From bd6879ac518431174a490ba42f7e6e822dcb3ee1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?P=C3=A9ter=20Szil=C3=A1gyi?= <peterke@gmail.com>
Date: Mon, 5 Mar 2018 14:33:45 +0200
Subject: core/vm, crypto/bn256: switch over to cloudflare library (#16203)

* core/vm, crypto/bn256: switch over to cloudflare library

* crypto/bn256: unmarshal constraint + start pure go impl

* crypto/bn256: combo cloudflare and google lib

* travis: drop 386 test job
---
 crypto/bn256/cloudflare/mul_bmi2.h | 112 +++++++++++++++++++++++++++++++++++++
 1 file changed, 112 insertions(+)
 create mode 100644 crypto/bn256/cloudflare/mul_bmi2.h

(limited to 'crypto/bn256/cloudflare/mul_bmi2.h')

diff --git a/crypto/bn256/cloudflare/mul_bmi2.h b/crypto/bn256/cloudflare/mul_bmi2.h
new file mode 100644
index 000000000..71ad0499a
--- /dev/null
+++ b/crypto/bn256/cloudflare/mul_bmi2.h
@@ -0,0 +1,112 @@
+#define mulBMI2(a0,a1,a2,a3, rb) \
+	MOVQ a0, DX \
+	MOVQ $0, R13 \
+	MULXQ 0+rb, R8, R9 \
+	MULXQ 8+rb, AX, R10 \
+	ADDQ AX, R9 \
+	MULXQ 16+rb, AX, R11 \
+	ADCQ AX, R10 \
+	MULXQ 24+rb, AX, R12 \
+	ADCQ AX, R11 \
+	ADCQ $0, R12 \
+	ADCQ $0, R13 \
+	\
+	MOVQ a1, DX \
+	MOVQ $0, R14 \
+	MULXQ 0+rb, AX, BX \
+	ADDQ AX, R9 \
+	ADCQ BX, R10 \
+	MULXQ 16+rb, AX, BX \
+	ADCQ AX, R11 \
+	ADCQ BX, R12 \
+	ADCQ $0, R13 \
+	MULXQ 8+rb, AX, BX \
+	ADDQ AX, R10 \
+	ADCQ BX, R11 \
+	MULXQ 24+rb, AX, BX \
+	ADCQ AX, R12 \
+	ADCQ BX, R13 \
+	ADCQ $0, R14 \
+	\
+	MOVQ a2, DX \
+	MOVQ $0, R15 \
+	MULXQ 0+rb, AX, BX \
+	ADDQ AX, R10 \
+	ADCQ BX, R11 \
+	MULXQ 16+rb, AX, BX \
+	ADCQ AX, R12 \
+	ADCQ BX, R13 \
+	ADCQ $0, R14 \
+	MULXQ 8+rb, AX, BX \
+	ADDQ AX, R11 \
+	ADCQ BX, R12 \
+	MULXQ 24+rb, AX, BX \
+	ADCQ AX, R13 \
+	ADCQ BX, R14 \
+	ADCQ $0, R15 \
+	\
+	MOVQ a3, DX \
+	MULXQ 0+rb, AX, BX \
+	ADDQ AX, R11 \
+	ADCQ BX, R12 \
+	MULXQ 16+rb, AX, BX \
+	ADCQ AX, R13 \
+	ADCQ BX, R14 \
+	ADCQ $0, R15 \
+	MULXQ 8+rb, AX, BX \
+	ADDQ AX, R12 \
+	ADCQ BX, R13 \
+	MULXQ 24+rb, AX, BX \
+	ADCQ AX, R14 \
+	ADCQ BX, R15
+
+#define gfpReduceBMI2() \
+	\ // m = (T * N') mod R, store m in R8:R9:R10:R11
+	MOVQ ·np+0(SB), DX \
+	MULXQ 0(SP), R8, R9 \
+	MULXQ 8(SP), AX, R10 \
+	ADDQ AX, R9 \
+	MULXQ 16(SP), AX, R11 \
+	ADCQ AX, R10 \
+	MULXQ 24(SP), AX, BX \
+	ADCQ AX, R11 \
+	\
+	MOVQ ·np+8(SB), DX \
+	MULXQ 0(SP), AX, BX \
+	ADDQ AX, R9 \
+	ADCQ BX, R10 \
+	MULXQ 16(SP), AX, BX \
+	ADCQ AX, R11 \
+	MULXQ 8(SP), AX, BX \
+	ADDQ AX, R10 \
+	ADCQ BX, R11 \
+	\
+	MOVQ ·np+16(SB), DX \
+	MULXQ 0(SP), AX, BX \
+	ADDQ AX, R10 \
+	ADCQ BX, R11 \
+	MULXQ 8(SP), AX, BX \
+	ADDQ AX, R11 \
+	\
+	MOVQ ·np+24(SB), DX \
+	MULXQ 0(SP), AX, BX \
+	ADDQ AX, R11 \
+	\
+	storeBlock(R8,R9,R10,R11, 64(SP)) \
+	\
+	\ // m * N
+	mulBMI2(·p2+0(SB),·p2+8(SB),·p2+16(SB),·p2+24(SB), 64(SP)) \
+	\
+	\ // Add the 512-bit intermediate to m*N
+	MOVQ $0, AX \
+	ADDQ 0(SP), R8 \
+	ADCQ 8(SP), R9 \
+	ADCQ 16(SP), R10 \
+	ADCQ 24(SP), R11 \
+	ADCQ 32(SP), R12 \
+	ADCQ 40(SP), R13 \
+	ADCQ 48(SP), R14 \
+	ADCQ 56(SP), R15 \
+	ADCQ $0, AX \
+	\
+	gfpCarry(R12,R13,R14,R15,AX, R8,R9,R10,R11,BX)
-- 
cgit v1.2.3