6366, 0xdf585d714902994c, 0x4cd54625f855fae7}},
+ {{0xe4426715d1cf99b2, 0x7352d51102a20d34, 0x23d1157b8b12109f, 0x794cc9277cb1f3a3}},
+ {{0x4af6c426c2ac5053, 0xbc9aedad32f67258, 0x2ad032f10a311021, 0x7008357b6fcc8e85}}},
+{{{0xd01b9fbb82584a34, 0x47ab6463d2b4792b, 0xb631639c48536202, 0x13a92a3669d6d428}},
+ {{0x0b88672738773f01, 0xb8ccc8fa95fbccfb, 0x8d2dd5a3b9ad29b6, 0x06ef7e9851ad0f6a}},
+ {{0xca93771cc0577de5, 0x7540e41e5035dc5c, 0x24680f01d802e071, 0x3c296ddf8a2af86a}}},
+{{{0xfceb4d2ebb1f2541, 0xb89510c740adb91f, 0xfc71a37dd0a1ad05, 0x0a892c700747717b}},
+ {{0xaead15f9d914a713, 0xa92f7bf98c8ff912, 0xaff823179f53d730, 0x7a99d393490c77ba}},
+ {{0x8f52ed2436bda3e8, 0x77a8c84157e80794, 0xa5a96563262f9ce0, 0x286762d28302f7d2}}},
+{{{0x7c558e2bce2ef5bd, 0xe4986cb46747bc63, 0x154a179f3bbb89b8, 0x7686f2a3d6f1767a}},
+ {{0x4e7836093ce35b25, 0x82e1181db26baa97, 0x0cc192d3cbc7b83f, 0x32f1da046a9d9d3a}},
+ {{0xaa8d12a66d597c6a, 0x8f11930304d3852b, 0x3f91dc73c209b022, 0x561305f8a9ad28a6}}},
+{{{0x6722cc28e7b0c0d5, 0x709de9bbdb075c53, 0xcaf68da7d7010a61, 0x030a1aef2c57cc6c}},
+ {{0x100c978dec92aed1, 0xca43d5434d6d73e5, 0x83131b22d847ba48, 0x00aaec53e35d4d2c}},
+ {{0x7bb1f773003ad2aa, 0x0b3f29802b216608, 0x7821dc86520ed23e, 0x20be9c1c24065480}}},
+{{{0x20e0e44ae2025e60, 0xb03b3b2fcbdcb938, 0x105d639cf95a0d1c, 0x69764c545067e311}},
+ {{0xe15387d8249673a6, 0x5943bc2df546e493, 0x1c7f9a81c36f63b5, 0x750ab3361f0ac1de}},
+ {{0x1e8a3283a2f81037, 0x6f2eda23bd7fcbf1, 0xb72fd15bac2e2563, 0x54f96b3fb7075040}}},
+{{{0x177dafc616b11ecd, 0x89764b9cfa576479, 0xb7a8a110e6ece785, 0x78e6839fbe85dbf0}},
+ {{0x0fadf20429669279, 0x3adda2047d7d724a, 0x6f3d94828c5760f1, 0x3d7fe9c52bb7539e}},
+ {{0x70332df737b8856b, 0x75d05d43041a178a, 0x320ff74aa0e59e22, 0x70f268f350088242}}},
+{{{0x2324112070dcf355, 0x380cc97ee7fce117, 0xb31ddeed3552b698, 0x404e56c039b8c4b9}},
+ {{0x66864583b1805f47, 0xf535c5d160dd7c19, 0xe9874eb71e4cb006, 0x7c0d345cfad889d9}},
+ {{0x591f1f4b8c78338a, 0xa0366ab167e0b5e1, 0x5cbc4152b45f3d44, 0x20d754762aaec777}}},
+{{{0x9d74feb135b9f543, 0x84b37df1de8c956c, 0xe9322b0757138ba9, 0x38b8ada8790b4ce1}},
+ {{0x5e8fc36fc73bb758, 0xace543a5363cbb9a, 0xa9934a7d903bc922, 0x2b8f1e46f3ceec62}},
+ {{0xb5c04a9cdf51f95d, 0x2b3952aecb1fdeac, 0x1d106d8b328b66da, 0x049aeb32ceba1953}}},
+{{{0xd7767d3c63dcfe7e, 0x209c594897856e40, 0xb6676861e14f7c13, 0x51c665e0c8d625fc}},
+ {{0xaa507d0b75fc7931, 0x0fef924b7a6725d3, 0x1d82542b396b3930, 0x795ee17530f674fc}},
+ {{0x254a5b0a52ecbd81, 0x5d411f6ee034afe7, 0xe6a24d0dcaee4a31, 0x6cd19bf49dc54477}}},
+{{{0x7e87619052179ca3, 0x571d0a060b2c9f85, 0x80a2baa88499711e, 0x7520f3db40b2e638}},
+ {{0x1ffe612165afc386, 0x082a2a88b8d51b10, 0x76f6627e20990baa, 0x5e01b3a7429e43e7}},
+ {{0x3db50be3d39357a1, 0x967b6cdd599e94a5, 0x1a309a64df311e6e, 0x71092c9ccef3c986}}},
+{{{0x53d8523f0364918c, 0xa2b404f43fab6b1c, 0x080b4a9e6681e5a4, 0x0ea15b03d0257ba7}},
+ {{0x856bd8ac74051dcf, 0x03f6a40855b7aa1e, 0x3a4ae7cbc9743ceb, 0x4173a5bb7137abde}},
+ {{0x17c56e31f0f9218a, 0x5a696e2b1afc4708, 0xf7931668f4b2f176, 0x5fc565614a4e3a67}}},
+{{{0x136e570dc46d7ae5, 0x0fd0aacc54f8dc8f, 0x59549f03310dad86, 0x62711c414c454aa1}},
+ {{0x4892e1e67790988e, 0x01d5950f1c5cd722, 0xe3b0819ae5923eed, 0x3214c7409d46651b}},
+ {{0x1329827406651770, 0x3ba4a0668a279436, 0xd9b6b8ec185d223c, 0x5bea94073ecb833c}}},
+{{{0x641dbf0912c89be4, 0xacf38b317d6e579c, 0xabfe9e02f697b065, 0x3aacd5c148f61eec}},
+ {{0xb470ce63f343d2f8, 0x0067ba8f0543e8f1, 0x35da51a1a2117b6f, 0x4ad0785944f1bd2f}},
+ {{0x858e3b34c3318301, 0xdc99c04707316826, 0x34085b2ed39da88c, 0x3aff0cb1d902853d}}},
+{{{0x87c5c7eb3a20405e, 0x8ee311efedad56c9, 0x29252e48ad29d5f9, 0x110e7e86f4cd251d}},
+ {{0x9226430bf4c53505, 0x68e49c13261f2283, 0x09ef33788fd327c6, 0x2ccf9f732bd99e7f}},
+ {{0x57c0d89ed603f5e4, 0x12888628f0b0200c, 0x53172709a02e3bb7, 0x05c557e0b9693a37}}},
+{{{0xd8f9ce311fc97e6f, 0x7a3f263011f9fdae, 0xe15b7ea08bed25dd, 0x6e154c178fe9875a}},
+ {{0xf776bbb089c20eb0, 0x61f85bf6fa0fd85c, 0xb6b93f4e634421fb, 0x289fef0841861205}},
+ {{0xcf616336fed69abf, 0x9b16e4e78335c94f, 0x13789765753a7fe7, 0x6afbf642a95ca319}}},
+{{{0x7da8de0c62f5d2c1, 0x98fc3da4b00e7b9a, 0x7deb6ada0dad70e0, 0x0db4b851b95038c4}},
+ {{0x5de55070f913a8cc, 0x7d1d167b2b0cf561, 0xda2956b690ead489, 0x12c093cedb801ed9}},
+ {{0xfc147f9308b8190f, 0x06969da0a11ae310, 0xcee75572dac7d7fd, 0x33aa8799c6635ce6}}},
+{{{0xaf0ff51ebd085cf2, 0x78f51a8967d33f1f, 0x6ec2bfe15060033c, 0x233c6f29e8e21a86}},
+ {{0x8348f588fc156cb1, 0x6da2ba9b1a0a6d27, 0xe2262d5c87ca5ab6, 0x212cd0c1c8d589a6}},
+ {{0xd2f4d5107f18c781, 0x122ecdf2527e9d28, 0xa70a862a3d3d3341, 0x1db7778911914ce3}}},
+{{{0xddf352397c6bc26f, 0x7a97e2cc53d50113, 0x7c74f43abf79a330, 0x31ad97ad26e2adfc}},
+ {{0xb3394769dd701ab6, 0xe2b8ded419cf8da5, 0x15df4161fd2ac852, 0x7ae2ca8a017d24be}},
+ {{0xb7e817ed0920b962, 0x1e8518cc3f19da9d, 0xe491c14f25560a64, 0x1ed1fc53a6622c83}}}

ext/ed25519-amd64-asm/ge25519_double.c

@@ -0,0 +1,8 @@
+#include "ge25519.h"
+
+void ge25519_double(ge25519_p3 *r, const ge25519_p3 *p)
+{
+  ge25519_p1p1 grp1p1;
+  ge25519_dbl_p1p1(&grp1p1, (ge25519_p2 *)p);
+  ge25519_p1p1_to_p3(r, &grp1p1);
+}

ext/ed25519-amd64-asm/ge25519_double_scalarmult.c

@@ -0,0 +1,102 @@
+#include "fe25519.h"
+#include "sc25519.h"
+#include "ge25519.h"
+
+#define S1_SWINDOWSIZE 5
+#define PRE1_SIZE (1<<(S1_SWINDOWSIZE-2))
+#define S2_SWINDOWSIZE 7
+#define PRE2_SIZE (1<<(S2_SWINDOWSIZE-2))
+
+ge25519_niels pre2[PRE2_SIZE] = {
+#include "ge25519_base_slide_multiples.data"
+};
+
+static const fe25519 ec2d = {{0xEBD69B9426B2F146, 0x00E0149A8283B156, 0x198E80F2EEF3D130, 0xA406D9DC56DFFCE7}};
+
+static void setneutral(ge25519 *r)
+{
+  fe25519_setint(&r->x,0);
+  fe25519_setint(&r->y,1);
+  fe25519_setint(&r->z,1);
+  fe25519_setint(&r->t,0);
+}
+
+/* computes [s1]p1 + [s2]p2 */
+void ge25519_double_scalarmult_vartime(ge25519_p3 *r, const ge25519_p3 *p1, const sc25519 *s1, const sc25519 *s2)
+{
+  signed char slide1[256], slide2[256];
+  ge25519_pniels pre1[PRE1_SIZE], neg;
+  ge25519_p3 d1;
+  ge25519_p1p1 t;
+  ge25519_niels nneg;
+  fe25519 d;
+  int i;
+  
+  sc25519_slide(slide1, s1, S1_SWINDOWSIZE);
+  sc25519_slide(slide2, s2, S2_SWINDOWSIZE);
+
+  /* precomputation */
+  pre1[0] = *(ge25519_pniels *)p1;                                                                         
+  ge25519_dbl_p1p1(&t,(ge25519_p2 *)pre1);      ge25519_p1p1_to_p3(&d1, &t);
+  /* Convert pre[0] to projective Niels representation */
+  d = pre1[0].ysubx;
+  fe25519_sub(&pre1[0].ysubx, &pre1[0].xaddy, &pre1[0].ysubx);
+  fe25519_add(&pre1[0].xaddy, &pre1[0].xaddy, &d);
+  fe25519_mul(&pre1[0].t2d, &pre1[0].t2d, &ec2d);
+
+  for(i=0;i<PRE1_SIZE-1;i++)
+  {
+    ge25519_pnielsadd_p1p1(&t, &d1, &pre1[i]);  ge25519_p1p1_to_p3((ge25519_p3 *)&pre1[i+1], &t);
+    /* Convert pre1[i+1] to projective Niels representation */
+    d = pre1[i+1].ysubx;
+    fe25519_sub(&pre1[i+1].ysubx, &pre1[i+1].xaddy, &pre1[i+1].ysubx);
+    fe25519_add(&pre1[i+1].xaddy, &pre1[i+1].xaddy, &d);
+    fe25519_mul(&pre1[i+1].t2d, &pre1[i+1].t2d, &ec2d);
+  }
+
+  setneutral(r);
+  for (i = 255;i >= 0;--i) {
+    if (slide1[i] || slide2[i]) goto firstbit;
+  }
+
+  for(;i>=0;i--)
+  {
+    firstbit:
+
+    ge25519_dbl_p1p1(&t, (ge25519_p2 *)r);
+
+    if(slide1[i]>0)
+    {
+      ge25519_p1p1_to_p3(r, &t);
+      ge25519_pnielsadd_p1p1(&t, r, &pre1[slide1[i]/2]);
+    }
+    else if(slide1[i]<0)
+    {
+      ge25519_p1p1_to_p3(r, &t);
+      neg = pre1[-slide1[i]/2];
+      d = neg.ysubx;
+      neg.ysubx = neg.xaddy;
+      neg.xaddy = d;
+      fe25519_neg(&neg.t2d, &neg.t2d);
+      ge25519_pnielsadd_p1p1(&t, r, &neg);
+    }
+
+    if(slide2[i]>0)
+    {
+      ge25519_p1p1_to_p3(r, &t);
+      ge25519_nielsadd_p1p1(&t, r, &pre2[slide2[i]/2]);
+    }
+    else if(slide2[i]<0)
+    {
+      ge25519_p1p1_to_p3(r, &t);
+      nneg = pre2[-slide2[i]/2];
+      d = nneg.ysubx;
+      nneg.ysubx = nneg.xaddy;
+      nneg.xaddy = d;
+      fe25519_neg(&nneg.t2d, &nneg.t2d);
+      ge25519_nielsadd_p1p1(&t, r, &nneg);
+    }
+
+    ge25519_p1p1_to_p2((ge25519_p2 *)r, &t);
+  }
+}

ext/ed25519-amd64-asm/ge25519_isneutral.c

@@ -0,0 +1,9 @@
+#include "fe25519.h"
+#include "ge25519.h"
+
+int ge25519_isneutral_vartime(const ge25519_p3 *p)
+{
+  if(!fe25519_iszero_vartime(&p->x)) return 0;
+  if(!fe25519_iseq_vartime(&p->y, &p->z)) return 0;
+  return 1;
+}

ext/ed25519-amd64-asm/ge25519_multi_scalarmult.c

@@ -0,0 +1,102 @@
+#include "fe25519.h"
+#include "sc25519.h"
+#include "ge25519.h"
+#include "index_heap.h"
+
+static void setneutral(ge25519 *r)
+{
+  fe25519_setint(&r->x,0);
+  fe25519_setint(&r->y,1);
+  fe25519_setint(&r->z,1);
+  fe25519_setint(&r->t,0);
+}
+
+static void ge25519_scalarmult_vartime_2limbs(ge25519 *r, ge25519 *p, sc25519 *s)
+{
+  if (s->v[1] == 0 && s->v[0] == 1) /* This will happen most of the time after Bos-Coster */
+    *r = *p; 
+  else if (s->v[1] == 0 && s->v[0] == 0) /* This won't ever happen, except for all scalars == 0 in Bos-Coster */
+    setneutral(r);
+  else
+  {
+    ge25519 d;
+    unsigned long long mask = (1ULL << 63);
+    int i = 1;
+    while(!(mask & s->v[1]) && mask != 0)
+      mask >>= 1;
+    if(mask == 0)
+    {
+      mask = (1ULL << 63);
+      i = 0;
+      while(!(mask & s->v[0]) && mask != 0)
+        mask >>= 1;
+    }
+    d = *p;
+    mask >>= 1;
+    for(;mask != 0;mask >>= 1)
+    {
+      ge25519_double(&d,&d);
+      if(s->v[i] & mask)
+        ge25519_add(&d,&d,p);
+    }
+    if(i==1)
+    {
+      mask = (1ULL << 63);
+      for(;mask != 0;mask >>= 1)
+      {
+        ge25519_double(&d,&d);
+        if(s->v[0] & mask)
+          ge25519_add(&d,&d,p);
+      }
+    }
+    *r = d;
+  }
+}
+
+/* caller's responsibility to ensure npoints >= 5 */
+void ge25519_multi_scalarmult_vartime(ge25519_p3 *r, ge25519_p3 *p, sc25519 *s, const unsigned long long npoints)
+{
+  unsigned long long pos[npoints];
+  unsigned long long hlen=((npoints+1)/2)|1;
+  unsigned long long max1, max2,i;
+
+  heap_init(pos, hlen, s);
+  
+  for(i=0;;i++)
+  {
+    heap_get2max(pos, &max1, &max2, s);
+    if((s[max1].v[3] == 0) || (sc25519_iszero_vartime(&s[max2]))) break;
+    sc25519_sub_nored(&s[max1],&s[max1],&s[max2]);
+    ge25519_add(&p[max2],&p[max2],&p[max1]);
+    heap_rootreplaced(pos, hlen, s);
+  }
+  for(;;i++)
+  {
+    heap_get2max(pos, &max1, &max2, s);
+    if((s[max1].v[2] == 0) || (sc25519_iszero_vartime(&s[max2]))) break;
+    sc25519_sub_nored(&s[max1],&s[max1],&s[max2]);
+    ge25519_add(&p[max2],&p[max2],&p[max1]);
+    heap_rootreplaced_3limbs(pos, hlen, s);
+  }
+  /* We know that (npoints-1)/2 scalars are only 128-bit scalars */
+  heap_extend(pos, hlen, npoints, s);
+  hlen = npoints;
+  for(;;i++)
+  {
+    heap_get2max(pos, &max1, &max2, s);
+    if((s[max1].v[1] == 0) || (sc25519_iszero_vartime(&s[max2]))) break;
+    sc25519_sub_nored(&s[max1],&s[max1],&s[max2]);
+    ge25519_add(&p[max2],&p[max2],&p[max1]);
+    heap_rootreplaced_2limbs(pos, hlen, s);
+  }
+  for(;;i++)
+  {
+    heap_get2max(pos, &max1, &max2, s);
+    if(sc25519_iszero_vartime(&s[max2])) break;
+    sc25519_sub_nored(&s[max1],&s[max1],&s[max2]);
+    ge25519_add(&p[max2],&p[max2],&p[max1]);
+    heap_rootreplaced_1limb(pos, hlen, s);
+  }
+
+  ge25519_scalarmult_vartime_2limbs(r, &p[max1], &s[max1]);
+}

ext/ed25519-amd64-asm/ge25519_pack.c

@@ -0,0 +1,13 @@
+#include "fe25519.h"
+#include "sc25519.h"
+#include "ge25519.h"
+
+void ge25519_pack(unsigned char r[32], const ge25519_p3 *p)
+{
+  fe25519 tx, ty, zi;
+  fe25519_invert(&zi, &p->z); 
+  fe25519_mul(&tx, &p->x, &zi);
+  fe25519_mul(&ty, &p->y, &zi);
+  fe25519_pack(r, &ty);
+  r[31] ^= fe25519_getparity(&tx) << 7;
+}

ext/ed25519-amd64-asm/ge25519_scalarmult_base.c

@@ -0,0 +1,68 @@
+#include "fe25519.h"
+#include "sc25519.h"
+#include "ge25519.h"
+
+/* Multiples of the base point in Niels' representation */
+static const ge25519_niels ge25519_base_multiples_niels[] = {
+#ifdef SMALLTABLES
+#include "ge25519_base_niels_smalltables.data"
+#else
+#include "ge25519_base_niels.data"
+#endif
+};
+
+/* d */
+static const fe25519 ecd = {{0x75EB4DCA135978A3, 0x00700A4D4141D8AB, 0x8CC740797779E898, 0x52036CEE2B6FFE73}};
+
+void ge25519_scalarmult_base(ge25519_p3 *r, const sc25519 *s)
+{
+  signed char b[64];
+  int i;
+  ge25519_niels t;
+  fe25519 d;
+
+  sc25519_window4(b,s);
+
+#ifdef SMALLTABLES
+  ge25519_p1p1 tp1p1;
+  choose_t((ge25519_niels *)r, 0, (signed long long) b[1], ge25519_base_multiples_niels);
+  fe25519_sub(&d, &r->y, &r->x);
+  fe25519_add(&r->y, &r->y, &r->x);
+  r->x = d;
+  r->t = r->z;
+  fe25519_setint(&r->z,2);
+  for(i=3;i<64;i+=2)
+  {
+    choose_t(&t, (unsigned long long) i/2, (signed long long) b[i], ge25519_base_multiples_niels);
+    ge25519_nielsadd2(r, &t);
+  }
+  ge25519_dbl_p1p1(&tp1p1,(ge25519_p2 *)r);
+  ge25519_p1p1_to_p2((ge25519_p2 *)r, &tp1p1);
+  ge25519_dbl_p1p1(&tp1p1,(ge25519_p2 *)r);
+  ge25519_p1p1_to_p2((ge25519_p2 *)r, &tp1p1);
+  ge25519_dbl_p1p1(&tp1p1,(ge25519_p2 *)r);
+  ge25519_p1p1_to_p2((ge25519_p2 *)r, &tp1p1);
+  ge25519_dbl_p1p1(&tp1p1,(ge25519_p2 *)r);
+  ge25519_p1p1_to_p3(r, &tp1p1);
+  choose_t(&t, (unsigned long long) 0, (signed long long) b[0], ge25519_base_multiples_niels);
+  fe25519_mul(&t.t2d, &t.t2d, &ecd);
+  ge25519_nielsadd2(r, &t);
+  for(i=2;i<64;i+=2)
+  {
+    choose_t(&t, (unsigned long long) i/2, (signed long long) b[i], ge25519_base_multiples_niels);
+    ge25519_nielsadd2(r, &t);
+  }
+#else 
+  choose_t((ge25519_niels *)r, 0, (signed long long) b[0], ge25519_base_multiples_niels);
+  fe25519_sub(&d, &r->y, &r->x);
+  fe25519_add(&r->y, &r->y, &r->x);
+  r->x = d;
+  r->t = r->z;
+  fe25519_setint(&r->z,2);
+  for(i=1;i<64;i++)
+  {
+    choose_t(&t, (unsigned long long) i, (signed long long) b[i], ge25519_base_multiples_niels);
+    ge25519_nielsadd2(r, &t);
+  }
+#endif 
+}

ext/ed25519-amd64-asm/ge25519_unpackneg.c

@@ -0,0 +1,60 @@
+#include "fe25519.h"
+#include "ge25519.h"
+
+/* d */
+static const fe25519 ecd = {{0x75EB4DCA135978A3, 0x00700A4D4141D8AB, 0x8CC740797779E898, 0x52036CEE2B6FFE73}};
+/* sqrt(-1) */
+static const fe25519 sqrtm1 = {{0xC4EE1B274A0EA0B0, 0x2F431806AD2FE478, 0x2B4D00993DFBD7A7, 0x2B8324804FC1DF0B}};
+
+/* return 0 on success, -1 otherwise */
+int ge25519_unpackneg_vartime(ge25519_p3 *r, const unsigned char p[32])
+{
+  fe25519 t, chk, num, den, den2, den4, den6;
+  unsigned char par = p[31] >> 7;
+
+  fe25519_setint(&r->z,1);
+  fe25519_unpack(&r->y, p); 
+  fe25519_square(&num, &r->y); /* x = y^2 */
+  fe25519_mul(&den, &num, &ecd); /* den = dy^2 */
+  fe25519_sub(&num, &num, &r->z); /* x = y^2-1 */
+  fe25519_add(&den, &r->z, &den); /* den = dy^2+1 */
+
+  /* Computation of sqrt(num/den)
+     1.: computation of num^((p-5)/8)*den^((7p-35)/8) = (num*den^7)^((p-5)/8)
+  */
+  fe25519_square(&den2, &den);
+  fe25519_square(&den4, &den2);
+  fe25519_mul(&den6, &den4, &den2);
+  fe25519_mul(&t, &den6, &num);
+  fe25519_mul(&t, &t, &den);
+
+  fe25519_pow2523(&t, &t);
+  /* 2. computation of r->x = t * num * den^3
+  */
+  fe25519_mul(&t, &t, &num);
+  fe25519_mul(&t, &t, &den);
+  fe25519_mul(&t, &t, &den);
+  fe25519_mul(&r->x, &t, &den);
+
+  /* 3. Check whether sqrt computation gave correct result, multiply by sqrt(-1) if not:
+  */
+  fe25519_square(&chk, &r->x);
+  fe25519_mul(&chk, &chk, &den);
+  if (!fe25519_iseq_vartime(&chk, &num))
+    fe25519_mul(&r->x, &r->x, &sqrtm1);
+
+  /* 4. Now we have one of the two square roots, except if input was not a square
+  */
+  fe25519_square(&chk, &r->x);
+  fe25519_mul(&chk, &chk, &den);
+  if (!fe25519_iseq_vartime(&chk, &num))
+    return -1;
+
+  /* 5. Choose the desired square root according to parity:
+  */
+  if(fe25519_getparity(&r->x) != (1-par))
+    fe25519_neg(&r->x, &r->x);
+
+  fe25519_mul(&r->t, &r->x, &r->y);
+  return 0;
+}

ext/ed25519-amd64-asm/heap_rootreplaced.s

@@ -0,0 +1,476 @@
+
+# qhasm: int64 hp
+
+# qhasm: int64 hlen
+
+# qhasm: int64 sp
+
+# qhasm: int64 pp
+
+# qhasm: input hp
+
+# qhasm: input hlen
+
+# qhasm: input sp
+
+# qhasm: int64 prc
+
+# qhasm: int64 plc
+
+# qhasm: int64 pc
+
+# qhasm: int64 d
+
+# qhasm: int64 spp
+
+# qhasm: int64 sprc
+
+# qhasm: int64 spc
+
+# qhasm: int64 c0
+
+# qhasm: int64 c1
+
+# qhasm: int64 c2
+
+# qhasm: int64 c3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm: int64 p0
+
+# qhasm: int64 p1
+
+# qhasm: int64 p2
+
+# qhasm: int64 p3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_heap_rootreplaced
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_heap_rootreplaced
+.globl crypto_sign_ed25519_amd64_64_heap_rootreplaced
+_crypto_sign_ed25519_amd64_64_heap_rootreplaced:
+crypto_sign_ed25519_amd64_64_heap_rootreplaced:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm: caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm: caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm: caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm: caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm: caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm: caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm: caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: pp = 0
+# asm 1: mov  $0,>pp=int64#4
+# asm 2: mov  $0,>pp=%rcx
+mov  $0,%rcx
+
+# qhasm: siftdownloop:
+._siftdownloop:
+
+# qhasm: prc = pp
+# asm 1: mov  <pp=int64#4,>prc=int64#5
+# asm 2: mov  <pp=%rcx,>prc=%r8
+mov  %rcx,%r8
+
+# qhasm: prc *= 2
+# asm 1: imulq  $2,<prc=int64#5,>prc=int64#5
+# asm 2: imulq  $2,<prc=%r8,>prc=%r8
+imulq  $2,%r8,%r8
+
+# qhasm: pc = prc
+# asm 1: mov  <prc=int64#5,>pc=int64#6
+# asm 2: mov  <prc=%r8,>pc=%r9
+mov  %r8,%r9
+
+# qhasm: prc += 2
+# asm 1: add  $2,<prc=int64#5
+# asm 2: add  $2,<prc=%r8
+add  $2,%r8
+
+# qhasm: pc += 1
+# asm 1: add  $1,<pc=int64#6
+# asm 2: add  $1,<pc=%r9
+add  $1,%r9
+
+# qhasm: unsigned>? hlen - prc
+# asm 1: cmp  <prc=int64#5,<hlen=int64#2
+# asm 2: cmp  <prc=%r8,<hlen=%rsi
+cmp  %r8,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop if !unsigned>
+jbe ._siftuploop
+
+# qhasm: sprc = *(uint64 *)(hp + prc * 8)
+# asm 1: movq   (<hp=int64#1,<prc=int64#5,8),>sprc=int64#7
+# asm 2: movq   (<hp=%rdi,<prc=%r8,8),>sprc=%rax
+movq   (%rdi,%r8,8),%rax
+
+# qhasm: sprc <<= 5
+# asm 1: shl  $5,<sprc=int64#7
+# asm 2: shl  $5,<sprc=%rax
+shl  $5,%rax
+
+# qhasm: sprc += sp
+# asm 1: add  <sp=int64#3,<sprc=int64#7
+# asm 2: add  <sp=%rdx,<sprc=%rax
+add  %rdx,%rax
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#6,8),>spc=int64#8
+# asm 2: movq   (<hp=%rdi,<pc=%r9,8),>spc=%r10
+movq   (%rdi,%r9,8),%r10
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#8
+# asm 2: shl  $5,<spc=%r10
+shl  $5,%r10
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#8
+# asm 2: add  <sp=%rdx,<spc=%r10
+add  %rdx,%r10
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#8),>c0=int64#9
+# asm 2: movq   0(<spc=%r10),>c0=%r11
+movq   0(%r10),%r11
+
+# qhasm: c1 = *(uint64 *)(spc +  8)
+# asm 1: movq   8(<spc=int64#8),>c1=int64#10
+# asm 2: movq   8(<spc=%r10),>c1=%r12
+movq   8(%r10),%r12
+
+# qhasm: c2 = *(uint64 *)(spc + 16)
+# asm 1: movq   16(<spc=int64#8),>c2=int64#11
+# asm 2: movq   16(<spc=%r10),>c2=%r13
+movq   16(%r10),%r13
+
+# qhasm: c3 = *(uint64 *)(spc + 24)
+# asm 1: movq   24(<spc=int64#8),>c3=int64#12
+# asm 2: movq   24(<spc=%r10),>c3=%r14
+movq   24(%r10),%r14
+
+# qhasm: carry? c0 -= *(uint64 *)(sprc +  0)
+# asm 1: subq 0(<sprc=int64#7),<c0=int64#9
+# asm 2: subq 0(<sprc=%rax),<c0=%r11
+subq 0(%rax),%r11
+
+# qhasm: carry? c1 -= *(uint64 *)(sprc +  8) - carry
+# asm 1: sbbq 8(<sprc=int64#7),<c1=int64#10
+# asm 2: sbbq 8(<sprc=%rax),<c1=%r12
+sbbq 8(%rax),%r12
+
+# qhasm: carry? c2 -= *(uint64 *)(sprc + 16) - carry
+# asm 1: sbbq 16(<sprc=int64#7),<c2=int64#11
+# asm 2: sbbq 16(<sprc=%rax),<c2=%r13
+sbbq 16(%rax),%r13
+
+# qhasm: carry? c3 -= *(uint64 *)(sprc + 24) - carry
+# asm 1: sbbq 24(<sprc=int64#7),<c3=int64#12
+# asm 2: sbbq 24(<sprc=%rax),<c3=%r14
+sbbq 24(%rax),%r14
+
+# qhasm: pc = prc if carry
+# asm 1: cmovc <prc=int64#5,<pc=int64#6
+# asm 2: cmovc <prc=%r8,<pc=%r9
+cmovc %r8,%r9
+
+# qhasm: spc = sprc if carry
+# asm 1: cmovc <sprc=int64#7,<spc=int64#8
+# asm 2: cmovc <sprc=%rax,<spc=%r10
+cmovc %rax,%r10
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#8
+# asm 2: sub  <sp=%rdx,<spc=%r10
+sub  %rdx,%r10
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#8
+# asm 2: shr  $5,<spc=%r10
+shr  $5,%r10
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#8,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r10,(<hp=%rdi,<pp=%rcx,8)
+movq  %r10,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#6,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%r9,8)
+movq  %r8,(%rdi,%r9,8)
+
+# qhasm: pp = pc
+# asm 1: mov  <pc=int64#6,>pp=int64#4
+# asm 2: mov  <pc=%r9,>pp=%rcx
+mov  %r9,%rcx
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftdownloop
+jmp ._siftdownloop
+
+# qhasm: siftuploop:
+._siftuploop:
+
+# qhasm: pc = pp
+# asm 1: mov  <pp=int64#4,>pc=int64#2
+# asm 2: mov  <pp=%rcx,>pc=%rsi
+mov  %rcx,%rsi
+
+# qhasm: pp -= 1
+# asm 1: sub  $1,<pp=int64#4
+# asm 2: sub  $1,<pp=%rcx
+sub  $1,%rcx
+
+# qhasm: (uint64) pp >>= 1
+# asm 1: shr  $1,<pp=int64#4
+# asm 2: shr  $1,<pp=%rcx
+shr  $1,%rcx
+
+# qhasm: unsigned>? pc - 0 
+# asm 1: cmp  $0,<pc=int64#2
+# asm 2: cmp  $0,<pc=%rsi
+cmp  $0,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if !unsigned>
+jbe ._end
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#2,8),>spc=int64#6
+# asm 2: movq   (<hp=%rdi,<pc=%rsi,8),>spc=%r9
+movq   (%rdi,%rsi,8),%r9
+
+# qhasm: spp <<= 5
+# asm 1: shl  $5,<spp=int64#5
+# asm 2: shl  $5,<spp=%r8
+shl  $5,%r8
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#6
+# asm 2: shl  $5,<spc=%r9
+shl  $5,%r9
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#6
+# asm 2: add  <sp=%rdx,<spc=%r9
+add  %rdx,%r9
+
+# qhasm: spp += sp
+# asm 1: add  <sp=int64#3,<spp=int64#5
+# asm 2: add  <sp=%rdx,<spp=%r8
+add  %rdx,%r8
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#6),>c0=int64#7
+# asm 2: movq   0(<spc=%r9),>c0=%rax
+movq   0(%r9),%rax
+
+# qhasm: c1 = *(uint64 *)(spc +  8)
+# asm 1: movq   8(<spc=int64#6),>c1=int64#8
+# asm 2: movq   8(<spc=%r9),>c1=%r10
+movq   8(%r9),%r10
+
+# qhasm: c2 = *(uint64 *)(spc + 16)
+# asm 1: movq   16(<spc=int64#6),>c2=int64#9
+# asm 2: movq   16(<spc=%r9),>c2=%r11
+movq   16(%r9),%r11
+
+# qhasm: c3 = *(uint64 *)(spc + 24)
+# asm 1: movq   24(<spc=int64#6),>c3=int64#10
+# asm 2: movq   24(<spc=%r9),>c3=%r12
+movq   24(%r9),%r12
+
+# qhasm: carry? c0 -= *(uint64 *)(spp +  0)
+# asm 1: subq 0(<spp=int64#5),<c0=int64#7
+# asm 2: subq 0(<spp=%r8),<c0=%rax
+subq 0(%r8),%rax
+
+# qhasm: carry? c1 -= *(uint64 *)(spp +  8) - carry
+# asm 1: sbbq 8(<spp=int64#5),<c1=int64#8
+# asm 2: sbbq 8(<spp=%r8),<c1=%r10
+sbbq 8(%r8),%r10
+
+# qhasm: carry? c2 -= *(uint64 *)(spp + 16) - carry
+# asm 1: sbbq 16(<spp=int64#5),<c2=int64#9
+# asm 2: sbbq 16(<spp=%r8),<c2=%r11
+sbbq 16(%r8),%r11
+
+# qhasm: carry? c3 -= *(uint64 *)(spp + 24) - carry
+# asm 1: sbbq 24(<spp=int64#5),<c3=int64#10
+# asm 2: sbbq 24(<spp=%r8),<c3=%r12
+sbbq 24(%r8),%r12
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if carry
+jc ._end
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#6
+# asm 2: sub  <sp=%rdx,<spc=%r9
+sub  %rdx,%r9
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#6
+# asm 2: shr  $5,<spc=%r9
+shr  $5,%r9
+
+# qhasm: spp -= sp
+# asm 1: sub  <sp=int64#3,<spp=int64#5
+# asm 2: sub  <sp=%rdx,<spp=%r8
+sub  %rdx,%r8
+
+# qhasm: (uint64) spp >>= 5
+# asm 1: shr  $5,<spp=int64#5
+# asm 2: shr  $5,<spp=%r8
+shr  $5,%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#6,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r9,(<hp=%rdi,<pp=%rcx,8)
+movq  %r9,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#2,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%rsi,8)
+movq  %r8,(%rdi,%rsi,8)
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop
+jmp ._siftuploop
+
+# qhasm: end:
+._end:
+
+# qhasm: caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm: caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm: caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm: caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm: caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm: caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm: caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/heap_rootreplaced_1limb.s

@@ -0,0 +1,416 @@
+
+# qhasm: int64 hp
+
+# qhasm: int64 hlen
+
+# qhasm: int64 sp
+
+# qhasm: int64 pp
+
+# qhasm: input hp
+
+# qhasm: input hlen
+
+# qhasm: input sp
+
+# qhasm: int64 prc
+
+# qhasm: int64 plc
+
+# qhasm: int64 pc
+
+# qhasm: int64 d
+
+# qhasm: int64 spp
+
+# qhasm: int64 sprc
+
+# qhasm: int64 spc
+
+# qhasm: int64 c0
+
+# qhasm: int64 c1
+
+# qhasm: int64 c2
+
+# qhasm: int64 c3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm: int64 p0
+
+# qhasm: int64 p1
+
+# qhasm: int64 p2
+
+# qhasm: int64 p3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_heap_rootreplaced_1limb
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_heap_rootreplaced_1limb
+.globl crypto_sign_ed25519_amd64_64_heap_rootreplaced_1limb
+_crypto_sign_ed25519_amd64_64_heap_rootreplaced_1limb:
+crypto_sign_ed25519_amd64_64_heap_rootreplaced_1limb:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm: caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm: caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm: caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm: caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm: caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm: caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm: caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: pp = 0
+# asm 1: mov  $0,>pp=int64#4
+# asm 2: mov  $0,>pp=%rcx
+mov  $0,%rcx
+
+# qhasm: siftdownloop:
+._siftdownloop:
+
+# qhasm: prc = pp
+# asm 1: mov  <pp=int64#4,>prc=int64#5
+# asm 2: mov  <pp=%rcx,>prc=%r8
+mov  %rcx,%r8
+
+# qhasm: prc *= 2
+# asm 1: imulq  $2,<prc=int64#5,>prc=int64#5
+# asm 2: imulq  $2,<prc=%r8,>prc=%r8
+imulq  $2,%r8,%r8
+
+# qhasm: pc = prc
+# asm 1: mov  <prc=int64#5,>pc=int64#6
+# asm 2: mov  <prc=%r8,>pc=%r9
+mov  %r8,%r9
+
+# qhasm: prc += 2
+# asm 1: add  $2,<prc=int64#5
+# asm 2: add  $2,<prc=%r8
+add  $2,%r8
+
+# qhasm: pc += 1
+# asm 1: add  $1,<pc=int64#6
+# asm 2: add  $1,<pc=%r9
+add  $1,%r9
+
+# qhasm: unsigned>? hlen - prc
+# asm 1: cmp  <prc=int64#5,<hlen=int64#2
+# asm 2: cmp  <prc=%r8,<hlen=%rsi
+cmp  %r8,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop if !unsigned>
+jbe ._siftuploop
+
+# qhasm: sprc = *(uint64 *)(hp + prc * 8)
+# asm 1: movq   (<hp=int64#1,<prc=int64#5,8),>sprc=int64#7
+# asm 2: movq   (<hp=%rdi,<prc=%r8,8),>sprc=%rax
+movq   (%rdi,%r8,8),%rax
+
+# qhasm: sprc <<= 5
+# asm 1: shl  $5,<sprc=int64#7
+# asm 2: shl  $5,<sprc=%rax
+shl  $5,%rax
+
+# qhasm: sprc += sp
+# asm 1: add  <sp=int64#3,<sprc=int64#7
+# asm 2: add  <sp=%rdx,<sprc=%rax
+add  %rdx,%rax
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#6,8),>spc=int64#8
+# asm 2: movq   (<hp=%rdi,<pc=%r9,8),>spc=%r10
+movq   (%rdi,%r9,8),%r10
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#8
+# asm 2: shl  $5,<spc=%r10
+shl  $5,%r10
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#8
+# asm 2: add  <sp=%rdx,<spc=%r10
+add  %rdx,%r10
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#8),>c0=int64#9
+# asm 2: movq   0(<spc=%r10),>c0=%r11
+movq   0(%r10),%r11
+
+# qhasm: carry? c0 -= *(uint64 *)(sprc +  0)
+# asm 1: subq 0(<sprc=int64#7),<c0=int64#9
+# asm 2: subq 0(<sprc=%rax),<c0=%r11
+subq 0(%rax),%r11
+
+# qhasm: pc = prc if carry
+# asm 1: cmovc <prc=int64#5,<pc=int64#6
+# asm 2: cmovc <prc=%r8,<pc=%r9
+cmovc %r8,%r9
+
+# qhasm: spc = sprc if carry
+# asm 1: cmovc <sprc=int64#7,<spc=int64#8
+# asm 2: cmovc <sprc=%rax,<spc=%r10
+cmovc %rax,%r10
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#8
+# asm 2: sub  <sp=%rdx,<spc=%r10
+sub  %rdx,%r10
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#8
+# asm 2: shr  $5,<spc=%r10
+shr  $5,%r10
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#8,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r10,(<hp=%rdi,<pp=%rcx,8)
+movq  %r10,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#6,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%r9,8)
+movq  %r8,(%rdi,%r9,8)
+
+# qhasm: pp = pc
+# asm 1: mov  <pc=int64#6,>pp=int64#4
+# asm 2: mov  <pc=%r9,>pp=%rcx
+mov  %r9,%rcx
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftdownloop
+jmp ._siftdownloop
+
+# qhasm: siftuploop:
+._siftuploop:
+
+# qhasm: pc = pp
+# asm 1: mov  <pp=int64#4,>pc=int64#2
+# asm 2: mov  <pp=%rcx,>pc=%rsi
+mov  %rcx,%rsi
+
+# qhasm: pp -= 1
+# asm 1: sub  $1,<pp=int64#4
+# asm 2: sub  $1,<pp=%rcx
+sub  $1,%rcx
+
+# qhasm: (uint64) pp >>= 1
+# asm 1: shr  $1,<pp=int64#4
+# asm 2: shr  $1,<pp=%rcx
+shr  $1,%rcx
+
+# qhasm: unsigned>? pc - 0 
+# asm 1: cmp  $0,<pc=int64#2
+# asm 2: cmp  $0,<pc=%rsi
+cmp  $0,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if !unsigned>
+jbe ._end
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#2,8),>spc=int64#6
+# asm 2: movq   (<hp=%rdi,<pc=%rsi,8),>spc=%r9
+movq   (%rdi,%rsi,8),%r9
+
+# qhasm: spp <<= 5
+# asm 1: shl  $5,<spp=int64#5
+# asm 2: shl  $5,<spp=%r8
+shl  $5,%r8
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#6
+# asm 2: shl  $5,<spc=%r9
+shl  $5,%r9
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#6
+# asm 2: add  <sp=%rdx,<spc=%r9
+add  %rdx,%r9
+
+# qhasm: spp += sp
+# asm 1: add  <sp=int64#3,<spp=int64#5
+# asm 2: add  <sp=%rdx,<spp=%r8
+add  %rdx,%r8
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#6),>c0=int64#7
+# asm 2: movq   0(<spc=%r9),>c0=%rax
+movq   0(%r9),%rax
+
+# qhasm: carry? c0 -= *(uint64 *)(spp +  0)
+# asm 1: subq 0(<spp=int64#5),<c0=int64#7
+# asm 2: subq 0(<spp=%r8),<c0=%rax
+subq 0(%r8),%rax
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if carry
+jc ._end
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#6
+# asm 2: sub  <sp=%rdx,<spc=%r9
+sub  %rdx,%r9
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#6
+# asm 2: shr  $5,<spc=%r9
+shr  $5,%r9
+
+# qhasm: spp -= sp
+# asm 1: sub  <sp=int64#3,<spp=int64#5
+# asm 2: sub  <sp=%rdx,<spp=%r8
+sub  %rdx,%r8
+
+# qhasm: (uint64) spp >>= 5
+# asm 1: shr  $5,<spp=int64#5
+# asm 2: shr  $5,<spp=%r8
+shr  $5,%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#6,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r9,(<hp=%rdi,<pp=%rcx,8)
+movq  %r9,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#2,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%rsi,8)
+movq  %r8,(%rdi,%rsi,8)
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop
+jmp ._siftuploop
+
+# qhasm: end:
+._end:
+
+# qhasm: caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm: caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm: caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm: caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm: caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm: caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm: caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/heap_rootreplaced_2limbs.s

@@ -0,0 +1,436 @@
+
+# qhasm: int64 hp
+
+# qhasm: int64 hlen
+
+# qhasm: int64 sp
+
+# qhasm: int64 pp
+
+# qhasm: input hp
+
+# qhasm: input hlen
+
+# qhasm: input sp
+
+# qhasm: int64 prc
+
+# qhasm: int64 plc
+
+# qhasm: int64 pc
+
+# qhasm: int64 d
+
+# qhasm: int64 spp
+
+# qhasm: int64 sprc
+
+# qhasm: int64 spc
+
+# qhasm: int64 c0
+
+# qhasm: int64 c1
+
+# qhasm: int64 c2
+
+# qhasm: int64 c3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm: int64 p0
+
+# qhasm: int64 p1
+
+# qhasm: int64 p2
+
+# qhasm: int64 p3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_heap_rootreplaced_2limbs
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_heap_rootreplaced_2limbs
+.globl crypto_sign_ed25519_amd64_64_heap_rootreplaced_2limbs
+_crypto_sign_ed25519_amd64_64_heap_rootreplaced_2limbs:
+crypto_sign_ed25519_amd64_64_heap_rootreplaced_2limbs:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm: caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm: caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm: caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm: caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm: caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm: caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm: caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: pp = 0
+# asm 1: mov  $0,>pp=int64#4
+# asm 2: mov  $0,>pp=%rcx
+mov  $0,%rcx
+
+# qhasm: siftdownloop:
+._siftdownloop:
+
+# qhasm: prc = pp
+# asm 1: mov  <pp=int64#4,>prc=int64#5
+# asm 2: mov  <pp=%rcx,>prc=%r8
+mov  %rcx,%r8
+
+# qhasm: prc *= 2
+# asm 1: imulq  $2,<prc=int64#5,>prc=int64#5
+# asm 2: imulq  $2,<prc=%r8,>prc=%r8
+imulq  $2,%r8,%r8
+
+# qhasm: pc = prc
+# asm 1: mov  <prc=int64#5,>pc=int64#6
+# asm 2: mov  <prc=%r8,>pc=%r9
+mov  %r8,%r9
+
+# qhasm: prc += 2
+# asm 1: add  $2,<prc=int64#5
+# asm 2: add  $2,<prc=%r8
+add  $2,%r8
+
+# qhasm: pc += 1
+# asm 1: add  $1,<pc=int64#6
+# asm 2: add  $1,<pc=%r9
+add  $1,%r9
+
+# qhasm: unsigned>? hlen - prc
+# asm 1: cmp  <prc=int64#5,<hlen=int64#2
+# asm 2: cmp  <prc=%r8,<hlen=%rsi
+cmp  %r8,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop if !unsigned>
+jbe ._siftuploop
+
+# qhasm: sprc = *(uint64 *)(hp + prc * 8)
+# asm 1: movq   (<hp=int64#1,<prc=int64#5,8),>sprc=int64#7
+# asm 2: movq   (<hp=%rdi,<prc=%r8,8),>sprc=%rax
+movq   (%rdi,%r8,8),%rax
+
+# qhasm: sprc <<= 5
+# asm 1: shl  $5,<sprc=int64#7
+# asm 2: shl  $5,<sprc=%rax
+shl  $5,%rax
+
+# qhasm: sprc += sp
+# asm 1: add  <sp=int64#3,<sprc=int64#7
+# asm 2: add  <sp=%rdx,<sprc=%rax
+add  %rdx,%rax
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#6,8),>spc=int64#8
+# asm 2: movq   (<hp=%rdi,<pc=%r9,8),>spc=%r10
+movq   (%rdi,%r9,8),%r10
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#8
+# asm 2: shl  $5,<spc=%r10
+shl  $5,%r10
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#8
+# asm 2: add  <sp=%rdx,<spc=%r10
+add  %rdx,%r10
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#8),>c0=int64#9
+# asm 2: movq   0(<spc=%r10),>c0=%r11
+movq   0(%r10),%r11
+
+# qhasm: c1 = *(uint64 *)(spc +  8)
+# asm 1: movq   8(<spc=int64#8),>c1=int64#10
+# asm 2: movq   8(<spc=%r10),>c1=%r12
+movq   8(%r10),%r12
+
+# qhasm: carry? c0 -= *(uint64 *)(sprc +  0)
+# asm 1: subq 0(<sprc=int64#7),<c0=int64#9
+# asm 2: subq 0(<sprc=%rax),<c0=%r11
+subq 0(%rax),%r11
+
+# qhasm: carry? c1 -= *(uint64 *)(sprc +  8) - carry
+# asm 1: sbbq 8(<sprc=int64#7),<c1=int64#10
+# asm 2: sbbq 8(<sprc=%rax),<c1=%r12
+sbbq 8(%rax),%r12
+
+# qhasm: pc = prc if carry
+# asm 1: cmovc <prc=int64#5,<pc=int64#6
+# asm 2: cmovc <prc=%r8,<pc=%r9
+cmovc %r8,%r9
+
+# qhasm: spc = sprc if carry
+# asm 1: cmovc <sprc=int64#7,<spc=int64#8
+# asm 2: cmovc <sprc=%rax,<spc=%r10
+cmovc %rax,%r10
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#8
+# asm 2: sub  <sp=%rdx,<spc=%r10
+sub  %rdx,%r10
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#8
+# asm 2: shr  $5,<spc=%r10
+shr  $5,%r10
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#8,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r10,(<hp=%rdi,<pp=%rcx,8)
+movq  %r10,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#6,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%r9,8)
+movq  %r8,(%rdi,%r9,8)
+
+# qhasm: pp = pc
+# asm 1: mov  <pc=int64#6,>pp=int64#4
+# asm 2: mov  <pc=%r9,>pp=%rcx
+mov  %r9,%rcx
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftdownloop
+jmp ._siftdownloop
+
+# qhasm: siftuploop:
+._siftuploop:
+
+# qhasm: pc = pp
+# asm 1: mov  <pp=int64#4,>pc=int64#2
+# asm 2: mov  <pp=%rcx,>pc=%rsi
+mov  %rcx,%rsi
+
+# qhasm: pp -= 1
+# asm 1: sub  $1,<pp=int64#4
+# asm 2: sub  $1,<pp=%rcx
+sub  $1,%rcx
+
+# qhasm: (uint64) pp >>= 1
+# asm 1: shr  $1,<pp=int64#4
+# asm 2: shr  $1,<pp=%rcx
+shr  $1,%rcx
+
+# qhasm: unsigned>? pc - 0 
+# asm 1: cmp  $0,<pc=int64#2
+# asm 2: cmp  $0,<pc=%rsi
+cmp  $0,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if !unsigned>
+jbe ._end
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#2,8),>spc=int64#6
+# asm 2: movq   (<hp=%rdi,<pc=%rsi,8),>spc=%r9
+movq   (%rdi,%rsi,8),%r9
+
+# qhasm: spp <<= 5
+# asm 1: shl  $5,<spp=int64#5
+# asm 2: shl  $5,<spp=%r8
+shl  $5,%r8
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#6
+# asm 2: shl  $5,<spc=%r9
+shl  $5,%r9
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#6
+# asm 2: add  <sp=%rdx,<spc=%r9
+add  %rdx,%r9
+
+# qhasm: spp += sp
+# asm 1: add  <sp=int64#3,<spp=int64#5
+# asm 2: add  <sp=%rdx,<spp=%r8
+add  %rdx,%r8
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#6),>c0=int64#7
+# asm 2: movq   0(<spc=%r9),>c0=%rax
+movq   0(%r9),%rax
+
+# qhasm: c1 = *(uint64 *)(spc +  8)
+# asm 1: movq   8(<spc=int64#6),>c1=int64#8
+# asm 2: movq   8(<spc=%r9),>c1=%r10
+movq   8(%r9),%r10
+
+# qhasm: carry? c0 -= *(uint64 *)(spp +  0)
+# asm 1: subq 0(<spp=int64#5),<c0=int64#7
+# asm 2: subq 0(<spp=%r8),<c0=%rax
+subq 0(%r8),%rax
+
+# qhasm: carry? c1 -= *(uint64 *)(spp +  8) - carry
+# asm 1: sbbq 8(<spp=int64#5),<c1=int64#8
+# asm 2: sbbq 8(<spp=%r8),<c1=%r10
+sbbq 8(%r8),%r10
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if carry
+jc ._end
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#6
+# asm 2: sub  <sp=%rdx,<spc=%r9
+sub  %rdx,%r9
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#6
+# asm 2: shr  $5,<spc=%r9
+shr  $5,%r9
+
+# qhasm: spp -= sp
+# asm 1: sub  <sp=int64#3,<spp=int64#5
+# asm 2: sub  <sp=%rdx,<spp=%r8
+sub  %rdx,%r8
+
+# qhasm: (uint64) spp >>= 5
+# asm 1: shr  $5,<spp=int64#5
+# asm 2: shr  $5,<spp=%r8
+shr  $5,%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#6,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r9,(<hp=%rdi,<pp=%rcx,8)
+movq  %r9,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#2,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%rsi,8)
+movq  %r8,(%rdi,%rsi,8)
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop
+jmp ._siftuploop
+
+# qhasm: end:
+._end:
+
+# qhasm: caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm: caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm: caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm: caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm: caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm: caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm: caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/heap_rootreplaced_3limbs.s

@@ -0,0 +1,456 @@
+
+# qhasm: int64 hp
+
+# qhasm: int64 hlen
+
+# qhasm: int64 sp
+
+# qhasm: int64 pp
+
+# qhasm: input hp
+
+# qhasm: input hlen
+
+# qhasm: input sp
+
+# qhasm: int64 prc
+
+# qhasm: int64 plc
+
+# qhasm: int64 pc
+
+# qhasm: int64 d
+
+# qhasm: int64 spp
+
+# qhasm: int64 sprc
+
+# qhasm: int64 spc
+
+# qhasm: int64 c0
+
+# qhasm: int64 c1
+
+# qhasm: int64 c2
+
+# qhasm: int64 c3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm: int64 p0
+
+# qhasm: int64 p1
+
+# qhasm: int64 p2
+
+# qhasm: int64 p3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_heap_rootreplaced_3limbs
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_heap_rootreplaced_3limbs
+.globl crypto_sign_ed25519_amd64_64_heap_rootreplaced_3limbs
+_crypto_sign_ed25519_amd64_64_heap_rootreplaced_3limbs:
+crypto_sign_ed25519_amd64_64_heap_rootreplaced_3limbs:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm: caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm: caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm: caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm: caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm: caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm: caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm: caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: pp = 0
+# asm 1: mov  $0,>pp=int64#4
+# asm 2: mov  $0,>pp=%rcx
+mov  $0,%rcx
+
+# qhasm: siftdownloop:
+._siftdownloop:
+
+# qhasm: prc = pp
+# asm 1: mov  <pp=int64#4,>prc=int64#5
+# asm 2: mov  <pp=%rcx,>prc=%r8
+mov  %rcx,%r8
+
+# qhasm: prc *= 2
+# asm 1: imulq  $2,<prc=int64#5,>prc=int64#5
+# asm 2: imulq  $2,<prc=%r8,>prc=%r8
+imulq  $2,%r8,%r8
+
+# qhasm: pc = prc
+# asm 1: mov  <prc=int64#5,>pc=int64#6
+# asm 2: mov  <prc=%r8,>pc=%r9
+mov  %r8,%r9
+
+# qhasm: prc += 2
+# asm 1: add  $2,<prc=int64#5
+# asm 2: add  $2,<prc=%r8
+add  $2,%r8
+
+# qhasm: pc += 1
+# asm 1: add  $1,<pc=int64#6
+# asm 2: add  $1,<pc=%r9
+add  $1,%r9
+
+# qhasm: unsigned>? hlen - prc
+# asm 1: cmp  <prc=int64#5,<hlen=int64#2
+# asm 2: cmp  <prc=%r8,<hlen=%rsi
+cmp  %r8,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop if !unsigned>
+jbe ._siftuploop
+
+# qhasm: sprc = *(uint64 *)(hp + prc * 8)
+# asm 1: movq   (<hp=int64#1,<prc=int64#5,8),>sprc=int64#7
+# asm 2: movq   (<hp=%rdi,<prc=%r8,8),>sprc=%rax
+movq   (%rdi,%r8,8),%rax
+
+# qhasm: sprc <<= 5
+# asm 1: shl  $5,<sprc=int64#7
+# asm 2: shl  $5,<sprc=%rax
+shl  $5,%rax
+
+# qhasm: sprc += sp
+# asm 1: add  <sp=int64#3,<sprc=int64#7
+# asm 2: add  <sp=%rdx,<sprc=%rax
+add  %rdx,%rax
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#6,8),>spc=int64#8
+# asm 2: movq   (<hp=%rdi,<pc=%r9,8),>spc=%r10
+movq   (%rdi,%r9,8),%r10
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#8
+# asm 2: shl  $5,<spc=%r10
+shl  $5,%r10
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#8
+# asm 2: add  <sp=%rdx,<spc=%r10
+add  %rdx,%r10
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#8),>c0=int64#9
+# asm 2: movq   0(<spc=%r10),>c0=%r11
+movq   0(%r10),%r11
+
+# qhasm: c1 = *(uint64 *)(spc +  8)
+# asm 1: movq   8(<spc=int64#8),>c1=int64#10
+# asm 2: movq   8(<spc=%r10),>c1=%r12
+movq   8(%r10),%r12
+
+# qhasm: c2 = *(uint64 *)(spc + 16)
+# asm 1: movq   16(<spc=int64#8),>c2=int64#11
+# asm 2: movq   16(<spc=%r10),>c2=%r13
+movq   16(%r10),%r13
+
+# qhasm: carry? c0 -= *(uint64 *)(sprc +  0)
+# asm 1: subq 0(<sprc=int64#7),<c0=int64#9
+# asm 2: subq 0(<sprc=%rax),<c0=%r11
+subq 0(%rax),%r11
+
+# qhasm: carry? c1 -= *(uint64 *)(sprc +  8) - carry
+# asm 1: sbbq 8(<sprc=int64#7),<c1=int64#10
+# asm 2: sbbq 8(<sprc=%rax),<c1=%r12
+sbbq 8(%rax),%r12
+
+# qhasm: carry? c2 -= *(uint64 *)(sprc + 16) - carry
+# asm 1: sbbq 16(<sprc=int64#7),<c2=int64#11
+# asm 2: sbbq 16(<sprc=%rax),<c2=%r13
+sbbq 16(%rax),%r13
+
+# qhasm: pc = prc if carry
+# asm 1: cmovc <prc=int64#5,<pc=int64#6
+# asm 2: cmovc <prc=%r8,<pc=%r9
+cmovc %r8,%r9
+
+# qhasm: spc = sprc if carry
+# asm 1: cmovc <sprc=int64#7,<spc=int64#8
+# asm 2: cmovc <sprc=%rax,<spc=%r10
+cmovc %rax,%r10
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#8
+# asm 2: sub  <sp=%rdx,<spc=%r10
+sub  %rdx,%r10
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#8
+# asm 2: shr  $5,<spc=%r10
+shr  $5,%r10
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#8,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r10,(<hp=%rdi,<pp=%rcx,8)
+movq  %r10,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#6,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%r9,8)
+movq  %r8,(%rdi,%r9,8)
+
+# qhasm: pp = pc
+# asm 1: mov  <pc=int64#6,>pp=int64#4
+# asm 2: mov  <pc=%r9,>pp=%rcx
+mov  %r9,%rcx
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftdownloop
+jmp ._siftdownloop
+
+# qhasm: siftuploop:
+._siftuploop:
+
+# qhasm: pc = pp
+# asm 1: mov  <pp=int64#4,>pc=int64#2
+# asm 2: mov  <pp=%rcx,>pc=%rsi
+mov  %rcx,%rsi
+
+# qhasm: pp -= 1
+# asm 1: sub  $1,<pp=int64#4
+# asm 2: sub  $1,<pp=%rcx
+sub  $1,%rcx
+
+# qhasm: (uint64) pp >>= 1
+# asm 1: shr  $1,<pp=int64#4
+# asm 2: shr  $1,<pp=%rcx
+shr  $1,%rcx
+
+# qhasm: unsigned>? pc - 0 
+# asm 1: cmp  $0,<pc=int64#2
+# asm 2: cmp  $0,<pc=%rsi
+cmp  $0,%rsi
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if !unsigned>
+jbe ._end
+
+# qhasm: spp = *(uint64 *)(hp + pp * 8)
+# asm 1: movq   (<hp=int64#1,<pp=int64#4,8),>spp=int64#5
+# asm 2: movq   (<hp=%rdi,<pp=%rcx,8),>spp=%r8
+movq   (%rdi,%rcx,8),%r8
+
+# qhasm: spc = *(uint64 *)(hp + pc * 8)
+# asm 1: movq   (<hp=int64#1,<pc=int64#2,8),>spc=int64#6
+# asm 2: movq   (<hp=%rdi,<pc=%rsi,8),>spc=%r9
+movq   (%rdi,%rsi,8),%r9
+
+# qhasm: spp <<= 5
+# asm 1: shl  $5,<spp=int64#5
+# asm 2: shl  $5,<spp=%r8
+shl  $5,%r8
+
+# qhasm: spc <<= 5
+# asm 1: shl  $5,<spc=int64#6
+# asm 2: shl  $5,<spc=%r9
+shl  $5,%r9
+
+# qhasm: spc += sp
+# asm 1: add  <sp=int64#3,<spc=int64#6
+# asm 2: add  <sp=%rdx,<spc=%r9
+add  %rdx,%r9
+
+# qhasm: spp += sp
+# asm 1: add  <sp=int64#3,<spp=int64#5
+# asm 2: add  <sp=%rdx,<spp=%r8
+add  %rdx,%r8
+
+# qhasm: c0 = *(uint64 *)(spc +  0)
+# asm 1: movq   0(<spc=int64#6),>c0=int64#7
+# asm 2: movq   0(<spc=%r9),>c0=%rax
+movq   0(%r9),%rax
+
+# qhasm: c1 = *(uint64 *)(spc +  8)
+# asm 1: movq   8(<spc=int64#6),>c1=int64#8
+# asm 2: movq   8(<spc=%r9),>c1=%r10
+movq   8(%r9),%r10
+
+# qhasm: c2 = *(uint64 *)(spc + 16)
+# asm 1: movq   16(<spc=int64#6),>c2=int64#9
+# asm 2: movq   16(<spc=%r9),>c2=%r11
+movq   16(%r9),%r11
+
+# qhasm: carry? c0 -= *(uint64 *)(spp +  0)
+# asm 1: subq 0(<spp=int64#5),<c0=int64#7
+# asm 2: subq 0(<spp=%r8),<c0=%rax
+subq 0(%r8),%rax
+
+# qhasm: carry? c1 -= *(uint64 *)(spp +  8) - carry
+# asm 1: sbbq 8(<spp=int64#5),<c1=int64#8
+# asm 2: sbbq 8(<spp=%r8),<c1=%r10
+sbbq 8(%r8),%r10
+
+# qhasm: carry? c2 -= *(uint64 *)(spp + 16) - carry
+# asm 1: sbbq 16(<spp=int64#5),<c2=int64#9
+# asm 2: sbbq 16(<spp=%r8),<c2=%r11
+sbbq 16(%r8),%r11
+# comment:fp stack unchanged by jump
+
+# qhasm: goto end if carry
+jc ._end
+
+# qhasm: spc -= sp
+# asm 1: sub  <sp=int64#3,<spc=int64#6
+# asm 2: sub  <sp=%rdx,<spc=%r9
+sub  %rdx,%r9
+
+# qhasm: (uint64) spc >>= 5
+# asm 1: shr  $5,<spc=int64#6
+# asm 2: shr  $5,<spc=%r9
+shr  $5,%r9
+
+# qhasm: spp -= sp
+# asm 1: sub  <sp=int64#3,<spp=int64#5
+# asm 2: sub  <sp=%rdx,<spp=%r8
+sub  %rdx,%r8
+
+# qhasm: (uint64) spp >>= 5
+# asm 1: shr  $5,<spp=int64#5
+# asm 2: shr  $5,<spp=%r8
+shr  $5,%r8
+
+# qhasm: *(uint64 *)(hp + pp * 8) = spc
+# asm 1: movq  <spc=int64#6,(<hp=int64#1,<pp=int64#4,8)
+# asm 2: movq  <spc=%r9,(<hp=%rdi,<pp=%rcx,8)
+movq  %r9,(%rdi,%rcx,8)
+
+# qhasm: *(uint64 *)(hp + pc * 8) = spp
+# asm 1: movq  <spp=int64#5,(<hp=int64#1,<pc=int64#2,8)
+# asm 2: movq  <spp=%r8,(<hp=%rdi,<pc=%rsi,8)
+movq  %r8,(%rdi,%rsi,8)
+# comment:fp stack unchanged by jump
+
+# qhasm: goto siftuploop
+jmp ._siftuploop
+
+# qhasm: end:
+._end:
+
+# qhasm: caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm: caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm: caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm: caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm: caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm: caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm: caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/hram.c

@@ -0,0 +1,16 @@
+/*#include "crypto_hash_sha512.h"*/
+#include "hram.h"
+
+extern void ZT_sha512internal(void *digest,const void *data,unsigned int len);
+
+void get_hram(unsigned char *hram, const unsigned char *sm, const unsigned char *pk, unsigned char *playground, unsigned long long smlen)
+{
+  unsigned long long i;
+
+  for (i =  0;i < 32;++i)    playground[i] = sm[i];
+  for (i = 32;i < 64;++i)    playground[i] = pk[i-32];
+  for (i = 64;i < smlen;++i) playground[i] = sm[i];
+
+  /*crypto_hash_sha512(hram,playground,smlen);*/
+  ZT_sha512internal(hram,playground,smlen);
+}

ext/ed25519-amd64-asm/hram.h

@@ -0,0 +1,8 @@
+#ifndef HRAM_H
+#define HRAM_H
+
+#define get_hram crypto_sign_ed25519_amd64_64_get_hram
+
+extern void get_hram(unsigned char *hram, const unsigned char *sm, const unsigned char *pk, unsigned char *playground, unsigned long long smlen);
+
+#endif

ext/ed25519-amd64-asm/implementors

@@ -0,0 +1,5 @@
+Daniel J. Bernstein
+Niels Duif
+Tanja Lange
+lead: Peter Schwabe
+Bo-Yin Yang

ext/ed25519-amd64-asm/index_heap.c

@@ -0,0 +1,58 @@
+#include "sc25519.h"
+#include "index_heap.h"
+
+/* caller's responsibility to ensure hlen>=3 */
+void heap_init(unsigned long long *h, unsigned long long hlen, sc25519 *scalars)
+{
+  h[0] = 0;
+  unsigned long long i=1;
+  while(i<hlen)
+    heap_push(h, &i, i, scalars);
+}
+
+void heap_extend(unsigned long long *h, unsigned long long oldlen, unsigned long long newlen, sc25519 *scalars)
+{
+  unsigned long long i=oldlen;
+  while(i<newlen)
+    heap_push(h, &i, i, scalars);
+}
+
+
+void heap_push(unsigned long long *h, unsigned long long *hlen, unsigned long long elem, sc25519 *scalars)
+{
+  /* Move up towards the root */
+  /* XXX: Check size of hlen, whether cast to signed value is ok */
+  signed long long pos = *hlen;
+  signed long long ppos = (pos-1)/2;
+  unsigned long long t;
+  h[*hlen] = elem;
+  while(pos > 0)
+  {
+    /* if(sc25519_lt_vartime(&scalars[h[ppos]], &scalars[h[pos]])) */
+    if(sc25519_lt(&scalars[h[ppos]], &scalars[h[pos]]))
+    {
+      t = h[ppos];
+      h[ppos] = h[pos];
+      h[pos] = t;
+      pos = ppos;
+      ppos = (pos-1)/2;
+    }
+    else break;
+  } 
+  (*hlen)++;
+}
+
+/* Put the largest value in the heap in max1, the second largest in max2 */
+void heap_get2max(unsigned long long *h, unsigned long long *max1, unsigned long long *max2, sc25519 *scalars)
+{
+  *max1 = h[0];
+  *max2 = h[1];
+  if(sc25519_lt(&scalars[h[1]],&scalars[h[2]]))
+    *max2 = h[2];
+}
+
+/* After the root has been replaced, restore heap property */
+/* extern void heap_rootreplaced(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+*/
+/* extern void heap_rootreplaced_shortscalars(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+*/

ext/ed25519-amd64-asm/index_heap.h

@@ -0,0 +1,31 @@
+#ifndef INDEX_HEAP_H
+#define INDEX_HEAP_H
+
+#include "sc25519.h"
+
+#define heap_init                crypto_sign_ed25519_amd64_64_heap_init
+#define heap_extend              crypto_sign_ed25519_amd64_64_heap_extend
+#define heap_pop                 crypto_sign_ed25519_amd64_64_heap_pop
+#define heap_push                crypto_sign_ed25519_amd64_64_heap_push
+#define heap_get2max             crypto_sign_ed25519_amd64_64_heap_get2max
+#define heap_rootreplaced        crypto_sign_ed25519_amd64_64_heap_rootreplaced
+#define heap_rootreplaced_3limbs crypto_sign_ed25519_amd64_64_heap_rootreplaced_3limbs
+#define heap_rootreplaced_2limbs crypto_sign_ed25519_amd64_64_heap_rootreplaced_2limbs
+#define heap_rootreplaced_1limb  crypto_sign_ed25519_amd64_64_heap_rootreplaced_1limb
+
+void heap_init(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+
+void heap_extend(unsigned long long *h, unsigned long long oldlen, unsigned long long newlen, sc25519 *scalars);
+
+unsigned long long heap_pop(unsigned long long *h, unsigned long long *hlen, sc25519 *scalars);
+
+void heap_push(unsigned long long *h, unsigned long long *hlen, unsigned long long elem, sc25519 *scalars);
+
+void heap_get2max(unsigned long long *h, unsigned long long *max1, unsigned long long *max2, sc25519 *scalars);
+
+void heap_rootreplaced(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+void heap_rootreplaced_3limbs(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+void heap_rootreplaced_2limbs(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+void heap_rootreplaced_1limb(unsigned long long *h, unsigned long long hlen, sc25519 *scalars);
+
+#endif

ext/ed25519-amd64-asm/keypair.c

@@ -0,0 +1,25 @@
+#include <string.h>
+#include "crypto_sign.h"
+#include "crypto_hash_sha512.h"
+#include "randombytes.h"
+#include "ge25519.h"
+
+int crypto_sign_keypair(unsigned char *pk,unsigned char *sk)
+{
+  unsigned char az[64];
+  sc25519 scsk;
+  ge25519 gepk;
+
+  randombytes(sk,32);
+  crypto_hash_sha512(az,sk,32);
+  az[0] &= 248;
+  az[31] &= 127;
+  az[31] |= 64;
+
+  sc25519_from32bytes(&scsk,az);
+  
+  ge25519_scalarmult_base(&gepk, &scsk);
+  ge25519_pack(pk, &gepk);
+  memmove(sk + 32,pk,32);
+  return 0;
+}

ext/ed25519-amd64-asm/open.c

@@ -0,0 +1,49 @@
+#include <string.h>
+#include "crypto_sign.h"
+#include "crypto_verify_32.h"
+#include "crypto_hash_sha512.h"
+#include "ge25519.h"
+
+int crypto_sign_open(
+    unsigned char *m,unsigned long long *mlen,
+    const unsigned char *sm,unsigned long long smlen,
+    const unsigned char *pk
+    )
+{
+  unsigned char pkcopy[32];
+  unsigned char rcopy[32];
+  unsigned char hram[64];
+  unsigned char rcheck[32];
+  ge25519 get1, get2;
+  sc25519 schram, scs;
+
+  if (smlen < 64) goto badsig;
+  if (sm[63] & 224) goto badsig;
+  if (ge25519_unpackneg_vartime(&get1,pk)) goto badsig;
+
+  memmove(pkcopy,pk,32);
+  memmove(rcopy,sm,32);
+
+  sc25519_from32bytes(&scs, sm+32);
+
+  memmove(m,sm,smlen);
+  memmove(m + 32,pkcopy,32);
+  crypto_hash_sha512(hram,m,smlen);
+
+  sc25519_from64bytes(&schram, hram);
+
+  ge25519_double_scalarmult_vartime(&get2, &get1, &schram, &scs);
+  ge25519_pack(rcheck, &get2);
+
+  if (crypto_verify_32(rcopy,rcheck) == 0) {
+    memmove(m,m + 64,smlen - 64);
+    memset(m + smlen - 64,0,64);
+    *mlen = smlen - 64;
+    return 0;
+  }
+
+badsig:
+  *mlen = (unsigned long long) -1;
+  memset(m,0,smlen);
+  return -1;
+}

ext/ed25519-amd64-asm/sc25519.h

@@ -0,0 +1,66 @@
+#ifndef SC25519_H
+#define SC25519_H
+
+#define sc25519                  crypto_sign_ed25519_amd64_64_sc25519
+#define shortsc25519             crypto_sign_ed25519_amd64_64_shortsc25519
+#define sc25519_from32bytes      crypto_sign_ed25519_amd64_64_sc25519_from32bytes
+#define shortsc25519_from16bytes crypto_sign_ed25519_amd64_64_shortsc25519_from16bytes
+#define sc25519_from64bytes      crypto_sign_ed25519_amd64_64_sc25519_from64bytes
+#define sc25519_from_shortsc     crypto_sign_ed25519_amd64_64_sc25519_from_shortsc
+#define sc25519_to32bytes        crypto_sign_ed25519_amd64_64_sc25519_to32bytes
+#define sc25519_iszero_vartime   crypto_sign_ed25519_amd64_64_sc25519_iszero_vartime
+#define sc25519_isshort_vartime  crypto_sign_ed25519_amd64_64_sc25519_isshort_vartime
+#define sc25519_lt               crypto_sign_ed25519_amd64_64_sc25519_lt
+#define sc25519_add              crypto_sign_ed25519_amd64_64_sc25519_add
+#define sc25519_sub_nored        crypto_sign_ed25519_amd64_64_sc25519_sub_nored
+#define sc25519_mul              crypto_sign_ed25519_amd64_64_sc25519_mul
+#define sc25519_mul_shortsc      crypto_sign_ed25519_amd64_64_sc25519_mul_shortsc
+#define sc25519_window4          crypto_sign_ed25519_amd64_64_sc25519_window4
+#define sc25519_slide          crypto_sign_ed25519_amd64_64_sc25519_slide
+#define sc25519_2interleave2     crypto_sign_ed25519_amd64_64_sc25519_2interleave2
+#define sc25519_barrett crypto_sign_ed25519_amd64_64_sc25519_barrett
+
+typedef struct 
+{
+  unsigned long long v[4]; 
+}
+sc25519;
+
+typedef struct 
+{
+  unsigned long long v[2]; 
+}
+shortsc25519;
+
+void sc25519_from32bytes(sc25519 *r, const unsigned char x[32]);
+
+void sc25519_from64bytes(sc25519 *r, const unsigned char x[64]);
+
+void sc25519_from_shortsc(sc25519 *r, const shortsc25519 *x);
+
+void sc25519_to32bytes(unsigned char r[32], const sc25519 *x);
+
+int sc25519_iszero_vartime(const sc25519 *x);
+
+int sc25519_lt(const sc25519 *x, const sc25519 *y);
+
+void sc25519_add(sc25519 *r, const sc25519 *x, const sc25519 *y);
+
+void sc25519_sub_nored(sc25519 *r, const sc25519 *x, const sc25519 *y);
+
+void sc25519_mul(sc25519 *r, const sc25519 *x, const sc25519 *y);
+
+void sc25519_mul_shortsc(sc25519 *r, const sc25519 *x, const shortsc25519 *y);
+
+/* Convert s into a representation of the form \sum_{i=0}^{63}r[i]2^(4*i)
+ * with r[i] in {-8,...,7}
+ */
+void sc25519_window4(signed char r[85], const sc25519 *s);
+
+void sc25519_slide(signed char r[256], const sc25519 *s, int swindowsize);
+
+void sc25519_2interleave2(unsigned char r[127], const sc25519 *s1, const sc25519 *s2);
+
+void sc25519_barrett(sc25519 *r, unsigned long long x[8]);
+
+#endif

ext/ed25519-amd64-asm/sc25519_add.s

@@ -0,0 +1,232 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_sc25519_add
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_sc25519_add
+.globl crypto_sign_ed25519_amd64_64_sc25519_add
+_crypto_sign_ed25519_amd64_64_sc25519_add:
+crypto_sign_ed25519_amd64_64_sc25519_add:
+mov %rsp,%r11
+and $31,%r11
+add $32,%r11
+sub %r11,%rsp
+
+# qhasm: caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#1
+# asm 2: movq <caller4=%r14,>caller4_stack=0(%rsp)
+movq %r14,0(%rsp)
+
+# qhasm: caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#2
+# asm 2: movq <caller5=%r15,>caller5_stack=8(%rsp)
+movq %r15,8(%rsp)
+
+# qhasm: caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#3
+# asm 2: movq <caller6=%rbx,>caller6_stack=16(%rsp)
+movq %rbx,16(%rsp)
+
+# qhasm: r0 = *(uint64 *)(xp +  0)
+# asm 1: movq   0(<xp=int64#2),>r0=int64#4
+# asm 2: movq   0(<xp=%rsi),>r0=%rcx
+movq   0(%rsi),%rcx
+
+# qhasm: r1 = *(uint64 *)(xp +  8)
+# asm 1: movq   8(<xp=int64#2),>r1=int64#5
+# asm 2: movq   8(<xp=%rsi),>r1=%r8
+movq   8(%rsi),%r8
+
+# qhasm: r2 = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>r2=int64#6
+# asm 2: movq   16(<xp=%rsi),>r2=%r9
+movq   16(%rsi),%r9
+
+# qhasm: r3 = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>r3=int64#2
+# asm 2: movq   24(<xp=%rsi),>r3=%rsi
+movq   24(%rsi),%rsi
+
+# qhasm: carry? r0 += *(uint64 *)(yp +  0)
+# asm 1: addq 0(<yp=int64#3),<r0=int64#4
+# asm 2: addq 0(<yp=%rdx),<r0=%rcx
+addq 0(%rdx),%rcx
+
+# qhasm: carry? r1 += *(uint64 *)(yp +  8) + carry
+# asm 1: adcq 8(<yp=int64#3),<r1=int64#5
+# asm 2: adcq 8(<yp=%rdx),<r1=%r8
+adcq 8(%rdx),%r8
+
+# qhasm: carry? r2 += *(uint64 *)(yp + 16) + carry
+# asm 1: adcq 16(<yp=int64#3),<r2=int64#6
+# asm 2: adcq 16(<yp=%rdx),<r2=%r9
+adcq 16(%rdx),%r9
+
+# qhasm: r3 += *(uint64 *)(yp + 24) + carry
+# asm 1: adcq 24(<yp=int64#3),<r3=int64#2
+# asm 2: adcq 24(<yp=%rdx),<r3=%rsi
+adcq 24(%rdx),%rsi
+
+# qhasm: t0 = r0
+# asm 1: mov  <r0=int64#4,>t0=int64#3
+# asm 2: mov  <r0=%rcx,>t0=%rdx
+mov  %rcx,%rdx
+
+# qhasm: t1 = r1
+# asm 1: mov  <r1=int64#5,>t1=int64#7
+# asm 2: mov  <r1=%r8,>t1=%rax
+mov  %r8,%rax
+
+# qhasm: t2 = r2
+# asm 1: mov  <r2=int64#6,>t2=int64#8
+# asm 2: mov  <r2=%r9,>t2=%r10
+mov  %r9,%r10
+
+# qhasm: t3 = r3
+# asm 1: mov  <r3=int64#2,>t3=int64#12
+# asm 2: mov  <r3=%rsi,>t3=%r14
+mov  %rsi,%r14
+
+# qhasm: carry? t0 -= *(uint64 *) &crypto_sign_ed25519_amd64_64_ORDER0
+# asm 1: sub  crypto_sign_ed25519_amd64_64_ORDER0,<t0=int64#3
+# asm 2: sub  crypto_sign_ed25519_amd64_64_ORDER0,<t0=%rdx
+sub  crypto_sign_ed25519_amd64_64_ORDER0,%rdx
+
+# qhasm: carry? t1 -= *(uint64 *) &crypto_sign_ed25519_amd64_64_ORDER1 - carry
+# asm 1: sbb  crypto_sign_ed25519_amd64_64_ORDER1,<t1=int64#7
+# asm 2: sbb  crypto_sign_ed25519_amd64_64_ORDER1,<t1=%rax
+sbb  crypto_sign_ed25519_amd64_64_ORDER1,%rax
+
+# qhasm: carry? t2 -= *(uint64 *) &crypto_sign_ed25519_amd64_64_ORDER2 - carry
+# asm 1: sbb  crypto_sign_ed25519_amd64_64_ORDER2,<t2=int64#8
+# asm 2: sbb  crypto_sign_ed25519_amd64_64_ORDER2,<t2=%r10
+sbb  crypto_sign_ed25519_amd64_64_ORDER2,%r10
+
+# qhasm: unsigned<? t3 -= *(uint64 *) &crypto_sign_ed25519_amd64_64_ORDER3 - carry
+# asm 1: sbb  crypto_sign_ed25519_amd64_64_ORDER3,<t3=int64#12
+# asm 2: sbb  crypto_sign_ed25519_amd64_64_ORDER3,<t3=%r14
+sbb  crypto_sign_ed25519_amd64_64_ORDER3,%r14
+
+# qhasm: r0 = t0 if !unsigned<
+# asm 1: cmovae <t0=int64#3,<r0=int64#4
+# asm 2: cmovae <t0=%rdx,<r0=%rcx
+cmovae %rdx,%rcx
+
+# qhasm: r1 = t1 if !unsigned<
+# asm 1: cmovae <t1=int64#7,<r1=int64#5
+# asm 2: cmovae <t1=%rax,<r1=%r8
+cmovae %rax,%r8
+
+# qhasm: r2 = t2 if !unsigned<
+# asm 1: cmovae <t2=int64#8,<r2=int64#6
+# asm 2: cmovae <t2=%r10,<r2=%r9
+cmovae %r10,%r9
+
+# qhasm: r3 = t3 if !unsigned<
+# asm 1: cmovae <t3=int64#12,<r3=int64#2
+# asm 2: cmovae <t3=%r14,<r3=%rsi
+cmovae %r14,%rsi
+
+# qhasm: *(uint64 *)(rp +  0) = r0
+# asm 1: movq   <r0=int64#4,0(<rp=int64#1)
+# asm 2: movq   <r0=%rcx,0(<rp=%rdi)
+movq   %rcx,0(%rdi)
+
+# qhasm: *(uint64 *)(rp +  8) = r1
+# asm 1: movq   <r1=int64#5,8(<rp=int64#1)
+# asm 2: movq   <r1=%r8,8(<rp=%rdi)
+movq   %r8,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#6,16(<rp=int64#1)
+# asm 2: movq   <r2=%r9,16(<rp=%rdi)
+movq   %r9,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#2,24(<rp=int64#1)
+# asm 2: movq   <r3=%rsi,24(<rp=%rdi)
+movq   %rsi,24(%rdi)
+
+# qhasm: caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#1,>caller4=int64#12
+# asm 2: movq <caller4_stack=0(%rsp),>caller4=%r14
+movq 0(%rsp),%r14
+
+# qhasm: caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#2,>caller5=int64#13
+# asm 2: movq <caller5_stack=8(%rsp),>caller5=%r15
+movq 8(%rsp),%r15
+
+# qhasm: caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#3,>caller6=int64#14
+# asm 2: movq <caller6_stack=16(%rsp),>caller6=%rbx
+movq 16(%rsp),%rbx
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/sc25519_from32bytes.c

@@ -0,0 +1,55 @@
+#include "sc25519.h"
+
+/*Arithmetic modulo the group order n = 2^252 + 27742317777372353535851937790883648493 
+ *                                    = 7237005577332262213973186563042994240857116359379907606001950938285454250989
+ */
+
+/* Contains order, 2*order, 4*order, 8*order, each represented in 4 consecutive unsigned long long */
+static const unsigned long long order[16] = {0x5812631A5CF5D3EDULL, 0x14DEF9DEA2F79CD6ULL, 
+                                             0x0000000000000000ULL, 0x1000000000000000ULL,
+                                             0xB024C634B9EBA7DAULL, 0x29BDF3BD45EF39ACULL,
+                                             0x0000000000000000ULL, 0x2000000000000000ULL,
+                                             0x60498C6973D74FB4ULL, 0x537BE77A8BDE7359ULL,
+                                             0x0000000000000000ULL, 0x4000000000000000ULL,
+                                             0xC09318D2E7AE9F68ULL, 0xA6F7CEF517BCE6B2ULL,
+                                             0x0000000000000000ULL, 0x8000000000000000ULL};
+
+static unsigned long long smaller(unsigned long long a,unsigned long long b)
+{
+  unsigned long long atop = a >> 32;
+  unsigned long long abot = a & 4294967295;
+  unsigned long long btop = b >> 32;
+  unsigned long long bbot = b & 4294967295;
+  unsigned long long atopbelowbtop = (atop - btop) >> 63;
+  unsigned long long atopeqbtop = ((atop ^ btop) - 1) >> 63;
+  unsigned long long abotbelowbbot = (abot - bbot) >> 63;
+  return atopbelowbtop | (atopeqbtop & abotbelowbbot);
+}
+
+void sc25519_from32bytes(sc25519 *r, const unsigned char x[32])
+{
+  unsigned long long t[4];
+  unsigned long long b;
+  unsigned long long mask;
+  int i, j;
+
+  /* assuming little-endian */
+  r->v[0] = *(unsigned long long *)x;
+  r->v[1] = *(((unsigned long long *)x)+1);
+  r->v[2] = *(((unsigned long long *)x)+2);
+  r->v[3] = *(((unsigned long long *)x)+3);
+  
+  for(j=3;j>=0;j--)
+  {
+    b=0;
+    for(i=0;i<4;i++) 
+    {
+      b += order[4*j+i]; /* no overflow for this particular order */
+      t[i] = r->v[i] - b;
+      b = smaller(r->v[i],b);
+    }
+    mask = b - 1;
+    for(i=0;i<4;i++) 
+      r->v[i] ^= mask & (r->v[i] ^ t[i]);
+  }
+}

ext/ed25519-amd64-asm/sc25519_from64bytes.c

@@ -0,0 +1,7 @@
+#include "sc25519.h"
+
+void sc25519_from64bytes(sc25519 *r, const unsigned char x[64])
+{
+  /* assuming little-endian representation of unsigned long long */
+  sc25519_barrett(r, (unsigned long long *)x); 
+}

ext/ed25519-amd64-asm/sc25519_from_shortsc.c

@@ -0,0 +1,9 @@
+#include "sc25519.h"
+
+void sc25519_from_shortsc(sc25519 *r, const shortsc25519 *x)
+{
+  r->v[0] = x->v[0];
+  r->v[1] = x->v[1];
+  r->v[2] = 0;
+  r->v[3] = 0;
+}

ext/ed25519-amd64-asm/sc25519_iszero.c

@@ -0,0 +1,10 @@
+#include "sc25519.h"
+
+int sc25519_iszero_vartime(const sc25519 *x)
+{
+  if(x->v[0] != 0) return 0;
+  if(x->v[1] != 0) return 0;
+  if(x->v[2] != 0) return 0;
+  if(x->v[3] != 0) return 0;
+  return 1;
+}

ext/ed25519-amd64-asm/sc25519_lt.s

@@ -0,0 +1,131 @@
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: int64 ret
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: output ret
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm: int64 doof
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_sc25519_lt
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_sc25519_lt
+.globl crypto_sign_ed25519_amd64_64_sc25519_lt
+_crypto_sign_ed25519_amd64_64_sc25519_lt:
+crypto_sign_ed25519_amd64_64_sc25519_lt:
+mov %rsp,%r11
+and $31,%r11
+add $0,%r11
+sub %r11,%rsp
+
+# qhasm: t0 = *(uint64 *)(xp +  0)
+# asm 1: movq   0(<xp=int64#1),>t0=int64#3
+# asm 2: movq   0(<xp=%rdi),>t0=%rdx
+movq   0(%rdi),%rdx
+
+# qhasm: t1 = *(uint64 *)(xp +  8)
+# asm 1: movq   8(<xp=int64#1),>t1=int64#4
+# asm 2: movq   8(<xp=%rdi),>t1=%rcx
+movq   8(%rdi),%rcx
+
+# qhasm: t2 = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#1),>t2=int64#5
+# asm 2: movq   16(<xp=%rdi),>t2=%r8
+movq   16(%rdi),%r8
+
+# qhasm: t3 = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#1),>t3=int64#1
+# asm 2: movq   24(<xp=%rdi),>t3=%rdi
+movq   24(%rdi),%rdi
+
+# qhasm: carry? t0 -= *(uint64 *)(yp +  0)
+# asm 1: subq 0(<yp=int64#2),<t0=int64#3
+# asm 2: subq 0(<yp=%rsi),<t0=%rdx
+subq 0(%rsi),%rdx
+
+# qhasm: carry? t1 -= *(uint64 *)(yp +  8) - carry
+# asm 1: sbbq 8(<yp=int64#2),<t1=int64#4
+# asm 2: sbbq 8(<yp=%rsi),<t1=%rcx
+sbbq 8(%rsi),%rcx
+
+# qhasm: carry? t2 -= *(uint64 *)(yp + 16) - carry
+# asm 1: sbbq 16(<yp=int64#2),<t2=int64#5
+# asm 2: sbbq 16(<yp=%rsi),<t2=%r8
+sbbq 16(%rsi),%r8
+
+# qhasm: carry? t3 -= *(uint64 *)(yp + 24) - carry
+# asm 1: sbbq 24(<yp=int64#2),<t3=int64#1
+# asm 2: sbbq 24(<yp=%rsi),<t3=%rdi
+sbbq 24(%rsi),%rdi
+
+# qhasm: ret = 0
+# asm 1: mov  $0,>ret=int64#1
+# asm 2: mov  $0,>ret=%rdi
+mov  $0,%rdi
+
+# qhasm: doof = 1
+# asm 1: mov  $1,>doof=int64#2
+# asm 2: mov  $1,>doof=%rsi
+mov  $1,%rsi
+
+# qhasm: ret = doof if carry
+# asm 1: cmovc <doof=int64#2,<ret=int64#1
+# asm 2: cmovc <doof=%rsi,<ret=%rdi
+cmovc %rsi,%rdi
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/sc25519_mul.c

@@ -0,0 +1,12 @@
+#include "sc25519.h"
+
+#define ull4_mul        crypto_sign_ed25519_amd64_64_ull4_mul
+
+extern void ull4_mul(unsigned long long r[8], const unsigned long long x[4], const unsigned long long y[4]);
+
+void sc25519_mul(sc25519 *r, const sc25519 *x, const sc25519 *y)
+{
+  unsigned long long t[8];
+  ull4_mul(t, x->v, y->v);
+  sc25519_barrett(r, t);
+}

ext/ed25519-amd64-asm/sc25519_mul_shortsc.c

@@ -0,0 +1,9 @@
+#include "sc25519.h"
+
+void sc25519_mul_shortsc(sc25519 *r, const sc25519 *x, const shortsc25519 *y)
+{
+  /* XXX: This wants to be faster */
+  sc25519 t;
+  sc25519_from_shortsc(&t, y);
+  sc25519_mul(r, x, &t);
+}

ext/ed25519-amd64-asm/sc25519_slide.c

@@ -0,0 +1,49 @@
+#include "sc25519.h"
+
+void sc25519_slide(signed char r[256], const sc25519 *s, int swindowsize)
+{
+  int i,j,k,b,m=(1<<(swindowsize-1))-1, soplen=256;
+  unsigned long long sv0 = s->v[0];
+  unsigned long long sv1 = s->v[1];
+  unsigned long long sv2 = s->v[2];
+  unsigned long long sv3 = s->v[3];
+
+  /* first put the binary expansion into r  */
+  for(i=0;i<64;i++) {
+    r[i] = sv0 & 1;
+    r[i+64] = sv1 & 1;
+    r[i+128] = sv2 & 1;
+    r[i+192] = sv3 & 1;
+    sv0 >>= 1;
+    sv1 >>= 1;
+    sv2 >>= 1;
+    sv3 >>= 1;
+  }
+
+  /* Making it sliding window */
+  for (j = 0;j < soplen;++j) 
+  {
+    if (r[j]) {
+      for (b = 1;b < soplen - j && b <= 6;++b) {
+        if (r[j] + (r[j + b] << b) <= m) 
+        {
+          r[j] += r[j + b] << b; r[j + b] = 0;
+        } 
+        else if (r[j] - (r[j + b] << b) >= -m) 
+        {
+          r[j] -= r[j + b] << b;
+          for (k = j + b;k < soplen;++k) 
+          {
+            if (!r[k]) {
+              r[k] = 1;
+              break;
+            }
+            r[k] = 0;
+          }
+        } 
+        else if (r[j + b])
+          break;
+      }
+    }
+  }
+}

ext/ed25519-amd64-asm/sc25519_sub_nored.s

@@ -0,0 +1,142 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm: int64 t0
+
+# qhasm: int64 t1
+
+# qhasm: int64 t2
+
+# qhasm: int64 t3
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_sc25519_sub_nored
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_sc25519_sub_nored
+.globl crypto_sign_ed25519_amd64_64_sc25519_sub_nored
+_crypto_sign_ed25519_amd64_64_sc25519_sub_nored:
+crypto_sign_ed25519_amd64_64_sc25519_sub_nored:
+mov %rsp,%r11
+and $31,%r11
+add $0,%r11
+sub %r11,%rsp
+
+# qhasm: r0 = *(uint64 *)(xp +  0)
+# asm 1: movq   0(<xp=int64#2),>r0=int64#4
+# asm 2: movq   0(<xp=%rsi),>r0=%rcx
+movq   0(%rsi),%rcx
+
+# qhasm: r1 = *(uint64 *)(xp +  8)
+# asm 1: movq   8(<xp=int64#2),>r1=int64#5
+# asm 2: movq   8(<xp=%rsi),>r1=%r8
+movq   8(%rsi),%r8
+
+# qhasm: r2 = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>r2=int64#6
+# asm 2: movq   16(<xp=%rsi),>r2=%r9
+movq   16(%rsi),%r9
+
+# qhasm: r3 = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>r3=int64#2
+# asm 2: movq   24(<xp=%rsi),>r3=%rsi
+movq   24(%rsi),%rsi
+
+# qhasm: carry? r0 -= *(uint64 *)(yp +  0)
+# asm 1: subq 0(<yp=int64#3),<r0=int64#4
+# asm 2: subq 0(<yp=%rdx),<r0=%rcx
+subq 0(%rdx),%rcx
+
+# qhasm: carry? r1 -= *(uint64 *)(yp +  8) - carry
+# asm 1: sbbq 8(<yp=int64#3),<r1=int64#5
+# asm 2: sbbq 8(<yp=%rdx),<r1=%r8
+sbbq 8(%rdx),%r8
+
+# qhasm: carry? r2 -= *(uint64 *)(yp + 16) - carry
+# asm 1: sbbq 16(<yp=int64#3),<r2=int64#6
+# asm 2: sbbq 16(<yp=%rdx),<r2=%r9
+sbbq 16(%rdx),%r9
+
+# qhasm: r3 -= *(uint64 *)(yp + 24) - carry
+# asm 1: sbbq 24(<yp=int64#3),<r3=int64#2
+# asm 2: sbbq 24(<yp=%rdx),<r3=%rsi
+sbbq 24(%rdx),%rsi
+
+# qhasm: *(uint64 *)(rp +  0) = r0
+# asm 1: movq   <r0=int64#4,0(<rp=int64#1)
+# asm 2: movq   <r0=%rcx,0(<rp=%rdi)
+movq   %rcx,0(%rdi)
+
+# qhasm: *(uint64 *)(rp +  8) = r1
+# asm 1: movq   <r1=int64#5,8(<rp=int64#1)
+# asm 2: movq   <r1=%r8,8(<rp=%rdi)
+movq   %r8,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#6,16(<rp=int64#1)
+# asm 2: movq   <r2=%r9,16(<rp=%rdi)
+movq   %r9,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#2,24(<rp=int64#1)
+# asm 2: movq   <r3=%rsi,24(<rp=%rdi)
+movq   %rsi,24(%rdi)
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

ext/ed25519-amd64-asm/sc25519_to32bytes.c

@@ -0,0 +1,8 @@
+#include "sc25519.h"
+
+void sc25519_to32bytes(unsigned char r[32], const sc25519 *x)
+{
+  /* assuming little-endian */
+  int i;
+  for(i=0;i<32;i++) r[i] = i[(unsigned char *)x->v]; 
+}

ext/ed25519-amd64-asm/sc25519_window4.c

@@ -0,0 +1,27 @@
+#include "sc25519.h"
+
+void sc25519_window4(signed char r[64], const sc25519 *s)
+{
+  char carry;
+  int i;
+  for(i=0;i<16;i++)
+    r[i] = (s->v[0] >> (4*i)) & 15;
+  for(i=0;i<16;i++)
+    r[i+16] = (s->v[1] >> (4*i)) & 15;
+  for(i=0;i<16;i++)
+    r[i+32] = (s->v[2] >> (4*i)) & 15;
+  for(i=0;i<16;i++)
+    r[i+48] = (s->v[3] >> (4*i)) & 15;
+
+  /* Making it signed */
+  carry = 0;
+  for(i=0;i<63;i++)
+  {
+    r[i] += carry;
+    r[i+1] += r[i] >> 4;
+    r[i] &= 15;
+    carry = r[i] >> 3;
+    r[i] -= carry << 4;
+  }
+  r[63] += carry;
+}

ext/ed25519-amd64-asm/sign.c

@@ -0,0 +1,165 @@
+#include <stdlib.h>
+#include <string.h>
+/*#include "crypto_sign.h"
+#include "crypto_hash_sha512.h"*/
+#include "ge25519.h"
+
+/* Original */
+#if 0
+int crypto_sign(
+    unsigned char *sm,unsigned long long *smlen,
+    const unsigned char *m,unsigned long long mlen,
+    const unsigned char *sk
+    )
+{
+  unsigned char pk[32];
+  unsigned char az[64];
+  unsigned char nonce[64];
+  unsigned char hram[64];
+  sc25519 sck, scs, scsk;
+  ge25519 ger;
+
+  memmove(pk,sk + 32,32);
+  /* pk: 32-byte public key A */
+
+  crypto_hash_sha512(az,sk,32);
+  az[0] &= 248;
+  az[31] &= 127;
+  az[31] |= 64;
+  /* az: 32-byte scalar a, 32-byte randomizer z */
+
+  *smlen = mlen + 64;
+  memmove(sm + 64,m,mlen);
+  memmove(sm + 32,az + 32,32);
+  /* sm: 32-byte uninit, 32-byte z, mlen-byte m */
+
+  crypto_hash_sha512(nonce, sm+32, mlen+32);
+  /* nonce: 64-byte H(z,m) */
+
+  sc25519_from64bytes(&sck, nonce);
+  ge25519_scalarmult_base(&ger, &sck);
+  ge25519_pack(sm, &ger);
+  /* sm: 32-byte R, 32-byte z, mlen-byte m */
+
+  memmove(sm + 32,pk,32);
+  /* sm: 32-byte R, 32-byte A, mlen-byte m */
+
+  crypto_hash_sha512(hram,sm,mlen + 64);
+  /* hram: 64-byte H(R,A,m) */
+
+  sc25519_from64bytes(&scs, hram);
+  sc25519_from32bytes(&scsk, az);
+  sc25519_mul(&scs, &scs, &scsk);
+  sc25519_add(&scs, &scs, &sck);
+  /* scs: S = nonce + H(R,A,m)a */
+
+  sc25519_to32bytes(sm + 32,&scs);
+  /* sm: 32-byte R, 32-byte S, mlen-byte m */
+
+  return 0;
+}
+#endif
+
+#if 0
+void C25519::sign(const C25519::Private &myPrivate,const C25519::Public &myPublic,const void *msg,unsigned int len,void *signature)
+{
+  sc25519 sck, scs, scsk;
+  ge25519 ger;
+  unsigned char r[32];
+  unsigned char s[32];
+  unsigned char extsk[64];
+  unsigned char hmg[crypto_hash_sha512_BYTES];
+  unsigned char hram[crypto_hash_sha512_BYTES];
+  unsigned char *sig = (unsigned char *)signature;
+  unsigned char digest[64]; // we sign the first 32 bytes of SHA-512(msg)
+
+  SHA512::hash(digest,msg,len);
+
+  SHA512::hash(extsk,myPrivate.data + 32,32);
+  extsk[0] &= 248;
+  extsk[31] &= 127;
+  extsk[31] |= 64;
+
+  for(unsigned int i=0;i<32;i++)
+    sig[32 + i] = extsk[32 + i];
+  for(unsigned int i=0;i<32;i++)
+    sig[64 + i] = digest[i];
+
+  SHA512::hash(hmg,sig + 32,64);
+
+  /* Computation of R */
+  sc25519_from64bytes(&sck, hmg);
+  ge25519_scalarmult_base(&ger, &sck);
+  ge25519_pack(r, &ger);
+
+  /* Computation of s */
+  for(unsigned int i=0;i<32;i++)
+    sig[i] = r[i];
+
+  get_hram(hram,sig,myPublic.data + 32,sig,96);
+
+  sc25519_from64bytes(&scs, hram);
+  sc25519_from32bytes(&scsk, extsk);
+  sc25519_mul(&scs, &scs, &scsk);
+
+  sc25519_add(&scs, &scs, &sck);
+
+  sc25519_to32bytes(s,&scs); /* cat s */
+  for(unsigned int i=0;i<32;i++)
+    sig[32 + i] = s[i];
+}
+
+void get_hram(unsigned char *hram, const unsigned char *sm, const unsigned char *pk, unsigned char *playground, unsigned long long smlen)
+{
+  unsigned long long i;
+
+  for (i =  0;i < 32;++i)    playground[i] = sm[i];
+  for (i = 32;i < 64;++i)    playground[i] = pk[i-32];
+  for (i = 64;i < smlen;++i) playground[i] = sm[i];
+
+  //crypto_hash_sha512(hram,playground,smlen);
+  ZeroTier::SHA512::hash(hram,playground,(unsigned int)smlen);
+}
+#endif
+
+extern void ZT_sha512internal(void *digest,const void *data,unsigned int len);
+
+extern void ed25519_amd64_asm_sign(const unsigned char *sk,const unsigned char *pk,const unsigned char *m,const unsigned int mlen,unsigned char *sig)
+{
+  unsigned char az[64];
+  unsigned char nonce[64];
+  unsigned char hram[64];
+  sc25519 sck, scs, scsk;
+  ge25519 ger;
+  unsigned char digest[64];
+  unsigned int i;
+
+  ZT_sha512internal(digest,m,mlen);
+
+  ZT_sha512internal(az,sk,32);
+  az[0] &= 248;
+  az[31] &= 127;
+  az[31] |= 64;
+
+  for(i=0;i<32;i++)
+    sig[32 + i] = az[32 + i];
+  for(i=0;i<32;i++)
+    sig[64 + i] = digest[i];
+
+  ZT_sha512internal(nonce,sig + 32,64);
+
+  sc25519_from64bytes(&sck, nonce);
+  ge25519_scalarmult_base(&ger, &sck);
+  ge25519_pack(sig, &ger);
+
+  memmove(sig + 32,pk,32);
+
+  ZT_sha512internal(hram,sig,96);
+
+  sc25519_from64bytes(&scs, hram);
+  sc25519_from32bytes(&scsk, az);
+  sc25519_mul(&scs, &scs, &scsk);
+  sc25519_add(&scs, &scs, &sck);
+
+  sc25519_to32bytes(sig + 32,&scs);
+}

ext/ed25519-amd64-asm/ull4_mul.s

@@ -0,0 +1,716 @@
+
+# qhasm: int64 rp
+
+# qhasm: int64 xp
+
+# qhasm: int64 yp
+
+# qhasm: input rp
+
+# qhasm: input xp
+
+# qhasm: input yp
+
+# qhasm: int64 r0
+
+# qhasm: int64 r1
+
+# qhasm: int64 r2
+
+# qhasm: int64 r3
+
+# qhasm: int64 r4
+
+# qhasm: int64 r5
+
+# qhasm: int64 r6
+
+# qhasm: int64 r7
+
+# qhasm: int64 c
+
+# qhasm: int64 zero
+
+# qhasm: int64 rax
+
+# qhasm: int64 rdx
+
+# qhasm:   int64 caller1
+
+# qhasm:   int64 caller2
+
+# qhasm:   int64 caller3
+
+# qhasm:   int64 caller4
+
+# qhasm:   int64 caller5
+
+# qhasm:   int64 caller6
+
+# qhasm:   int64 caller7
+
+# qhasm:   caller caller1
+
+# qhasm:   caller caller2
+
+# qhasm:   caller caller3
+
+# qhasm:   caller caller4
+
+# qhasm:   caller caller5
+
+# qhasm:   caller caller6
+
+# qhasm:   caller caller7
+
+# qhasm:   stack64 caller1_stack
+
+# qhasm:   stack64 caller2_stack
+
+# qhasm:   stack64 caller3_stack
+
+# qhasm:   stack64 caller4_stack
+
+# qhasm:   stack64 caller5_stack
+
+# qhasm:   stack64 caller6_stack
+
+# qhasm:   stack64 caller7_stack
+
+# qhasm: enter crypto_sign_ed25519_amd64_64_ull4_mul 
+.text
+.p2align 5
+.globl _crypto_sign_ed25519_amd64_64_ull4_mul
+.globl crypto_sign_ed25519_amd64_64_ull4_mul
+_crypto_sign_ed25519_amd64_64_ull4_mul:
+crypto_sign_ed25519_amd64_64_ull4_mul:
+mov %rsp,%r11
+and $31,%r11
+add $64,%r11
+sub %r11,%rsp
+
+# qhasm:   caller1_stack = caller1
+# asm 1: movq <caller1=int64#9,>caller1_stack=stack64#1
+# asm 2: movq <caller1=%r11,>caller1_stack=0(%rsp)
+movq %r11,0(%rsp)
+
+# qhasm:   caller2_stack = caller2
+# asm 1: movq <caller2=int64#10,>caller2_stack=stack64#2
+# asm 2: movq <caller2=%r12,>caller2_stack=8(%rsp)
+movq %r12,8(%rsp)
+
+# qhasm:   caller3_stack = caller3
+# asm 1: movq <caller3=int64#11,>caller3_stack=stack64#3
+# asm 2: movq <caller3=%r13,>caller3_stack=16(%rsp)
+movq %r13,16(%rsp)
+
+# qhasm:   caller4_stack = caller4
+# asm 1: movq <caller4=int64#12,>caller4_stack=stack64#4
+# asm 2: movq <caller4=%r14,>caller4_stack=24(%rsp)
+movq %r14,24(%rsp)
+
+# qhasm:   caller5_stack = caller5
+# asm 1: movq <caller5=int64#13,>caller5_stack=stack64#5
+# asm 2: movq <caller5=%r15,>caller5_stack=32(%rsp)
+movq %r15,32(%rsp)
+
+# qhasm:   caller6_stack = caller6
+# asm 1: movq <caller6=int64#14,>caller6_stack=stack64#6
+# asm 2: movq <caller6=%rbx,>caller6_stack=40(%rsp)
+movq %rbx,40(%rsp)
+
+# qhasm:   caller7_stack = caller7
+# asm 1: movq <caller7=int64#15,>caller7_stack=stack64#7
+# asm 2: movq <caller7=%rbp,>caller7_stack=48(%rsp)
+movq %rbp,48(%rsp)
+
+# qhasm: yp = yp
+# asm 1: mov  <yp=int64#3,>yp=int64#4
+# asm 2: mov  <yp=%rdx,>yp=%rcx
+mov  %rdx,%rcx
+
+# qhasm: r4 = 0
+# asm 1: mov  $0,>r4=int64#5
+# asm 2: mov  $0,>r4=%r8
+mov  $0,%r8
+
+# qhasm: r5 = 0
+# asm 1: mov  $0,>r5=int64#6
+# asm 2: mov  $0,>r5=%r9
+mov  $0,%r9
+
+# qhasm: r6 = 0
+# asm 1: mov  $0,>r6=int64#8
+# asm 2: mov  $0,>r6=%r10
+mov  $0,%r10
+
+# qhasm: r7 = 0
+# asm 1: mov  $0,>r7=int64#9
+# asm 2: mov  $0,>r7=%r11
+mov  $0,%r11
+
+# qhasm: zero = 0
+# asm 1: mov  $0,>zero=int64#10
+# asm 2: mov  $0,>zero=%r12
+mov  $0,%r12
+
+# qhasm: rax = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>rax=int64#7
+# asm 2: movq   0(<xp=%rsi),>rax=%rax
+movq   0(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  0)
+# asm 1: mulq  0(<yp=int64#4)
+# asm 2: mulq  0(<yp=%rcx)
+mulq  0(%rcx)
+
+# qhasm: r0 = rax
+# asm 1: mov  <rax=int64#7,>r0=int64#11
+# asm 2: mov  <rax=%rax,>r0=%r13
+mov  %rax,%r13
+
+# qhasm: c = rdx
+# asm 1: mov  <rdx=int64#3,>c=int64#12
+# asm 2: mov  <rdx=%rdx,>c=%r14
+mov  %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>rax=int64#7
+# asm 2: movq   0(<xp=%rsi),>rax=%rax
+movq   0(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  8)
+# asm 1: mulq  8(<yp=int64#4)
+# asm 2: mulq  8(<yp=%rcx)
+mulq  8(%rcx)
+
+# qhasm: r1 = rax
+# asm 1: mov  <rax=int64#7,>r1=int64#13
+# asm 2: mov  <rax=%rax,>r1=%r15
+mov  %rax,%r15
+
+# qhasm: carry? r1 += c
+# asm 1: add  <c=int64#12,<r1=int64#13
+# asm 2: add  <c=%r14,<r1=%r15
+add  %r14,%r15
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>rax=int64#7
+# asm 2: movq   0(<xp=%rsi),>rax=%rax
+movq   0(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 16)
+# asm 1: mulq  16(<yp=int64#4)
+# asm 2: mulq  16(<yp=%rcx)
+mulq  16(%rcx)
+
+# qhasm: r2 = rax
+# asm 1: mov  <rax=int64#7,>r2=int64#14
+# asm 2: mov  <rax=%rax,>r2=%rbx
+mov  %rax,%rbx
+
+# qhasm: carry? r2 += c
+# asm 1: add  <c=int64#12,<r2=int64#14
+# asm 2: add  <c=%r14,<r2=%rbx
+add  %r14,%rbx
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 0)
+# asm 1: movq   0(<xp=int64#2),>rax=int64#7
+# asm 2: movq   0(<xp=%rsi),>rax=%rax
+movq   0(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 24)
+# asm 1: mulq  24(<yp=int64#4)
+# asm 2: mulq  24(<yp=%rcx)
+mulq  24(%rcx)
+
+# qhasm: r3 = rax
+# asm 1: mov  <rax=int64#7,>r3=int64#15
+# asm 2: mov  <rax=%rax,>r3=%rbp
+mov  %rax,%rbp
+
+# qhasm: carry? r3 += c
+# asm 1: add  <c=int64#12,<r3=int64#15
+# asm 2: add  <c=%r14,<r3=%rbp
+add  %r14,%rbp
+
+# qhasm: r4 += rdx + carry
+# asm 1: adc <rdx=int64#3,<r4=int64#5
+# asm 2: adc <rdx=%rdx,<r4=%r8
+adc %rdx,%r8
+
+# qhasm: rax = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>rax=int64#7
+# asm 2: movq   8(<xp=%rsi),>rax=%rax
+movq   8(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  0)
+# asm 1: mulq  0(<yp=int64#4)
+# asm 2: mulq  0(<yp=%rcx)
+mulq  0(%rcx)
+
+# qhasm: carry? r1 += rax
+# asm 1: add  <rax=int64#7,<r1=int64#13
+# asm 2: add  <rax=%rax,<r1=%r15
+add  %rax,%r15
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>rax=int64#7
+# asm 2: movq   8(<xp=%rsi),>rax=%rax
+movq   8(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  8)
+# asm 1: mulq  8(<yp=int64#4)
+# asm 2: mulq  8(<yp=%rcx)
+mulq  8(%rcx)
+
+# qhasm: carry? r2 += rax
+# asm 1: add  <rax=int64#7,<r2=int64#14
+# asm 2: add  <rax=%rax,<r2=%rbx
+add  %rax,%rbx
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r2 += c
+# asm 1: add  <c=int64#12,<r2=int64#14
+# asm 2: add  <c=%r14,<r2=%rbx
+add  %r14,%rbx
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>rax=int64#7
+# asm 2: movq   8(<xp=%rsi),>rax=%rax
+movq   8(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 16)
+# asm 1: mulq  16(<yp=int64#4)
+# asm 2: mulq  16(<yp=%rcx)
+mulq  16(%rcx)
+
+# qhasm: carry? r3 += rax 
+# asm 1: add  <rax=int64#7,<r3=int64#15
+# asm 2: add  <rax=%rax,<r3=%rbp
+add  %rax,%rbp
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r3 += c
+# asm 1: add  <c=int64#12,<r3=int64#15
+# asm 2: add  <c=%r14,<r3=%rbp
+add  %r14,%rbp
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 8)
+# asm 1: movq   8(<xp=int64#2),>rax=int64#7
+# asm 2: movq   8(<xp=%rsi),>rax=%rax
+movq   8(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 24)
+# asm 1: mulq  24(<yp=int64#4)
+# asm 2: mulq  24(<yp=%rcx)
+mulq  24(%rcx)
+
+# qhasm: carry? r4 += rax
+# asm 1: add  <rax=int64#7,<r4=int64#5
+# asm 2: add  <rax=%rax,<r4=%r8
+add  %rax,%r8
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r4 += c
+# asm 1: add  <c=int64#12,<r4=int64#5
+# asm 2: add  <c=%r14,<r4=%r8
+add  %r14,%r8
+
+# qhasm: r5 += rdx + carry
+# asm 1: adc <rdx=int64#3,<r5=int64#6
+# asm 2: adc <rdx=%rdx,<r5=%r9
+adc %rdx,%r9
+
+# qhasm: rax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>rax=int64#7
+# asm 2: movq   16(<xp=%rsi),>rax=%rax
+movq   16(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  0)
+# asm 1: mulq  0(<yp=int64#4)
+# asm 2: mulq  0(<yp=%rcx)
+mulq  0(%rcx)
+
+# qhasm: carry? r2 += rax
+# asm 1: add  <rax=int64#7,<r2=int64#14
+# asm 2: add  <rax=%rax,<r2=%rbx
+add  %rax,%rbx
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>rax=int64#7
+# asm 2: movq   16(<xp=%rsi),>rax=%rax
+movq   16(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  8)
+# asm 1: mulq  8(<yp=int64#4)
+# asm 2: mulq  8(<yp=%rcx)
+mulq  8(%rcx)
+
+# qhasm: carry? r3 += rax
+# asm 1: add  <rax=int64#7,<r3=int64#15
+# asm 2: add  <rax=%rax,<r3=%rbp
+add  %rax,%rbp
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r3 += c
+# asm 1: add  <c=int64#12,<r3=int64#15
+# asm 2: add  <c=%r14,<r3=%rbp
+add  %r14,%rbp
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>rax=int64#7
+# asm 2: movq   16(<xp=%rsi),>rax=%rax
+movq   16(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 16)
+# asm 1: mulq  16(<yp=int64#4)
+# asm 2: mulq  16(<yp=%rcx)
+mulq  16(%rcx)
+
+# qhasm: carry? r4 += rax
+# asm 1: add  <rax=int64#7,<r4=int64#5
+# asm 2: add  <rax=%rax,<r4=%r8
+add  %rax,%r8
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r4 += c
+# asm 1: add  <c=int64#12,<r4=int64#5
+# asm 2: add  <c=%r14,<r4=%r8
+add  %r14,%r8
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 16)
+# asm 1: movq   16(<xp=int64#2),>rax=int64#7
+# asm 2: movq   16(<xp=%rsi),>rax=%rax
+movq   16(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 24)
+# asm 1: mulq  24(<yp=int64#4)
+# asm 2: mulq  24(<yp=%rcx)
+mulq  24(%rcx)
+
+# qhasm: carry? r5 += rax
+# asm 1: add  <rax=int64#7,<r5=int64#6
+# asm 2: add  <rax=%rax,<r5=%r9
+add  %rax,%r9
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r5 += c
+# asm 1: add  <c=int64#12,<r5=int64#6
+# asm 2: add  <c=%r14,<r5=%r9
+add  %r14,%r9
+
+# qhasm: r6 += rdx + carry
+# asm 1: adc <rdx=int64#3,<r6=int64#8
+# asm 2: adc <rdx=%rdx,<r6=%r10
+adc %rdx,%r10
+
+# qhasm: rax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>rax=int64#7
+# asm 2: movq   24(<xp=%rsi),>rax=%rax
+movq   24(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  0)
+# asm 1: mulq  0(<yp=int64#4)
+# asm 2: mulq  0(<yp=%rcx)
+mulq  0(%rcx)
+
+# qhasm: carry? r3 += rax
+# asm 1: add  <rax=int64#7,<r3=int64#15
+# asm 2: add  <rax=%rax,<r3=%rbp
+add  %rax,%rbp
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>rax=int64#7
+# asm 2: movq   24(<xp=%rsi),>rax=%rax
+movq   24(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp +  8)
+# asm 1: mulq  8(<yp=int64#4)
+# asm 2: mulq  8(<yp=%rcx)
+mulq  8(%rcx)
+
+# qhasm: carry? r4 += rax
+# asm 1: add  <rax=int64#7,<r4=int64#5
+# asm 2: add  <rax=%rax,<r4=%r8
+add  %rax,%r8
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r4 += c
+# asm 1: add  <c=int64#12,<r4=int64#5
+# asm 2: add  <c=%r14,<r4=%r8
+add  %r14,%r8
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>rax=int64#7
+# asm 2: movq   24(<xp=%rsi),>rax=%rax
+movq   24(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 16)
+# asm 1: mulq  16(<yp=int64#4)
+# asm 2: mulq  16(<yp=%rcx)
+mulq  16(%rcx)
+
+# qhasm: carry? r5 += rax
+# asm 1: add  <rax=int64#7,<r5=int64#6
+# asm 2: add  <rax=%rax,<r5=%r9
+add  %rax,%r9
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r5 += c
+# asm 1: add  <c=int64#12,<r5=int64#6
+# asm 2: add  <c=%r14,<r5=%r9
+add  %r14,%r9
+
+# qhasm: c = 0
+# asm 1: mov  $0,>c=int64#12
+# asm 2: mov  $0,>c=%r14
+mov  $0,%r14
+
+# qhasm: c += rdx + carry
+# asm 1: adc <rdx=int64#3,<c=int64#12
+# asm 2: adc <rdx=%rdx,<c=%r14
+adc %rdx,%r14
+
+# qhasm: rax = *(uint64 *)(xp + 24)
+# asm 1: movq   24(<xp=int64#2),>rax=int64#7
+# asm 2: movq   24(<xp=%rsi),>rax=%rax
+movq   24(%rsi),%rax
+
+# qhasm: (uint128) rdx rax = rax * *(uint64 *)(yp + 24)
+# asm 1: mulq  24(<yp=int64#4)
+# asm 2: mulq  24(<yp=%rcx)
+mulq  24(%rcx)
+
+# qhasm: carry? r6 += rax
+# asm 1: add  <rax=int64#7,<r6=int64#8
+# asm 2: add  <rax=%rax,<r6=%r10
+add  %rax,%r10
+
+# qhasm: rdx += zero + carry
+# asm 1: adc <zero=int64#10,<rdx=int64#3
+# asm 2: adc <zero=%r12,<rdx=%rdx
+adc %r12,%rdx
+
+# qhasm: carry? r6 += c
+# asm 1: add  <c=int64#12,<r6=int64#8
+# asm 2: add  <c=%r14,<r6=%r10
+add  %r14,%r10
+
+# qhasm: r7 += rdx + carry
+# asm 1: adc <rdx=int64#3,<r7=int64#9
+# asm 2: adc <rdx=%rdx,<r7=%r11
+adc %rdx,%r11
+
+# qhasm: *(uint64 *)(rp +  0) = r0
+# asm 1: movq   <r0=int64#11,0(<rp=int64#1)
+# asm 2: movq   <r0=%r13,0(<rp=%rdi)
+movq   %r13,0(%rdi)
+
+# qhasm: *(uint64 *)(rp +  8) = r1
+# asm 1: movq   <r1=int64#13,8(<rp=int64#1)
+# asm 2: movq   <r1=%r15,8(<rp=%rdi)
+movq   %r15,8(%rdi)
+
+# qhasm: *(uint64 *)(rp + 16) = r2
+# asm 1: movq   <r2=int64#14,16(<rp=int64#1)
+# asm 2: movq   <r2=%rbx,16(<rp=%rdi)
+movq   %rbx,16(%rdi)
+
+# qhasm: *(uint64 *)(rp + 24) = r3
+# asm 1: movq   <r3=int64#15,24(<rp=int64#1)
+# asm 2: movq   <r3=%rbp,24(<rp=%rdi)
+movq   %rbp,24(%rdi)
+
+# qhasm: *(uint64 *)(rp + 32) = r4
+# asm 1: movq   <r4=int64#5,32(<rp=int64#1)
+# asm 2: movq   <r4=%r8,32(<rp=%rdi)
+movq   %r8,32(%rdi)
+
+# qhasm: *(uint64 *)(rp + 40) = r5
+# asm 1: movq   <r5=int64#6,40(<rp=int64#1)
+# asm 2: movq   <r5=%r9,40(<rp=%rdi)
+movq   %r9,40(%rdi)
+
+# qhasm: *(uint64 *)(rp + 48) = r6
+# asm 1: movq   <r6=int64#8,48(<rp=int64#1)
+# asm 2: movq   <r6=%r10,48(<rp=%rdi)
+movq   %r10,48(%rdi)
+
+# qhasm: *(uint64 *)(rp + 56) = r7
+# asm 1: movq   <r7=int64#9,56(<rp=int64#1)
+# asm 2: movq   <r7=%r11,56(<rp=%rdi)
+movq   %r11,56(%rdi)
+
+# qhasm:   caller1 = caller1_stack
+# asm 1: movq <caller1_stack=stack64#1,>caller1=int64#9
+# asm 2: movq <caller1_stack=0(%rsp),>caller1=%r11
+movq 0(%rsp),%r11
+
+# qhasm:   caller2 = caller2_stack
+# asm 1: movq <caller2_stack=stack64#2,>caller2=int64#10
+# asm 2: movq <caller2_stack=8(%rsp),>caller2=%r12
+movq 8(%rsp),%r12
+
+# qhasm:   caller3 = caller3_stack
+# asm 1: movq <caller3_stack=stack64#3,>caller3=int64#11
+# asm 2: movq <caller3_stack=16(%rsp),>caller3=%r13
+movq 16(%rsp),%r13
+
+# qhasm:   caller4 = caller4_stack
+# asm 1: movq <caller4_stack=stack64#4,>caller4=int64#12
+# asm 2: movq <caller4_stack=24(%rsp),>caller4=%r14
+movq 24(%rsp),%r14
+
+# qhasm:   caller5 = caller5_stack
+# asm 1: movq <caller5_stack=stack64#5,>caller5=int64#13
+# asm 2: movq <caller5_stack=32(%rsp),>caller5=%r15
+movq 32(%rsp),%r15
+
+# qhasm:   caller6 = caller6_stack
+# asm 1: movq <caller6_stack=stack64#6,>caller6=int64#14
+# asm 2: movq <caller6_stack=40(%rsp),>caller6=%rbx
+movq 40(%rsp),%rbx
+
+# qhasm:   caller7 = caller7_stack
+# asm 1: movq <caller7_stack=stack64#7,>caller7=int64#15
+# asm 2: movq <caller7_stack=48(%rsp),>caller7=%rbp
+movq 48(%rsp),%rbp
+
+# qhasm: leave
+add %r11,%rsp
+mov %rdi,%rax
+mov %rsi,%rdx
+ret

make-linux.mk

@@ -64,7 +64,6 @@ endif
 ifeq ($(ZT_DEBUG),1)
 	override CFLAGS+=-Wall -Wno-deprecated -Werror -g -pthread $(INCLUDES) $(DEFS)
 	override CXXFLAGS+=-Wall -Wno-deprecated -Werror -g -std=c++11 -pthread $(INCLUDES) $(DEFS)
-	override LDFLAGS+=
 	ZT_TRACE=1
 	STRIP?=echo
 	# The following line enables optimization for the crypto code, since
@@ -72,10 +71,9 @@ ifeq ($(ZT_DEBUG),1)
 node/Salsa20.o node/SHA512.o node/C25519.o node/Poly1305.o: CXXFLAGS=-Wall -O2 -g -pthread $(INCLUDES) $(DEFS)
 else
 	CFLAGS?=-O3 -fstack-protector
-	override CFLAGS+=-Wall -Wno-deprecated -fPIE -pthread $(INCLUDES) -DNDEBUG $(DEFS)
+	override CFLAGS+=-Wall -Wno-deprecated -pthread $(INCLUDES) -DNDEBUG $(DEFS)
 	CXXFLAGS?=-O3 -fstack-protector
-	override CXXFLAGS+=-Wall -Wno-deprecated -fPIE -std=c++11 -pthread $(INCLUDES) -DNDEBUG $(DEFS)
+	override CXXFLAGS+=-Wall -Wno-deprecated -std=c++11 -pthread $(INCLUDES) -DNDEBUG $(DEFS)
-	override LDFLAGS+=-pie -Wl,-z,relro,-z,now
 	STRIP?=strip
 	STRIP+=--strip-all
 endif
@@ -99,11 +97,11 @@ CC_MACH=$(shell $(CC) -dumpmachine | cut -d '-' -f 1)
 ZT_ARCHITECTURE=999
 ifeq ($(CC_MACH),x86_64)
 	ZT_ARCHITECTURE=2
-	ZT_USE_X64_ASM_SALSA2012=1
+	ZT_USE_X64_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),amd64)
 	ZT_ARCHITECTURE=2
-	ZT_USE_X64_ASM_SALSA2012=1
+	ZT_USE_X64_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),powerpc64le)
 	ZT_ARCHITECTURE=8
@@ -129,42 +127,42 @@ endif
 ifeq ($(CC_MACH),arm)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armel)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armhf)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armv6)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armv6zk)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armv6kz)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armv7)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),armv7l)
 	ZT_ARCHITECTURE=3
 	override DEFS+=-DZT_NO_TYPE_PUNNING
-	ZT_USE_ARM32_NEON_ASM_SALSA2012=1
+	ZT_USE_ARM32_NEON_ASM_CRYPTO=1
 endif
 ifeq ($(CC_MACH),arm64)
 	ZT_ARCHITECTURE=4
@@ -220,7 +218,7 @@ ifeq ($(ZT_ARCHITECTURE),3)
 	ifeq ($(shell if [ -e /usr/bin/dpkg ]; then dpkg --print-architecture; fi),armel)
 		override CFLAGS+=-march=armv5 -mfloat-abi=soft -msoft-float -mno-unaligned-access -marm
 		override CXXFLAGS+=-march=armv5 -mfloat-abi=soft -msoft-float -mno-unaligned-access -marm
-		ZT_USE_ARM32_NEON_ASM_SALSA2012=0
+		ZT_USE_ARM32_NEON_ASM_CRYPTO=0
 	else
 		override CFLAGS+=-march=armv5 -mno-unaligned-access -marm
 		override CXXFLAGS+=-march=armv5 -mno-unaligned-access -marm
@@ -228,11 +226,11 @@ ifeq ($(ZT_ARCHITECTURE),3)
 endif
 
 # Build faster crypto on some targets
-ifeq ($(ZT_USE_X64_ASM_SALSA2012),1)
+ifeq ($(ZT_USE_X64_ASM_CRYPTO),1)
-	override DEFS+=-DZT_USE_X64_ASM_SALSA2012
+	override DEFS+=-DZT_USE_X64_ASM_SALSA2012 -DZT_USE_FAST_X64_ED25519
-	override CORE_OBJS+=ext/x64-salsa2012-asm/salsa2012.o
+	override CORE_OBJS+=ext/x64-salsa2012-asm/salsa2012.o ext/ed25519-amd64-asm/choose_t.o ext/ed25519-amd64-asm/consts.o ext/ed25519-amd64-asm/fe25519_add.o ext/ed25519-amd64-asm/fe25519_freeze.o ext/ed25519-amd64-asm/fe25519_mul.o ext/ed25519-amd64-asm/fe25519_square.o ext/ed25519-amd64-asm/fe25519_sub.o ext/ed25519-amd64-asm/ge25519_add_p1p1.o ext/ed25519-amd64-asm/ge25519_dbl_p1p1.o ext/ed25519-amd64-asm/ge25519_nielsadd2.o ext/ed25519-amd64-asm/ge25519_nielsadd_p1p1.o ext/ed25519-amd64-asm/ge25519_p1p1_to_p2.o ext/ed25519-amd64-asm/ge25519_p1p1_to_p3.o ext/ed25519-amd64-asm/ge25519_pnielsadd_p1p1.o ext/ed25519-amd64-asm/heap_rootreplaced.o ext/ed25519-amd64-asm/heap_rootreplaced_1limb.o ext/ed25519-amd64-asm/heap_rootreplaced_2limbs.o ext/ed25519-amd64-asm/heap_rootreplaced_3limbs.o ext/ed25519-amd64-asm/sc25519_add.o ext/ed25519-amd64-asm/sc25519_barrett.o ext/ed25519-amd64-asm/sc25519_lt.o ext/ed25519-amd64-asm/sc25519_sub_nored.o ext/ed25519-amd64-asm/ull4_mul.o ext/ed25519-amd64-asm/fe25519_getparity.o ext/ed25519-amd64-asm/fe25519_invert.o ext/ed25519-amd64-asm/fe25519_iseq.o ext/ed25519-amd64-asm/fe25519_iszero.o ext/ed25519-amd64-asm/fe25519_neg.o ext/ed25519-amd64-asm/fe25519_pack.o ext/ed25519-amd64-asm/fe25519_pow2523.o ext/ed25519-amd64-asm/fe25519_setint.o ext/ed25519-amd64-asm/fe25519_unpack.o ext/ed25519-amd64-asm/ge25519_add.o ext/ed25519-amd64-asm/ge25519_base.o ext/ed25519-amd64-asm/ge25519_double.o ext/ed25519-amd64-asm/ge25519_double_scalarmult.o ext/ed25519-amd64-asm/ge25519_isneutral.o ext/ed25519-amd64-asm/ge25519_multi_scalarmult.o ext/ed25519-amd64-asm/ge25519_pack.o ext/ed25519-amd64-asm/ge25519_scalarmult_base.o ext/ed25519-amd64-asm/ge25519_unpackneg.o ext/ed25519-amd64-asm/hram.o ext/ed25519-amd64-asm/index_heap.o ext/ed25519-amd64-asm/sc25519_from32bytes.o ext/ed25519-amd64-asm/sc25519_from64bytes.o ext/ed25519-amd64-asm/sc25519_from_shortsc.o ext/ed25519-amd64-asm/sc25519_iszero.o ext/ed25519-amd64-asm/sc25519_mul.o ext/ed25519-amd64-asm/sc25519_mul_shortsc.o ext/ed25519-amd64-asm/sc25519_slide.o ext/ed25519-amd64-asm/sc25519_to32bytes.o ext/ed25519-amd64-asm/sc25519_window4.o ext/ed25519-amd64-asm/sign.o
 endif
-ifeq ($(ZT_USE_ARM32_NEON_ASM_SALSA2012),1)
+ifeq ($(ZT_USE_ARM32_NEON_ASM_CRYPTO),1)
 	override DEFS+=-DZT_USE_ARM32_NEON_ASM_SALSA2012
 	override CORE_OBJS+=ext/arm32-neon-salsa2012-asm/salsa2012.o
 endif

node/C25519.cpp

@@ -295,10 +295,7 @@ int crypto_scalarmult_base(unsigned char *q,const unsigned char *n)
 //////////////////////////////////////////////////////////////////////////////
 //////////////////////////////////////////////////////////////////////////////
 
-// This is the Ed25519 stuff from SUPERCOP:
+// Ed25519 ref from: http://bench.cr.yp.to/supercop.html
-// http://bench.cr.yp.to/supercop.html
-
-// Also public domain, newer version than the Ed25519 found in NaCl
 
 typedef struct
 {
@@ -306,6 +303,49 @@ typedef struct
 }
 fe25519;
 
+typedef struct
+{
+	crypto_uint32 v[32];
+}
+sc25519;
+
+typedef struct
+{
+	crypto_uint32 v[16];
+}
+shortsc25519;
+
+typedef struct
+{
+	fe25519 x;
+	fe25519 y;
+	fe25519 z;
+	fe25519 t;
+} ge25519;
+
+#define ge25519_p3 ge25519
+
+typedef struct
+{
+	fe25519 x;
+	fe25519 z;
+	fe25519 y;
+	fe25519 t;
+} ge25519_p1p1;
+
+typedef struct
+{
+	fe25519 x;
+	fe25519 y;
+	fe25519 z;
+} ge25519_p2;
+
+typedef struct
+{
+	fe25519 x;
+	fe25519 y;
+} ge25519_aff;
+
 static void fe25519_sub(fe25519 *r, const fe25519 *x, const fe25519 *y);
 
 crypto_uint32 equal(crypto_uint32 a,crypto_uint32 b) /* 16-bit inputs */
@@ -619,18 +659,6 @@ void fe25519_pow2523(fe25519 *r, const fe25519 *x)
 	/* 2^252 - 3 */ fe25519_mul(r,&t,x);
 }
 
-typedef struct
-{
-	crypto_uint32 v[32];
-}
-sc25519;
-
-typedef struct
-{
-	crypto_uint32 v[16];
-}
-shortsc25519;
-
 static const crypto_uint32 m[32] = {0xED, 0xD3, 0xF5, 0x5C, 0x1A, 0x63, 0x12, 0x58, 0xD6, 0x9C, 0xF7, 0xA2, 0xDE, 0xF9, 0xDE, 0x14,
 																		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x10};
 
@@ -827,14 +855,6 @@ void sc25519_2interleave2(unsigned char r[127], const sc25519 *s1, const sc25519
 	r[126] = ((s1->v[31] >> 4) & 3) ^ (((s2->v[31] >> 4) & 3) << 2);
 }
 
-typedef struct
-{
-	fe25519 x;
-	fe25519 y;
-	fe25519 z;
-	fe25519 t;
-} ge25519;
-
 /* d */
 static const fe25519 ge25519_ecd = {{0xA3, 0x78, 0x59, 0x13, 0xCA, 0x4D, 0xEB, 0x75, 0xAB, 0xD8, 0x41, 0x41, 0x4D, 0x0A, 0x70, 0x00,
 											0x98, 0xE8, 0x79, 0x77, 0x79, 0x40, 0xC7, 0x8C, 0x73, 0xFE, 0x6F, 0x2B, 0xEE, 0x6C, 0x03, 0x52}};
@@ -845,30 +865,6 @@ static const fe25519 ge25519_ec2d = {{0x59, 0xF1, 0xB2, 0x26, 0x94, 0x9B, 0xD6,
 static const fe25519 ge25519_sqrtm1 = {{0xB0, 0xA0, 0x0E, 0x4A, 0x27, 0x1B, 0xEE, 0xC4, 0x78, 0xE4, 0x2F, 0xAD, 0x06, 0x18, 0x43, 0x2F,
 												 0xA7, 0xD7, 0xFB, 0x3D, 0x99, 0x00, 0x4D, 0x2B, 0x0B, 0xDF, 0xC1, 0x4F, 0x80, 0x24, 0x83, 0x2B}};
 
-#define ge25519_p3 ge25519
-
-typedef struct
-{
-	fe25519 x;
-	fe25519 z;
-	fe25519 y;
-	fe25519 t;
-} ge25519_p1p1;
-
-typedef struct
-{
-	fe25519 x;
-	fe25519 y;
-	fe25519 z;
-} ge25519_p2;
-
-typedef struct
-{
-	fe25519 x;
-	fe25519 y;
-} ge25519_aff;
-
-
 /* Packed coordinates of the base point */
 static const ge25519 ge25519_base = {{{0x1A, 0xD5, 0x25, 0x8F, 0x60, 0x2D, 0x56, 0xC9, 0xB2, 0xA7, 0x25, 0x95, 0x60, 0xC7, 0x2C, 0x69,
 																0x5C, 0xDC, 0xD6, 0xFD, 0x31, 0xE2, 0xA4, 0xC0, 0xFE, 0x53, 0x6E, 0xCD, 0xD3, 0x36, 0x69, 0x21}},
@@ -1999,6 +1995,10 @@ void get_hram(unsigned char *hram, const unsigned char *sm, const unsigned char
 
 } // anonymous namespace
 
+#ifdef ZT_USE_FAST_X64_ED25519
+extern "C" void ed25519_amd64_asm_sign(const unsigned char *sk,const unsigned char *pk,const unsigned char *m,const unsigned int mlen,unsigned char *sig);
+#endif
+
 namespace ZeroTier {
 
 void C25519::agree(const C25519::Private &mine,const C25519::Public &their,void *keybuf,unsigned int keylen)
@@ -2019,6 +2019,9 @@ void C25519::agree(const C25519::Private &mine,const C25519::Public &their,void
 
 void C25519::sign(const C25519::Private &myPrivate,const C25519::Public &myPublic,const void *msg,unsigned int len,void *signature)
 {
+#ifdef ZT_USE_FAST_X64_ED25519
+	ed25519_amd64_asm_sign(myPrivate.data + 32,myPublic.data + 32,(const unsigned char *)msg,len,(unsigned char *)signature);
+#else
 	sc25519 sck, scs, scsk;
 	ge25519 ger;
 	unsigned char r[32];
@@ -2063,6 +2066,7 @@ void C25519::sign(const C25519::Private &myPrivate,const C25519::Public &myPubli
 	sc25519_to32bytes(s,&scs); /* cat s */
 	for(unsigned int i=0;i<32;i++)
 		sig[32 + i] = s[i];
+#endif
 }
 
 bool C25519::verify(const C25519::Public &their,const void *msg,unsigned int len,const void *signature)

node/SHA512.cpp

@@ -357,4 +357,11 @@ void SHA512::hash(void *digest,const void *data,unsigned int len)
 
 } // namespace ZeroTier
 
-#endif
+#endif // !ZT_HAVE_NATIVE_SHA512
+
+// Internally re-export to included C code, which includes some fast crypto code ported in on some platforms.
+// This eliminates the need to link against a third party SHA512() from this code
+extern "C" void ZT_sha512internal(void *digest,const void *data,unsigned int len)
+{
+	ZeroTier::SHA512::hash(digest,data,len);
+}