diff mbox series

[FFmpeg-devel] lavc/riscv: don't set vxrm if unnecessary

Message ID 20240725143517.12683-1-remi@remlab.net
State New
Headers show
Series [FFmpeg-devel] lavc/riscv: don't set vxrm if unnecessary | expand

Checks

Context Check Description
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished

Commit Message

Rémi Denis-Courmont July 25, 2024, 2:35 p.m. UTC 
While narrowing clip is nominally a rounding operation, the rounding mode
has no arithmetic consequence if the right shift is by zero bits.
---
 libavcodec/riscv/h263dsp_rvv.S | 5 ++---
 libavcodec/riscv/vp8dsp_rvv.S  | 1 -
 2 files changed, 2 insertions(+), 4 deletions(-)
diff mbox series

Patch

diff --git a/libavcodec/riscv/h263dsp_rvv.S b/libavcodec/riscv/h263dsp_rvv.S
index 97503d527c..2a57284e4b 100644
--- a/libavcodec/riscv/h263dsp_rvv.S
+++ b/libavcodec/riscv/h263dsp_rvv.S
@@ -30,14 +30,13 @@  func ff_h263_h_loop_filter_rvv, zve32x
         vssseg4e8.v v8, (a0), a1
         ret
 1:
-        csrwi       vxrm, 0
-2:      auipc       t1, %pcrel_hi(ff_h263_loop_filter_strength)
+        auipc       t1, %pcrel_hi(ff_h263_loop_filter_strength)
         vwsubu.vv   v14, v10, v9       # p2 - p1
         add         t1, t1, a2
         vwsubu.vv   v12, v8, v11       # p0 - p3
         vsetvli     zero, zero, e16, m1, ta, mu
         vsll.vi     v14, v14, 2
-        lbu         t1, %pcrel_lo(2b)(t1) # strength
+        lbu         t1, %pcrel_lo(1b)(t1) # strength
         vadd.vv     v16, v12, v14
         # Divide by 8 toward 0. v16 is a signed 10-bit value at this point.
         vsrl.vi     v18, v16, 16 - 3   # v18 = (v16 < 0) ? 7 : 0
diff --git a/libavcodec/riscv/vp8dsp_rvv.S b/libavcodec/riscv/vp8dsp_rvv.S
index 7e062d9f13..e50e27ce82 100644
--- a/libavcodec/riscv/vp8dsp_rvv.S
+++ b/libavcodec/riscv/vp8dsp_rvv.S
@@ -166,7 +166,6 @@  endfunc
 
 # a3 = DC
 func ff_vp78_idct_dc_add_rvv, zve32x
-        csrwi      vxrm, 0
         vsetivli   zero, 4, e8, mf4, ta, ma
         sh         zero, (a1)
         vlse32.v   v8, (a0), a2