replace switch_s_plus_2a with branch macro #130
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Uses a macro with destinations as args to create a high bytes and a low bytes
table for each destination offset by -1, then uses the stack and rts for the
actual branching.
Macro expands to 11-12 bytes per branch with nothing centralized, compared to
the previous 5-6 bytes per branch to a centralized routine.
Reduces cycle count from 54-57 cycles per branch to 23-26, with the variance
from possible page boundary crossing and if the destination is in zero page.
Normal game logic involves 16 branches per frame.
Supports up to 16 destinations but can be easily extended, the most in use is
the playstate branch with 12 destinations.
Previous:
lda dest ; 3/4
jsr switch_s_plus_2a ; 6
switch_s_plus_2a:
asl a ; 2
tay ; 2
iny ; 2
pla ; 4
sta switchTmp1 ; 3
pla ; 4
sta switchTmp2 ; 3
lda (switchTmp1),y ; 5/6
tax ; 2
iny ; 2
lda (switchTmp1),y ; 5/6
sta switchTmp2 ; 3
stx switchTmp1 ; 3
jmp (switchTmp1) ; 5
; 54-57
New:
ldx dest ; 3/4
lda hiBytes,x ; 4/5
pha ; 3
lda loBytes,x ; 4/5
pha ; 3
rts ; 6
; 23-26
|
Forgot to mention, if the macro is modified to be padded with at least 4 nops, the test will also pass. |
Throws warning when single destination is defined but optimizes to jmp
instruction anyway.
For 2 to 4 destinations, uses a decrementing x register, branch
instructions and jmp instructions to save a minimum of 7 cycles, up to 18.
Rom space is saved for 1-3 destinations. In the case of 4 destinations,
it uses an additional 2 bytes compared to the rts method.
How each option expands:
1 destination: 3 cycles
jmp a1
2 destinations: maximum 11 cycles to a0
ldx dest
beq addr0
jmp a1
addr0:
jmp a0
3 destinations: maximum 15 cycles to a1
ldx dest
beq addr0
dex
beq addr1
jmp a2
addr1:
jmp a1
addr0:
jmp a0
4 destinations: maximum 19 cycles to a2
ldx dest
beq addr0
dex
beq addr1
dex
beq addr2
jmp a3
addr2:
jmp a2
addr1:
jmp a1
addr0:
jmp a0
5+ destinations: 23-26 cycles
ldx dest
lda hyBytes,x
pha
lda loBytes,x
pha
rts
hiBytes:
.byte >a0,>a1,>a2,>a3,>a4
hiBytes:
.byte <a0,<a1,<a2,<a3,<a4
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
See commit notes for details.
Interestingly this change causes this test to panic:
TetrisGYM/tests/src/nmi.rs
Lines 59 to 65 in a222aa4
The first anomaly is that the the
nmi + 25points to line 15 here, not past oam staging as described.TetrisGYM/src/nmi/nmi.asm
Lines 11 to 17 in a222aa4
The second anomaly is that
sleepCounterappears to be affected by whereemu.run_until_vblank()stops. It appears as though the decrement happens in one frame but the following increment doesn't happen, causing subsequent frames to have a decrementing sleep counter. Ifrun_until_vblankis replaced with a custom function that stops at scanline 243 instead, the test will pass.