[Feature] Add AVX2 version of siphash

This version has been taken from the google code at
https://github.com/google/highwayhash/

The final code is obtained by compiling c++ using clang-3.8 compiler.

1 files changed, 201 insertions, 0 deletions

diff --git a/src/libcryptobox/siphash/avx2.S b/src/libcryptobox/siphash/avx2.S
new file mode 100644
index 000000000..995fc7636
--- /dev/null
+++ b/src/libcryptobox/siphash/avx2.S
@@ -0,0 +1,201 @@
+/*-
+ * Copyright 2015 Google Inc. All Rights Reserved.
+ * Copyright 2016 Vsevolod Stakhov
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include "../macro.S"
+#include "constants.S"
+
+/*
+ * Generated by clang-3.8 from siphash avx2 implementation written by
+ * Jan Wassenberg and Jyrki Alakuijala
+ */
+
+SECTION_TEXT
+
+GLOBAL_HIDDEN_FN siphash_avx2
+siphash_avx2_local:
+	.cfi_startproc
+## BB#0:                                ## %entry
+	pushq	%rbp
+Ltmp0:
+	.cfi_def_cfa_offset 16
+Ltmp1:
+	.cfi_offset %rbp, -16
+	movq	%rsp, %rbp
+Ltmp2:
+	.cfi_def_cfa_register %rbp
+	pushq	%rbx
+	subq	$40, %rsp
+Ltmp3:
+	.cfi_offset %rbx, -24
+	movq	%rdx, %rbx
+	vmovdqu	(%rdi), %xmm0
+	vpxor	LCPI0_0(%rip), %xmm0, %xmm1
+	vpxor	LCPI0_1(%rip), %xmm0, %xmm0
+	vpunpcklqdq	%xmm0, %xmm1, %xmm6 ## xmm6 = xmm1[0],xmm0[0]
+	vpunpckhqdq	%xmm0, %xmm1, %xmm7 ## xmm7 = xmm1[1],xmm0[1]
+	movq	%rbx, %rax
+	andq	$-8, %rax
+	je	LBB0_1
+## BB#2:                                ## %for.body.preheader
+	xorl	%ecx, %ecx
+	vmovdqa	LCPI0_2(%rip), %xmm0    ## xmm0 = [13,16]
+	vmovdqa	LCPI0_3(%rip), %xmm1    ## xmm1 = [51,48]
+	vmovdqa	LCPI0_4(%rip), %xmm2    ## xmm2 = [17,21]
+	vmovdqa	LCPI0_5(%rip), %xmm3    ## xmm3 = [47,43]
+	.align	4, 0x90
+LBB0_3:                                 ## %for.body
+                                        ## =>This Inner Loop Header: Depth=1
+	vmovq	(%rsi,%rcx), %xmm4      ## xmm4 = mem[0],zero
+	vpslldq	$8, %xmm4, %xmm5        ## xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2,3,4,5,6,7]
+	vpxor	%xmm5, %xmm7, %xmm5
+	vpaddq	%xmm6, %xmm5, %xmm6
+	vpsllvq	%xmm0, %xmm5, %xmm7
+	vpsrlvq	%xmm1, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm5
+	vpshufd	$30, %xmm6, %xmm6       ## xmm6 = xmm6[2,3,1,0]
+	vpaddq	%xmm5, %xmm6, %xmm6
+	vpsllvq	%xmm2, %xmm5, %xmm7
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm5
+	vpshufd	$30, %xmm6, %xmm6       ## xmm6 = xmm6[2,3,1,0]
+	vpaddq	%xmm5, %xmm6, %xmm6
+	vpsllvq	%xmm0, %xmm5, %xmm7
+	vpsrlvq	%xmm1, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm5
+	vpshufd	$30, %xmm6, %xmm6       ## xmm6 = xmm6[2,3,1,0]
+	vpaddq	%xmm5, %xmm6, %xmm6
+	vpsllvq	%xmm2, %xmm5, %xmm7
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm7
+	vpshufd	$30, %xmm6, %xmm5       ## xmm5 = xmm6[2,3,1,0]
+	vpxor	%xmm5, %xmm4, %xmm6
+	addq	$8, %rcx
+	cmpq	%rax, %rcx
+	jb	LBB0_3
+## BB#4:                                ## %for.end.loopexit
+	vmovdqa	%xmm7, -48(%rbp)        ## 16-byte Spill
+	vmovdqa	%xmm6, -32(%rbp)        ## 16-byte Spill
+	addq	%rax, %rsi
+	jmp	LBB0_5
+LBB0_1:
+	vmovdqa	%xmm7, -48(%rbp)        ## 16-byte Spill
+	vmovdqa	%xmm6, -32(%rbp)        ## 16-byte Spill
+	xorl	%eax, %eax
+LBB0_5:                                 ## %for.end
+	movq	$0, -16(%rbp)
+	movq	%rbx, %rdx
+	subq	%rax, %rdx
+	leaq	-16(%rbp), %rdi
+	callq	_memcpy
+	movb	%bl, -9(%rbp)
+	vmovq	-16(%rbp), %xmm4        ## xmm4 = mem[0],zero
+	vpslldq	$8, %xmm4, %xmm0        ## xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2,3,4,5,6,7]
+	vpxor	-48(%rbp), %xmm0, %xmm2 ## 16-byte Folded Reload
+	vpaddq	-32(%rbp), %xmm2, %xmm3 ## 16-byte Folded Reload
+	vmovdqa	LCPI0_2(%rip), %xmm0    ## xmm0 = [13,16]
+	vpsllvq	%xmm0, %xmm2, %xmm5
+	vmovdqa	LCPI0_3(%rip), %xmm1    ## xmm1 = [51,48]
+	vpsrlvq	%xmm1, %xmm2, %xmm2
+	vpor	%xmm5, %xmm2, %xmm2
+	vpxor	%xmm3, %xmm2, %xmm5
+	vpshufd	$30, %xmm3, %xmm2       ## xmm2 = xmm3[2,3,1,0]
+	vpaddq	%xmm5, %xmm2, %xmm6
+	vmovdqa	LCPI0_4(%rip), %xmm2    ## xmm2 = [17,21]
+	vpsllvq	%xmm2, %xmm5, %xmm7
+	vmovdqa	LCPI0_5(%rip), %xmm3    ## xmm3 = [47,43]
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm5
+	vpshufd	$30, %xmm6, %xmm6       ## xmm6 = xmm6[2,3,1,0]
+	vpaddq	%xmm5, %xmm6, %xmm6
+	vpsllvq	%xmm0, %xmm5, %xmm7
+	vpsrlvq	%xmm1, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm5
+	vpshufd	$30, %xmm6, %xmm6       ## xmm6 = xmm6[2,3,1,0]
+	vpaddq	%xmm5, %xmm6, %xmm6
+	vpsllvq	%xmm2, %xmm5, %xmm7
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm7, %xmm5, %xmm5
+	vpxor	%xmm6, %xmm5, %xmm5
+	vpshufd	$30, %xmm6, %xmm6       ## xmm6 = xmm6[2,3,1,0]
+	movl	$255, %eax
+	vmovq	%rax, %xmm7
+	vpslldq	$8, %xmm7, %xmm7        ## xmm7 = zero,zero,zero,zero,zero,zero,zero,zero,xmm7[0,1,2,3,4,5,6,7]
+	vpxor	%xmm7, %xmm4, %xmm4
+	vpxor	%xmm4, %xmm6, %xmm4
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm0, %xmm5, %xmm6
+	vpsrlvq	%xmm1, %xmm5, %xmm5
+	vpor	%xmm6, %xmm5, %xmm5
+	vpxor	%xmm4, %xmm5, %xmm5
+	vpshufd	$30, %xmm4, %xmm4       ## xmm4 = xmm4[2,3,1,0]
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm2, %xmm5, %xmm6
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm6, %xmm5, %xmm5
+	vpxor	%xmm4, %xmm5, %xmm5
+	vpshufd	$30, %xmm4, %xmm4       ## xmm4 = xmm4[2,3,1,0]
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm0, %xmm5, %xmm6
+	vpsrlvq	%xmm1, %xmm5, %xmm5
+	vpor	%xmm6, %xmm5, %xmm5
+	vpxor	%xmm4, %xmm5, %xmm5
+	vpshufd	$30, %xmm4, %xmm4       ## xmm4 = xmm4[2,3,1,0]
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm2, %xmm5, %xmm6
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm6, %xmm5, %xmm5
+	vpxor	%xmm4, %xmm5, %xmm5
+	vpshufd	$30, %xmm4, %xmm4       ## xmm4 = xmm4[2,3,1,0]
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm0, %xmm5, %xmm6
+	vpsrlvq	%xmm1, %xmm5, %xmm5
+	vpor	%xmm6, %xmm5, %xmm5
+	vpxor	%xmm4, %xmm5, %xmm5
+	vpshufd	$30, %xmm4, %xmm4       ## xmm4 = xmm4[2,3,1,0]
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm2, %xmm5, %xmm6
+	vpsrlvq	%xmm3, %xmm5, %xmm5
+	vpor	%xmm6, %xmm5, %xmm5
+	vpxor	%xmm4, %xmm5, %xmm5
+	vpshufd	$30, %xmm4, %xmm4       ## xmm4 = xmm4[2,3,1,0]
+	vpaddq	%xmm5, %xmm4, %xmm4
+	vpsllvq	%xmm0, %xmm5, %xmm0
+	vpsrlvq	%xmm1, %xmm5, %xmm1
+	vpor	%xmm0, %xmm1, %xmm0
+	vpxor	%xmm4, %xmm0, %xmm0
+	vpshufd	$30, %xmm4, %xmm1       ## xmm1 = xmm4[2,3,1,0]
+	vpaddq	%xmm0, %xmm1, %xmm1
+	vpsllvq	%xmm2, %xmm0, %xmm2
+	vpsrlvq	%xmm3, %xmm0, %xmm0
+	vpor	%xmm2, %xmm0, %xmm0
+	vpshufd	$30, %xmm1, %xmm2       ## xmm2 = xmm1[2,3,1,0]
+	vpxor	%xmm2, %xmm1, %xmm1
+	vpxor	%xmm1, %xmm0, %xmm0
+	vpshufd	$78, %xmm0, %xmm1       ## xmm1 = xmm0[2,3,0,1]
+	vpxor	%xmm1, %xmm0, %xmm0
+	vmovq	%xmm0, %rax
+	addq	$40, %rsp
+	popq	%rbx
+	popq	%rbp
+	retq
+	.cfi_endproc
+FN_END siphash_avx2