... | @@ -84,9 +84,7 @@ operand* 用来存放算术中的参数值,如 a+b=c+overflow 指令中的 a,b |
... | @@ -84,9 +84,7 @@ operand* 用来存放算术中的参数值,如 a+b=c+overflow 指令中的 a,b |
|
- define carry_hi = (t2 + (t3 << 64) + c_hi + carry_lo).saturating_sub(d_hi) >> 128
|
|
- define carry_hi = (t2 + (t3 << 64) + c_hi + carry_lo).saturating_sub(d_hi) >> 128
|
|
```
|
|
```
|
|
- 如果是 0 行,约束 num_row is 10,并且约束 cnt 自增的有效性
|
|
- 如果是 0 行,约束 num_row is 10,并且约束 cnt 自增的有效性
|
|
- a_lo = u16 sum(rotation cur)
|
|
- b_lo = u16 sum(rotation -2) //请注意因为a是输入值,我们可以不对a_hi,a_lo进行range约束
|
|
- a_hi = u16 sum(rotation -1)
|
|
|
|
- b_lo = u16 sum(rotation -2)
|
|
|
|
- b_hi = u16 sum(rotation -3)
|
|
- b_hi = u16 sum(rotation -3)
|
|
- c_lo = u16 sum(rotation -4)
|
|
- c_lo = u16 sum(rotation -4)
|
|
- c_hi = u16 sum(rotation -5)
|
|
- c_hi = u16 sum(rotation -5)
|
... | @@ -153,30 +151,52 @@ operand* 用来存放算术中的参数值,如 a+b=c+overflow 指令中的 a,b |
... | @@ -153,30 +151,52 @@ operand* 用来存放算术中的参数值,如 a+b=c+overflow 指令中的 a,b |
|
- d = pop1
|
|
- d = pop1
|
|
```
|
|
```
|
|
- a_abs,b_abs,c_abs,d_abs
|
|
- a_abs,b_abs,c_abs,d_abs
|
|
``
|
|
```
|
|
关于补码的相关约束如下
|
|
关于补码的相关约束如下
|
|
- x_abs_lo == lo when x >= 0
|
|
- x_abs_lo == lo when x >= 0
|
|
- x_abs_hi == hi when x < 0
|
|
- x_abs_hi == hi when x >= 0
|
|
- sum == 0 when x < 0 (小于0证明存在补码,同时我们有x+x_abs = 1 <<256。x + x_abs 约束请参考add部分)
|
|
- sum == 0 when x < 0 (小于0证明存在补码,同时我们有x+x_abs = 1 <<256。x + x_abs 约束请参考add部分)
|
|
- carry_hi == 1 when x < 0
|
|
- carry_hi == 1 when x < 0
|
|
``
|
|
```
|
|
- mul_add_words (请参考mul部分约束内容)
|
|
- mul_add_words (请参考mul部分约束内容)
|
|
- abs(remainder) < abs(divisor) when divisor != 0
|
|
- abs(remainder) < abs(divisor) when divisor != 0
|
|
- mul_add_words中的carry_hi == 0
|
|
- mul_add_words中的carry_hi == 0
|
|
- sign(dividend) == sign(remainder) when quotient, divisor and remainder are all non-zero。 这里主要是有mod的值与被除数的符号相同
|
|
- sign(dividend) == sign(remainder) when quotient, divisor and remainder are all non-zero。 这里主要是有mod的值与被除数的符号相同
|
|
- quotient_abs_word.is_neg().expr() + divisor_abs_word.is_neg().expr()
|
|
- quotient_abs_word.is_neg().expr() + divisor_abs_word.is_neg().expr() (如果被除数的补码也是负数,与余数和除数等于零,则排除约束情况) 这里说到的符号都是非补码的符号
|
|
- dividend_abs_word.is_neg().expr()
|
|
- dividend_abs_word.is_neg().expr()
|
|
- 2.expr()
|
|
- 2.expr()
|
|
* quotient_abs_word.is_neg().expr()
|
|
* quotient_abs_word.is_neg().expr()
|
|
* divisor_abs_word.is_neg().expr(),
|
|
* divisor_abs_word.is_neg().expr(), 这里计算的都是div情况下的变化
|
|
|
|
|
|
|
|
|
|
- AddMod 计算addMod操作码我们等于验证a,b,n,r 其中n是mod值,r是余数。我们有(a+b)%n = r。我们可以将这个约束转化为(a+b) = n * q + r。为了约束简单我们可以有
|
|
- AddMod 计算addMod操作码我们等于验证a,b,n,r 其中n是mod值,r是余数。我们有(a+b)%n = r。我们可以将这个约束转化为(a+b) = n * q + r。为了约束简单我们可以有
|
|
a % n= a_div_n + a_remainder
|
|
a % n= a_div_n + a_remainder a = n * a_div_n + a_remainder
|
|
(a_remainder + b) = (a_remainder_plus_b +a_remainder_plus_b_overflow << 256 ) % n= b_div_n + r
|
|
(a_remainder + b) = (a_remainder_plus_b +a_remainder_plus_b_overflow << 256 )
|
|
其中a_remainder+b 大于256位,我们可以有以下约束
|
|
(a_remainder_plus_b + a_remainder_plus_b_overflow << 256 ) % n= b_div_n + r
|
|
|
|
note: 其中a_remainder+b 大于256位. 我们可以有以下约束
|
|
|
|
```
|
|
|
|
/// Construct the gadget that checks a * b + c == d * 2**256 + e
|
|
|
|
/// where a, b, c, d, e are 256-bit words.
|
|
|
|
///
|
|
|
|
/// We execute a multi-limb multiplication as follows:
|
|
|
|
/// a and b is divided into 4 64-bit limbs, denoted as a0~a3 and b0~b3
|
|
|
|
/// defined t0, t1, t2, t3, t4, t5, t6:
|
|
|
|
/// t0 = a0 * b0, // 0 - 128bit
|
|
|
|
/// t1 = a0 * b1 + a1 * b0, //64 - 193bit 两数相加可能存在进位
|
|
|
|
/// t2 = a0 * b2 + a2 * b0 + a1 * b1, //128 - 258bit
|
|
|
|
/// t3 = a0 * b3 + a3 * b0 + a2 * b1 + a1 * b2, //192 - 322bit
|
|
|
|
/// t4 = a1 * b3 + a2 * b2 + a3 * b1,
|
|
|
|
/// t5 = a2 * b3 + a3 * b2,
|
|
|
|
/// t6 = a3 * b3,
|
|
|
|
|
|
|
|
/// Finally we just prove:
|
|
|
|
/// t0 + t1 * 2^64 + c_lo = e_lo + carry_0 * 2^128 // carry_0 is 65bit
|
|
|
|
/// t2 + t3 * 2^64 + c_hi + carry_0 = e_hi + carry_1 * 2^128
|
|
|
|
/// t4 + t5 * 2^64 + carry_1 = d_lo + carry_2 * 2^128
|
|
|
|
/// t6 + carry_2 = d_hi
|
|
|
|
```
|
|
- a,n,a_remainder,a_div_n 存在mul_add_words约束 a_div_n * n + a_remainder = a
|
|
- a,n,a_remainder,a_div_n 存在mul_add_words约束 a_div_n * n + a_remainder = a
|
|
- b,a_remainder,a_remainder_plus_b 存在add_words约束 a_remainder + b = a_remainder_plus_b
|
|
- b,a_remainder,a_remainder_plus_b 存在add_words约束 a_remainder + b = a_remainder_plus_b + a_remainder_plus_b_overflow << 256
|
|
- b_div_n,n,b_remainder,a_reduced_plus_b_overflow 存在mul_add_words约束 b_div_n * n + r = a_remainder_plus_b + a_remainder_plus_b_overflow << 256 (mul_add_512_gadget)
|
|
- b_div_n,n,b_remainder,a_reduced_plus_b_overflow 存在mul_add_words约束 b_div_n * n + r = a_remainder_plus_b + a_remainder_plus_b_overflow << 256 (mul_add_512_gadget)
|
|
- a_remainder < n 约束
|
|
- a_remainder < n 约束
|
|
- r < n 约束
|
|
- r < n 约束
|
... | @@ -184,16 +204,19 @@ operand* 用来存放算术中的参数值,如 a+b=c+overflow 指令中的 a,b |
... | @@ -184,16 +204,19 @@ operand* 用来存放算术中的参数值,如 a+b=c+overflow 指令中的 a,b |
|
|
|
|
|
|
|
|
|
- MulMod 计算mulMod操作码我们等于验证a,b,n,r 其中n是mod值,r是余数。我们有a*b%n = r。我们可以将这个约束转化为a*b = n * q + r。为了约束简单我们可以有
|
|
- MulMod 计算mulMod操作码我们等于验证a,b,n,r 其中n是mod值,r是余数。我们有a*b%n = r。我们可以将这个约束转化为a*b = n * q + r。为了约束简单我们可以有
|
|
a % n= a_div_n + a_remainder
|
|
a % n= a_div_n + a_remainder -> a = n * a_div_n + a_remainder
|
|
a_remainder * b =pordut & n= b_div_n + r
|
|
a_remainder * b =product & n= b_div_n + r
|
|
pordut是512bit,我们将它分为两个256bit的数,pordut_hi,pordut_lo。我们可以有以下约束
|
|
product是512bit,我们将它分为两个256bit的数,product_hi,product_lo。我们可以有以下约束
|
|
- a,n,a_remainder,a_div_n 存在mul_add_words约束
|
|
- a,n,a_remainder,a_div_n 存在mul_words约束
|
|
- b,a_remainder,pordut_hi,pordut_lo 存在mul_add_words约束,请注意这里b*a_remainder的结果等于pordut是512bit
|
|
- b,a_remainder,product_hi,product_lo 存在mul_add_words约束,请注意这里b*a_remainder的结果等于port是512bit.我们最后有product_hi * 2^256 + product_lo = b * a_remainder
|
|
- pordut_hi,pordut_lo,n,b_div_n,r 存在mul_add_words约束,同样需要注意这里n * b_div_n + r 的结果等于pordut是512bit
|
|
- product_hi,product_lo,n,b_div_n,r 存在mul_add_words约束,同样需要注意这里n * b_div_n + r 的结果等于product是512bit.我们最后有product_hi * 2^256 + product_lo = n * b_div_n + r
|
|
|
|
- a_remainder < n 约束
|
|
- r < n 约束
|
|
- r < n 约束
|
|
- n_is_zero 约束
|
|
- n_is_zero 约束
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
- Length
|
|
- Length
|
|
arithmetic中的布局:
|
|
arithmetic中的布局:
|
|
|
|
|
... | | ... | |