Lephenixnoir
/
gintctl
4
4
Fork 2
Code Issues 1 Pull Requests Releases Wiki Activity

perf/cpu: remove tests that don't fit on fx

This commit is contained in:
Lephenixnoir 2022-05-16 21:58:42 +01:00
parent 7e9b26f42c
commit fb13d985ff
Signed by: Lephenixnoir
GPG Key ID: 1BBA026E13FC0495
2 changed files with 0 additions and 95 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

93
src/perf/cpu.S
View File

@ -495,101 +495,8 @@ bench double_write, 1024
2: mov.l r1, @r5
end
/* [2D image rendering]
This section is used to investigate the performance of Azur's built-in image
shader. Most of the core loops must perform at 5-10 cycles per iteration,
which is fairly easy to validate here. */
bench azur_p8_rgb565, 512
mov.l .buffer2, r3
mov.l .buffer, r5
mov.l r10, @-r15
mov #0, r0
mov.l r8, @-r15
mov #0, r10
mov r3, r8
nop
/* 2-unrolled 2-stage main loop */
1: mov.b @r3+, r6
shll r10
mov.w @(r0,r8), r0
nop
mov.w r0, @(4,r5)
mov r10, r0
mov.b @r3+, r10
add #4, r5
mov.w @(r0,r8), r0
shll r6
mov.w r0, @(2,r5)
2: mov r6, r0
mov.l @r15+, r8
mov.l @r15+, r10
end
bench azur_p8_rgb565a, 512
mov.l .buffer2, r3
mov.l .buffer, r5
mov.l r10, @-r15
mov #0, r6
mov.l r8, @-r15
mov #0, r10
mov r3, r8
nop
/* 2-unrolled 2-stage main loop */
1: add r6, r6
mov r6, r0
add r10, r10
bt.s 5f
tst r10, r10
mov.w @(r0,r8), r0
mov.w r0, @(4,r5)
5: mov.b @r3+, r6
mov r10, r0
bt.s 6f
add #4, r5
mov.w @(r0,r8), r0
mov.w r0, @(2,r5)
6: mov.b @r3+, r10
2: tst r6, r6
mov.l @r15+, r8
mov.l @r15+, r10
end
/* XRAM buffer */
.align 4
.buffer:
.long _cpu_perf_xram_buffer
/* Secondary buffer in RAM */
.buffer2:
.long _buffer2
.section .data, "aw"
_buffer2:
.zero 2048

2
src/perf/cpu.c
View File

@ -55,8 +55,6 @@
MACRO(double_read, 1024, "Double read") \
MACRO(double_incr_read, 1024, "Double increment read") \
MACRO(double_write, 1024, "Double write") \
MACRO(azur_p8_rgb565, 512, "Azur: P8_RGB565 loop") \
MACRO(azur_p8_rgb565a, 512, "Azur: P8_RGB565A loop") \
GXRAM uint32_t cpu_perf_xram_buffer[512];