From 9c3d34345c58ad954386433af8fcc88ee0f8b672 Mon Sep 17 00:00:00 2001 From: bunnie Date: Thu, 20 Aug 2020 15:47:02 +0800 Subject: [PATCH] add hardware acceleration to differential point add and double hacky implementation, but it'll let us do a performance benchmark --- Cargo.toml | 2 + src/field.rs | 133 ++++++++++++++++++++++++++++++++++++++ src/lib.rs | 2 + src/montgomery.rs | 159 +++++++++++++++++++++++++++++++++++++++++++++- 4 files changed, 294 insertions(+), 2 deletions(-) diff --git a/Cargo.toml b/Cargo.toml index 84ce1bc..ea6dd72 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -49,6 +49,8 @@ packed_simd = { version = "0.3", features = ["into_bits"], optional = true } zeroize = { version = "1", default-features = false } engine25519-as = { path="../engine25519-as", default-features = false, features = [] } rand = { version = "0.7" } +volatile = "0.2.6" +pac = { path = "../../../../sim_support/rust/pac" } [features] nightly = ["subtle/nightly"] diff --git a/src/field.rs b/src/field.rs index f997777..581d5f4 100644 --- a/src/field.rs +++ b/src/field.rs @@ -910,10 +910,143 @@ mod test { } } + + fn test_diff_add_and_double(mut file: &mut File) { + use montgomery::ProjectivePoint; + + // test cswap. three input registers: (r0, r1) to swap, (r2) to control swap, one output register (r31). + let num_src_regs = 5; + let reg_window = 0; + let num_tests = 8; + let loading_address = 0; // microcode loading address + + let mcode = assemble_engine25519!( + start: + // test preamble + psa %20, %0 + psa %21, %1 + psa %22, %2 + psa %23, %3 + psa %24, %4 + + // P.U in %20 + // P.W in %21 + // Q.U in %22 + // Q.W in %23 + // affine_PmQ in %24 + // %30 is the TRD scratch register + // %29 is the subtraction temporary value register + + // let t0 = &P.U + &P.W; + add %0, %20, %21 + trd %30, %0 + sub %0, %0, %30 + // let t1 = &P.U - &P.W; + sub %21, #3, %21 // negate &P.W using #FIELDPRIME (#3) + add %1, %20, %21 + trd %30, %1 + sub %1, %1, %30 + // let t2 = &Q.U + &Q.W; + add %2, %22, %23 + trd %30, %2 + sub %2, %2, %30 + // let t3 = &Q.U - &Q.W; + sub %23, #3, %23 + add %3, %22, %23 + trd %30, %3 + sub %3, %3, %30 + // let t4 = t0.square(); // (U_P + W_P)^2 = U_P^2 + 2 U_P W_P + W_P^2 + mul %4, %0, %0 + // let t5 = t1.square(); // (U_P - W_P)^2 = U_P^2 - 2 U_P W_P + W_P^2 + mul %5, %1, %1 + // let t6 = &t4 - &t5; // 4 U_P W_P + sub %29, #3, %5 + add %6, %4, %29 + trd %30, %6 + sub %6, %6, %30 + // let t7 = &t0 * &t3; // (U_P + W_P) (U_Q - W_Q) = U_P U_Q + W_P U_Q - U_P W_Q - W_P W_Q + mul %7, %0, %3 + // let t8 = &t1 * &t2; // (U_P - W_P) (U_Q + W_Q) = U_P U_Q - W_P U_Q + U_P W_Q - W_P W_Q + mul %8, %1, %2 + // let t9 = &t7 + &t8; // 2 (U_P U_Q - W_P W_Q) + add %9, %7, %8 + trd %30, %9 + sub %9, %9, %30 + // let t10 = &t7 - &t8; // 2 (W_P U_Q - U_P W_Q) + sub %29, #3, %8 + add %10, %7, %29 + trd %30, %10 + sub %10, %10, %30 + // let t11 = t9.square(); // 4 (U_P U_Q - W_P W_Q)^2 + mul %11, %9, %9 + // let t12 = t10.square(); // 4 (W_P U_Q - U_P W_Q)^2 + mul %12, %10, %10 + // let t13 = &APLUS2_OVER_FOUR * &t6; // (A + 2) U_P U_Q + mul %13, #4, %6 // #4 is A+2/4 + // let t14 = &t4 * &t5; // ((U_P + W_P)(U_P - W_P))^2 = (U_P^2 - W_P^2)^2 + mul %14, %4, %5 + // let t15 = &t13 + &t5; // (U_P - W_P)^2 + (A + 2) U_P W_P + add %15, %13, %5 + trd %30, %15 + sub %15, %15, %30 + // let t16 = &t6 * &t15; // 4 (U_P W_P) ((U_P - W_P)^2 + (A + 2) U_P W_P) + mul %16, %6, %15 + // let t17 = affine_PmQ * &t12; // U_D * 4 (W_P U_Q - U_P W_Q)^2 + mul %17, %24, %12 // affine_PmQ loaded into %24 + + ///// these can be eliminated down the road, but included for 1:1 algorithm correspodence to reference in early testing + // let t18 = t11; // W_D * 4 (U_P U_Q - W_P W_Q)^2 + psa %18, %11 + + // P.U = t14; // U_{P'} = (U_P + W_P)^2 (U_P - W_P)^2 + psa %20, %14 + // P.W = t16; // W_{P'} = (4 U_P W_P) ((U_P - W_P)^2 + ((A + 2)/4) 4 U_P W_P) + psa %21, %16 + // Q.U = t18; // U_{Q'} = W_D * 4 (U_P U_Q - W_P W_Q)^2 + psa %22, %18 + // Q.W = t17; // W_{Q'} = U_D * 4 (W_P U_Q - U_P W_Q)^2 + psa %23, %17 + + // test postamble -- sum together the points to create a single composite test output + add %31, %20, %21 + trd %30, %31 + sub %31, %31, %30 + add %31, %31, %22 + trd %30, %31 + sub %31, %31, %30 + add %31, %31, %23 + trd %30, %31 + sub %31, %31, %30 // leave result in r31 + + fin // finish execution + ); write_test_header(&mut file, loading_address, &mcode, num_src_regs, reg_window, num_tests); + + use montgomery::differential_add_and_double; + + // test vectors + for _ in 0..8 { + let pu = FieldElement::from_bytes(&rand::thread_rng().gen::<[u8; 32]>()); + let pw = FieldElement::from_bytes(&rand::thread_rng().gen::<[u8; 32]>()); + let qu = FieldElement::from_bytes(&rand::thread_rng().gen::<[u8; 32]>()); + let qw = FieldElement::from_bytes(&rand::thread_rng().gen::<[u8; 32]>()); + let pmq = FieldElement::from_bytes(&rand::thread_rng().gen::<[u8; 32]>()); + write_helper(&mut file, pu); + write_helper(&mut file, pw); + write_helper(&mut file, qu); + write_helper(&mut file, qw); + write_helper(&mut file, pmq); + let mut P: ProjectivePoint = ProjectivePoint{U:pu, W:pw}; + let mut Q: ProjectivePoint = ProjectivePoint{U:qu, W:qw}; + differential_add_and_double(&mut P, &mut Q, &pmq); + write_helper(&mut file, &(&P.U + &P.W) + &(&Q.U + &Q.W)); + } + } + test_add(&mut file); test_loop(&mut file); test_cswap(&mut file); test_mul(&mut file); + test_diff_add_and_double(&mut file); // end sequence let _ = file.write(&(0xFFFF_FFFF as u32).to_le_bytes()); diff --git a/src/lib.rs b/src/lib.rs index c1d809a..0930a76 100644 --- a/src/lib.rs +++ b/src/lib.rs @@ -62,6 +62,8 @@ pub(crate) mod macros; #[macro_use] extern crate engine25519_as; extern crate rand; +extern crate volatile; +extern crate pac; //------------------------------------------------------------------------ // curve25519-dalek public modules diff --git a/src/montgomery.rs b/src/montgomery.rs index c3676c2..87514d2 100644 --- a/src/montgomery.rs +++ b/src/montgomery.rs @@ -160,7 +160,7 @@ impl MontgomeryPoint { /// \\( \mathbb P(\mathbb F\_p) \\), which we identify with the Kummer /// line of the Montgomery curve. #[derive(Copy, Clone, Debug)] -struct ProjectivePoint { +pub struct ProjectivePoint { pub U: FieldElement, pub W: FieldElement, } @@ -220,7 +220,7 @@ impl ProjectivePoint { /// $$ /// (U\_Q : W\_Q) \gets u(P + Q). /// $$ -fn differential_add_and_double( +pub fn differential_add_and_double( P: &mut ProjectivePoint, Q: &mut ProjectivePoint, affine_PmQ: &FieldElement, @@ -260,6 +260,161 @@ fn differential_add_and_double( Q.W = t17; // W_{Q'} = U_D * 4 (W_P U_Q - U_P W_Q)^2 } +pub fn differential_add_and_double_hw( + P: &mut ProjectivePoint, + Q: &mut ProjectivePoint, + affine_PmQ: &FieldElement, +) { + use volatile::Volatile; + let p = unsafe { pac::Peripherals::steal() }; + let mcode = assemble_engine25519!( + start: + // P.U in %20 + // P.W in %21 + // Q.U in %22 + // Q.W in %23 + // affine_PmQ in %24 + // %30 is the TRD scratch register + // %29 is the subtraction temporary value register + + // let t0 = &P.U + &P.W; + add %0, %20, %21 + trd %30, %0 + sub %0, %0, %30 + // let t1 = &P.U - &P.W; + sub %21, #3, %21 // negate &P.W using #FIELDPRIME (#3) + add %1, %20, %21 + trd %30, %1 + sub %1, %1, %30 + // let t2 = &Q.U + &Q.W; + add %2, %22, %23 + trd %30, %2 + sub %2, %2, %30 + // let t3 = &Q.U - &Q.W; + sub %23, #3, %23 + add %3, %22, %23 + trd %30, %3 + sub %3, %3, %30 + // let t4 = t0.square(); // (U_P + W_P)^2 = U_P^2 + 2 U_P W_P + W_P^2 + mul %4, %0, %0 + // let t5 = t1.square(); // (U_P - W_P)^2 = U_P^2 - 2 U_P W_P + W_P^2 + mul %5, %1, %1 + // let t6 = &t4 - &t5; // 4 U_P W_P + sub %29, #3, %5 + add %6, %4, %29 + trd %30, %6 + sub %6, %6, %30 + // let t7 = &t0 * &t3; // (U_P + W_P) (U_Q - W_Q) = U_P U_Q + W_P U_Q - U_P W_Q - W_P W_Q + mul %7, %0, %3 + // let t8 = &t1 * &t2; // (U_P - W_P) (U_Q + W_Q) = U_P U_Q - W_P U_Q + U_P W_Q - W_P W_Q + mul %8, %1, %2 + // let t9 = &t7 + &t8; // 2 (U_P U_Q - W_P W_Q) + add %9, %7, %8 + trd %30, %9 + sub %9, %9, %30 + // let t10 = &t7 - &t8; // 2 (W_P U_Q - U_P W_Q) + sub %29, #3, %8 + add %10, %7, %29 + trd %30, %10 + sub %10, %10, %30 + // let t11 = t9.square(); // 4 (U_P U_Q - W_P W_Q)^2 + mul %11, %9, %9 + // let t12 = t10.square(); // 4 (W_P U_Q - U_P W_Q)^2 + mul %12, %10, %10 + // let t13 = &APLUS2_OVER_FOUR * &t6; // (A + 2) U_P U_Q + mul %13, #4, %6 // #4 is A+2/4 + // let t14 = &t4 * &t5; // ((U_P + W_P)(U_P - W_P))^2 = (U_P^2 - W_P^2)^2 + mul %14, %4, %5 + // let t15 = &t13 + &t5; // (U_P - W_P)^2 + (A + 2) U_P W_P + add %15, %13, %5 + trd %30, %15 + sub %15, %15, %30 + // let t16 = &t6 * &t15; // 4 (U_P W_P) ((U_P - W_P)^2 + (A + 2) U_P W_P) + mul %16, %6, %15 + // let t17 = affine_PmQ * &t12; // U_D * 4 (W_P U_Q - U_P W_Q)^2 + mul %17, %24, %12 // affine_PmQ loaded into %24 + + ///// these can be eliminated down the road, but included for 1:1 algorithm correspodence to reference in early testing + // let t18 = t11; // W_D * 4 (U_P U_Q - W_P W_Q)^2 + psa %18, %11 + // P.U = t14; // U_{P'} = (U_P + W_P)^2 (U_P - W_P)^2 + psa %20, %14 + // P.W = t16; // W_{P'} = (4 U_P W_P) ((U_P - W_P)^2 + ((A + 2)/4) 4 U_P W_P) + psa %21, %16 + // Q.U = t18; // U_{Q'} = W_D * 4 (U_P U_Q - W_P W_Q)^2 + psa %22, %18 + // Q.W = t17; // W_{Q'} = U_D * 4 (W_P U_Q - U_P W_Q)^2 + psa %23, %17 + + fin // finish execution + ); + let microcode_ptr: *mut u32 = 0xe002_0000 as *mut u32; + let microcode = microcode_ptr as *mut Volatile; + + // copy the microcode in -- later on we can optimize this so it's only done once? + for i in 0..mcode.len() as usize { + unsafe { (*(microcode.add(i))).write( mcode[i] ); } + } + // setup the engine microcode parameters + unsafe{ + p.ENGINE.window.write(|w| w.bits(0)); + p.ENGINE.mpstart.write(|w| w.bits(0)); + p.ENGINE.mplen.write(|w| w.bits(mcode.len() as u32)); + } + + fn copy_to_rf(bytes: [u8; 32], register: usize) { + let rf_ptr: *mut u32 = 0xe003_0000 as *mut u32; + let rf = rf_ptr as *mut Volatile; + for word in 0..8 { + let mut temp: [u8; 4] = [0; 4]; + for i in 0..4 { + temp[i] = bytes[word*4 + i]; + } + unsafe { (*( rf.add( (register * 8 + word) as usize )) ).write( u32::from_le_bytes(temp) ); } + } + } + + // P.U in %20 + // P.W in %21 + // Q.U in %22 + // Q.W in %23 + // affine_PmQ in %24 + copy_to_rf(P.U.to_bytes(), 20); + copy_to_rf(P.W.to_bytes(), 21); + copy_to_rf(Q.U.to_bytes(), 22); + copy_to_rf(Q.W.to_bytes(), 23); + copy_to_rf(affine_PmQ.to_bytes(), 24); + + // start the run + p.ENGINE.control.write(|w| w.go().set_bit()); + loop { + let status = p.ENGINE.status.read().bits(); + if (status & 1) == 0 { + break; + } + } + + fn copy_from_rf(register: usize) -> [u8; 32] { + let rf_ptr: *mut u32 = 0xe003_0000 as *mut u32; + let rf = rf_ptr as *mut Volatile; + let mut bytes: [u8; 32] = [0; 32]; + for word in 0..8 { + unsafe{ + let value: u32 = (*( rf.add( (register * 8 + word) as usize ))).read(); + let b = value.to_le_bytes(); + for i in 0..4 { + bytes[word*4 + i] = b[i]; + } + } + } + bytes + } + P.U = FieldElement::from_bytes(©_from_rf(20)); + P.W = FieldElement::from_bytes(©_from_rf(21)); + Q.U = FieldElement::from_bytes(©_from_rf(22)); + Q.W = FieldElement::from_bytes(©_from_rf(23)); +} + define_mul_assign_variants!(LHS = MontgomeryPoint, RHS = Scalar); define_mul_variants!(LHS = MontgomeryPoint, RHS = Scalar, Output = MontgomeryPoint);