Skip to content

perf: optimize cmp zero and jump with test #462

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.

Sign up for GitHub

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

Merged
merged 1 commit into from
Mar 9, 2025
Merged

perf: optimize cmp zero and jump with test #462

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 21 additions & 21 deletions src/machine/asm/execute_x64.S
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -194,7 +194,7 @@
movq TEMP1, CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_LAST_READ_FRAME(MACHINE); \
movq CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_FRAMES_PTR(MACHINE), TEMP3; \
movzbl (TEMP3, TEMP1), TEMP2d; \
cmp $0, TEMP2d; \
test TEMP2d, TEMP2d; \
jne 1f; \
movb $1, (TEMP3, TEMP1); \
PREPCALL; \
Expand All @@ -209,7 +209,7 @@
shr $CKB_VM_ASM_MEMORY_FRAME_SHIFTS, TEMP1; \
movq CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_FRAMES_PTR(MACHINE), TEMP3; \
movzbl (TEMP3, TEMP1), TEMP2d; \
cmp $0, TEMP2d; \
test TEMP2d, TEMP2d; \
jne 2f; \
movb $1, (TEMP3, TEMP1);\
PREPCALL; \
Expand Down Expand Up @@ -288,7 +288,7 @@
shr $CKB_VM_ASM_MEMORY_FRAME_PAGE_SHIFTS, TEMP1; \
movq CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_FRAMES_PTR(MACHINE), TEMP3; \
movzbl (TEMP3, TEMP1), temp_regd; \
cmp $0, temp_regd; \
test temp_regd, temp_regd; \
jne 1f; \
movb $1, (TEMP3, TEMP1); \
PREPCALL; \
Expand Down Expand Up @@ -319,7 +319,7 @@
shr $CKB_VM_ASM_MEMORY_FRAME_PAGE_SHIFTS, TEMP1; \
movq CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_FRAMES_PTR(MACHINE), TEMP3; \
movzbl (TEMP3, TEMP1), temp_regd; \
cmp $0, temp_regd; \
test temp_regd, temp_regd; \
jne 2f; \
movb $1, (TEMP3, TEMP1); \
PREPCALL; \
Expand Down Expand Up @@ -443,7 +443,7 @@ ckb_vm_x64_execute:
cmp %rcx, %rdx
jne .exit_trace
mov CKB_VM_ASM_TRACE_OFFSET_LENGTH(TRACE), %edx
cmp $0, %rdx
test %rdx, %rdx
je .exit_trace
movq CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_CYCLES(MACHINE), %rax
addq CKB_VM_ASM_TRACE_OFFSET_CYCLES(TRACE), %rax
Expand Down Expand Up @@ -485,7 +485,7 @@ ckb_vm_x64_execute:
.prepare_trace:
movq CKB_VM_ASM_INVOKE_DATA_OFFSET_PAUSE(INVOKE_DATA), %rax
movzbl 0(%rax), %eax
cmp $0, %rax
test %rax, %rax
jnz .exit_pause
LOAD_PC(%rax, %eax, %rcx, %ecx, TEMP3d)
shr $2, %eax
Expand All @@ -497,7 +497,7 @@ ckb_vm_x64_execute:
cmp %rcx, %rdx
jne .exit_trace
mov CKB_VM_ASM_TRACE_OFFSET_LENGTH(TRACE), %edx
cmp $0, %rdx
test %rdx, %rdx
je .exit_trace
movq CKB_VM_ASM_ASM_CORE_MACHINE_OFFSET_CYCLES(MACHINE), %rax
addq CKB_VM_ASM_TRACE_OFFSET_CYCLES(TRACE), %rax
Expand Down Expand Up @@ -634,7 +634,7 @@ ckb_vm_x64_execute:
jne .div_branch1
jmp .div_branch3
.div_branch1:
cmp $0, RS2r
test RS2r, RS2r
jne .div_branch2
movq $UINT64_MAX, RS1
jmp .div_branch3
Expand All @@ -651,7 +651,7 @@ ckb_vm_x64_execute:
.CKB_VM_ASM_LABEL_OP_DIVU:
DECODE_R
movq REGISTER_ADDRESS(RS2r), RS2r
cmp $0, RS2r
test RS2r, RS2r
jne .divu_branch1
WRITE_RD_VALUE($UINT64_MAX, RS2r)
NEXT_INST
Expand All @@ -671,7 +671,7 @@ ckb_vm_x64_execute:
DECODE_R
movq REGISTER_ADDRESS(RS2r), RS2r
mov RS2rd, RS2rd
cmp $0, RS2r
test RS2r, RS2r
jne .divuw_branch1
WRITE_RD_VALUE($UINT64_MAX, RS2r)
NEXT_INST
Expand Down Expand Up @@ -703,7 +703,7 @@ ckb_vm_x64_execute:
jne .divw_branch1
jmp .divw_branch3
.divw_branch1:
cmp $0, RS2r
test RS2r, RS2r
jne .divw_branch2
movq $UINT64_MAX, RS1
jmp .divw_branch3
Expand Down Expand Up @@ -1022,7 +1022,7 @@ ckb_vm_x64_execute:
xorq RS1, RS1
jmp .rem_branch3
.rem_branch1:
cmp $0, RS2r
test RS2r, RS2r
jne .rem_branch2
jmp .rem_branch3
.rem_branch2:
Expand All @@ -1038,7 +1038,7 @@ ckb_vm_x64_execute:
.CKB_VM_ASM_LABEL_OP_REMU:
DECODE_R
movq REGISTER_ADDRESS(RS2r), RS2r
cmp $0, RS2r
test RS2r, RS2r
jne .remu_branch2
movq REGISTER_ADDRESS(RS1), RS1
WRITE_RD(RS1)
Expand All @@ -1059,7 +1059,7 @@ ckb_vm_x64_execute:
DECODE_R
movq REGISTER_ADDRESS(RS2r), RS2r
mov RS2rd, RS2rd
cmp $0, RS2r
test RS2r, RS2r
jne .remuw_branch2
movq REGISTER_ADDRESS(RS1), RS1
movslq RS1d, RS1
Expand Down Expand Up @@ -1094,7 +1094,7 @@ ckb_vm_x64_execute:
xorq RS1, RS1
jmp .remw_branch3
.remw_branch1:
cmp $0, RS2r
test RS2r, RS2r
jne .remw_branch2
jmp .remw_branch3
.remw_branch2:
Expand Down Expand Up @@ -1761,7 +1761,7 @@ ckb_vm_x64_execute:
.CKB_VM_ASM_LABEL_OP_CLZ:
DECODE_R
movq REGISTER_ADDRESS(RS1), RS1
cmp $0, RS1
test RS1, RS1
je .clz_branch
bsr RS1, RS1
neg RS1
Expand All @@ -1775,7 +1775,7 @@ ckb_vm_x64_execute:
.CKB_VM_ASM_LABEL_OP_CLZW:
DECODE_R
movq REGISTER_ADDRESS(RS1), RS1
cmp $0, RS1d
test RS1d, RS1d
je .clzw_branch
bsr RS1d, RS1d
neg RS1
Expand Down Expand Up @@ -1872,7 +1872,7 @@ ckb_vm_x64_execute:
.CKB_VM_ASM_LABEL_OP_CTZ:
DECODE_R
movq REGISTER_ADDRESS(RS1), RS1
cmp $0, RS1
test RS1, RS1
je .ctz_branch
bsf RS1, RS1
WRITE_RD(RS1)
Expand All @@ -1884,7 +1884,7 @@ ckb_vm_x64_execute:
.CKB_VM_ASM_LABEL_OP_CTZW:
DECODE_R
movq REGISTER_ADDRESS(RS1), RS1
cmp $0, RS1d
test RS1d, RS1d
je .ctzw_branch
bsf RS1d, RS1d
WRITE_RD(RS1)
Expand Down Expand Up @@ -2286,7 +2286,7 @@ ckb_vm_x64_execute:
xorq TEMP1, TEMP1
jmp .wide_div_branch3
.wide_div_branch1:
cmp $0, RS2r
test RS2r, RS2r
jne .wide_div_branch2
movq RS1, TEMP1
movq $UINT64_MAX, RS1
Expand All @@ -2307,7 +2307,7 @@ ckb_vm_x64_execute:
DECODE_R4
movq REGISTER_ADDRESS(RS1), RS1
movq REGISTER_ADDRESS(RS2r), RS2r
cmp $0, RS2r
test RS2r, RS2r
jne .wide_divu_branch1
WRITE_RS3(RS1)
WRITE_RD_VALUE($UINT64_MAX, RS2r)
Expand Down