-rw-r--r-- 90559 lib25519-20230630/crypto_mGnP/ed25519/amd64-64/ge25519_pnielsadd_p1p1.S raw
#include "crypto_asm_hidden.h" // linker define ge25519_pnielsadd_p1p1 // linker use 38 # qhasm: int64 rp # qhasm: int64 pp # qhasm: int64 qp # qhasm: input rp # qhasm: input pp # qhasm: input qp # 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 a0 # qhasm: int64 a1 # qhasm: int64 a2 # qhasm: int64 a3 # qhasm: stack64 a0_stack # qhasm: stack64 a1_stack # qhasm: stack64 a2_stack # qhasm: stack64 a3_stack # qhasm: int64 b0 # qhasm: int64 b1 # qhasm: int64 b2 # qhasm: int64 b3 # qhasm: stack64 b0_stack # qhasm: stack64 b1_stack # qhasm: stack64 b2_stack # qhasm: stack64 b3_stack # qhasm: int64 c0 # qhasm: int64 c1 # qhasm: int64 c2 # qhasm: int64 c3 # qhasm: stack64 c0_stack # qhasm: stack64 c1_stack # qhasm: stack64 c2_stack # qhasm: stack64 c3_stack # qhasm: int64 d0 # qhasm: int64 d1 # qhasm: int64 d2 # qhasm: int64 d3 # qhasm: stack64 d0_stack # qhasm: stack64 d1_stack # qhasm: stack64 d2_stack # qhasm: stack64 d3_stack # qhasm: int64 t10 # qhasm: int64 t11 # qhasm: int64 t12 # qhasm: int64 t13 # qhasm: stack64 t10_stack # qhasm: stack64 t11_stack # qhasm: stack64 t12_stack # qhasm: stack64 t13_stack # qhasm: int64 t20 # qhasm: int64 t21 # qhasm: int64 t22 # qhasm: int64 t23 # qhasm: stack64 t20_stack # qhasm: stack64 t21_stack # qhasm: stack64 t22_stack # qhasm: stack64 t23_stack # qhasm: int64 rx0 # qhasm: int64 rx1 # qhasm: int64 rx2 # qhasm: int64 rx3 # qhasm: int64 ry0 # qhasm: int64 ry1 # qhasm: int64 ry2 # qhasm: int64 ry3 # qhasm: int64 rz0 # qhasm: int64 rz1 # qhasm: int64 rz2 # qhasm: int64 rz3 # qhasm: int64 rt0 # qhasm: int64 rt1 # qhasm: int64 rt2 # qhasm: int64 rt3 # qhasm: int64 x0 # qhasm: int64 x1 # qhasm: int64 x2 # qhasm: int64 x3 # 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: int64 addt0 # qhasm: int64 addt1 # qhasm: int64 subt0 # qhasm: int64 subt1 # qhasm: enter CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1) .text .p2align 5 ASM_HIDDEN _CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1) .globl _CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1) ASM_HIDDEN CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1) .globl CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1) _CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1): CRYPTO_SHARED_NAMESPACE(ge25519_pnielsadd_p1p1): mov %rsp,%r11 and $31,%r11 add $128,%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: qp = qp # asm 1: mov <qp=int64#3,>qp=int64#4 # asm 2: mov <qp=%rdx,>qp=%rcx mov %rdx,%rcx # qhasm: a0 = *(uint64 *)(pp + 32) # asm 1: movq 32(<pp=int64#2),>a0=int64#3 # asm 2: movq 32(<pp=%rsi),>a0=%rdx movq 32(%rsi),%rdx # qhasm: a1 = *(uint64 *)(pp + 40) # asm 1: movq 40(<pp=int64#2),>a1=int64#5 # asm 2: movq 40(<pp=%rsi),>a1=%r8 movq 40(%rsi),%r8 # qhasm: a2 = *(uint64 *)(pp + 48) # asm 1: movq 48(<pp=int64#2),>a2=int64#6 # asm 2: movq 48(<pp=%rsi),>a2=%r9 movq 48(%rsi),%r9 # qhasm: a3 = *(uint64 *)(pp + 56) # asm 1: movq 56(<pp=int64#2),>a3=int64#7 # asm 2: movq 56(<pp=%rsi),>a3=%rax movq 56(%rsi),%rax # qhasm: b0 = a0 # asm 1: mov <a0=int64#3,>b0=int64#8 # asm 2: mov <a0=%rdx,>b0=%r10 mov %rdx,%r10 # qhasm: b1 = a1 # asm 1: mov <a1=int64#5,>b1=int64#9 # asm 2: mov <a1=%r8,>b1=%r11 mov %r8,%r11 # qhasm: b2 = a2 # asm 1: mov <a2=int64#6,>b2=int64#10 # asm 2: mov <a2=%r9,>b2=%r12 mov %r9,%r12 # qhasm: b3 = a3 # asm 1: mov <a3=int64#7,>b3=int64#11 # asm 2: mov <a3=%rax,>b3=%r13 mov %rax,%r13 # qhasm: carry? a0 -= *(uint64 *)(pp + 0) # asm 1: subq 0(<pp=int64#2),<a0=int64#3 # asm 2: subq 0(<pp=%rsi),<a0=%rdx subq 0(%rsi),%rdx # qhasm: carry? a1 -= *(uint64 *)(pp + 8) - carry # asm 1: sbbq 8(<pp=int64#2),<a1=int64#5 # asm 2: sbbq 8(<pp=%rsi),<a1=%r8 sbbq 8(%rsi),%r8 # qhasm: carry? a2 -= *(uint64 *)(pp + 16) - carry # asm 1: sbbq 16(<pp=int64#2),<a2=int64#6 # asm 2: sbbq 16(<pp=%rsi),<a2=%r9 sbbq 16(%rsi),%r9 # qhasm: carry? a3 -= *(uint64 *)(pp + 24) - carry # asm 1: sbbq 24(<pp=int64#2),<a3=int64#7 # asm 2: sbbq 24(<pp=%rsi),<a3=%rax sbbq 24(%rsi),%rax # qhasm: subt0 = 0 # asm 1: mov $0,>subt0=int64#12 # asm 2: mov $0,>subt0=%r14 mov $0,%r14 # qhasm: subt1 = 38 # asm 1: mov $38,>subt1=int64#13 # asm 2: mov $38,>subt1=%r15 mov $38,%r15 # qhasm: subt1 = subt0 if !carry # asm 1: cmovae <subt0=int64#12,<subt1=int64#13 # asm 2: cmovae <subt0=%r14,<subt1=%r15 cmovae %r14,%r15 # qhasm: carry? a0 -= subt1 # asm 1: sub <subt1=int64#13,<a0=int64#3 # asm 2: sub <subt1=%r15,<a0=%rdx sub %r15,%rdx # qhasm: carry? a1 -= subt0 - carry # asm 1: sbb <subt0=int64#12,<a1=int64#5 # asm 2: sbb <subt0=%r14,<a1=%r8 sbb %r14,%r8 # qhasm: carry? a2 -= subt0 - carry # asm 1: sbb <subt0=int64#12,<a2=int64#6 # asm 2: sbb <subt0=%r14,<a2=%r9 sbb %r14,%r9 # qhasm: carry? a3 -= subt0 - carry # asm 1: sbb <subt0=int64#12,<a3=int64#7 # asm 2: sbb <subt0=%r14,<a3=%rax sbb %r14,%rax # qhasm: subt0 = subt1 if carry # asm 1: cmovc <subt1=int64#13,<subt0=int64#12 # asm 2: cmovc <subt1=%r15,<subt0=%r14 cmovc %r15,%r14 # qhasm: a0 -= subt0 # asm 1: sub <subt0=int64#12,<a0=int64#3 # asm 2: sub <subt0=%r14,<a0=%rdx sub %r14,%rdx # qhasm: carry? b0 += *(uint64 *)(pp + 0) # asm 1: addq 0(<pp=int64#2),<b0=int64#8 # asm 2: addq 0(<pp=%rsi),<b0=%r10 addq 0(%rsi),%r10 # qhasm: carry? b1 += *(uint64 *)(pp + 8) + carry # asm 1: adcq 8(<pp=int64#2),<b1=int64#9 # asm 2: adcq 8(<pp=%rsi),<b1=%r11 adcq 8(%rsi),%r11 # qhasm: carry? b2 += *(uint64 *)(pp + 16) + carry # asm 1: adcq 16(<pp=int64#2),<b2=int64#10 # asm 2: adcq 16(<pp=%rsi),<b2=%r12 adcq 16(%rsi),%r12 # qhasm: carry? b3 += *(uint64 *)(pp + 24) + carry # asm 1: adcq 24(<pp=int64#2),<b3=int64#11 # asm 2: adcq 24(<pp=%rsi),<b3=%r13 adcq 24(%rsi),%r13 # qhasm: addt0 = 0 # asm 1: mov $0,>addt0=int64#12 # asm 2: mov $0,>addt0=%r14 mov $0,%r14 # qhasm: addt1 = 38 # asm 1: mov $38,>addt1=int64#13 # asm 2: mov $38,>addt1=%r15 mov $38,%r15 # qhasm: addt1 = addt0 if !carry # asm 1: cmovae <addt0=int64#12,<addt1=int64#13 # asm 2: cmovae <addt0=%r14,<addt1=%r15 cmovae %r14,%r15 # qhasm: carry? b0 += addt1 # asm 1: add <addt1=int64#13,<b0=int64#8 # asm 2: add <addt1=%r15,<b0=%r10 add %r15,%r10 # qhasm: carry? b1 += addt0 + carry # asm 1: adc <addt0=int64#12,<b1=int64#9 # asm 2: adc <addt0=%r14,<b1=%r11 adc %r14,%r11 # qhasm: carry? b2 += addt0 + carry # asm 1: adc <addt0=int64#12,<b2=int64#10 # asm 2: adc <addt0=%r14,<b2=%r12 adc %r14,%r12 # qhasm: carry? b3 += addt0 + carry # asm 1: adc <addt0=int64#12,<b3=int64#11 # asm 2: adc <addt0=%r14,<b3=%r13 adc %r14,%r13 # qhasm: addt0 = addt1 if carry # asm 1: cmovc <addt1=int64#13,<addt0=int64#12 # asm 2: cmovc <addt1=%r15,<addt0=%r14 cmovc %r15,%r14 # qhasm: b0 += addt0 # asm 1: add <addt0=int64#12,<b0=int64#8 # asm 2: add <addt0=%r14,<b0=%r10 add %r14,%r10 # qhasm: a0_stack = a0 # asm 1: movq <a0=int64#3,>a0_stack=stack64#8 # asm 2: movq <a0=%rdx,>a0_stack=56(%rsp) movq %rdx,56(%rsp) # qhasm: a1_stack = a1 # asm 1: movq <a1=int64#5,>a1_stack=stack64#9 # asm 2: movq <a1=%r8,>a1_stack=64(%rsp) movq %r8,64(%rsp) # qhasm: a2_stack = a2 # asm 1: movq <a2=int64#6,>a2_stack=stack64#10 # asm 2: movq <a2=%r9,>a2_stack=72(%rsp) movq %r9,72(%rsp) # qhasm: a3_stack = a3 # asm 1: movq <a3=int64#7,>a3_stack=stack64#11 # asm 2: movq <a3=%rax,>a3_stack=80(%rsp) movq %rax,80(%rsp) # qhasm: b0_stack = b0 # asm 1: movq <b0=int64#8,>b0_stack=stack64#12 # asm 2: movq <b0=%r10,>b0_stack=88(%rsp) movq %r10,88(%rsp) # qhasm: b1_stack = b1 # asm 1: movq <b1=int64#9,>b1_stack=stack64#13 # asm 2: movq <b1=%r11,>b1_stack=96(%rsp) movq %r11,96(%rsp) # qhasm: b2_stack = b2 # asm 1: movq <b2=int64#10,>b2_stack=stack64#14 # asm 2: movq <b2=%r12,>b2_stack=104(%rsp) movq %r12,104(%rsp) # qhasm: b3_stack = b3 # asm 1: movq <b3=int64#11,>b3_stack=stack64#15 # asm 2: movq <b3=%r13,>b3_stack=112(%rsp) movq %r13,112(%rsp) # 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 = a0_stack # asm 1: movq <a0_stack=stack64#8,>mulx0=int64#10 # asm 2: movq <a0_stack=56(%rsp),>mulx0=%r12 movq 56(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 0) # asm 1: movq 0(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 0(<qp=%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: a0 = mulrax # asm 1: mov <mulrax=int64#7,>a0=int64#11 # asm 2: mov <mulrax=%rax,>a0=%r13 mov %rax,%r13 # qhasm: a1 = mulrdx # asm 1: mov <mulrdx=int64#3,>a1=int64#12 # asm 2: mov <mulrdx=%rdx,>a1=%r14 mov %rdx,%r14 # qhasm: mulrax = *(uint64 *)(qp + 8) # asm 1: movq 8(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 8(<qp=%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? a1 += mulrax # asm 1: add <mulrax=int64#7,<a1=int64#12 # asm 2: add <mulrax=%rax,<a1=%r14 add %rax,%r14 # qhasm: a2 = 0 # asm 1: mov $0,>a2=int64#13 # asm 2: mov $0,>a2=%r15 mov $0,%r15 # qhasm: a2 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<a2=int64#13 # asm 2: adc <mulrdx=%rdx,<a2=%r15 adc %rdx,%r15 # qhasm: mulrax = *(uint64 *)(qp + 16) # asm 1: movq 16(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 16(<qp=%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? a2 += mulrax # asm 1: add <mulrax=int64#7,<a2=int64#13 # asm 2: add <mulrax=%rax,<a2=%r15 add %rax,%r15 # qhasm: a3 = 0 # asm 1: mov $0,>a3=int64#14 # asm 2: mov $0,>a3=%rbx mov $0,%rbx # qhasm: a3 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<a3=int64#14 # asm 2: adc <mulrdx=%rdx,<a3=%rbx adc %rdx,%rbx # qhasm: mulrax = *(uint64 *)(qp + 24) # asm 1: movq 24(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 24(<qp=%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? a3 += mulrax # asm 1: add <mulrax=int64#7,<a3=int64#14 # asm 2: add <mulrax=%rax,<a3=%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 = a1_stack # asm 1: movq <a1_stack=stack64#9,>mulx1=int64#10 # asm 2: movq <a1_stack=64(%rsp),>mulx1=%r12 movq 64(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 0) # asm 1: movq 0(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 0(<qp=%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? a1 += mulrax # asm 1: add <mulrax=int64#7,<a1=int64#12 # asm 2: add <mulrax=%rax,<a1=%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 *)(qp + 8) # asm 1: movq 8(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 8(<qp=%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? a2 += mulrax # asm 1: add <mulrax=int64#7,<a2=int64#13 # asm 2: add <mulrax=%rax,<a2=%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? a2 += mulc # asm 1: add <mulc=int64#15,<a2=int64#13 # asm 2: add <mulc=%rbp,<a2=%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 *)(qp + 16) # asm 1: movq 16(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 16(<qp=%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? a3 += mulrax # asm 1: add <mulrax=int64#7,<a3=int64#14 # asm 2: add <mulrax=%rax,<a3=%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? a3 += mulc # asm 1: add <mulc=int64#15,<a3=int64#14 # asm 2: add <mulc=%rbp,<a3=%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 *)(qp + 24) # asm 1: movq 24(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 24(<qp=%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 = a2_stack # asm 1: movq <a2_stack=stack64#10,>mulx2=int64#10 # asm 2: movq <a2_stack=72(%rsp),>mulx2=%r12 movq 72(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 0) # asm 1: movq 0(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 0(<qp=%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? a2 += mulrax # asm 1: add <mulrax=int64#7,<a2=int64#13 # asm 2: add <mulrax=%rax,<a2=%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 *)(qp + 8) # asm 1: movq 8(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 8(<qp=%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? a3 += mulrax # asm 1: add <mulrax=int64#7,<a3=int64#14 # asm 2: add <mulrax=%rax,<a3=%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? a3 += mulc # asm 1: add <mulc=int64#15,<a3=int64#14 # asm 2: add <mulc=%rbp,<a3=%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 *)(qp + 16) # asm 1: movq 16(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 16(<qp=%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 *)(qp + 24) # asm 1: movq 24(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 24(<qp=%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 = a3_stack # asm 1: movq <a3_stack=stack64#11,>mulx3=int64#10 # asm 2: movq <a3_stack=80(%rsp),>mulx3=%r12 movq 80(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 0) # asm 1: movq 0(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 0(<qp=%rcx),>mulrax=%rax movq 0(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%r12 mul %r12 # qhasm: carry? a3 += mulrax # asm 1: add <mulrax=int64#7,<a3=int64#14 # asm 2: add <mulrax=%rax,<a3=%rbx add %rax,%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 *)(qp + 8) # asm 1: movq 8(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 8(<qp=%rcx),>mulrax=%rax movq 8(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%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 *)(qp + 16) # asm 1: movq 16(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 16(<qp=%rcx),>mulrax=%rax movq 16(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%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: 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 *)(qp + 24) # asm 1: movq 24(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 24(<qp=%rcx),>mulrax=%rax movq 24(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%r12 mul %r12 # 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#15,<mulr6=int64#8 # asm 2: add <mulc=%rbp,<mulr6=%r10 add %rbp,%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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: mulr4 = mulrax # asm 1: mov <mulrax=int64#7,>mulr4=int64#5 # asm 2: mov <mulrax=%rax,>mulr4=%r8 mov %rax,%r8 # 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#6 # asm 2: mov <mulrdx=%rdx,>mulr5=%r9 mov %rdx,%r9 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr5 += mulrax # asm 1: add <mulrax=int64#7,<mulr5=int64#6 # asm 2: add <mulrax=%rax,<mulr5=%r9 add %rax,%r9 # 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#8 # asm 2: mov $0,>mulr6=%r10 mov $0,%r10 # qhasm: mulr6 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<mulr6=int64#8 # asm 2: adc <mulrdx=%rdx,<mulr6=%r10 adc %rdx,%r10 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr6 += mulrax # asm 1: add <mulrax=int64#7,<mulr6=int64#8 # asm 2: add <mulrax=%rax,<mulr6=%r10 add %rax,%r10 # 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#9 # asm 2: mov $0,>mulr7=%r11 mov $0,%r11 # qhasm: mulr7 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<mulr7=int64#9 # asm 2: adc <mulrdx=%rdx,<mulr7=%r11 adc %rdx,%r11 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr7 += mulrax # asm 1: add <mulrax=int64#7,<mulr7=int64#9 # asm 2: add <mulrax=%rax,<mulr7=%r11 add %rax,%r11 # 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? a0 += mulr4 # asm 1: add <mulr4=int64#5,<a0=int64#11 # asm 2: add <mulr4=%r8,<a0=%r13 add %r8,%r13 # qhasm: carry? a1 += mulr5 + carry # asm 1: adc <mulr5=int64#6,<a1=int64#12 # asm 2: adc <mulr5=%r9,<a1=%r14 adc %r9,%r14 # qhasm: carry? a2 += mulr6 + carry # asm 1: adc <mulr6=int64#8,<a2=int64#13 # asm 2: adc <mulr6=%r10,<a2=%r15 adc %r10,%r15 # qhasm: carry? a3 += mulr7 + carry # asm 1: adc <mulr7=int64#9,<a3=int64#14 # asm 2: adc <mulr7=%r11,<a3=%rbx adc %r11,%rbx # qhasm: mulzero = 0 # asm 1: mov $0,>mulzero=int64#3 # asm 2: mov $0,>mulzero=%rdx mov $0,%rdx # qhasm: mulr8 += mulzero + carry # asm 1: adc <mulzero=int64#3,<mulr8=int64#7 # asm 2: adc <mulzero=%rdx,<mulr8=%rax adc %rdx,%rax # qhasm: mulr8 *= 38 # asm 1: imulq $38,<mulr8=int64#7,>mulr8=int64#5 # asm 2: imulq $38,<mulr8=%rax,>mulr8=%r8 imulq $38,%rax,%r8 # qhasm: carry? a0 += mulr8 # asm 1: add <mulr8=int64#5,<a0=int64#11 # asm 2: add <mulr8=%r8,<a0=%r13 add %r8,%r13 # qhasm: carry? a1 += mulzero + carry # asm 1: adc <mulzero=int64#3,<a1=int64#12 # asm 2: adc <mulzero=%rdx,<a1=%r14 adc %rdx,%r14 # qhasm: carry? a2 += mulzero + carry # asm 1: adc <mulzero=int64#3,<a2=int64#13 # asm 2: adc <mulzero=%rdx,<a2=%r15 adc %rdx,%r15 # qhasm: carry? a3 += mulzero + carry # asm 1: adc <mulzero=int64#3,<a3=int64#14 # asm 2: adc <mulzero=%rdx,<a3=%rbx adc %rdx,%rbx # qhasm: mulzero += mulzero + carry # asm 1: adc <mulzero=int64#3,<mulzero=int64#3 # asm 2: adc <mulzero=%rdx,<mulzero=%rdx adc %rdx,%rdx # qhasm: mulzero *= 38 # asm 1: imulq $38,<mulzero=int64#3,>mulzero=int64#3 # asm 2: imulq $38,<mulzero=%rdx,>mulzero=%rdx imulq $38,%rdx,%rdx # qhasm: a0 += mulzero # asm 1: add <mulzero=int64#3,<a0=int64#11 # asm 2: add <mulzero=%rdx,<a0=%r13 add %rdx,%r13 # qhasm: a0_stack = a0 # asm 1: movq <a0=int64#11,>a0_stack=stack64#8 # asm 2: movq <a0=%r13,>a0_stack=56(%rsp) movq %r13,56(%rsp) # qhasm: a1_stack = a1 # asm 1: movq <a1=int64#12,>a1_stack=stack64#9 # asm 2: movq <a1=%r14,>a1_stack=64(%rsp) movq %r14,64(%rsp) # qhasm: a2_stack = a2 # asm 1: movq <a2=int64#13,>a2_stack=stack64#10 # asm 2: movq <a2=%r15,>a2_stack=72(%rsp) movq %r15,72(%rsp) # qhasm: a3_stack = a3 # asm 1: movq <a3=int64#14,>a3_stack=stack64#11 # asm 2: movq <a3=%rbx,>a3_stack=80(%rsp) movq %rbx,80(%rsp) # 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 = b0_stack # asm 1: movq <b0_stack=stack64#12,>mulx0=int64#10 # asm 2: movq <b0_stack=88(%rsp),>mulx0=%r12 movq 88(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 32) # asm 1: movq 32(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 32(<qp=%rcx),>mulrax=%rax movq 32(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: rx0 = mulrax # asm 1: mov <mulrax=int64#7,>rx0=int64#11 # asm 2: mov <mulrax=%rax,>rx0=%r13 mov %rax,%r13 # qhasm: rx1 = mulrdx # asm 1: mov <mulrdx=int64#3,>rx1=int64#12 # asm 2: mov <mulrdx=%rdx,>rx1=%r14 mov %rdx,%r14 # qhasm: mulrax = *(uint64 *)(qp + 40) # asm 1: movq 40(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 40(<qp=%rcx),>mulrax=%rax movq 40(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? rx1 += mulrax # asm 1: add <mulrax=int64#7,<rx1=int64#12 # asm 2: add <mulrax=%rax,<rx1=%r14 add %rax,%r14 # qhasm: rx2 = 0 # asm 1: mov $0,>rx2=int64#13 # asm 2: mov $0,>rx2=%r15 mov $0,%r15 # qhasm: rx2 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<rx2=int64#13 # asm 2: adc <mulrdx=%rdx,<rx2=%r15 adc %rdx,%r15 # qhasm: mulrax = *(uint64 *)(qp + 48) # asm 1: movq 48(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 48(<qp=%rcx),>mulrax=%rax movq 48(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? rx2 += mulrax # asm 1: add <mulrax=int64#7,<rx2=int64#13 # asm 2: add <mulrax=%rax,<rx2=%r15 add %rax,%r15 # qhasm: rx3 = 0 # asm 1: mov $0,>rx3=int64#14 # asm 2: mov $0,>rx3=%rbx mov $0,%rbx # qhasm: rx3 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<rx3=int64#14 # asm 2: adc <mulrdx=%rdx,<rx3=%rbx adc %rdx,%rbx # qhasm: mulrax = *(uint64 *)(qp + 56) # asm 1: movq 56(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 56(<qp=%rcx),>mulrax=%rax movq 56(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? rx3 += mulrax # asm 1: add <mulrax=int64#7,<rx3=int64#14 # asm 2: add <mulrax=%rax,<rx3=%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 = b1_stack # asm 1: movq <b1_stack=stack64#13,>mulx1=int64#10 # asm 2: movq <b1_stack=96(%rsp),>mulx1=%r12 movq 96(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 32) # asm 1: movq 32(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 32(<qp=%rcx),>mulrax=%rax movq 32(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? rx1 += mulrax # asm 1: add <mulrax=int64#7,<rx1=int64#12 # asm 2: add <mulrax=%rax,<rx1=%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 *)(qp + 40) # asm 1: movq 40(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 40(<qp=%rcx),>mulrax=%rax movq 40(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? rx2 += mulrax # asm 1: add <mulrax=int64#7,<rx2=int64#13 # asm 2: add <mulrax=%rax,<rx2=%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? rx2 += mulc # asm 1: add <mulc=int64#15,<rx2=int64#13 # asm 2: add <mulc=%rbp,<rx2=%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 *)(qp + 48) # asm 1: movq 48(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 48(<qp=%rcx),>mulrax=%rax movq 48(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? rx3 += mulrax # asm 1: add <mulrax=int64#7,<rx3=int64#14 # asm 2: add <mulrax=%rax,<rx3=%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? rx3 += mulc # asm 1: add <mulc=int64#15,<rx3=int64#14 # asm 2: add <mulc=%rbp,<rx3=%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 *)(qp + 56) # asm 1: movq 56(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 56(<qp=%rcx),>mulrax=%rax movq 56(%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 = b2_stack # asm 1: movq <b2_stack=stack64#14,>mulx2=int64#10 # asm 2: movq <b2_stack=104(%rsp),>mulx2=%r12 movq 104(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 32) # asm 1: movq 32(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 32(<qp=%rcx),>mulrax=%rax movq 32(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx2 # asm 1: mul <mulx2=int64#10 # asm 2: mul <mulx2=%r12 mul %r12 # qhasm: carry? rx2 += mulrax # asm 1: add <mulrax=int64#7,<rx2=int64#13 # asm 2: add <mulrax=%rax,<rx2=%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 *)(qp + 40) # asm 1: movq 40(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 40(<qp=%rcx),>mulrax=%rax movq 40(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx2 # asm 1: mul <mulx2=int64#10 # asm 2: mul <mulx2=%r12 mul %r12 # qhasm: carry? rx3 += mulrax # asm 1: add <mulrax=int64#7,<rx3=int64#14 # asm 2: add <mulrax=%rax,<rx3=%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? rx3 += mulc # asm 1: add <mulc=int64#15,<rx3=int64#14 # asm 2: add <mulc=%rbp,<rx3=%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 *)(qp + 48) # asm 1: movq 48(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 48(<qp=%rcx),>mulrax=%rax movq 48(%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 *)(qp + 56) # asm 1: movq 56(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 56(<qp=%rcx),>mulrax=%rax movq 56(%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 = b3_stack # asm 1: movq <b3_stack=stack64#15,>mulx3=int64#10 # asm 2: movq <b3_stack=112(%rsp),>mulx3=%r12 movq 112(%rsp),%r12 # qhasm: mulrax = *(uint64 *)(qp + 32) # asm 1: movq 32(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 32(<qp=%rcx),>mulrax=%rax movq 32(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%r12 mul %r12 # qhasm: carry? rx3 += mulrax # asm 1: add <mulrax=int64#7,<rx3=int64#14 # asm 2: add <mulrax=%rax,<rx3=%rbx add %rax,%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 *)(qp + 40) # asm 1: movq 40(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 40(<qp=%rcx),>mulrax=%rax movq 40(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%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 *)(qp + 48) # asm 1: movq 48(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 48(<qp=%rcx),>mulrax=%rax movq 48(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%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: 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 *)(qp + 56) # asm 1: movq 56(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 56(<qp=%rcx),>mulrax=%rax movq 56(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%r12 mul %r12 # 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#15,<mulr6=int64#8 # asm 2: add <mulc=%rbp,<mulr6=%r10 add %rbp,%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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: mulr4 = mulrax # asm 1: mov <mulrax=int64#7,>mulr4=int64#5 # asm 2: mov <mulrax=%rax,>mulr4=%r8 mov %rax,%r8 # 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#6 # asm 2: mov <mulrdx=%rdx,>mulr5=%r9 mov %rdx,%r9 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr5 += mulrax # asm 1: add <mulrax=int64#7,<mulr5=int64#6 # asm 2: add <mulrax=%rax,<mulr5=%r9 add %rax,%r9 # 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#8 # asm 2: mov $0,>mulr6=%r10 mov $0,%r10 # qhasm: mulr6 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<mulr6=int64#8 # asm 2: adc <mulrdx=%rdx,<mulr6=%r10 adc %rdx,%r10 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr6 += mulrax # asm 1: add <mulrax=int64#7,<mulr6=int64#8 # asm 2: add <mulrax=%rax,<mulr6=%r10 add %rax,%r10 # 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#9 # asm 2: mov $0,>mulr7=%r11 mov $0,%r11 # qhasm: mulr7 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<mulr7=int64#9 # asm 2: adc <mulrdx=%rdx,<mulr7=%r11 adc %rdx,%r11 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr7 += mulrax # asm 1: add <mulrax=int64#7,<mulr7=int64#9 # asm 2: add <mulrax=%rax,<mulr7=%r11 add %rax,%r11 # 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? rx0 += mulr4 # asm 1: add <mulr4=int64#5,<rx0=int64#11 # asm 2: add <mulr4=%r8,<rx0=%r13 add %r8,%r13 # qhasm: carry? rx1 += mulr5 + carry # asm 1: adc <mulr5=int64#6,<rx1=int64#12 # asm 2: adc <mulr5=%r9,<rx1=%r14 adc %r9,%r14 # qhasm: carry? rx2 += mulr6 + carry # asm 1: adc <mulr6=int64#8,<rx2=int64#13 # asm 2: adc <mulr6=%r10,<rx2=%r15 adc %r10,%r15 # qhasm: carry? rx3 += mulr7 + carry # asm 1: adc <mulr7=int64#9,<rx3=int64#14 # asm 2: adc <mulr7=%r11,<rx3=%rbx adc %r11,%rbx # qhasm: mulzero = 0 # asm 1: mov $0,>mulzero=int64#3 # asm 2: mov $0,>mulzero=%rdx mov $0,%rdx # qhasm: mulr8 += mulzero + carry # asm 1: adc <mulzero=int64#3,<mulr8=int64#7 # asm 2: adc <mulzero=%rdx,<mulr8=%rax adc %rdx,%rax # qhasm: mulr8 *= 38 # asm 1: imulq $38,<mulr8=int64#7,>mulr8=int64#5 # asm 2: imulq $38,<mulr8=%rax,>mulr8=%r8 imulq $38,%rax,%r8 # qhasm: carry? rx0 += mulr8 # asm 1: add <mulr8=int64#5,<rx0=int64#11 # asm 2: add <mulr8=%r8,<rx0=%r13 add %r8,%r13 # qhasm: carry? rx1 += mulzero + carry # asm 1: adc <mulzero=int64#3,<rx1=int64#12 # asm 2: adc <mulzero=%rdx,<rx1=%r14 adc %rdx,%r14 # qhasm: carry? rx2 += mulzero + carry # asm 1: adc <mulzero=int64#3,<rx2=int64#13 # asm 2: adc <mulzero=%rdx,<rx2=%r15 adc %rdx,%r15 # qhasm: carry? rx3 += mulzero + carry # asm 1: adc <mulzero=int64#3,<rx3=int64#14 # asm 2: adc <mulzero=%rdx,<rx3=%rbx adc %rdx,%rbx # qhasm: mulzero += mulzero + carry # asm 1: adc <mulzero=int64#3,<mulzero=int64#3 # asm 2: adc <mulzero=%rdx,<mulzero=%rdx adc %rdx,%rdx # qhasm: mulzero *= 38 # asm 1: imulq $38,<mulzero=int64#3,>mulzero=int64#3 # asm 2: imulq $38,<mulzero=%rdx,>mulzero=%rdx imulq $38,%rdx,%rdx # qhasm: rx0 += mulzero # asm 1: add <mulzero=int64#3,<rx0=int64#11 # asm 2: add <mulzero=%rdx,<rx0=%r13 add %rdx,%r13 # qhasm: ry0 = rx0 # asm 1: mov <rx0=int64#11,>ry0=int64#3 # asm 2: mov <rx0=%r13,>ry0=%rdx mov %r13,%rdx # qhasm: ry1 = rx1 # asm 1: mov <rx1=int64#12,>ry1=int64#5 # asm 2: mov <rx1=%r14,>ry1=%r8 mov %r14,%r8 # qhasm: ry2 = rx2 # asm 1: mov <rx2=int64#13,>ry2=int64#6 # asm 2: mov <rx2=%r15,>ry2=%r9 mov %r15,%r9 # qhasm: ry3 = rx3 # asm 1: mov <rx3=int64#14,>ry3=int64#7 # asm 2: mov <rx3=%rbx,>ry3=%rax mov %rbx,%rax # qhasm: carry? ry0 += a0_stack # asm 1: addq <a0_stack=stack64#8,<ry0=int64#3 # asm 2: addq <a0_stack=56(%rsp),<ry0=%rdx addq 56(%rsp),%rdx # qhasm: carry? ry1 += a1_stack + carry # asm 1: adcq <a1_stack=stack64#9,<ry1=int64#5 # asm 2: adcq <a1_stack=64(%rsp),<ry1=%r8 adcq 64(%rsp),%r8 # qhasm: carry? ry2 += a2_stack + carry # asm 1: adcq <a2_stack=stack64#10,<ry2=int64#6 # asm 2: adcq <a2_stack=72(%rsp),<ry2=%r9 adcq 72(%rsp),%r9 # qhasm: carry? ry3 += a3_stack + carry # asm 1: adcq <a3_stack=stack64#11,<ry3=int64#7 # asm 2: adcq <a3_stack=80(%rsp),<ry3=%rax adcq 80(%rsp),%rax # qhasm: addt0 = 0 # asm 1: mov $0,>addt0=int64#8 # asm 2: mov $0,>addt0=%r10 mov $0,%r10 # qhasm: addt1 = 38 # asm 1: mov $38,>addt1=int64#9 # asm 2: mov $38,>addt1=%r11 mov $38,%r11 # qhasm: addt1 = addt0 if !carry # asm 1: cmovae <addt0=int64#8,<addt1=int64#9 # asm 2: cmovae <addt0=%r10,<addt1=%r11 cmovae %r10,%r11 # qhasm: carry? ry0 += addt1 # asm 1: add <addt1=int64#9,<ry0=int64#3 # asm 2: add <addt1=%r11,<ry0=%rdx add %r11,%rdx # qhasm: carry? ry1 += addt0 + carry # asm 1: adc <addt0=int64#8,<ry1=int64#5 # asm 2: adc <addt0=%r10,<ry1=%r8 adc %r10,%r8 # qhasm: carry? ry2 += addt0 + carry # asm 1: adc <addt0=int64#8,<ry2=int64#6 # asm 2: adc <addt0=%r10,<ry2=%r9 adc %r10,%r9 # qhasm: carry? ry3 += addt0 + carry # asm 1: adc <addt0=int64#8,<ry3=int64#7 # asm 2: adc <addt0=%r10,<ry3=%rax adc %r10,%rax # qhasm: addt0 = addt1 if carry # asm 1: cmovc <addt1=int64#9,<addt0=int64#8 # asm 2: cmovc <addt1=%r11,<addt0=%r10 cmovc %r11,%r10 # qhasm: ry0 += addt0 # asm 1: add <addt0=int64#8,<ry0=int64#3 # asm 2: add <addt0=%r10,<ry0=%rdx add %r10,%rdx # qhasm: carry? rx0 -= a0_stack # asm 1: subq <a0_stack=stack64#8,<rx0=int64#11 # asm 2: subq <a0_stack=56(%rsp),<rx0=%r13 subq 56(%rsp),%r13 # qhasm: carry? rx1 -= a1_stack - carry # asm 1: sbbq <a1_stack=stack64#9,<rx1=int64#12 # asm 2: sbbq <a1_stack=64(%rsp),<rx1=%r14 sbbq 64(%rsp),%r14 # qhasm: carry? rx2 -= a2_stack - carry # asm 1: sbbq <a2_stack=stack64#10,<rx2=int64#13 # asm 2: sbbq <a2_stack=72(%rsp),<rx2=%r15 sbbq 72(%rsp),%r15 # qhasm: carry? rx3 -= a3_stack - carry # asm 1: sbbq <a3_stack=stack64#11,<rx3=int64#14 # asm 2: sbbq <a3_stack=80(%rsp),<rx3=%rbx sbbq 80(%rsp),%rbx # qhasm: subt0 = 0 # asm 1: mov $0,>subt0=int64#8 # asm 2: mov $0,>subt0=%r10 mov $0,%r10 # qhasm: subt1 = 38 # asm 1: mov $38,>subt1=int64#9 # asm 2: mov $38,>subt1=%r11 mov $38,%r11 # qhasm: subt1 = subt0 if !carry # asm 1: cmovae <subt0=int64#8,<subt1=int64#9 # asm 2: cmovae <subt0=%r10,<subt1=%r11 cmovae %r10,%r11 # qhasm: carry? rx0 -= subt1 # asm 1: sub <subt1=int64#9,<rx0=int64#11 # asm 2: sub <subt1=%r11,<rx0=%r13 sub %r11,%r13 # qhasm: carry? rx1 -= subt0 - carry # asm 1: sbb <subt0=int64#8,<rx1=int64#12 # asm 2: sbb <subt0=%r10,<rx1=%r14 sbb %r10,%r14 # qhasm: carry? rx2 -= subt0 - carry # asm 1: sbb <subt0=int64#8,<rx2=int64#13 # asm 2: sbb <subt0=%r10,<rx2=%r15 sbb %r10,%r15 # qhasm: carry? rx3 -= subt0 - carry # asm 1: sbb <subt0=int64#8,<rx3=int64#14 # asm 2: sbb <subt0=%r10,<rx3=%rbx sbb %r10,%rbx # qhasm: subt0 = subt1 if carry # asm 1: cmovc <subt1=int64#9,<subt0=int64#8 # asm 2: cmovc <subt1=%r11,<subt0=%r10 cmovc %r11,%r10 # qhasm: rx0 -= subt0 # asm 1: sub <subt0=int64#8,<rx0=int64#11 # asm 2: sub <subt0=%r10,<rx0=%r13 sub %r10,%r13 # qhasm: *(uint64 *) (rp + 0) = rx0 # asm 1: movq <rx0=int64#11,0(<rp=int64#1) # asm 2: movq <rx0=%r13,0(<rp=%rdi) movq %r13,0(%rdi) # qhasm: *(uint64 *) (rp + 8) = rx1 # asm 1: movq <rx1=int64#12,8(<rp=int64#1) # asm 2: movq <rx1=%r14,8(<rp=%rdi) movq %r14,8(%rdi) # qhasm: *(uint64 *) (rp + 16) = rx2 # asm 1: movq <rx2=int64#13,16(<rp=int64#1) # asm 2: movq <rx2=%r15,16(<rp=%rdi) movq %r15,16(%rdi) # qhasm: *(uint64 *) (rp + 24) = rx3 # asm 1: movq <rx3=int64#14,24(<rp=int64#1) # asm 2: movq <rx3=%rbx,24(<rp=%rdi) movq %rbx,24(%rdi) # qhasm: *(uint64 *) (rp + 64) = ry0 # asm 1: movq <ry0=int64#3,64(<rp=int64#1) # asm 2: movq <ry0=%rdx,64(<rp=%rdi) movq %rdx,64(%rdi) # qhasm: *(uint64 *) (rp + 72) = ry1 # asm 1: movq <ry1=int64#5,72(<rp=int64#1) # asm 2: movq <ry1=%r8,72(<rp=%rdi) movq %r8,72(%rdi) # qhasm: *(uint64 *) (rp + 80) = ry2 # asm 1: movq <ry2=int64#6,80(<rp=int64#1) # asm 2: movq <ry2=%r9,80(<rp=%rdi) movq %r9,80(%rdi) # qhasm: *(uint64 *) (rp + 88) = ry3 # asm 1: movq <ry3=int64#7,88(<rp=int64#1) # asm 2: movq <ry3=%rax,88(<rp=%rdi) movq %rax,88(%rdi) # 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 *)(pp + 96) # asm 1: movq 96(<pp=int64#2),>mulx0=int64#10 # asm 2: movq 96(<pp=%rsi),>mulx0=%r12 movq 96(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 96) # asm 1: movq 96(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 96(<qp=%rcx),>mulrax=%rax movq 96(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: c0 = mulrax # asm 1: mov <mulrax=int64#7,>c0=int64#11 # asm 2: mov <mulrax=%rax,>c0=%r13 mov %rax,%r13 # qhasm: c1 = mulrdx # asm 1: mov <mulrdx=int64#3,>c1=int64#12 # asm 2: mov <mulrdx=%rdx,>c1=%r14 mov %rdx,%r14 # qhasm: mulrax = *(uint64 *)(qp + 104) # asm 1: movq 104(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 104(<qp=%rcx),>mulrax=%rax movq 104(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? c1 += mulrax # asm 1: add <mulrax=int64#7,<c1=int64#12 # asm 2: add <mulrax=%rax,<c1=%r14 add %rax,%r14 # qhasm: c2 = 0 # asm 1: mov $0,>c2=int64#13 # asm 2: mov $0,>c2=%r15 mov $0,%r15 # qhasm: c2 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<c2=int64#13 # asm 2: adc <mulrdx=%rdx,<c2=%r15 adc %rdx,%r15 # qhasm: mulrax = *(uint64 *)(qp + 112) # asm 1: movq 112(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 112(<qp=%rcx),>mulrax=%rax movq 112(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? c2 += mulrax # asm 1: add <mulrax=int64#7,<c2=int64#13 # asm 2: add <mulrax=%rax,<c2=%r15 add %rax,%r15 # qhasm: c3 = 0 # asm 1: mov $0,>c3=int64#14 # asm 2: mov $0,>c3=%rbx mov $0,%rbx # qhasm: c3 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<c3=int64#14 # asm 2: adc <mulrdx=%rdx,<c3=%rbx adc %rdx,%rbx # qhasm: mulrax = *(uint64 *)(qp + 120) # asm 1: movq 120(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 120(<qp=%rcx),>mulrax=%rax movq 120(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? c3 += mulrax # asm 1: add <mulrax=int64#7,<c3=int64#14 # asm 2: add <mulrax=%rax,<c3=%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 *)(pp + 104) # asm 1: movq 104(<pp=int64#2),>mulx1=int64#10 # asm 2: movq 104(<pp=%rsi),>mulx1=%r12 movq 104(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 96) # asm 1: movq 96(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 96(<qp=%rcx),>mulrax=%rax movq 96(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? c1 += mulrax # asm 1: add <mulrax=int64#7,<c1=int64#12 # asm 2: add <mulrax=%rax,<c1=%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 *)(qp + 104) # asm 1: movq 104(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 104(<qp=%rcx),>mulrax=%rax movq 104(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? c2 += mulrax # asm 1: add <mulrax=int64#7,<c2=int64#13 # asm 2: add <mulrax=%rax,<c2=%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? c2 += mulc # asm 1: add <mulc=int64#15,<c2=int64#13 # asm 2: add <mulc=%rbp,<c2=%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 *)(qp + 112) # asm 1: movq 112(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 112(<qp=%rcx),>mulrax=%rax movq 112(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? c3 += mulrax # asm 1: add <mulrax=int64#7,<c3=int64#14 # asm 2: add <mulrax=%rax,<c3=%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? c3 += mulc # asm 1: add <mulc=int64#15,<c3=int64#14 # asm 2: add <mulc=%rbp,<c3=%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 *)(qp + 120) # asm 1: movq 120(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 120(<qp=%rcx),>mulrax=%rax movq 120(%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 *)(pp + 112) # asm 1: movq 112(<pp=int64#2),>mulx2=int64#10 # asm 2: movq 112(<pp=%rsi),>mulx2=%r12 movq 112(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 96) # asm 1: movq 96(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 96(<qp=%rcx),>mulrax=%rax movq 96(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx2 # asm 1: mul <mulx2=int64#10 # asm 2: mul <mulx2=%r12 mul %r12 # qhasm: carry? c2 += mulrax # asm 1: add <mulrax=int64#7,<c2=int64#13 # asm 2: add <mulrax=%rax,<c2=%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 *)(qp + 104) # asm 1: movq 104(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 104(<qp=%rcx),>mulrax=%rax movq 104(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx2 # asm 1: mul <mulx2=int64#10 # asm 2: mul <mulx2=%r12 mul %r12 # qhasm: carry? c3 += mulrax # asm 1: add <mulrax=int64#7,<c3=int64#14 # asm 2: add <mulrax=%rax,<c3=%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? c3 += mulc # asm 1: add <mulc=int64#15,<c3=int64#14 # asm 2: add <mulc=%rbp,<c3=%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 *)(qp + 112) # asm 1: movq 112(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 112(<qp=%rcx),>mulrax=%rax movq 112(%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 *)(qp + 120) # asm 1: movq 120(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 120(<qp=%rcx),>mulrax=%rax movq 120(%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 *)(pp + 120) # asm 1: movq 120(<pp=int64#2),>mulx3=int64#10 # asm 2: movq 120(<pp=%rsi),>mulx3=%r12 movq 120(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 96) # asm 1: movq 96(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 96(<qp=%rcx),>mulrax=%rax movq 96(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%r12 mul %r12 # qhasm: carry? c3 += mulrax # asm 1: add <mulrax=int64#7,<c3=int64#14 # asm 2: add <mulrax=%rax,<c3=%rbx add %rax,%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 *)(qp + 104) # asm 1: movq 104(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 104(<qp=%rcx),>mulrax=%rax movq 104(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%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 *)(qp + 112) # asm 1: movq 112(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 112(<qp=%rcx),>mulrax=%rax movq 112(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%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: 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 *)(qp + 120) # asm 1: movq 120(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 120(<qp=%rcx),>mulrax=%rax movq 120(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#10 # asm 2: mul <mulx3=%r12 mul %r12 # 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#15,<mulr6=int64#8 # asm 2: add <mulc=%rbp,<mulr6=%r10 add %rbp,%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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: mulr4 = mulrax # asm 1: mov <mulrax=int64#7,>mulr4=int64#5 # asm 2: mov <mulrax=%rax,>mulr4=%r8 mov %rax,%r8 # 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#6 # asm 2: mov <mulrdx=%rdx,>mulr5=%r9 mov %rdx,%r9 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr5 += mulrax # asm 1: add <mulrax=int64#7,<mulr5=int64#6 # asm 2: add <mulrax=%rax,<mulr5=%r9 add %rax,%r9 # 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#8 # asm 2: mov $0,>mulr6=%r10 mov $0,%r10 # qhasm: mulr6 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<mulr6=int64#8 # asm 2: adc <mulrdx=%rdx,<mulr6=%r10 adc %rdx,%r10 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr6 += mulrax # asm 1: add <mulrax=int64#7,<mulr6=int64#8 # asm 2: add <mulrax=%rax,<mulr6=%r10 add %rax,%r10 # 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#9 # asm 2: mov $0,>mulr7=%r11 mov $0,%r11 # qhasm: mulr7 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<mulr7=int64#9 # asm 2: adc <mulrdx=%rdx,<mulr7=%r11 adc %rdx,%r11 # qhasm: (uint128) mulrdx mulrax = mulrax * *(uint64 *)&CRYPTO_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # qhasm: carry? mulr7 += mulrax # asm 1: add <mulrax=int64#7,<mulr7=int64#9 # asm 2: add <mulrax=%rax,<mulr7=%r11 add %rax,%r11 # 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? c0 += mulr4 # asm 1: add <mulr4=int64#5,<c0=int64#11 # asm 2: add <mulr4=%r8,<c0=%r13 add %r8,%r13 # qhasm: carry? c1 += mulr5 + carry # asm 1: adc <mulr5=int64#6,<c1=int64#12 # asm 2: adc <mulr5=%r9,<c1=%r14 adc %r9,%r14 # qhasm: carry? c2 += mulr6 + carry # asm 1: adc <mulr6=int64#8,<c2=int64#13 # asm 2: adc <mulr6=%r10,<c2=%r15 adc %r10,%r15 # qhasm: carry? c3 += mulr7 + carry # asm 1: adc <mulr7=int64#9,<c3=int64#14 # asm 2: adc <mulr7=%r11,<c3=%rbx adc %r11,%rbx # qhasm: mulzero = 0 # asm 1: mov $0,>mulzero=int64#3 # asm 2: mov $0,>mulzero=%rdx mov $0,%rdx # qhasm: mulr8 += mulzero + carry # asm 1: adc <mulzero=int64#3,<mulr8=int64#7 # asm 2: adc <mulzero=%rdx,<mulr8=%rax adc %rdx,%rax # qhasm: mulr8 *= 38 # asm 1: imulq $38,<mulr8=int64#7,>mulr8=int64#5 # asm 2: imulq $38,<mulr8=%rax,>mulr8=%r8 imulq $38,%rax,%r8 # qhasm: carry? c0 += mulr8 # asm 1: add <mulr8=int64#5,<c0=int64#11 # asm 2: add <mulr8=%r8,<c0=%r13 add %r8,%r13 # qhasm: carry? c1 += mulzero + carry # asm 1: adc <mulzero=int64#3,<c1=int64#12 # asm 2: adc <mulzero=%rdx,<c1=%r14 adc %rdx,%r14 # qhasm: carry? c2 += mulzero + carry # asm 1: adc <mulzero=int64#3,<c2=int64#13 # asm 2: adc <mulzero=%rdx,<c2=%r15 adc %rdx,%r15 # qhasm: carry? c3 += mulzero + carry # asm 1: adc <mulzero=int64#3,<c3=int64#14 # asm 2: adc <mulzero=%rdx,<c3=%rbx adc %rdx,%rbx # qhasm: mulzero += mulzero + carry # asm 1: adc <mulzero=int64#3,<mulzero=int64#3 # asm 2: adc <mulzero=%rdx,<mulzero=%rdx adc %rdx,%rdx # qhasm: mulzero *= 38 # asm 1: imulq $38,<mulzero=int64#3,>mulzero=int64#3 # asm 2: imulq $38,<mulzero=%rdx,>mulzero=%rdx imulq $38,%rdx,%rdx # qhasm: c0 += mulzero # asm 1: add <mulzero=int64#3,<c0=int64#11 # asm 2: add <mulzero=%rdx,<c0=%r13 add %rdx,%r13 # qhasm: c0_stack = c0 # asm 1: movq <c0=int64#11,>c0_stack=stack64#8 # asm 2: movq <c0=%r13,>c0_stack=56(%rsp) movq %r13,56(%rsp) # qhasm: c1_stack = c1 # asm 1: movq <c1=int64#12,>c1_stack=stack64#9 # asm 2: movq <c1=%r14,>c1_stack=64(%rsp) movq %r14,64(%rsp) # qhasm: c2_stack = c2 # asm 1: movq <c2=int64#13,>c2_stack=stack64#10 # asm 2: movq <c2=%r15,>c2_stack=72(%rsp) movq %r15,72(%rsp) # qhasm: c3_stack = c3 # asm 1: movq <c3=int64#14,>c3_stack=stack64#11 # asm 2: movq <c3=%rbx,>c3_stack=80(%rsp) movq %rbx,80(%rsp) # 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 *)(pp + 64) # asm 1: movq 64(<pp=int64#2),>mulx0=int64#10 # asm 2: movq 64(<pp=%rsi),>mulx0=%r12 movq 64(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 64) # asm 1: movq 64(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 64(<qp=%rcx),>mulrax=%rax movq 64(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: rt0 = mulrax # asm 1: mov <mulrax=int64#7,>rt0=int64#11 # asm 2: mov <mulrax=%rax,>rt0=%r13 mov %rax,%r13 # qhasm: rt1 = mulrdx # asm 1: mov <mulrdx=int64#3,>rt1=int64#12 # asm 2: mov <mulrdx=%rdx,>rt1=%r14 mov %rdx,%r14 # qhasm: mulrax = *(uint64 *)(qp + 72) # asm 1: movq 72(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 72(<qp=%rcx),>mulrax=%rax movq 72(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? rt1 += mulrax # asm 1: add <mulrax=int64#7,<rt1=int64#12 # asm 2: add <mulrax=%rax,<rt1=%r14 add %rax,%r14 # qhasm: rt2 = 0 # asm 1: mov $0,>rt2=int64#13 # asm 2: mov $0,>rt2=%r15 mov $0,%r15 # qhasm: rt2 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<rt2=int64#13 # asm 2: adc <mulrdx=%rdx,<rt2=%r15 adc %rdx,%r15 # qhasm: mulrax = *(uint64 *)(qp + 80) # asm 1: movq 80(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 80(<qp=%rcx),>mulrax=%rax movq 80(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? rt2 += mulrax # asm 1: add <mulrax=int64#7,<rt2=int64#13 # asm 2: add <mulrax=%rax,<rt2=%r15 add %rax,%r15 # qhasm: rt3 = 0 # asm 1: mov $0,>rt3=int64#14 # asm 2: mov $0,>rt3=%rbx mov $0,%rbx # qhasm: rt3 += mulrdx + carry # asm 1: adc <mulrdx=int64#3,<rt3=int64#14 # asm 2: adc <mulrdx=%rdx,<rt3=%rbx adc %rdx,%rbx # qhasm: mulrax = *(uint64 *)(qp + 88) # asm 1: movq 88(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 88(<qp=%rcx),>mulrax=%rax movq 88(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx0 # asm 1: mul <mulx0=int64#10 # asm 2: mul <mulx0=%r12 mul %r12 # qhasm: carry? rt3 += mulrax # asm 1: add <mulrax=int64#7,<rt3=int64#14 # asm 2: add <mulrax=%rax,<rt3=%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 *)(pp + 72) # asm 1: movq 72(<pp=int64#2),>mulx1=int64#10 # asm 2: movq 72(<pp=%rsi),>mulx1=%r12 movq 72(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 64) # asm 1: movq 64(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 64(<qp=%rcx),>mulrax=%rax movq 64(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? rt1 += mulrax # asm 1: add <mulrax=int64#7,<rt1=int64#12 # asm 2: add <mulrax=%rax,<rt1=%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 *)(qp + 72) # asm 1: movq 72(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 72(<qp=%rcx),>mulrax=%rax movq 72(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? rt2 += mulrax # asm 1: add <mulrax=int64#7,<rt2=int64#13 # asm 2: add <mulrax=%rax,<rt2=%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? rt2 += mulc # asm 1: add <mulc=int64#15,<rt2=int64#13 # asm 2: add <mulc=%rbp,<rt2=%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 *)(qp + 80) # asm 1: movq 80(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 80(<qp=%rcx),>mulrax=%rax movq 80(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx1 # asm 1: mul <mulx1=int64#10 # asm 2: mul <mulx1=%r12 mul %r12 # qhasm: carry? rt3 += mulrax # asm 1: add <mulrax=int64#7,<rt3=int64#14 # asm 2: add <mulrax=%rax,<rt3=%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? rt3 += mulc # asm 1: add <mulc=int64#15,<rt3=int64#14 # asm 2: add <mulc=%rbp,<rt3=%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 *)(qp + 88) # asm 1: movq 88(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 88(<qp=%rcx),>mulrax=%rax movq 88(%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 *)(pp + 80) # asm 1: movq 80(<pp=int64#2),>mulx2=int64#10 # asm 2: movq 80(<pp=%rsi),>mulx2=%r12 movq 80(%rsi),%r12 # qhasm: mulrax = *(uint64 *)(qp + 64) # asm 1: movq 64(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 64(<qp=%rcx),>mulrax=%rax movq 64(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx2 # asm 1: mul <mulx2=int64#10 # asm 2: mul <mulx2=%r12 mul %r12 # qhasm: carry? rt2 += mulrax # asm 1: add <mulrax=int64#7,<rt2=int64#13 # asm 2: add <mulrax=%rax,<rt2=%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 *)(qp + 72) # asm 1: movq 72(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 72(<qp=%rcx),>mulrax=%rax movq 72(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx2 # asm 1: mul <mulx2=int64#10 # asm 2: mul <mulx2=%r12 mul %r12 # qhasm: carry? rt3 += mulrax # asm 1: add <mulrax=int64#7,<rt3=int64#14 # asm 2: add <mulrax=%rax,<rt3=%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? rt3 += mulc # asm 1: add <mulc=int64#15,<rt3=int64#14 # asm 2: add <mulc=%rbp,<rt3=%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 *)(qp + 80) # asm 1: movq 80(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 80(<qp=%rcx),>mulrax=%rax movq 80(%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 *)(qp + 88) # asm 1: movq 88(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 88(<qp=%rcx),>mulrax=%rax movq 88(%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 *)(pp + 88) # asm 1: movq 88(<pp=int64#2),>mulx3=int64#2 # asm 2: movq 88(<pp=%rsi),>mulx3=%rsi movq 88(%rsi),%rsi # qhasm: mulrax = *(uint64 *)(qp + 64) # asm 1: movq 64(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 64(<qp=%rcx),>mulrax=%rax movq 64(%rcx),%rax # qhasm: (uint128) mulrdx mulrax = mulrax * mulx3 # asm 1: mul <mulx3=int64#2 # asm 2: mul <mulx3=%rsi mul %rsi # qhasm: carry? rt3 += mulrax # asm 1: add <mulrax=int64#7,<rt3=int64#14 # asm 2: add <mulrax=%rax,<rt3=%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 *)(qp + 72) # asm 1: movq 72(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 72(<qp=%rcx),>mulrax=%rax movq 72(%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 *)(qp + 80) # asm 1: movq 80(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 80(<qp=%rcx),>mulrax=%rax movq 80(%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 *)(qp + 88) # asm 1: movq 88(<qp=int64#4),>mulrax=int64#7 # asm 2: movq 88(<qp=%rcx),>mulrax=%rax movq 88(%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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # 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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # 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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # 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_SHARED_NAMESPACE(38) mulq CRYPTO_SHARED_NAMESPACE(38)(%rip) # 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? rt0 += mulr4 # asm 1: add <mulr4=int64#2,<rt0=int64#11 # asm 2: add <mulr4=%rsi,<rt0=%r13 add %rsi,%r13 # qhasm: carry? rt1 += mulr5 + carry # asm 1: adc <mulr5=int64#4,<rt1=int64#12 # asm 2: adc <mulr5=%rcx,<rt1=%r14 adc %rcx,%r14 # qhasm: carry? rt2 += mulr6 + carry # asm 1: adc <mulr6=int64#5,<rt2=int64#13 # asm 2: adc <mulr6=%r8,<rt2=%r15 adc %r8,%r15 # qhasm: carry? rt3 += mulr7 + carry # asm 1: adc <mulr7=int64#6,<rt3=int64#14 # asm 2: adc <mulr7=%r9,<rt3=%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? rt0 += mulr8 # asm 1: add <mulr8=int64#3,<rt0=int64#11 # asm 2: add <mulr8=%rdx,<rt0=%r13 add %rdx,%r13 # qhasm: carry? rt1 += mulzero + carry # asm 1: adc <mulzero=int64#2,<rt1=int64#12 # asm 2: adc <mulzero=%rsi,<rt1=%r14 adc %rsi,%r14 # qhasm: carry? rt2 += mulzero + carry # asm 1: adc <mulzero=int64#2,<rt2=int64#13 # asm 2: adc <mulzero=%rsi,<rt2=%r15 adc %rsi,%r15 # qhasm: carry? rt3 += mulzero + carry # asm 1: adc <mulzero=int64#2,<rt3=int64#14 # asm 2: adc <mulzero=%rsi,<rt3=%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: rt0 += mulzero # asm 1: add <mulzero=int64#2,<rt0=int64#11 # asm 2: add <mulzero=%rsi,<rt0=%r13 add %rsi,%r13 # qhasm: carry? rt0 += rt0 # asm 1: add <rt0=int64#11,<rt0=int64#11 # asm 2: add <rt0=%r13,<rt0=%r13 add %r13,%r13 # qhasm: carry? rt1 += rt1 + carry # asm 1: adc <rt1=int64#12,<rt1=int64#12 # asm 2: adc <rt1=%r14,<rt1=%r14 adc %r14,%r14 # qhasm: carry? rt2 += rt2 + carry # asm 1: adc <rt2=int64#13,<rt2=int64#13 # asm 2: adc <rt2=%r15,<rt2=%r15 adc %r15,%r15 # qhasm: carry? rt3 += rt3 + carry # asm 1: adc <rt3=int64#14,<rt3=int64#14 # asm 2: adc <rt3=%rbx,<rt3=%rbx adc %rbx,%rbx # qhasm: addt0 = 0 # asm 1: mov $0,>addt0=int64#2 # asm 2: mov $0,>addt0=%rsi mov $0,%rsi # qhasm: addt1 = 38 # asm 1: mov $38,>addt1=int64#3 # asm 2: mov $38,>addt1=%rdx mov $38,%rdx # qhasm: addt1 = addt0 if !carry # asm 1: cmovae <addt0=int64#2,<addt1=int64#3 # asm 2: cmovae <addt0=%rsi,<addt1=%rdx cmovae %rsi,%rdx # qhasm: carry? rt0 += addt1 # asm 1: add <addt1=int64#3,<rt0=int64#11 # asm 2: add <addt1=%rdx,<rt0=%r13 add %rdx,%r13 # qhasm: carry? rt1 += addt0 + carry # asm 1: adc <addt0=int64#2,<rt1=int64#12 # asm 2: adc <addt0=%rsi,<rt1=%r14 adc %rsi,%r14 # qhasm: carry? rt2 += addt0 + carry # asm 1: adc <addt0=int64#2,<rt2=int64#13 # asm 2: adc <addt0=%rsi,<rt2=%r15 adc %rsi,%r15 # qhasm: carry? rt3 += addt0 + carry # asm 1: adc <addt0=int64#2,<rt3=int64#14 # asm 2: adc <addt0=%rsi,<rt3=%rbx adc %rsi,%rbx # qhasm: addt0 = addt1 if carry # asm 1: cmovc <addt1=int64#3,<addt0=int64#2 # asm 2: cmovc <addt1=%rdx,<addt0=%rsi cmovc %rdx,%rsi # qhasm: rt0 += addt0 # asm 1: add <addt0=int64#2,<rt0=int64#11 # asm 2: add <addt0=%rsi,<rt0=%r13 add %rsi,%r13 # qhasm: rz0 = rt0 # asm 1: mov <rt0=int64#11,>rz0=int64#2 # asm 2: mov <rt0=%r13,>rz0=%rsi mov %r13,%rsi # qhasm: rz1 = rt1 # asm 1: mov <rt1=int64#12,>rz1=int64#3 # asm 2: mov <rt1=%r14,>rz1=%rdx mov %r14,%rdx # qhasm: rz2 = rt2 # asm 1: mov <rt2=int64#13,>rz2=int64#4 # asm 2: mov <rt2=%r15,>rz2=%rcx mov %r15,%rcx # qhasm: rz3 = rt3 # asm 1: mov <rt3=int64#14,>rz3=int64#5 # asm 2: mov <rt3=%rbx,>rz3=%r8 mov %rbx,%r8 # qhasm: carry? rz0 += c0_stack # asm 1: addq <c0_stack=stack64#8,<rz0=int64#2 # asm 2: addq <c0_stack=56(%rsp),<rz0=%rsi addq 56(%rsp),%rsi # qhasm: carry? rz1 += c1_stack + carry # asm 1: adcq <c1_stack=stack64#9,<rz1=int64#3 # asm 2: adcq <c1_stack=64(%rsp),<rz1=%rdx adcq 64(%rsp),%rdx # qhasm: carry? rz2 += c2_stack + carry # asm 1: adcq <c2_stack=stack64#10,<rz2=int64#4 # asm 2: adcq <c2_stack=72(%rsp),<rz2=%rcx adcq 72(%rsp),%rcx # qhasm: carry? rz3 += c3_stack + carry # asm 1: adcq <c3_stack=stack64#11,<rz3=int64#5 # asm 2: adcq <c3_stack=80(%rsp),<rz3=%r8 adcq 80(%rsp),%r8 # qhasm: addt0 = 0 # asm 1: mov $0,>addt0=int64#6 # asm 2: mov $0,>addt0=%r9 mov $0,%r9 # 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#6,<addt1=int64#7 # asm 2: cmovae <addt0=%r9,<addt1=%rax cmovae %r9,%rax # qhasm: carry? rz0 += addt1 # asm 1: add <addt1=int64#7,<rz0=int64#2 # asm 2: add <addt1=%rax,<rz0=%rsi add %rax,%rsi # qhasm: carry? rz1 += addt0 + carry # asm 1: adc <addt0=int64#6,<rz1=int64#3 # asm 2: adc <addt0=%r9,<rz1=%rdx adc %r9,%rdx # qhasm: carry? rz2 += addt0 + carry # asm 1: adc <addt0=int64#6,<rz2=int64#4 # asm 2: adc <addt0=%r9,<rz2=%rcx adc %r9,%rcx # qhasm: carry? rz3 += addt0 + carry # asm 1: adc <addt0=int64#6,<rz3=int64#5 # asm 2: adc <addt0=%r9,<rz3=%r8 adc %r9,%r8 # qhasm: addt0 = addt1 if carry # asm 1: cmovc <addt1=int64#7,<addt0=int64#6 # asm 2: cmovc <addt1=%rax,<addt0=%r9 cmovc %rax,%r9 # qhasm: rz0 += addt0 # asm 1: add <addt0=int64#6,<rz0=int64#2 # asm 2: add <addt0=%r9,<rz0=%rsi add %r9,%rsi # qhasm: carry? rt0 -= c0_stack # asm 1: subq <c0_stack=stack64#8,<rt0=int64#11 # asm 2: subq <c0_stack=56(%rsp),<rt0=%r13 subq 56(%rsp),%r13 # qhasm: carry? rt1 -= c1_stack - carry # asm 1: sbbq <c1_stack=stack64#9,<rt1=int64#12 # asm 2: sbbq <c1_stack=64(%rsp),<rt1=%r14 sbbq 64(%rsp),%r14 # qhasm: carry? rt2 -= c2_stack - carry # asm 1: sbbq <c2_stack=stack64#10,<rt2=int64#13 # asm 2: sbbq <c2_stack=72(%rsp),<rt2=%r15 sbbq 72(%rsp),%r15 # qhasm: carry? rt3 -= c3_stack - carry # asm 1: sbbq <c3_stack=stack64#11,<rt3=int64#14 # asm 2: sbbq <c3_stack=80(%rsp),<rt3=%rbx sbbq 80(%rsp),%rbx # qhasm: subt0 = 0 # asm 1: mov $0,>subt0=int64#6 # asm 2: mov $0,>subt0=%r9 mov $0,%r9 # 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#6,<subt1=int64#7 # asm 2: cmovae <subt0=%r9,<subt1=%rax cmovae %r9,%rax # qhasm: carry? rt0 -= subt1 # asm 1: sub <subt1=int64#7,<rt0=int64#11 # asm 2: sub <subt1=%rax,<rt0=%r13 sub %rax,%r13 # qhasm: carry? rt1 -= subt0 - carry # asm 1: sbb <subt0=int64#6,<rt1=int64#12 # asm 2: sbb <subt0=%r9,<rt1=%r14 sbb %r9,%r14 # qhasm: carry? rt2 -= subt0 - carry # asm 1: sbb <subt0=int64#6,<rt2=int64#13 # asm 2: sbb <subt0=%r9,<rt2=%r15 sbb %r9,%r15 # qhasm: carry? rt3 -= subt0 - carry # asm 1: sbb <subt0=int64#6,<rt3=int64#14 # asm 2: sbb <subt0=%r9,<rt3=%rbx sbb %r9,%rbx # qhasm: subt0 = subt1 if carry # asm 1: cmovc <subt1=int64#7,<subt0=int64#6 # asm 2: cmovc <subt1=%rax,<subt0=%r9 cmovc %rax,%r9 # qhasm: rt0 -= subt0 # asm 1: sub <subt0=int64#6,<rt0=int64#11 # asm 2: sub <subt0=%r9,<rt0=%r13 sub %r9,%r13 # qhasm: *(uint64 *)(rp + 32) = rz0 # asm 1: movq <rz0=int64#2,32(<rp=int64#1) # asm 2: movq <rz0=%rsi,32(<rp=%rdi) movq %rsi,32(%rdi) # qhasm: *(uint64 *)(rp + 40) = rz1 # asm 1: movq <rz1=int64#3,40(<rp=int64#1) # asm 2: movq <rz1=%rdx,40(<rp=%rdi) movq %rdx,40(%rdi) # qhasm: *(uint64 *)(rp + 48) = rz2 # asm 1: movq <rz2=int64#4,48(<rp=int64#1) # asm 2: movq <rz2=%rcx,48(<rp=%rdi) movq %rcx,48(%rdi) # qhasm: *(uint64 *)(rp + 56) = rz3 # asm 1: movq <rz3=int64#5,56(<rp=int64#1) # asm 2: movq <rz3=%r8,56(<rp=%rdi) movq %r8,56(%rdi) # qhasm: *(uint64 *)(rp + 96) = rt0 # asm 1: movq <rt0=int64#11,96(<rp=int64#1) # asm 2: movq <rt0=%r13,96(<rp=%rdi) movq %r13,96(%rdi) # qhasm: *(uint64 *)(rp + 104) = rt1 # asm 1: movq <rt1=int64#12,104(<rp=int64#1) # asm 2: movq <rt1=%r14,104(<rp=%rdi) movq %r14,104(%rdi) # qhasm: *(uint64 *)(rp + 112) = rt2 # asm 1: movq <rt2=int64#13,112(<rp=int64#1) # asm 2: movq <rt2=%r15,112(<rp=%rdi) movq %r15,112(%rdi) # qhasm: *(uint64 *)(rp + 120) = rt3 # asm 1: movq <rt3=int64#14,120(<rp=int64#1) # asm 2: movq <rt3=%rbx,120(<rp=%rdi) movq %rbx,120(%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