-
Notifications
You must be signed in to change notification settings - Fork 285
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
EIP-2929: Gas cost increases for state access opcodes #289
Conversation
lib/evmone/instructions_calls.cpp
Outdated
@@ -20,6 +20,13 @@ evmc_status_code call(ExecutionState& state) noexcept | |||
|
|||
state.stack.push(0); // Assume failure. | |||
|
|||
if (state.rev >= EVMC_BERLIN && state.host.access_account(dst) == EVMC_COLD_ACCESS) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Minimal improvement here would be to merge access_account()
and account_exists()
. I believe the account can have 3 statuses then: hot, cold, non-existing.
BTW, can a non-existing account become hot?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
SELFDESTRUCT
puts non-existing beneficiary into a warm set, the same I thihk with calls, so warm/cold and existing/non-existing seem to be orthogonal.
One could also think that account_exists
puts non-existing account into a warm set, it's used only in SELFDESTRUCT
and calls which do it, and it would reflect reality of updating non-existence caches in clients.
return EVMC_OUT_OF_GAS; | ||
} | ||
|
||
x = intx::be::load<uint256>(state.host.get_balance(addr)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Here and with other methods accessing information about accounts the get_balance()
can be extended to return both account status and the balance. The gas check should be done after the call, although I'm sure @holiman would be happy with such implementation. But my reasoning that even "access account" makes it hot so it can also return the balance in the same time.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That was my thinking at first with sload
, to return both the value and the access status. But then I realised that it complicates the API. I believe that having a separate method specifically to record access in the sense of EIP-2929 leads to a simpler API compared to reporting access status in various methods. The performance overhead shouldn't be too big, just an extra function call.
Codecov Report
@@ Coverage Diff @@
## master #289 +/- ##
==========================================
- Coverage 99.75% 99.07% -0.68%
==========================================
Files 24 24
Lines 3659 3694 +35
==========================================
+ Hits 3650 3660 +10
- Misses 9 34 +25
Flags with carried forward coverage won't be shown. Click here to find out more.
|
c53634e
to
915c089
Compare
Consensus tests are going to be updated later in #298. |
6f18e82
to
c63312d
Compare
|
||
#include <evmc/instructions.h> | ||
#include <array> | ||
|
||
namespace evmone::instr | ||
{ | ||
/// EIP-2929 constants (https://eips.ethereum.org/EIPS/eip-2929). | ||
inline constexpr auto cold_sload_cost = 2100; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not sure what we're going to do with these constants if they change in future forks... Mabye just prepend with berlin_
. But I guess it's ok to leave like this now.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
True. But it would be easy to move them to constexpr array. E.g. instr::constants[EVMC_BERLIN].cold_sload_cost
.
inline constexpr auto cold_sload_cost = 2100; | ||
inline constexpr auto cold_account_access_cost = 2600; | ||
inline constexpr auto warm_storage_read_cost = 100; | ||
inline constexpr auto additional_cold_account_access_cost = |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe introduce also additional_sload_cost
and use it in sload
for consistency.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I used the name, but only locally where it is used. Also explained the global definition.
|
||
const auto& r = host.recorded_account_accesses; | ||
ASSERT_EQ(r.size(), 24); | ||
EXPECT_EQ(r[0], msg.sender); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe could be better-looking with some fancy assert
EXPECT_THAT(r, ElementsAre(msg.sender, msg.destination, ...));
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is fine so I don't need to count these in my head. And I don't need to import Google Mock.
state.host.get_storage(state.msg->destination, intx::be::store<evmc::bytes32>(x))); | ||
const auto key = intx::be::store<evmc::bytes32>(x); | ||
|
||
if (state.rev >= EVMC_BERLIN && |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is there any measureable performance diff for these revision checks? Should we templetize based on version at some point?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There are not benchmarks for state access. I could add some, but these are may not be "portable" because depending on the pre-state initialization.
I don't think it is worth duplicating the code with templates for these. However, it may be good to priorities the most recent revision in the code.
lib/evmone/instructions.cpp
Outdated
@@ -316,7 +316,7 @@ constexpr std::array<instruction_exec_fn, 256> instruction_implementations = []( | |||
table[OP_SHL] = op<shl>; | |||
table[OP_SHR] = op<shr>; | |||
table[OP_SAR] = op<sar>; | |||
table[OP_EXTCODEHASH] = op<extcodehash>; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why is this removed here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh yeah, I was wandering the same @yperbasis.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, I simply removed duplicated code; see line 323.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah ok. I will sort these in separate PR then.
This uses temporary Silkworm version matching EVMC 8.0.0-alpha.1 ABI.
This was merged under #301. |
Keeping as draft until ethereum/evmc#571 is merged.