fzorb/REDACTED-rig
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

RandomX: rewrote dataset read code

Browse source 
Unified code for AMD and Intel
1% faster on Intel
0.15% faster on AMD Ryzen
This commit is contained in:
SChernykh 2021-05-20 12:45:42 +02:00
parent 7f7b1fb073
commit ff82ca57f2
10 changed files with 16 additions and 69 deletions
Download patch file Download diff file Expand all files Collapse all files

1
src/crypto/randomx/asm/program_prologue_linux.inc
View file

@ -22,6 +22,7 @@
mov rsi, rdx ;# uint8_t* scratchpad
mov rax, rbp
ror rbp, 32
;# zero integer registers
xor r8, r8

1
src/crypto/randomx/asm/program_prologue_win64.inc
View file

@ -35,6 +35,7 @@
mov rbx, r9 ;# loop counter
mov rax, rbp
ror rbp, 32
;# zero integer registers
xor r8, r8

23
src/crypto/randomx/asm/program_read_dataset.inc
View file

@ -1,17 +1,16 @@
mov ecx, ebp ;# ecx = ma
and ecx, RANDOMX_DATASET_BASE_MASK
xor r8, qword ptr [rdi+rcx]
ror rbp, 32 ;# swap "ma" and "mx"
xor rbp, rax ;# modify "mx"
mov edx, ebp ;# edx = mx
and edx, RANDOMX_DATASET_BASE_MASK
prefetchnta byte ptr [rdi+rdx]
ror rbp, 32 ;# swap "ma" and "mx"
mov edx, ebp ;# edx = ma
and edx, RANDOMX_DATASET_BASE_MASK
lea rcx, [rdi+rdx] ;# dataset cache line
xor r8, qword ptr [rcx+0]
xor r9, qword ptr [rcx+8]
xor r10, qword ptr [rcx+16]
xor r11, qword ptr [rcx+24]
xor r12, qword ptr [rcx+32]
xor r13, qword ptr [rcx+40]
xor r14, qword ptr [rcx+48]
xor r15, qword ptr [rcx+56]
xor r9, qword ptr [rdi+rcx+8]
xor r10, qword ptr [rdi+rcx+16]
xor r11, qword ptr [rdi+rcx+24]
xor r12, qword ptr [rdi+rcx+32]
xor r13, qword ptr [rdi+rcx+40]
xor r14, qword ptr [rdi+rcx+48]
xor r15, qword ptr [rdi+rcx+56]

17
src/crypto/randomx/asm/program_read_dataset_ryzen.inc
View file

@ -1,17 +0,0 @@
mov rcx, rbp ;# ecx = ma
shr rcx, 32
and ecx, RANDOMX_DATASET_BASE_MASK
xor r8, qword ptr [rdi+rcx]
xor rbp, rax ;# modify "mx"
mov edx, ebp ;# edx = mx
and edx, RANDOMX_DATASET_BASE_MASK
prefetchnta byte ptr [rdi+rdx]
ror rbp, 32 ;# swap "ma" and "mx"
xor r9, qword ptr [rdi+rcx+8]
xor r10, qword ptr [rdi+rcx+16]
xor r11, qword ptr [rdi+rcx+24]
xor r12, qword ptr [rdi+rcx+32]
xor r13, qword ptr [rdi+rcx+40]
xor r14, qword ptr [rdi+rcx+48]
xor r15, qword ptr [rdi+rcx+56]