diff options
| author | Xiaolin Zhao <zhaoxiaolin@loongson.cn> | 2025-08-29 16:20:16 +0800 |
|---|---|---|
| committer | Gopher Robot <gobot@golang.org> | 2025-09-04 09:22:33 -0700 |
| commit | b8cc907425c4b851d2b941cf689cf8177ea8a153 (patch) | |
| tree | c6d99ae0cff79fbfa55dcaa69928a1c24ffc474a /src/cmd/asm/internal | |
| parent | 8c27a808905b0611b0a7b7bbff08819206be3b86 (diff) | |
| download | go-b8cc907425c4b851d2b941cf689cf8177ea8a153.tar.xz | |
cmd/internal/obj/loong64: fix the usage of offset in the instructions [X]VLDREPL.{B/H/W/D}
The previously defined usage of offset was ambiguous and not easy to understand.
For example, to fetch 4 bytes of data from the address base+8 and
broadcast it to each word element of vector register V5, the assembly
implementation is as follows:
previous: VMOVQ 2(base), V5.W4
current: VMOVQ 8(base), V5.W4
Change-Id: I8bc84e35033ab63bd10f4c61618789f94314f78c
Reviewed-on: https://go-review.googlesource.com/c/go/+/699875
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
Auto-Submit: Michael Pratt <mpratt@google.com>
Reviewed-by: Meidan Li <limeidan@loongson.cn>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Diffstat (limited to 'src/cmd/asm/internal')
| -rw-r--r-- | src/cmd/asm/internal/asm/testdata/loong64enc1.s | 28 |
1 files changed, 22 insertions, 6 deletions
diff --git a/src/cmd/asm/internal/asm/testdata/loong64enc1.s b/src/cmd/asm/internal/asm/testdata/loong64enc1.s index 63676cc785..c5c6a4479a 100644 --- a/src/cmd/asm/internal/asm/testdata/loong64enc1.s +++ b/src/cmd/asm/internal/asm/testdata/loong64enc1.s @@ -538,13 +538,29 @@ lable2: // Load data from memory and broadcast to each element of a vector register: VMOVQ offset(Rj), <Vd>.<T> VMOVQ (R4), V0.B16 // 80008030 - VMOVQ 1(R4), V1.H8 // 81044030 - VMOVQ 2(R4), V2.W4 // 82082030 - VMOVQ 3(R4), V3.V2 // 830c1030 + VMOVQ 1(R4), V0.B16 // 80048030 + VMOVQ -3(R4), V0.B16 // 80f4bf30 + VMOVQ (R4), V1.H8 // 81004030 + VMOVQ 2(R4), V1.H8 // 81044030 + VMOVQ -6(R4), V1.H8 // 81f45f30 + VMOVQ (R4), V2.W4 // 82002030 + VMOVQ 8(R4), V2.W4 // 82082030 + VMOVQ -12(R4), V2.W4 // 82f42f30 + VMOVQ (R4), V3.V2 // 83001030 + VMOVQ 24(R4), V3.V2 // 830c1030 + VMOVQ -16(R4), V3.V2 // 83f81730 XVMOVQ (R4), X0.B32 // 80008032 - XVMOVQ 1(R4), X1.H16 // 81044032 - XVMOVQ 2(R4), X2.W8 // 82082032 - XVMOVQ 3(R4), X3.V4 // 830c1032 + XVMOVQ 1(R4), X0.B32 // 80048032 + XVMOVQ -5(R4), X0.B32 // 80ecbf32 + XVMOVQ (R4), X1.H16 // 81004032 + XVMOVQ 2(R4), X1.H16 // 81044032 + XVMOVQ -10(R4), X1.H16 // 81ec5f32 + XVMOVQ (R4), X2.W8 // 82002032 + XVMOVQ 8(R4), X2.W8 // 82082032 + XVMOVQ -20(R4), X2.W8 // 82ec2f32 + XVMOVQ (R4), X3.V4 // 83001032 + XVMOVQ 24(R4), X3.V4 // 830c1032 + XVMOVQ -24(R4), X3.V4 // 83f41732 // VSEQ{B,H,W,V}, XVSEQ{B,H,W,V} instruction VSEQB V1, V2, V3 // 43040070 |