From 4f18477db6d6c67abb9dfbf0c9077fecfd7e5ba8 Mon Sep 17 00:00:00 2001
From: Joel Sing <joel@sing.id.au>
Date: Thu, 27 Jun 2024 20:02:57 +1000
Subject: [PATCH] math/big: implement subVW in riscv64 assembly
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This provides an assembly implementation of subVW for riscv64,
processing up to four words per loop, resulting in a significant
performance gain.
On a StarFive VisionFive 2:
│ subvw.1 │ subvw.2 │
│ sec/op │ sec/op vs base │
SubVW/1-4 57.43n ± 0% 41.45n ± 0% -27.82% (p=0.000 n=10)
SubVW/2-4 69.31n ± 0% 48.15n ± 0% -30.53% (p=0.000 n=10)
SubVW/3-4 76.12n ± 0% 54.87n ± 0% -27.92% (p=0.000 n=10)
SubVW/4-4 85.47n ± 0% 56.14n ± 0% -34.32% (p=0.000 n=10)
SubVW/5-4 96.15n ± 0% 62.83n ± 0% -34.65% (p=0.000 n=10)
SubVW/10-4 149.60n ± 0% 89.55n ± 0% -40.14% (p=0.000 n=10)
SubVW/100-4 1115.0n ± 0% 549.3n ± 0% -50.74% (p=0.000 n=10)
SubVW/1000-4 10.732µ ± 0% 5.071µ ± 0% -52.75% (p=0.000 n=10)
SubVW/10000-4 153.0µ ± 0% 103.7µ ± 0% -32.21% (p=0.000 n=10)
SubVW/100000-4 1.542m ± 0% 1.046m ± 0% -32.13% (p=0.000 n=10)
SubVWext/1-4 57.42n ± 0% 41.45n ± 0% -27.81% (p=0.000 n=10)
SubVWext/2-4 69.33n ± 0% 48.15n ± 0% -30.55% (p=0.000 n=10)
SubVWext/3-4 76.12n ± 0% 54.93n ± 0% -27.84% (p=0.000 n=10)
SubVWext/4-4 85.47n ± 0% 56.14n ± 0% -34.32% (p=0.000 n=10)
SubVWext/5-4 96.15n ± 0% 62.83n ± 0% -34.65% (p=0.000 n=10)
SubVWext/10-4 149.60n ± 0% 89.56n ± 0% -40.14% (p=0.000 n=10)
SubVWext/100-4 1115.0n ± 0% 549.3n ± 0% -50.74% (p=0.000 n=10)
SubVWext/1000-4 10.732µ ± 0% 5.061µ ± 0% -52.84% (p=0.000 n=10)
SubVWext/10000-4 152.5µ ± 0% 103.7µ ± 0% -32.02% (p=0.000 n=10)
SubVWext/100000-4 1.533m ± 0% 1.046m ± 0% -31.75% (p=0.000 n=10)
geomean 1.005µ 633.7n -36.92%
│ subvw.1 │ subvw.2 │
│ B/s │ B/s vs base │
SubVW/1-4 132.9Mi ± 0% 184.1Mi ± 0% +38.54% (p=0.000 n=10)
SubVW/2-4 220.1Mi ± 0% 316.9Mi ± 0% +43.95% (p=0.000 n=10)
SubVW/3-4 300.7Mi ± 0% 417.1Mi ± 0% +38.72% (p=0.000 n=10)
SubVW/4-4 357.1Mi ± 0% 543.6Mi ± 0% +52.24% (p=0.000 n=10)
SubVW/5-4 396.7Mi ± 0% 607.2Mi ± 0% +53.03% (p=0.000 n=10)
SubVW/10-4 510.1Mi ± 0% 851.9Mi ± 0% +67.01% (p=0.000 n=10)
SubVW/100-4 684.2Mi ± 0% 1388.9Mi ± 0% +102.99% (p=0.000 n=10)
SubVW/1000-4 710.9Mi ± 0% 1504.5Mi ± 0% +111.63% (p=0.000 n=10)
SubVW/10000-4 498.7Mi ± 0% 735.7Mi ± 0% +47.52% (p=0.000 n=10)
SubVW/100000-4 494.8Mi ± 0% 729.1Mi ± 0% +47.34% (p=0.000 n=10)
SubVWext/1-4 132.9Mi ± 0% 184.1Mi ± 0% +38.53% (p=0.000 n=10)
SubVWext/2-4 220.1Mi ± 0% 316.9Mi ± 0% +44.00% (p=0.000 n=10)
SubVWext/3-4 300.7Mi ± 0% 416.7Mi ± 0% +38.57% (p=0.000 n=10)
SubVWext/4-4 357.1Mi ± 0% 543.6Mi ± 0% +52.24% (p=0.000 n=10)
SubVWext/5-4 396.7Mi ± 0% 607.2Mi ± 0% +53.04% (p=0.000 n=10)
SubVWext/10-4 510.1Mi ± 0% 851.9Mi ± 0% +67.01% (p=0.000 n=10)
SubVWext/100-4 684.2Mi ± 0% 1388.9Mi ± 0% +102.99% (p=0.000 n=10)
SubVWext/1000-4 710.9Mi ± 0% 1507.6Mi ± 0% +112.07% (p=0.000 n=10)
SubVWext/10000-4 500.1Mi ± 0% 735.7Mi ± 0% +47.10% (p=0.000 n=10)
SubVWext/100000-4 497.8Mi ± 0% 729.4Mi ± 0% +46.52% (p=0.000 n=10)
geomean 387.6Mi 614.5Mi +58.51%
Change-Id: I9d7fac719e977710ad9db9121fa298db6df605de
Reviewed-on: https://go-review.googlesource.com/c/go/+/595398
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
---
src/math/big/arith_riscv64.s | 59 +++++++++++++++++++++++++++++++++++-
1 file changed, 58 insertions(+), 1 deletion(-)
diff --git a/src/math/big/arith_riscv64.s b/src/math/big/arith_riscv64.s
index 0029a5b6b44..e856d2e9c0e 100644
--- a/src/math/big/arith_riscv64.s
+++ b/src/math/big/arith_riscv64.s
@@ -234,7 +234,64 @@ done:
RET
TEXT ·subVW(SB),NOSPLIT,$0
- JMP ·subVW_g(SB)
+ MOV x+24(FP), X5
+ MOV y+48(FP), X6
+ MOV z+0(FP), X7
+ MOV z_len+8(FP), X30
+
+ MOV $4, X28
+ MOV X6, X29 // b = y
+
+ BEQZ X30, done
+ BLTU X30, X28, loop1
+
+loop4:
+ MOV 0(X5), X8 // x[0]
+ MOV 8(X5), X11 // x[1]
+ MOV 16(X5), X14 // x[2]
+ MOV 24(X5), X17 // x[3]
+
+ SUB X29, X8, X10 // z[0] = x[0] - b
+ SLTU X10, X8, X29 // next b
+
+ SUB X29, X11, X13 // z[1] = x[1] - b
+ SLTU X13, X11, X29 // next b
+
+ SUB X29, X14, X16 // z[2] = x[2] - b
+ SLTU X16, X14, X29 // next b
+
+ SUB X29, X17, X19 // z[3] = x[3] - b
+ SLTU X19, X17, X29 // next b
+
+ MOV X10, 0(X7) // z[0]
+ MOV X13, 8(X7) // z[1]
+ MOV X16, 16(X7) // z[2]
+ MOV X19, 24(X7) // z[3]
+
+ ADD $32, X5
+ ADD $32, X7
+ SUB $4, X30
+
+ BGEU X30, X28, loop4
+ BEQZ X30, done
+
+loop1:
+ MOV 0(X5), X10 // x
+
+ SUB X29, X10, X12 // z = x - b
+ SLTU X12, X10, X29 // next b
+
+ MOV X12, 0(X7) // z
+
+ ADD $8, X5
+ ADD $8, X7
+ SUB $1, X30
+
+ BNEZ X30, loop1
+
+done:
+ MOV X29, c+56(FP) // return b
+ RET
TEXT ·shlVU(SB),NOSPLIT,$0
JMP ·shlVU_g(SB)
--
GitLab