pqc/crypto_kem/mceliece460896/sse/vec_reduce_asm.S
Thom Wiggers b3f9d4f8d6
Classic McEliece (#259)
* Add McEliece reference implementations

* Add Vec implementations of McEliece

* Add sse implementations

* Add AVX2 implementations

* Get rid of stuff not supported by Mac ABI

* restrict to two cores

* Ditch .data files

* Remove .hidden from all .S files

* speed up duplicate consistency tests by batching

* make cpuinfo more robust

* Hope to stabilize macos cpuinfo without ccache

* Revert "Hope to stabilize macos cpuinfo without ccache"

This reverts commit 6129c3cabe1abbc8b956bc87e902a698e32bf322.

* Just hardcode what's available at travis

* Fixed-size types in api.h

* namespace all header files in mceliece

* Ditch operations.h

* Get rid of static inline functions

* fixup! Ditch operations.h
2020-02-05 13:09:56 +01:00

514 lines
11 KiB
ArmAsm

# qhasm: int64 input_0
# qhasm: int64 input_1
# qhasm: int64 input_2
# qhasm: int64 input_3
# qhasm: int64 input_4
# qhasm: int64 input_5
# qhasm: stack64 input_6
# qhasm: stack64 input_7
# qhasm: int64 caller_r11
# qhasm: int64 caller_r12
# qhasm: int64 caller_r13
# qhasm: int64 caller_r14
# qhasm: int64 caller_r15
# qhasm: int64 caller_rbx
# qhasm: int64 caller_rbp
# qhasm: int64 t0
# qhasm: int64 t1
# qhasm: int64 c
# qhasm: int64 r
# qhasm: enter vec_reduce_asm
.p2align 5
.global _PQCLEAN_MCELIECE460896_SSE_vec_reduce_asm
.global PQCLEAN_MCELIECE460896_SSE_vec_reduce_asm
_PQCLEAN_MCELIECE460896_SSE_vec_reduce_asm:
PQCLEAN_MCELIECE460896_SSE_vec_reduce_asm:
mov %rsp,%r11
and $31,%r11
add $0,%r11
sub %r11,%rsp
# qhasm: r = 0
# asm 1: mov $0,>r=int64#7
# asm 2: mov $0,>r=%rax
mov $0,%rax
# qhasm: t0 = mem64[ input_0 + 192 ]
# asm 1: movq 192(<input_0=int64#1),>t0=int64#2
# asm 2: movq 192(<input_0=%rdi),>t0=%rsi
movq 192(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 200 ]
# asm 1: movq 200(<input_0=int64#1),>t1=int64#3
# asm 2: movq 200(<input_0=%rdi),>t1=%rdx
movq 200(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 176 ]
# asm 1: movq 176(<input_0=int64#1),>t0=int64#2
# asm 2: movq 176(<input_0=%rdi),>t0=%rsi
movq 176(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 184 ]
# asm 1: movq 184(<input_0=int64#1),>t1=int64#3
# asm 2: movq 184(<input_0=%rdi),>t1=%rdx
movq 184(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 160 ]
# asm 1: movq 160(<input_0=int64#1),>t0=int64#2
# asm 2: movq 160(<input_0=%rdi),>t0=%rsi
movq 160(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 168 ]
# asm 1: movq 168(<input_0=int64#1),>t1=int64#3
# asm 2: movq 168(<input_0=%rdi),>t1=%rdx
movq 168(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 144 ]
# asm 1: movq 144(<input_0=int64#1),>t0=int64#2
# asm 2: movq 144(<input_0=%rdi),>t0=%rsi
movq 144(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 152 ]
# asm 1: movq 152(<input_0=int64#1),>t1=int64#3
# asm 2: movq 152(<input_0=%rdi),>t1=%rdx
movq 152(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 128 ]
# asm 1: movq 128(<input_0=int64#1),>t0=int64#2
# asm 2: movq 128(<input_0=%rdi),>t0=%rsi
movq 128(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 136 ]
# asm 1: movq 136(<input_0=int64#1),>t1=int64#3
# asm 2: movq 136(<input_0=%rdi),>t1=%rdx
movq 136(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 112 ]
# asm 1: movq 112(<input_0=int64#1),>t0=int64#2
# asm 2: movq 112(<input_0=%rdi),>t0=%rsi
movq 112(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 120 ]
# asm 1: movq 120(<input_0=int64#1),>t1=int64#3
# asm 2: movq 120(<input_0=%rdi),>t1=%rdx
movq 120(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 96 ]
# asm 1: movq 96(<input_0=int64#1),>t0=int64#2
# asm 2: movq 96(<input_0=%rdi),>t0=%rsi
movq 96(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 104 ]
# asm 1: movq 104(<input_0=int64#1),>t1=int64#3
# asm 2: movq 104(<input_0=%rdi),>t1=%rdx
movq 104(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 80 ]
# asm 1: movq 80(<input_0=int64#1),>t0=int64#2
# asm 2: movq 80(<input_0=%rdi),>t0=%rsi
movq 80(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 88 ]
# asm 1: movq 88(<input_0=int64#1),>t1=int64#3
# asm 2: movq 88(<input_0=%rdi),>t1=%rdx
movq 88(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 64 ]
# asm 1: movq 64(<input_0=int64#1),>t0=int64#2
# asm 2: movq 64(<input_0=%rdi),>t0=%rsi
movq 64(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 72 ]
# asm 1: movq 72(<input_0=int64#1),>t1=int64#3
# asm 2: movq 72(<input_0=%rdi),>t1=%rdx
movq 72(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 48 ]
# asm 1: movq 48(<input_0=int64#1),>t0=int64#2
# asm 2: movq 48(<input_0=%rdi),>t0=%rsi
movq 48(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 56 ]
# asm 1: movq 56(<input_0=int64#1),>t1=int64#3
# asm 2: movq 56(<input_0=%rdi),>t1=%rdx
movq 56(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 32 ]
# asm 1: movq 32(<input_0=int64#1),>t0=int64#2
# asm 2: movq 32(<input_0=%rdi),>t0=%rsi
movq 32(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 40 ]
# asm 1: movq 40(<input_0=int64#1),>t1=int64#3
# asm 2: movq 40(<input_0=%rdi),>t1=%rdx
movq 40(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 16 ]
# asm 1: movq 16(<input_0=int64#1),>t0=int64#2
# asm 2: movq 16(<input_0=%rdi),>t0=%rsi
movq 16(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 24 ]
# asm 1: movq 24(<input_0=int64#1),>t1=int64#3
# asm 2: movq 24(<input_0=%rdi),>t1=%rdx
movq 24(%rdi),%rdx
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#3,<t0=int64#2
# asm 2: xor <t1=%rdx,<t0=%rsi
xor %rdx,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#2
# asm 2: popcnt <t0=%rsi, >c=%rsi
popcnt %rsi, %rsi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#2d
# asm 2: and $1,<c=%esi
and $1,%esi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#2,<r=int64#7
# asm 2: or <c=%rsi,<r=%rax
or %rsi,%rax
# qhasm: t0 = mem64[ input_0 + 0 ]
# asm 1: movq 0(<input_0=int64#1),>t0=int64#2
# asm 2: movq 0(<input_0=%rdi),>t0=%rsi
movq 0(%rdi),%rsi
# qhasm: t1 = mem64[ input_0 + 8 ]
# asm 1: movq 8(<input_0=int64#1),>t1=int64#1
# asm 2: movq 8(<input_0=%rdi),>t1=%rdi
movq 8(%rdi),%rdi
# qhasm: t0 ^= t1
# asm 1: xor <t1=int64#1,<t0=int64#2
# asm 2: xor <t1=%rdi,<t0=%rsi
xor %rdi,%rsi
# qhasm: c = count(t0)
# asm 1: popcnt <t0=int64#2, >c=int64#1
# asm 2: popcnt <t0=%rsi, >c=%rdi
popcnt %rsi, %rdi
# qhasm: (uint32) c &= 1
# asm 1: and $1,<c=int64#1d
# asm 2: and $1,<c=%edi
and $1,%edi
# qhasm: r <<= 1
# asm 1: shl $1,<r=int64#7
# asm 2: shl $1,<r=%rax
shl $1,%rax
# qhasm: r |= c
# asm 1: or <c=int64#1,<r=int64#7
# asm 2: or <c=%rdi,<r=%rax
or %rdi,%rax
# qhasm: return r
add %r11,%rsp
ret