Decrease the frequency of FADD/FSUB in favor of FMUL (#77)

* this better matches CPU capabilities since execution ports are usually split 1:1 between fadd and fmul
* the frequency of FSWAP_R decreased from 8 to 4 (it's ASIC-friendly)
* activate IROL_R instruction
This commit is contained in:
tevador 2019-06-22 16:05:22 +02:00 committed by GitHub
parent 83498cddf2
commit 91cd35ff13
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 13 additions and 13 deletions

View file

@ -29,16 +29,16 @@ RANDOMX_FREQ_IMUL_RCP EQU 8t
RANDOMX_FREQ_INEG_R EQU 2t
RANDOMX_FREQ_IXOR_R EQU 15t
RANDOMX_FREQ_IXOR_M EQU 5t
RANDOMX_FREQ_IROR_R EQU 10t
RANDOMX_FREQ_IROL_R EQU 0t
RANDOMX_FREQ_IROR_R EQU 8t
RANDOMX_FREQ_IROL_R EQU 2t
RANDOMX_FREQ_ISWAP_R EQU 4t
RANDOMX_FREQ_FSWAP_R EQU 8t
RANDOMX_FREQ_FADD_R EQU 20t
RANDOMX_FREQ_FSWAP_R EQU 4t
RANDOMX_FREQ_FADD_R EQU 16t
RANDOMX_FREQ_FADD_M EQU 5t
RANDOMX_FREQ_FSUB_R EQU 20t
RANDOMX_FREQ_FSUB_R EQU 16t
RANDOMX_FREQ_FSUB_M EQU 5t
RANDOMX_FREQ_FSCAL_R EQU 6t
RANDOMX_FREQ_FMUL_R EQU 20t
RANDOMX_FREQ_FMUL_R EQU 32t
RANDOMX_FREQ_FDIV_M EQU 4t
RANDOMX_FREQ_FSQRT_R EQU 6t
RANDOMX_FREQ_CBRANCH EQU 16t

View file

@ -61,7 +61,7 @@ namespace randomx {
constexpr int wtSum = RANDOMX_FREQ_IADD_RS + RANDOMX_FREQ_IADD_M + RANDOMX_FREQ_ISUB_R + \
RANDOMX_FREQ_ISUB_M + RANDOMX_FREQ_IMUL_R + RANDOMX_FREQ_IMUL_M + RANDOMX_FREQ_IMULH_R + \
RANDOMX_FREQ_IMULH_M + RANDOMX_FREQ_ISMULH_R + RANDOMX_FREQ_ISMULH_M + RANDOMX_FREQ_IMUL_RCP + \
RANDOMX_FREQ_INEG_R + RANDOMX_FREQ_IXOR_R + RANDOMX_FREQ_IXOR_M + RANDOMX_FREQ_IROR_R + RANDOMX_FREQ_ISWAP_R + \
RANDOMX_FREQ_INEG_R + RANDOMX_FREQ_IXOR_R + RANDOMX_FREQ_IXOR_M + RANDOMX_FREQ_IROR_R + RANDOMX_FREQ_IROL_R + RANDOMX_FREQ_ISWAP_R + \
RANDOMX_FREQ_FSWAP_R + RANDOMX_FREQ_FADD_R + RANDOMX_FREQ_FADD_M + RANDOMX_FREQ_FSUB_R + RANDOMX_FREQ_FSUB_M + \
RANDOMX_FREQ_FSCAL_R + RANDOMX_FREQ_FMUL_R + RANDOMX_FREQ_FDIV_M + RANDOMX_FREQ_FSQRT_R + RANDOMX_FREQ_CBRANCH + \
RANDOMX_FREQ_CFROUND + RANDOMX_FREQ_ISTORE + RANDOMX_FREQ_NOP;

View file

@ -96,18 +96,18 @@ Total sum of frequencies must be 256
#define RANDOMX_FREQ_INEG_R 2
#define RANDOMX_FREQ_IXOR_R 15
#define RANDOMX_FREQ_IXOR_M 5
#define RANDOMX_FREQ_IROR_R 10
#define RANDOMX_FREQ_IROL_R 0
#define RANDOMX_FREQ_IROR_R 8
#define RANDOMX_FREQ_IROL_R 2
#define RANDOMX_FREQ_ISWAP_R 4
//Floating point instructions
#define RANDOMX_FREQ_FSWAP_R 8
#define RANDOMX_FREQ_FADD_R 20
#define RANDOMX_FREQ_FSWAP_R 4
#define RANDOMX_FREQ_FADD_R 16
#define RANDOMX_FREQ_FADD_M 5
#define RANDOMX_FREQ_FSUB_R 20
#define RANDOMX_FREQ_FSUB_R 16
#define RANDOMX_FREQ_FSUB_M 5
#define RANDOMX_FREQ_FSCAL_R 6
#define RANDOMX_FREQ_FMUL_R 20
#define RANDOMX_FREQ_FMUL_R 32
#define RANDOMX_FREQ_FDIV_M 4
#define RANDOMX_FREQ_FSQRT_R 6