From 93abd35ffb0abea517f93646955ca0dabde75f63 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 11:06:49 -0800
Subject: [PATCH 01/12] optimizations

---
 .../kernel/asm/hash/blake2b/compression.asm   |  29 +--
 .../kernel/asm/hash/blake2b/g_functions.asm   | 205 +++++++++++++-----
 2 files changed, 154 insertions(+), 80 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
index a25158d9..840d8c54 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
@@ -181,30 +181,15 @@ compression_loop:
     POP
     POP
     // stack: cur_block, retdest
-    %blake2b_internal_state_addr
-    // stack: start, cur_block, retdest
-    PUSH 0
-    // stack: round=0, start, cur_block, retdest
 
     // Run 12 rounds of G functions.
-    %rep 12
-        // stack: round, start, cur_block, retdest
-        %call_blake2b_g_function(0, 4, 8, 12, 0, 1)
-        %call_blake2b_g_function(1, 5, 9, 13, 2, 3)
-        %call_blake2b_g_function(2, 6, 10, 14, 4, 5)
-        %call_blake2b_g_function(3, 7, 11, 15, 6, 7)
-        %call_blake2b_g_function(0, 5, 10, 15, 8, 9)
-        %call_blake2b_g_function(1, 6, 11, 12, 10, 11)
-        %call_blake2b_g_function(2, 7, 8, 13, 12, 13)
-        %call_blake2b_g_function(3, 4, 9, 14, 14, 15)
-        // stack: round, start, cur_block, retdest
-        %increment
-        // stack: round + 1, start, cur_block, retdest
-    %endrep
-    // stack: 12, start, cur_block, retdest
-    POP
-    POP
-
+    PUSH g_functions_return
+    // stack: g_functions_return, cur_block, retdest
+    %blake2b_internal_state_addr
+    // stack: start, g_functions_return, cur_block, retdest
+    %jump(run_12_rounds_g_function)
+g_functions_return:
+    
     // Finalize hash value.
     // stack: cur_block, retdest
     %blake2b_generate_new_hash_value(7)
diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
index 11e879fc..831841c5 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
@@ -1,89 +1,89 @@
-%macro blake2b_g_function
+blake2b_g_function:
     // Function to mix two input words, x and y, into the four words indexed by a, b, c, d (which
     // are in the range 0..16) in the internal state.
     // The internal state is stored in memory starting at the address start.
-    // stack: a, b, c, d, x, y, start
+    // stack: a, b, c, d, x, y, start, retdest
     %stack (indices: 4) -> (indices, indices)
-    // stack: a, b, c, d, a, b, c, d, x, y, start
+    // stack: a, b, c, d, a, b, c, d, x, y, start, retdest
     DUP11
-    // stack: start, a, b, c, d, a, b, c, d, x, y, start
+    // stack: start, a, b, c, d, a, b, c, d, x, y, start, retdest
     %stack (start, a, b, c, d) -> (d, start, c, start, b, start, a, start)
-    // stack: d, start, c, start, b, start, a, start, a, b, c, d, x, y, start
+    // stack: d, start, c, start, b, start, a, start, a, b, c, d, x, y, start, retdest
     ADD
     %mload_kernel_general
-    // stack: v[d], c, start, b, start, a, start, a, b, c, d, x, y, start
+    // stack: v[d], c, start, b, start, a, start, a, b, c, d, x, y, start, retdest
     %stack (vd, remaining: 6) -> (remaining, vd)
-    // stack: c, start, b, start, a, start, v[d], a, b, c, d, x, y, start
+    // stack: c, start, b, start, a, start, v[d], a, b, c, d, x, y, start, retdest
     ADD
     %mload_kernel_general
     %stack (vc, remaining: 4) -> (remaining, vc)
-    // stack: b, start, a, start, v[c], v[d], a, b, c, d, x, y, start
+    // stack: b, start, a, start, v[c], v[d], a, b, c, d, x, y, start, retdest
     ADD
     %mload_kernel_general
-    // stack: v[b], a, start, v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[b], a, start, v[c], v[d], a, b, c, d, x, y, start, retdest
     %stack (vb, remaining: 2) -> (remaining, vb)
-    // stack: a, start, v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: a, start, v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     ADD
     %mload_kernel_general
-    // stack: v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[a], v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     DUP2
-    // stack: v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     DUP10
-    // stack: x, v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: x, v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     ADD
     ADD
     %as_u64
-    // stack: v[a]' = (v[a] + v[b] + x) % 2^64, v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[a]' = (v[a] + v[b] + x) % 2^64, v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     %stack (a, b, c, d) -> (a, d, a, b, c, d)
-    // stack: v[a]', v[d], v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[a]', v[d], v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     XOR
     %rotr_64(32)
-    // stack: v[d]' = (v[d] ^ v[a]') >>> 32, v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[d]' = (v[d] ^ v[a]') >>> 32, v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
     %stack (top: 4, vd) -> (top)
-    // stack: v[d]', v[a]', v[b], v[c], a, b, c, d, x, y, start
+    // stack: v[d]', v[a]', v[b], v[c], a, b, c, d, x, y, start, retdest
     %stack (d, a, b, c) -> (c, d, a, b, d)
-    // stack: v[c], v[d]', v[a]', v[b], v[d]', a, b, c, d, x, y, start
+    // stack: v[c], v[d]', v[a]', v[b], v[d]', a, b, c, d, x, y, start, retdest
     ADD
     %as_u64
-    // stack: v[c]' = (v[c] + v[d]') % 2^64, v[a]', v[b], v[d]', a, b, c, d, x, y, start
+    // stack: v[c]' = (v[c] + v[d]') % 2^64, v[a]', v[b], v[d]', a, b, c, d, x, y, start, retdest
     %stack (c, a, b, d) -> (b, c, a, c, d)
-    // stack: v[b], v[c]', v[a]', v[c]', v[d]', a, b, c, d, x, y, start
+    // stack: v[b], v[c]', v[a]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
     XOR
     %rotr_64(24)
-    // stack: v[b]' = (v[b] ^ v[c]') >>> 24, v[a]', v[c]', v[d]', a, b, c, d, x, y, start
+    // stack: v[b]' = (v[b] ^ v[c]') >>> 24, v[a]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
     SWAP1
-    // stack: v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start
+    // stack: v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
     DUP2
-    // stack: v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start
+    // stack: v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
     DUP11
-    // stack: y, v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start
+    // stack: y, v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
     ADD
     ADD
     %as_u64
-    // stack: v[a]'' = (v[a]' + v[b]' + y) % 2^64, v[b]', v[c]', v[d]', a, b, c, d, x, y, start
+    // stack: v[a]'' = (v[a]' + v[b]' + y) % 2^64, v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
     SWAP3
-    // stack: v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start
+    // stack: v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start, retdest
     DUP4
-    // stack: v[a]'', v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start
+    // stack: v[a]'', v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start, retdest
     XOR
     %rotr_64(16)
-    // stack: v[d]'' = (v[a]'' ^ v[d]') >>> 8, v[b]', v[c]', v[a]'', a, b, c, d, x, y, start
+    // stack: v[d]'' = (v[a]'' ^ v[d]') >>> 8, v[b]', v[c]', v[a]'', a, b, c, d, x, y, start, retdest
     SWAP2
-    // stack: v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
+    // stack: v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
     DUP3
-    // stack: v[d]'', v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
+    // stack: v[d]'', v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
     ADD
     %as_u64
-    // stack: v[c]'' = (v[c]' + v[d]'') % 2^64, v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
+    // stack: v[c]'' = (v[c]' + v[d]'') % 2^64, v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
     DUP1
-    // stack: v[c]'', v[c]'', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
+    // stack: v[c]'', v[c]'', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
     SWAP2
-    // stack: v[b]', v[c]'', v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start
+    // stack: v[b]', v[c]'', v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
     XOR
     %rotr_64(63)
-    // stack: v[b]'' = (v[b]' ^ v[c]'') >>> 7, v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start
+    // stack: v[b]'' = (v[b]' ^ v[c]'') >>> 7, v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
     %stack (vb, vc, vd, va, a, b, c, d, x, y, start) -> (start, a, va, start, b, vb, start, c, vc, start, d, vd)
-    // stack: start, a, v[a]'', start, b, v[b]'', start, c, v[c]'', start, d, v[d]''
+    // stack: start, a, v[a]'', start, b, v[b]'', start, c, v[c]'', start, d, v[d]'', retdest
     ADD
     %mstore_kernel_general
     ADD
@@ -92,35 +92,124 @@
     %mstore_kernel_general
     ADD
     %mstore_kernel_general
-%endmacro
+    // stack: retdest
+    JUMP
 
-%macro call_blake2b_g_function(a, b, c, d, x_idx, y_idx)
-    // stack: round, start
-    PUSH $y_idx
-    DUP2
-    // stack: round, y_idx, round, start
+call_blake2b_g_function:
+    // stack: a, b, c, d, x_idx, y_idx, round, start, retdest
+    DUP6
+    // stack: y_idx, a, b, c, d, x_idx, y_idx, round, start, retdest
+    DUP8
+    // stack: round, y_idx, a, b, c, d, x_idx, y_idx, round, start, retdest
     %blake2b_permutation
-    // stack: s[y_idx], round, start
+    // stack: s[y_idx], a, b, c, d, x_idx, y_idx, round, start, retdest
     %blake2b_message_addr
     ADD
     %mload_kernel_general
-    // stack: m[s[y_idx]], round, start
-    PUSH $x_idx
-    DUP3
-    // stack: round, 2, m[s[y_idx]], round, start
+    // stack: m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
+    DUP6
+    // stack: x_idx, m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
+    DUP9
+    // stack: round, x_idx, m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
     %blake2b_permutation
-    // stack: s[x_idx], m[s[y_idx]], round, start
+    // stack: s[x_idx], m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
     %blake2b_message_addr
     ADD
     %mload_kernel_general
-    // stack: m[s[x_idx]], m[s[y_idx]], round, start
-    %stack (ss: 2, r, s) -> (ss, s, r, s)
-    // stack: m[s[x_idx]], m[s[y_idx]], start, round, start
-    PUSH $d
-    PUSH $c
-    PUSH $b
-    PUSH $a
-    // stack: a, b, c, d, m[s[x_idx]], m[s[y_idx]], start, round, start
-    %blake2b_g_function
-    // stack: round, start
-%endmacro
+    // stack: m[s[x_idx]], m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
+    %stack (mm: 2, abcd: 4, xy: 2, r, s) -> (abcd, mm, s, r, s)
+    // stack: a, b, c, d, m[s[x_idx]], m[s[y_idx]], start, round, start, retdest
+    %jump(blake2b_g_function)
+
+global run_g_function_round:
+    // stack: round, start, retdest
+    PUSH g_function_return_1
+    // stack: g_function_return_1, round, start, retdest
+    %stack (ret, r, s) -> (0, 4, 8, 12, 0, 1, r, s, ret, r, s)
+    // stack: a=0, b=4, c=8, d=12, x_idx=0, y_idx=1, round, start, g_function_return_1, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_1:
+    // stack: round, start, retdest
+    PUSH g_function_return_2
+    // stack: g_function_return_2, round, start, retdest
+    %stack (ret, r, s) -> (1, 5, 9, 13, 2, 3, r, s, ret, r, s)
+    // stack: a=1, b=5, c=9, d=13, x_idx=2, y_idx=3, round, start, g_function_return_2, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_2:
+    // stack: round, start, retdest
+    PUSH g_function_return_3
+    // stack: g_function_return_3, round, start, retdest
+    %stack (ret, r, s) -> (2, 6, 10, 14, 4, 5, r, s, ret, r, s)
+    // stack: a=2, b=6, c=10, d=14, x_idx=4, y_idx=5, round, start, g_function_return_3, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_3:
+    // stack: round, start, retdest
+    PUSH g_function_return_4
+    // stack: g_function_return_4, round, start, retdest
+    %stack (ret, r, s) -> (3, 7, 11, 15, 6, 7, r, s, ret, r, s)
+    // stack: a=3, b=7, c=11, d=15, x_idx=6, y_idx=7, round, start, g_function_return_4, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_4:
+    // stack: round, start, retdest
+    PUSH g_function_return_5
+    // stack: g_function_return_5, round, start, retdest
+    %stack (ret, r, s) -> (0, 5, 10, 15, 8, 9, r, s, ret, r, s)
+    // stack: a=0, b=5, c=10, d=15, x_idx=8, y_idx=9, round, start, g_function_return_5, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_5:
+    // stack: round, start, retdest
+    PUSH g_function_return_6
+    // stack: g_function_return_6, round, start, retdest
+    %stack (ret, r, s) -> (1, 6, 11, 12, 10, 11, r, s, ret, r, s)
+    // stack: a=1, b=6, c=11, d=12, x_idx=10, y_idx=11, round, start, g_function_return_6, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_6:
+    // stack: round, start, retdest
+    PUSH g_function_return_7
+    // stack: g_function_return_7, round, start, retdest
+    %stack (ret, r, s) -> (2, 7, 8, 13, 12, 13, r, s, ret, r, s)
+    // stack: a=2, b=7, c=8, d=13, x_idx=12, y_idx=13, round, start, g_function_return_7, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_7:
+    // stack: round, start, retdest
+    PUSH g_function_return_8
+    // stack: g_function_return_8, round, start, retdest
+    %stack (ret, r, s) -> (3, 4, 9, 14, 14, 15, r, s, ret, r, s)
+    // stack: a=3, b=4, c=9, d=14, x_idx=14, y_idx=15, round, start, g_function_return_8, round, start, retdest
+    %jump(call_blake2b_g_function)
+g_function_return_8:
+    // stack: round, start, retdest
+    SWAP1
+    // stack: start, round, retdest
+    SWAP2
+    // stack: retdest, round, start
+    JUMP
+
+
+global run_12_rounds_g_function:
+    // stack: start, retdest
+    PUSH 0
+    // stack: round=0, start, retdest
+run_next_round_g_function:
+    // stack: round, start, retdest
+    PUSH run_g_function_round_return
+    // stack: run_g_function_round_return, round, start, retdest
+    SWAP2
+    // stack: start, round, run_g_function_round_return, retdest
+    SWAP1
+    // stack: round, start, run_g_function_round_return, retdest
+    %jump(run_g_function_round)
+run_next_round_g_function_return:
+    // stack: round, start, retdest
+    %increment
+    // stack: round+1, start, retdest
+    DUP1
+    // stack: round+1, round+1, start, retdest
+    %lt_const(12)
+    // stack: round+1 < 12, round+1, start, retdest
+    %jumpi(run_next_round_g_function)
+    // stack: round+1, start, retdest
+    %pop2
+    // stack: retdest
+    JUMP
+

From 5f592e60dc1248a80d54ed39139800eb97e9f95a Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 12:25:34 -0800
Subject: [PATCH 02/12] fixes

---
 evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm | 12 ++++++------
 evm/src/cpu/kernel/tests/hash.rs                    |  2 ++
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
index 831841c5..ff3ee43a 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
@@ -117,8 +117,8 @@ call_blake2b_g_function:
     ADD
     %mload_kernel_general
     // stack: m[s[x_idx]], m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
-    %stack (mm: 2, abcd: 4, xy: 2, r, s) -> (abcd, mm, s, r, s)
-    // stack: a, b, c, d, m[s[x_idx]], m[s[y_idx]], start, round, start, retdest
+    %stack (mm: 2, abcd: 4, xy: 2, r, s) -> (abcd, mm, s)
+    // stack: a, b, c, d, m[s[x_idx]], m[s[y_idx]], start, retdest
     %jump(blake2b_g_function)
 
 global run_g_function_round:
@@ -192,12 +192,12 @@ global run_12_rounds_g_function:
     // stack: round=0, start, retdest
 run_next_round_g_function:
     // stack: round, start, retdest
-    PUSH run_g_function_round_return
-    // stack: run_g_function_round_return, round, start, retdest
+    PUSH run_next_round_g_function_return
+    // stack: run_next_round_g_function_return, round, start, retdest
     SWAP2
-    // stack: start, round, run_g_function_round_return, retdest
+    // stack: start, round, run_next_round_g_function_return, retdest
     SWAP1
-    // stack: round, start, run_g_function_round_return, retdest
+    // stack: round, start, run_next_round_g_function_return, retdest
     %jump(run_g_function_round)
 run_next_round_g_function_return:
     // stack: round, start, retdest
diff --git a/evm/src/cpu/kernel/tests/hash.rs b/evm/src/cpu/kernel/tests/hash.rs
index bc73ecd5..cc2c96eb 100644
--- a/evm/src/cpu/kernel/tests/hash.rs
+++ b/evm/src/cpu/kernel/tests/hash.rs
@@ -79,6 +79,8 @@ fn prepare_test<T>(
     // Run the interpeter
     let result = run_interpreter_with_memory(interpreter_setup).unwrap();
 
+    dbg!(result.stack().to_vec());
+
     Ok((expected, result.stack().to_vec()))
 }
 

From 70475a5a87bc3b2d9611bb4b66617dc636a43a1d Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 12:27:08 -0800
Subject: [PATCH 03/12] cleanup

---
 evm/src/cpu/kernel/tests/hash.rs | 2 --
 1 file changed, 2 deletions(-)

diff --git a/evm/src/cpu/kernel/tests/hash.rs b/evm/src/cpu/kernel/tests/hash.rs
index cc2c96eb..bc73ecd5 100644
--- a/evm/src/cpu/kernel/tests/hash.rs
+++ b/evm/src/cpu/kernel/tests/hash.rs
@@ -79,8 +79,6 @@ fn prepare_test<T>(
     // Run the interpeter
     let result = run_interpreter_with_memory(interpreter_setup).unwrap();
 
-    dbg!(result.stack().to_vec());
-
     Ok((expected, result.stack().to_vec()))
 }
 

From 40f90d8312f51e81407c48d7e464924b71d8f0d1 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 15:45:20 -0800
Subject: [PATCH 04/12] hash function optimization

---
 .../kernel/asm/hash/blake2b/compression.asm   | 12 +--
 evm/src/cpu/kernel/asm/hash/blake2b/hash.asm  | 91 +++++++++++++++++--
 2 files changed, 86 insertions(+), 17 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
index 840d8c54..fdf02d69 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
@@ -192,14 +192,10 @@ g_functions_return:
     
     // Finalize hash value.
     // stack: cur_block, retdest
-    %blake2b_generate_new_hash_value(7)
-    %blake2b_generate_new_hash_value(6)
-    %blake2b_generate_new_hash_value(5)
-    %blake2b_generate_new_hash_value(4)
-    %blake2b_generate_new_hash_value(3)
-    %blake2b_generate_new_hash_value(2)
-    %blake2b_generate_new_hash_value(1)
-    %blake2b_generate_new_hash_value(0)
+    PUSH hash_generate_return
+    // stack: hash_generate_return, cur_block, retdest
+    %jump(blake2b_generate_all_hash_values)
+hash_generate_return:
     // stack: h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7', cur_block, retdest
     DUP9
     // stack: cur_block, h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7', cur_block, retdest
diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm b/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
index 712a97c0..91a5530e 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
@@ -1,18 +1,91 @@
-%macro blake2b_generate_new_hash_value(i)
+blake2b_generate_new_hash_value:
+    // stack: i, retdest
     %blake2b_hash_value_addr
-    %add_const($i)
+    // stack: addr, i, retdest
+    DUP2
+    ADD
     %mload_kernel_general
-    // stack: h_i, ...
+    // stack: h_i, i, retdest
     %blake2b_internal_state_addr
-    %add_const($i)
+    // stack: addr, h_i, i, retdest
+    DUP3
+    ADD
     %mload_kernel_general
-    // stack: v_i, h_i, ...
+    // stack: v_i, h_i, i, retdest
     %blake2b_internal_state_addr
-    %add_const($i)
+    // stack: addr, v_i, h_i, i, retdest
+    DUP4
+    ADD
     %add_const(8)
     %mload_kernel_general
-    // stack: v_(i+8), v_i, h_i, ...
+    // stack: v_(i+8), v_i, h_i, i, retdest
     XOR
     XOR
-    // stack: h_i' = v_(i+8) ^ v_i ^ h_i, ...
-%endmacro
+    // stack: h_i' = v_(i+8) ^ v_i ^ h_i, i, retdest
+    SWAP1
+    POP
+    // stack: h_i', retdest
+    SWAP1
+    JUMP
+
+global blake2b_generate_all_hash_values:
+    // stack: retdest
+    PUSH blake2b_generate_hash_return_7
+    // stack: blake2b_generate_hash_return_7, retdest
+    PUSH 7
+    // stack: 7, blake2b_generate_hash_return_7, retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_7:
+    // stack: h_7', retdest
+    PUSH blake2b_generate_hash_return_6
+    // stack: blake2b_generate_hash_return_6, h_7', retdest
+    PUSH 6
+    // stack: 6, blake2b_generate_hash_return_6, h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_6:
+    // stack: h_6', h_7', retdest
+    PUSH blake2b_generate_hash_return_5
+    // stack: blake2b_generate_hash_return_5, h_6', h_7', retdest
+    PUSH 5
+    // stack: 5, blake2b_generate_hash_return_5, h_6', h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_5:
+    // stack: h_5', h_6', h_7', retdest
+    PUSH blake2b_generate_hash_return_4
+    // stack: blake2b_generate_hash_return_4, h_5', h_6', h_7', retdest
+    PUSH 4
+    // stack: 4, blake2b_generate_hash_return_4, h_5', h_6', h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_4:
+    // stack: h_4', h_5', h_6', h_7', retdest
+    PUSH blake2b_generate_hash_return_3
+    // stack: blake2b_generate_hash_return_3, h_4', h_5', h_6', h_7', retdest
+    PUSH 3
+    // stack: 3, blake2b_generate_hash_return_3, h_4', h_5', h_6', h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_3:
+    // stack: h_3', h_4', h_5', h_6', h_7', retdest
+    PUSH blake2b_generate_hash_return_2
+    // stack: blake2b_generate_hash_return_2, h_3', h_4', h_5', h_6', h_7', retdest
+    PUSH 2
+    // stack: 2, blake2b_generate_hash_return_2, h_3', h_4', h_5', h_6', h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_2:
+    // stack: h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    PUSH blake2b_generate_hash_return_1
+    // stack: blake2b_generate_hash_return_1, h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    PUSH 1
+    // stack: 1, blake2b_generate_hash_return_1, h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_1:
+    // stack: h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    PUSH blake2b_generate_hash_return_0
+    // stack: blake2b_generate_hash_return_0, h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    PUSH 0
+    // stack: 0, blake2b_generate_hash_return_0, h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    %jump(blake2b_generate_new_hash_value)
+blake2b_generate_hash_return_0:
+    // stack: h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
+    %stack (h: 8, ret) -> (ret, h)
+    // stack: retdest, h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7'
+    JUMP
\ No newline at end of file

From c37d1e25fcf4697c5ef3033b7a7390b37b3043b1 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 15:45:52 -0800
Subject: [PATCH 05/12] optimize hash generation further

---
 evm/src/cpu/kernel/asm/hash/blake2b/hash.asm | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm b/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
index 91a5530e..945c7fb0 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
@@ -14,17 +14,17 @@ blake2b_generate_new_hash_value:
     // stack: v_i, h_i, i, retdest
     %blake2b_internal_state_addr
     // stack: addr, v_i, h_i, i, retdest
-    DUP4
+    SWAP1
+    // stack: v_i, addr, h_i, i, retdest
+    SWAP3
+    // stack: i, addr, h_i, v_i, retdest
     ADD
     %add_const(8)
     %mload_kernel_general
-    // stack: v_(i+8), v_i, h_i, i, retdest
+    // stack: v_(i+8), h_i, v_i, retdest
     XOR
     XOR
-    // stack: h_i' = v_(i+8) ^ v_i ^ h_i, i, retdest
-    SWAP1
-    POP
-    // stack: h_i', retdest
+    // stack: h_i' = v_(i+8) ^ v_i ^ h_i, retdest
     SWAP1
     JUMP
 
@@ -88,4 +88,4 @@ blake2b_generate_hash_return_0:
     // stack: h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
     %stack (h: 8, ret) -> (ret, h)
     // stack: retdest, h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7'
-    JUMP
\ No newline at end of file
+    JUMP

From 9d8d81b4b0e4f5570edbda8414d13b4fa14d07b5 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 16:19:15 -0800
Subject: [PATCH 06/12] optimize hash generation further further

---
 evm/src/cpu/kernel/asm/hash/blake2b/hash.asm | 77 +++++---------------
 evm/src/cpu/kernel/tests/hash.rs             |  2 +
 2 files changed, 22 insertions(+), 57 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm b/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
index 945c7fb0..654b51b3 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/hash.asm
@@ -30,62 +30,25 @@ blake2b_generate_new_hash_value:
 
 global blake2b_generate_all_hash_values:
     // stack: retdest
-    PUSH blake2b_generate_hash_return_7
-    // stack: blake2b_generate_hash_return_7, retdest
-    PUSH 7
-    // stack: 7, blake2b_generate_hash_return_7, retdest
+    PUSH 8
+    // stack: i=8, retdest
+blake2b_generate_hash_loop:
+    // stack: i, h_i', ..., h_7', retdest
+    %decrement
+    // stack: i-1, h_i', ..., h_7', retdest
+    PUSH blake2b_generate_hash_return
+    // stack: blake2b_generate_hash_return, i-1, h_i', ..., h_7', retdest
+    DUP2
+    // stack: i-1, blake2b_generate_hash_return, i-1, h_i', ..., h_7', retdest
     %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_7:
-    // stack: h_7', retdest
-    PUSH blake2b_generate_hash_return_6
-    // stack: blake2b_generate_hash_return_6, h_7', retdest
-    PUSH 6
-    // stack: 6, blake2b_generate_hash_return_6, h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_6:
-    // stack: h_6', h_7', retdest
-    PUSH blake2b_generate_hash_return_5
-    // stack: blake2b_generate_hash_return_5, h_6', h_7', retdest
-    PUSH 5
-    // stack: 5, blake2b_generate_hash_return_5, h_6', h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_5:
-    // stack: h_5', h_6', h_7', retdest
-    PUSH blake2b_generate_hash_return_4
-    // stack: blake2b_generate_hash_return_4, h_5', h_6', h_7', retdest
-    PUSH 4
-    // stack: 4, blake2b_generate_hash_return_4, h_5', h_6', h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_4:
-    // stack: h_4', h_5', h_6', h_7', retdest
-    PUSH blake2b_generate_hash_return_3
-    // stack: blake2b_generate_hash_return_3, h_4', h_5', h_6', h_7', retdest
-    PUSH 3
-    // stack: 3, blake2b_generate_hash_return_3, h_4', h_5', h_6', h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_3:
-    // stack: h_3', h_4', h_5', h_6', h_7', retdest
-    PUSH blake2b_generate_hash_return_2
-    // stack: blake2b_generate_hash_return_2, h_3', h_4', h_5', h_6', h_7', retdest
-    PUSH 2
-    // stack: 2, blake2b_generate_hash_return_2, h_3', h_4', h_5', h_6', h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_2:
-    // stack: h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    PUSH blake2b_generate_hash_return_1
-    // stack: blake2b_generate_hash_return_1, h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    PUSH 1
-    // stack: 1, blake2b_generate_hash_return_1, h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_1:
-    // stack: h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    PUSH blake2b_generate_hash_return_0
-    // stack: blake2b_generate_hash_return_0, h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    PUSH 0
-    // stack: 0, blake2b_generate_hash_return_0, h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    %jump(blake2b_generate_new_hash_value)
-blake2b_generate_hash_return_0:
-    // stack: h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7', retdest
-    %stack (h: 8, ret) -> (ret, h)
-    // stack: retdest, h_0', h_1', h_2', h_3', h_4', h_5', h_6', h_7'
+blake2b_generate_hash_return:
+    // stack: h_(i-1)', i-1, h_i', ..., h_7', retdest
+    SWAP1
+    // stack: i-1, h_(i-1)', h_i', ..., h_7', retdest
+    DUP1
+    // stack: i-1, i-1, h_(i-1)', ..., h_7', retdest
+    %jumpi(blake2b_generate_hash_loop)
+    // stack: i-1=0, h_0', ..., h_7', retdest
+    %stack (i, h: 8, ret) -> (ret, h)
+    // stack: retdest, h_0'...h_7'
     JUMP
diff --git a/evm/src/cpu/kernel/tests/hash.rs b/evm/src/cpu/kernel/tests/hash.rs
index bc73ecd5..cc2c96eb 100644
--- a/evm/src/cpu/kernel/tests/hash.rs
+++ b/evm/src/cpu/kernel/tests/hash.rs
@@ -79,6 +79,8 @@ fn prepare_test<T>(
     // Run the interpeter
     let result = run_interpreter_with_memory(interpreter_setup).unwrap();
 
+    dbg!(result.stack().to_vec());
+
     Ok((expected, result.stack().to_vec()))
 }
 

From 5994f4d932e538f69cd5904b07e22b4c66456614 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 16:49:28 -0800
Subject: [PATCH 07/12] cleanup

---
 evm/src/cpu/kernel/asm/hash/blake2b/compression.asm | 1 -
 1 file changed, 1 deletion(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
index fdf02d69..11e5389b 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
@@ -189,7 +189,6 @@ compression_loop:
     // stack: start, g_functions_return, cur_block, retdest
     %jump(run_12_rounds_g_function)
 g_functions_return:
-    
     // Finalize hash value.
     // stack: cur_block, retdest
     PUSH hash_generate_return

From 0f55956ade0ef43f99dd07daf6807e5791f7eb20 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 7 Mar 2023 17:17:23 -0800
Subject: [PATCH 08/12] optimized initial hash value generation

---
 .../cpu/kernel/asm/hash/blake2b/addresses.asm | 16 -----
 .../kernel/asm/hash/blake2b/compression.asm   |  4 +-
 evm/src/cpu/kernel/asm/hash/blake2b/iv.asm    | 61 ++++++++++++++-----
 evm/src/cpu/kernel/tests/hash.rs              |  2 -
 4 files changed, 50 insertions(+), 33 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/addresses.asm b/evm/src/cpu/kernel/asm/hash/blake2b/addresses.asm
index 9d65b9ed..f1d7c3e9 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/addresses.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/addresses.asm
@@ -1,19 +1,3 @@
-// Load the initial hash value (the IV, but with params XOR'd into the first word).
-%macro blake2b_initial_hash_value
-    %blake2b_iv_i(7)
-    %blake2b_iv_i(6)
-    %blake2b_iv_i(5)
-    %blake2b_iv_i(4)
-    %blake2b_iv_i(3)
-    %blake2b_iv_i(2)
-    %blake2b_iv_i(1)
-    // stack: IV_1, IV_2, IV_3, IV_4, IV_5, IV_6, IV_7
-    PUSH 0x01010040 // params: key = 00, digest_size = 64 = 0x40
-    %blake2b_iv_i(0)
-    XOR
-    // stack: IV_0 ^ params, IV_1, IV_2, IV_3, IV_4, IV_5, IV_6, IV_7
-%endmacro
-
 // Address where the working version of the hash value is stored.
 %macro blake2b_hash_value_addr
     PUSH 0
diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
index 11e5389b..cd1f6a80 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/compression.asm
@@ -2,7 +2,9 @@ global blake2b_compression:
     // stack: retdest
     PUSH 0
     // stack: cur_block = 0, retdest
-    %blake2b_initial_hash_value
+    PUSH compression_loop
+    // stack: compression_loop, cur_block, retdest
+    %jump(blake2b_initial_hash_value)
 compression_loop:
     // stack: h_0, ..., h_7, cur_block, retdest
     
diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm b/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm
index 174afd33..6236fbef 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm
@@ -33,30 +33,63 @@ global blake2b_iv_const:
     BYTES 91, 224, 205, 25
     BYTES 19, 126, 33, 121
 
-%macro blake2b_iv
-    // stack: i, ...
+global blake2b_iv:
+    // stack: i, retdest
     PUSH blake2b_iv_const
-    // stack: blake2b_iv_const, i, ...
+    // stack: blake2b_iv_const, i, retdest
     SWAP1
-    // stack: i, blake2b_iv_const, ...
+    // stack: i, blake2b_iv_const, retdest
     %mul_const(8)
     ADD
-    // stack: blake2b_iv_const + 2 * i, ...
+    // stack: blake2b_iv_const + 2 * i, retdest
     DUP1
-    // stack: blake2b_iv_const + 2 * i, blake2b_iv_const + 2 * i, ...
+    // stack: blake2b_iv_const + 2 * i, blake2b_iv_const + 2 * i, retdest
     %add_const(4)
-    // stack: blake2b_iv_const + 2 * i + 1, blake2b_iv_const + 2 * i, ...
+    // stack: blake2b_iv_const + 2 * i + 1, blake2b_iv_const + 2 * i, retdest
     %mload_kernel_code_u32
     SWAP1
     %mload_kernel_code_u32
-    // stack: IV_i[32:], IV_i[:32], ...
+    // stack: IV_i[32:], IV_i[:32], retdest
     %shl_const(32)
-    // stack: IV_i[32:] << 32, IV_i[:32], ...
+    // stack: IV_i[32:] << 32, IV_i[:32], retdest
     OR
-    // stack: IV_i, ...
+    // stack: IV_i, retdest
+    SWAP1
+    JUMP
+
+%macro blake2b_iv
+    %stack (i) -> (i, %%after)
+    %jump(blake2b_iv)
+%%after:
 %endmacro
 
-%macro blake2b_iv_i(i)
-    PUSH $i
-    %blake2b_iv
-%endmacro
+// Load the initial hash value (the IV, but with params XOR'd into the first word).
+global blake2b_initial_hash_value:
+    // stack: retdest
+    PUSH 8
+    // stack: i=8, retdest
+blake2b_initial_hash_loop:
+    // stack: i, IV_i, ..., IV_7, retdest
+    %decrement
+    // stack: i-1, IV_i, ..., IV_7, retdest
+    PUSH blake2b_initial_hash_return
+    // stack: blake2b_initial_hash_return, i-1, IV_i, ..., IV_7, retdest
+    DUP2
+    // stack: i-1, blake2b_initial_hash_return, i-1, IV_i, ..., IV_7, retdest
+    %jump(blake2b_iv)
+blake2b_initial_hash_return:
+    // stack: IV_(i-1), i-1, IV_i, ..., IV_7, retdest
+    SWAP1
+    // stack: i-1, IV_(i-1), IV_i, ..., IV_7, retdest
+    DUP1
+    // stack: i-1, i-1, IV_(i-1), ..., IV_7, retdest
+    %jumpi(blake2b_initial_hash_loop)
+    // stack: i-1=0, IV_0, ..., IV_7, retdest
+    POP
+    // stack: IV_0, ..., IV_7, retdest
+    PUSH 0x01010040 // params: key = 00, digest_size = 64 = 0x40
+    XOR
+    // stack: IV_0 ^ params, IV_1, IV_2, IV_3, IV_4, IV_5, IV_6, IV_7, retdest
+    %stack(iv: 8, ret) -> (ret, iv)
+    JUMP
+
diff --git a/evm/src/cpu/kernel/tests/hash.rs b/evm/src/cpu/kernel/tests/hash.rs
index cc2c96eb..bc73ecd5 100644
--- a/evm/src/cpu/kernel/tests/hash.rs
+++ b/evm/src/cpu/kernel/tests/hash.rs
@@ -79,8 +79,6 @@ fn prepare_test<T>(
     // Run the interpeter
     let result = run_interpreter_with_memory(interpreter_setup).unwrap();
 
-    dbg!(result.stack().to_vec());
-
     Ok((expected, result.stack().to_vec()))
 }
 

From fda2e19054e01c344e214d4ac6c580e97b3fe597 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Wed, 8 Mar 2023 12:13:38 -0800
Subject: [PATCH 09/12] restored blake2b_g_function and call_blake2b_g_function
 macros

---
 .../kernel/asm/hash/blake2b/g_functions.asm   | 183 +++++++-----------
 1 file changed, 68 insertions(+), 115 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
index ff3ee43a..782393de 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
@@ -1,89 +1,89 @@
-blake2b_g_function:
+%macro blake2b_g_function
     // Function to mix two input words, x and y, into the four words indexed by a, b, c, d (which
     // are in the range 0..16) in the internal state.
     // The internal state is stored in memory starting at the address start.
-    // stack: a, b, c, d, x, y, start, retdest
+    // stack: a, b, c, d, x, y, start
     %stack (indices: 4) -> (indices, indices)
-    // stack: a, b, c, d, a, b, c, d, x, y, start, retdest
+    // stack: a, b, c, d, a, b, c, d, x, y, start
     DUP11
-    // stack: start, a, b, c, d, a, b, c, d, x, y, start, retdest
+    // stack: start, a, b, c, d, a, b, c, d, x, y, start
     %stack (start, a, b, c, d) -> (d, start, c, start, b, start, a, start)
-    // stack: d, start, c, start, b, start, a, start, a, b, c, d, x, y, start, retdest
+    // stack: d, start, c, start, b, start, a, start, a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
-    // stack: v[d], c, start, b, start, a, start, a, b, c, d, x, y, start, retdest
+    // stack: v[d], c, start, b, start, a, start, a, b, c, d, x, y, start
     %stack (vd, remaining: 6) -> (remaining, vd)
-    // stack: c, start, b, start, a, start, v[d], a, b, c, d, x, y, start, retdest
+    // stack: c, start, b, start, a, start, v[d], a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
     %stack (vc, remaining: 4) -> (remaining, vc)
-    // stack: b, start, a, start, v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: b, start, a, start, v[c], v[d], a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
-    // stack: v[b], a, start, v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: v[b], a, start, v[c], v[d], a, b, c, d, x, y, start
     %stack (vb, remaining: 2) -> (remaining, vb)
-    // stack: a, start, v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: a, start, v[b], v[c], v[d], a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
-    // stack: v[a], v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
     DUP2
-    // stack: v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
     DUP10
-    // stack: x, v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: x, v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
     ADD
     ADD
     %as_u64
-    // stack: v[a]' = (v[a] + v[b] + x) % 2^64, v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: v[a]' = (v[a] + v[b] + x) % 2^64, v[b], v[c], v[d], a, b, c, d, x, y, start
     %stack (a, b, c, d) -> (a, d, a, b, c, d)
-    // stack: v[a]', v[d], v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: v[a]', v[d], v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start
     XOR
     %rotr_64(32)
-    // stack: v[d]' = (v[d] ^ v[a]') >>> 32, v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start, retdest
+    // stack: v[d]' = (v[d] ^ v[a]') >>> 32, v[a]', v[b], v[c], v[d], a, b, c, d, x, y, start
     %stack (top: 4, vd) -> (top)
-    // stack: v[d]', v[a]', v[b], v[c], a, b, c, d, x, y, start, retdest
+    // stack: v[d]', v[a]', v[b], v[c], a, b, c, d, x, y, start
     %stack (d, a, b, c) -> (c, d, a, b, d)
-    // stack: v[c], v[d]', v[a]', v[b], v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[c], v[d]', v[a]', v[b], v[d]', a, b, c, d, x, y, start
     ADD
     %as_u64
-    // stack: v[c]' = (v[c] + v[d]') % 2^64, v[a]', v[b], v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[c]' = (v[c] + v[d]') % 2^64, v[a]', v[b], v[d]', a, b, c, d, x, y, start
     %stack (c, a, b, d) -> (b, c, a, c, d)
-    // stack: v[b], v[c]', v[a]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[b], v[c]', v[a]', v[c]', v[d]', a, b, c, d, x, y, start
     XOR
     %rotr_64(24)
-    // stack: v[b]' = (v[b] ^ v[c]') >>> 24, v[a]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[b]' = (v[b] ^ v[c]') >>> 24, v[a]', v[c]', v[d]', a, b, c, d, x, y, start
     SWAP1
-    // stack: v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start
     DUP2
-    // stack: v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start
     DUP11
-    // stack: y, v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
+    // stack: y, v[b]', v[a]', v[b]', v[c]', v[d]', a, b, c, d, x, y, start
     ADD
     ADD
     %as_u64
-    // stack: v[a]'' = (v[a]' + v[b]' + y) % 2^64, v[b]', v[c]', v[d]', a, b, c, d, x, y, start, retdest
+    // stack: v[a]'' = (v[a]' + v[b]' + y) % 2^64, v[b]', v[c]', v[d]', a, b, c, d, x, y, start
     SWAP3
-    // stack: v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start
     DUP4
-    // stack: v[a]'', v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[a]'', v[d]', v[b]', v[c]', v[a]'', a, b, c, d, x, y, start
     XOR
     %rotr_64(16)
-    // stack: v[d]'' = (v[a]'' ^ v[d]') >>> 8, v[b]', v[c]', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[d]'' = (v[a]'' ^ v[d]') >>> 8, v[b]', v[c]', v[a]'', a, b, c, d, x, y, start
     SWAP2
-    // stack: v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
     DUP3
-    // stack: v[d]'', v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[d]'', v[c]', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
     ADD
     %as_u64
-    // stack: v[c]'' = (v[c]' + v[d]'') % 2^64, v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[c]'' = (v[c]' + v[d]'') % 2^64, v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
     DUP1
-    // stack: v[c]'', v[c]'', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[c]'', v[c]'', v[b]', v[d]'', v[a]'', a, b, c, d, x, y, start
     SWAP2
-    // stack: v[b]', v[c]'', v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[b]', v[c]'', v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start
     XOR
     %rotr_64(63)
-    // stack: v[b]'' = (v[b]' ^ v[c]'') >>> 7, v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start, retdest
+    // stack: v[b]'' = (v[b]' ^ v[c]'') >>> 7, v[c]'', v[d]'', v[a]'', a, b, c, d, x, y, start
     %stack (vb, vc, vd, va, a, b, c, d, x, y, start) -> (start, a, va, start, b, vb, start, c, vc, start, d, vd)
-    // stack: start, a, v[a]'', start, b, v[b]'', start, c, v[c]'', start, d, v[d]'', retdest
+    // stack: start, a, v[a]'', start, b, v[b]'', start, c, v[c]'', start, d, v[d]''
     ADD
     %mstore_kernel_general
     ADD
@@ -92,100 +92,53 @@ blake2b_g_function:
     %mstore_kernel_general
     ADD
     %mstore_kernel_general
-    // stack: retdest
-    JUMP
+%endmacro
 
-call_blake2b_g_function:
-    // stack: a, b, c, d, x_idx, y_idx, round, start, retdest
-    DUP6
-    // stack: y_idx, a, b, c, d, x_idx, y_idx, round, start, retdest
-    DUP8
-    // stack: round, y_idx, a, b, c, d, x_idx, y_idx, round, start, retdest
+%macro call_blake2b_g_function(a, b, c, d, x_idx, y_idx)
+    // stack: round, start
+    PUSH $y_idx
+    DUP2
+    // stack: round, y_idx, round, start
     %blake2b_permutation
-    // stack: s[y_idx], a, b, c, d, x_idx, y_idx, round, start, retdest
+    // stack: s[y_idx], round, start
     %blake2b_message_addr
     ADD
     %mload_kernel_general
-    // stack: m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
-    DUP6
-    // stack: x_idx, m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
-    DUP9
-    // stack: round, x_idx, m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
+    // stack: m[s[y_idx]], round, start
+    PUSH $x_idx
+    DUP3
+    // stack: round, 2, m[s[y_idx]], round, start
     %blake2b_permutation
-    // stack: s[x_idx], m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
+    // stack: s[x_idx], m[s[y_idx]], round, start
     %blake2b_message_addr
     ADD
     %mload_kernel_general
-    // stack: m[s[x_idx]], m[s[y_idx]], a, b, c, d, x_idx, y_idx, round, start, retdest
-    %stack (mm: 2, abcd: 4, xy: 2, r, s) -> (abcd, mm, s)
-    // stack: a, b, c, d, m[s[x_idx]], m[s[y_idx]], start, retdest
-    %jump(blake2b_g_function)
+    // stack: m[s[x_idx]], m[s[y_idx]], round, start
+    %stack (ss: 2, r, s) -> (ss, s, r, s)
+    // stack: m[s[x_idx]], m[s[y_idx]], start, round, start
+    PUSH $d
+    PUSH $c
+    PUSH $b
+    PUSH $a
+    // stack: a, b, c, d, m[s[x_idx]], m[s[y_idx]], start, round, start
+    %blake2b_g_function
+    // stack: round, start
+%endmacro
 
-global run_g_function_round:
+run_g_function_round:
     // stack: round, start, retdest
-    PUSH g_function_return_1
-    // stack: g_function_return_1, round, start, retdest
-    %stack (ret, r, s) -> (0, 4, 8, 12, 0, 1, r, s, ret, r, s)
-    // stack: a=0, b=4, c=8, d=12, x_idx=0, y_idx=1, round, start, g_function_return_1, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_1:
-    // stack: round, start, retdest
-    PUSH g_function_return_2
-    // stack: g_function_return_2, round, start, retdest
-    %stack (ret, r, s) -> (1, 5, 9, 13, 2, 3, r, s, ret, r, s)
-    // stack: a=1, b=5, c=9, d=13, x_idx=2, y_idx=3, round, start, g_function_return_2, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_2:
-    // stack: round, start, retdest
-    PUSH g_function_return_3
-    // stack: g_function_return_3, round, start, retdest
-    %stack (ret, r, s) -> (2, 6, 10, 14, 4, 5, r, s, ret, r, s)
-    // stack: a=2, b=6, c=10, d=14, x_idx=4, y_idx=5, round, start, g_function_return_3, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_3:
-    // stack: round, start, retdest
-    PUSH g_function_return_4
-    // stack: g_function_return_4, round, start, retdest
-    %stack (ret, r, s) -> (3, 7, 11, 15, 6, 7, r, s, ret, r, s)
-    // stack: a=3, b=7, c=11, d=15, x_idx=6, y_idx=7, round, start, g_function_return_4, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_4:
-    // stack: round, start, retdest
-    PUSH g_function_return_5
-    // stack: g_function_return_5, round, start, retdest
-    %stack (ret, r, s) -> (0, 5, 10, 15, 8, 9, r, s, ret, r, s)
-    // stack: a=0, b=5, c=10, d=15, x_idx=8, y_idx=9, round, start, g_function_return_5, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_5:
-    // stack: round, start, retdest
-    PUSH g_function_return_6
-    // stack: g_function_return_6, round, start, retdest
-    %stack (ret, r, s) -> (1, 6, 11, 12, 10, 11, r, s, ret, r, s)
-    // stack: a=1, b=6, c=11, d=12, x_idx=10, y_idx=11, round, start, g_function_return_6, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_6:
-    // stack: round, start, retdest
-    PUSH g_function_return_7
-    // stack: g_function_return_7, round, start, retdest
-    %stack (ret, r, s) -> (2, 7, 8, 13, 12, 13, r, s, ret, r, s)
-    // stack: a=2, b=7, c=8, d=13, x_idx=12, y_idx=13, round, start, g_function_return_7, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_7:
-    // stack: round, start, retdest
-    PUSH g_function_return_8
-    // stack: g_function_return_8, round, start, retdest
-    %stack (ret, r, s) -> (3, 4, 9, 14, 14, 15, r, s, ret, r, s)
-    // stack: a=3, b=4, c=9, d=14, x_idx=14, y_idx=15, round, start, g_function_return_8, round, start, retdest
-    %jump(call_blake2b_g_function)
-g_function_return_8:
-    // stack: round, start, retdest
-    SWAP1
-    // stack: start, round, retdest
-    SWAP2
+    %call_blake2b_g_function(0, 4, 8, 12, 0, 1)
+    %call_blake2b_g_function(1, 5, 9, 13, 2, 3)
+    %call_blake2b_g_function(2, 6, 10, 14, 4, 5)
+    %call_blake2b_g_function(3, 7, 11, 15, 6, 7)
+    %call_blake2b_g_function(0, 5, 10, 15, 8, 9)
+    %call_blake2b_g_function(1, 6, 11, 12, 10, 11)
+    %call_blake2b_g_function(2, 7, 8, 13, 12, 13)
+    %call_blake2b_g_function(3, 4, 9, 14, 14, 15)
+    %stack (r, s, ret) -> (ret, r, s)
     // stack: retdest, round, start
     JUMP
 
-
 global run_12_rounds_g_function:
     // stack: start, retdest
     PUSH 0

From 33ccf8987ea4bd346f2619f35ac1f71b05276781 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Wed, 8 Mar 2023 12:34:50 -0800
Subject: [PATCH 10/12] small optimizations

---
 .../kernel/asm/hash/blake2b/g_functions.asm   | 32 ++++++++++++-------
 .../kernel/asm/hash/blake2b/permutations.asm  | 25 +++++++++++----
 2 files changed, 39 insertions(+), 18 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
index 782393de..f2d3b1d2 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/g_functions.asm
@@ -3,28 +3,38 @@
     // are in the range 0..16) in the internal state.
     // The internal state is stored in memory starting at the address start.
     // stack: a, b, c, d, x, y, start
-    %stack (indices: 4) -> (indices, indices)
+    DUP4
+    DUP4
+    DUP4
+    DUP4
     // stack: a, b, c, d, a, b, c, d, x, y, start
     DUP11
     // stack: start, a, b, c, d, a, b, c, d, x, y, start
-    %stack (start, a, b, c, d) -> (d, start, c, start, b, start, a, start)
-    // stack: d, start, c, start, b, start, a, start, a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
-    // stack: v[d], c, start, b, start, a, start, a, b, c, d, x, y, start
-    %stack (vd, remaining: 6) -> (remaining, vd)
-    // stack: c, start, b, start, a, start, v[d], a, b, c, d, x, y, start
+    // stack: v[a], b, c, d, a, b, c, d, x, y, start
+    SWAP1
+    // stack: b, v[a], c, d, a, b, c, d, x, y, start
+    DUP11
+    // stack: start, b, v[a], c, d, a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
-    %stack (vc, remaining: 4) -> (remaining, vc)
-    // stack: b, start, a, start, v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[b], v[a], c, d, a, b, c, d, x, y, start
+    SWAP2
+    // stack: c, v[a], v[b], d, a, b, c, d, x, y, start
+    DUP11
+    // stack: start, c, v[a], v[b], d, a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
-    // stack: v[b], a, start, v[c], v[d], a, b, c, d, x, y, start
-    %stack (vb, remaining: 2) -> (remaining, vb)
-    // stack: a, start, v[b], v[c], v[d], a, b, c, d, x, y, start
+    // stack: v[c], v[a], v[b], d, a, b, c, d, x, y, start
+    SWAP3
+    // stack: d, v[a], v[b], v[c], a, b, c, d, x, y, start
+    DUP11
+    // stack: start, d, v[a], v[b], v[c], a, b, c, d, x, y, start
     ADD
     %mload_kernel_general
+    // stack: v[d], v[a], v[b], v[c], a, b, c, d, x, y, start
+    %stack (vd, vs: 3) -> (vs, vd)
     // stack: v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
     DUP2
     // stack: v[b], v[a], v[b], v[c], v[d], a, b, c, d, x, y, start
diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/permutations.asm b/evm/src/cpu/kernel/asm/hash/blake2b/permutations.asm
index 5277e611..1ef455f1 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/permutations.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/permutations.asm
@@ -58,17 +58,28 @@ global permutation_9_constants:
     BYTES 15, 11, 9, 14
     BYTES 3, 12, 13, 0
 
-%macro blake2b_permutation
-    // stack: round, i
+global blake2b_permutation:
+    // stack: i, round, retdest
     PUSH permutation_0_constants
-    // stack: permutation_0_constants, round, i
-    SWAP1
-    // stack: round, permutation_1_constants, i
+    // stack: permutation_0_constants, i, round, retdest
+    SWAP2
+    // stack: round, i, permutation_0_constants, retdest
     %mod_const(10)
-    // stack: round % 10, permutation_1_constants, i
+    // stack: round % 10, i, permutation_0_constants, retdest
     %mul_const(16)
     ADD
-    // stack: permutation_(round)_constants, i
     ADD
     %mload_kernel_code
+    // stack: permutation_(round%10)_constants[i], retdest
+    SWAP1
+    JUMP
+
+%macro blake2b_permutation
+    // stack: round, i
+    PUSH %%after
+    // stack: %%after, round, i
+    SWAP2
+    // stack: i, round, %%after
+    %jump(blake2b_permutation)
+%%after:
 %endmacro

From 476a554aea93dc1f399cf87ea4900b6fb780cf42 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 14 Mar 2023 12:29:30 -0700
Subject: [PATCH 11/12] OR -> ADD

---
 .../kernel/asm/curve/secp256k1/ecrecover.asm  | 14 ++++------
 evm/src/cpu/kernel/asm/hash/blake2b/iv.asm    |  2 +-
 .../cpu/kernel/asm/hash/ripemd/functions.asm  |  2 +-
 evm/src/cpu/kernel/asm/hash/ripemd/main.asm   |  8 +++---
 .../cpu/kernel/asm/hash/sha2/compression.asm  | 14 +++++-----
 evm/src/cpu/kernel/asm/memory/core.asm        | 28 +++++++++----------
 evm/src/cpu/kernel/asm/util/basic_macros.asm  | 10 +++----
 7 files changed, 37 insertions(+), 41 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm b/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm
index dd2b86f6..c243a748 100644
--- a/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm
+++ b/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm
@@ -134,17 +134,13 @@ pubkey_to_addr:
     // stack: hash, v, r, s, retdest
     DUP2
     // stack: v, hash, v, r, s, retdest
-    PUSH 27
-    // stack: 27, v, hash, v, r, s, retdest
-    EQ
+    %eq_consts(27)
     // stack: v==27, hash, v, r, s, retdest
     DUP3
     // stack: v, v==27, hash, v, r, s, retdest
-    PUSH 28
-    // stack: 28, v, v==27, hash, v, r, s, retdest
-    EQ
+    %eq_consts(28)
     // stack: v==28, v==27, hash, v, r, s, retdest
-    OR
+    ADD // OR
     // stack: (v==28 || v==27), hash, v, r, s, retdest
     DUP5
     // stack: s, (v==28 || v==27), hash, v, r, s, retdest
@@ -154,7 +150,7 @@ pubkey_to_addr:
     // stack: r, (s >= N || s==0), (v==28 || v==27), hash, v, r, s, retdest
     %secp_is_out_of_bounds
     // stack: (r >= N || r==0), (s >= N || s==0), (v==28 || v==27), hash, v, r, s, retdest
-    OR
+    ADD // OR
     // stack: (r >= N || r==0 || s >= N || s==0), (v==28 || v==27), hash, v, r, s, retdest
     ISZERO
     // stack: (r < N & r!=0 & s < N & s!=0), (v==28 || v==27), hash, v, r, s, retdest
@@ -178,7 +174,7 @@ pubkey_to_addr:
     // stack: x < N, x==0
     ISZERO
     // stack: x >= N, x==0
-    OR
+    ADD // OR
     // stack: x >= N || x==0
 %endmacro
 
diff --git a/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm b/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm
index 174afd33..94e9ba27 100644
--- a/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm
+++ b/evm/src/cpu/kernel/asm/hash/blake2b/iv.asm
@@ -52,7 +52,7 @@ global blake2b_iv_const:
     // stack: IV_i[32:], IV_i[:32], ...
     %shl_const(32)
     // stack: IV_i[32:] << 32, IV_i[:32], ...
-    OR
+    ADD // OR
     // stack: IV_i, ...
 %endmacro
 
diff --git a/evm/src/cpu/kernel/asm/hash/ripemd/functions.asm b/evm/src/cpu/kernel/asm/hash/ripemd/functions.asm
index ac111215..de2fdcf6 100644
--- a/evm/src/cpu/kernel/asm/hash/ripemd/functions.asm
+++ b/evm/src/cpu/kernel/asm/hash/ripemd/functions.asm
@@ -18,7 +18,7 @@ global rol:
     // stack:         x << n, x >> (32-n), retdest
     %as_u32
     // stack:    u32(x << n), x >> (32-n), retdest
-    OR
+    ADD // OR
     // stack: u32(x << n) | (x >> (32-n)), retdest
     SWAP1  
     JUMP
diff --git a/evm/src/cpu/kernel/asm/hash/ripemd/main.asm b/evm/src/cpu/kernel/asm/hash/ripemd/main.asm
index 4aa6a6ad..0ad1e4e5 100644
--- a/evm/src/cpu/kernel/asm/hash/ripemd/main.asm
+++ b/evm/src/cpu/kernel/asm/hash/ripemd/main.asm
@@ -81,21 +81,21 @@ process:
     SWAP1
     %reverse_bytes_u32
     %shl_const(96)
-    OR
+    ADD // OR
     // stack: b' a', c, d, e, VARS
     SWAP1
     %reverse_bytes_u32
     %shl_const(64)
-    OR
+    ADD // OR
     // stack: c' b' a', d, e, VARS
     SWAP1
     %reverse_bytes_u32
     %shl_const(32)
-    OR 
+    ADD // OR
     // stack: d' c' b' a', e, VARS
     SWAP1
     %reverse_bytes_u32
-    OR 
+    ADD // OR
     // stack: e' d' c' b' a', VARS
     %stack (result, VARS: 3, retdest) -> (retdest, result)
     // stack: 0xdeadbeef, result
diff --git a/evm/src/cpu/kernel/asm/hash/sha2/compression.asm b/evm/src/cpu/kernel/asm/hash/sha2/compression.asm
index 8850c1c8..8c219ebb 100644
--- a/evm/src/cpu/kernel/asm/hash/sha2/compression.asm
+++ b/evm/src/cpu/kernel/asm/hash/sha2/compression.asm
@@ -263,19 +263,19 @@ compression_end:
     POP
     // stack: a[0]+a[64], b[0]+b[64], c[0]+c[64], d[0]+d[64], e[0]+e[64], f[0]+f[64], g[0]+g[64], h[0]+h[64], scratch_space_addr, message_schedule_addr, i, retdest
     %shl_const(32)
-    OR
+    ADD // OR
     %shl_const(32)
-    OR
+    ADD // OR
     %shl_const(32)
-    OR
+    ADD // OR
     %shl_const(32)
-    OR
+    ADD // OR
     %shl_const(32)
-    OR
+    ADD // OR
     %shl_const(32)
-    OR
+    ADD // OR
     %shl_const(32)
-    OR
+    ADD // OR
     // stack: concat(a[0]+a[64], b[0]+b[64], c[0]+c[64], d[0]+d[64], e[0]+e[64], f[0]+f[64], g[0]+g[64], h[0]+h[64]), scratch_space_addr, message_schedule_addr, i, retdest
     SWAP3
     // stack: i, scratch_space_addr, message_schedule_addr, concat(a[0]+a[64], b[0]+b[64], c[0]+c[64], d[0]+d[64], e[0]+e[64], f[0]+f[64], g[0]+g[64], h[0]+h[64]), retdest
diff --git a/evm/src/cpu/kernel/asm/memory/core.asm b/evm/src/cpu/kernel/asm/memory/core.asm
index a979f930..8f59a128 100644
--- a/evm/src/cpu/kernel/asm/memory/core.asm
+++ b/evm/src/cpu/kernel/asm/memory/core.asm
@@ -79,21 +79,21 @@
     DUP2
     %increment
     %mload_kernel($segment)
-    OR
+    ADD // OR
     // stack: (c_3 << 8) | c_2, offset
     %shl_const(8)
     // stack: ((c_3 << 8) | c_2) << 8, offset
     DUP2
     %add_const(2)
     %mload_kernel($segment)
-    OR
+    ADD // OR
     // stack: (((c_3 << 8) | c_2) << 8) | c_1, offset
     %shl_const(8)
     // stack: ((((c_3 << 8) | c_2) << 8) | c_1) << 8, offset
     SWAP1
     %add_const(3)
     %mload_kernel($segment)
-    OR
+    ADD // OR
     // stack: (((((c_3 << 8) | c_2) << 8) | c_1) << 8) | c_0
 %endmacro
 
@@ -107,19 +107,19 @@
     %increment
     %mload_kernel($segment)
     %shl_const(8)
-    OR
+    ADD
     // stack: c0 | (c1 << 8)             , offset
     DUP2
     %add_const(2)
     %mload_kernel($segment)
     %shl_const(16)
-    OR
+    ADD
     // stack: c0 | (c1 << 8) | (c2 << 16), offset
     SWAP1
     %add_const(3)
     %mload_kernel($segment)
     %shl_const(24)
-    OR
+    ADD // OR
     // stack: c0 | (c1 << 8) | (c2 << 16) | (c3 << 24)
 %endmacro
 
@@ -137,7 +137,7 @@
     // stack: hi, lo
     %shl_const(32)
     // stack: hi << 32, lo
-    OR
+    ADD // OR
     // stack: (hi << 32) | lo
 %endmacro
 
@@ -152,49 +152,49 @@
     DUP2
     %add_const(4)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 32) | c_6, offset
     %shl_const(32)
     // stack: ((c_7 << 32) | c_6) << 32, offset
     DUP2
     %add_const(8)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 64) | (c_6 << 32) | c_5, offset
     %shl_const(32)
     // stack: ((c_7 << 64) | (c_6 << 32) | c_5) << 32, offset
     DUP2
     %add_const(12)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 96) | (c_6 << 64) | (c_5 << 32) | c_4, offset
     %shl_const(32)
     // stack: ((c_7 << 96) | (c_6 << 64) | (c_5 << 32) | c_4) << 32, offset
     DUP2
     %add_const(16)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 128) | (c_6 << 96) | (c_5 << 64) | (c_4 << 32) | c_3, offset
     %shl_const(32)
     // stack: ((c_7 << 128) | (c_6 << 96) | (c_5 << 64) | (c_4 << 32) | c_3) << 32, offset
     DUP2
     %add_const(20)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 160) | (c_6 << 128) | (c_5 << 96) | (c_4 << 64) | (c_3 << 32) | c_2, offset
     %shl_const(32)
     // stack: ((c_7 << 160) | (c_6 << 128) | (c_5 << 96) | (c_4 << 64) | (c_3 << 32) | c_2) << 32, offset
     DUP2
     %add_const(24)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 192) | (c_6 << 160) | (c_5 << 128) | (c_4 << 96) | (c_3 << 64) | (c_2 << 32) | c_1, offset
     %shl_const(32)
     // stack: ((c_7 << 192) | (c_6 << 160) | (c_5 << 128) | (c_4 << 96) | (c_3 << 64) | (c_2 << 32) | c_1) << 32, offset
     DUP2
     %add_const(28)
     %mload_kernel_u32($segment)
-    OR
+    ADD // OR
     // stack: (c_7 << 224) | (c_6 << 192) | (c_5 << 160) | (c_4 << 128) | (c_3 << 96) | (c_2 << 64) | (c_1 << 32) | c_0, offset
     SWAP1
     POP
diff --git a/evm/src/cpu/kernel/asm/util/basic_macros.asm b/evm/src/cpu/kernel/asm/util/basic_macros.asm
index 5ee4a1e7..52556d3f 100644
--- a/evm/src/cpu/kernel/asm/util/basic_macros.asm
+++ b/evm/src/cpu/kernel/asm/util/basic_macros.asm
@@ -309,9 +309,9 @@
     BYTE
     %shl_const(24)
     // stack:       d000, b0, a, c00
-    OR 
-    OR
-    OR
+    ADD // OR
+    ADD // OR
+    ADD // OR
     // stack:                   dcba
 %endmacro
 
@@ -332,7 +332,7 @@
     %reverse_bytes_u32
     // stack: word_lo_inverted, word_hi_inverted
     %shl_const(32)
-    OR
+    ADD // OR
     // stack: word_inverted
 %endmacro
 
@@ -341,7 +341,7 @@
     // stack: a, b, c, d
     %rep 3
         %shl_const(64)
-        OR
+        ADD // OR
     %endrep
     // stack: a || b || c || d
 %endmacro

From 676a483c06c4ec004def5f9ed45eacddb7d34310 Mon Sep 17 00:00:00 2001
From: Nicholas Ward <npward@berkeley.edu>
Date: Tue, 14 Mar 2023 12:34:51 -0700
Subject: [PATCH 12/12] fix

---
 evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm b/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm
index c243a748..11ec27c8 100644
--- a/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm
+++ b/evm/src/cpu/kernel/asm/curve/secp256k1/ecrecover.asm
@@ -134,11 +134,11 @@ pubkey_to_addr:
     // stack: hash, v, r, s, retdest
     DUP2
     // stack: v, hash, v, r, s, retdest
-    %eq_consts(27)
+    %eq_const(27)
     // stack: v==27, hash, v, r, s, retdest
     DUP3
     // stack: v, v==27, hash, v, r, s, retdest
-    %eq_consts(28)
+    %eq_const(28)
     // stack: v==28, v==27, hash, v, r, s, retdest
     ADD // OR
     // stack: (v==28 || v==27), hash, v, r, s, retdest