[llvm] [NFC][AMDGPU] Align all gfx1250 VOP1 MC tests with downstream (PR #149567)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Jul 18 12:15:25 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-amdgpu
Author: Shilei Tian (shiltian)
<details>
<summary>Changes</summary>
This PR adds all VOP1 tests that haven't yet been upstreamed by copying the relevant test files directly from downstream. Afterward, the auto-generation script is run with the `--unique` option to deduplicate any redundant tests that may have been introduced during the downstream merge.
Co-authored-by: Mekhanoshin, Stanislav <Stanislav.Mekhanoshin@<!-- -->amd.com>
---
Patch is 671.16 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/149567.diff
9 Files Affected:
- (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop1.s (+39)
- (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1-fake16.s (+3681-108)
- (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1.s (+3827-86)
- (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1_dpp8-fake16.s (-4)
- (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1_dpp8.s (+8-12)
- (modified) llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop1.txt (+49-4)
- (modified) llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3_from_vop1.txt (+4094-167)
- (modified) llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3_from_vop1_dpp16.txt (+180-180)
- (modified) llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3_from_vop1_dpp8.txt (+60-60)
``````````diff
diff --git a/llvm/test/MC/AMDGPU/gfx1250_asm_vop1.s b/llvm/test/MC/AMDGPU/gfx1250_asm_vop1.s
index 3ddbc365224db..a313741ffe22d 100644
--- a/llvm/test/MC/AMDGPU/gfx1250_asm_vop1.s
+++ b/llvm/test/MC/AMDGPU/gfx1250_asm_vop1.s
@@ -664,6 +664,45 @@ v_cvt_f32_fp8_e32 v1, 3
v_cvt_f32_fp8_e32 v1, v3
// GFX1250: v_cvt_f32_fp8_e32 v1, v3 ; encoding: [0x03,0xd9,0x02,0x7e]
+v_cvt_pk_f32_bf8_e32 v[2:3], s3
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[2:3], s3 ; encoding: [0x03,0xde,0x04,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[4:5], s5
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[4:5], s5 ; encoding: [0x05,0xde,0x08,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[2:3], 3
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[2:3], 3 ; encoding: [0x83,0xde,0x04,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[4:5], 3
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[4:5], 3 ; encoding: [0x83,0xde,0x08,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[2:3], v3
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[2:3], v3 ; encoding: [0x03,0xdf,0x04,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[4:5], v3
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[4:5], v3 ; encoding: [0x03,0xdf,0x08,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[4:5], v127.h
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[4:5], v127.h ; encoding: [0xff,0xdf,0x08,0x7e]
+
+v_cvt_pk_f32_bf8_e32 v[4:5], v127.l
+// GFX1250: v_cvt_pk_f32_bf8_e32 v[4:5], v127.l ; encoding: [0x7f,0xdf,0x08,0x7e]
+
+v_cvt_pk_f32_fp8_e32 v[2:3], s3
+// GFX1250: v_cvt_pk_f32_fp8_e32 v[2:3], s3 ; encoding: [0x03,0xdc,0x04,0x7e]
+
+v_cvt_pk_f32_fp8_e32 v[2:3], 3
+// GFX1250: v_cvt_pk_f32_fp8_e32 v[2:3], 3 ; encoding: [0x83,0xdc,0x04,0x7e]
+
+v_cvt_pk_f32_fp8_e32 v[2:3], v3
+// GFX1250: v_cvt_pk_f32_fp8_e32 v[2:3], v3 ; encoding: [0x03,0xdd,0x04,0x7e]
+
+v_cvt_pk_f32_fp8_e32 v[4:5], v127.h
+// GFX1250: v_cvt_pk_f32_fp8_e32 v[4:5], v127.h ; encoding: [0xff,0xdd,0x08,0x7e]
+
+v_cvt_pk_f32_fp8_e32 v[4:5], v127.l
+// GFX1250: v_cvt_pk_f32_fp8_e32 v[4:5], v127.l ; encoding: [0x7f,0xdd,0x08,0x7e]
+
v_sat_pk4_i4_i8 v1, v2
// GFX1250: v_sat_pk4_i4_i8_e32 v1, v2 ; encoding: [0x02,0xe7,0x02,0x7e]
diff --git a/llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1-fake16.s b/llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1-fake16.s
index aa4e49d85f1ff..0931523bbf40c 100644
--- a/llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1-fake16.s
+++ b/llvm/test/MC/AMDGPU/gfx1250_asm_vop3_from_vop1-fake16.s
@@ -1,167 +1,3659 @@
// NOTE: Assertions have been autogenerated by utils/update_mc_test_checks.py UTC_ARGS: --version 5
// RUN: llvm-mc -triple=amdgcn -mcpu=gfx1250 -mattr=-real-true16 -show-encoding %s | FileCheck --check-prefix=GFX1250 %s
-v_tanh_bf16_e64 v5, v1
-// GFX1250: v_tanh_bf16_e64 v5, v1 ; encoding: [0x05,0x00,0xca,0xd5,0x01,0x01,0x00,0x00]
+v_bfrev_b32_e64 v5, v1
+// GFX1250: v_bfrev_b32_e64 v5, v1 ; encoding: [0x05,0x00,0xb8,0xd5,0x01,0x01,0x00,0x00]
+
+v_bfrev_b32_e64 v5, v255
+// GFX1250: v_bfrev_b32_e64 v5, v255 ; encoding: [0x05,0x00,0xb8,0xd5,0xff,0x01,0x00,0x00]
+
+v_bfrev_b32_e64 v5, s1
+// GFX1250: v_bfrev_b32_e64 v5, s1 ; encoding: [0x05,0x00,0xb8,0xd5,0x01,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, s105
+// GFX1250: v_bfrev_b32_e64 v5, s105 ; encoding: [0x05,0x00,0xb8,0xd5,0x69,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, vcc_lo
+// GFX1250: v_bfrev_b32_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xb8,0xd5,0x6a,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, vcc_hi
+// GFX1250: v_bfrev_b32_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xb8,0xd5,0x6b,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, ttmp15
+// GFX1250: v_bfrev_b32_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xb8,0xd5,0x7b,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, m0
+// GFX1250: v_bfrev_b32_e64 v5, m0 ; encoding: [0x05,0x00,0xb8,0xd5,0x7d,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, exec_lo
+// GFX1250: v_bfrev_b32_e64 v5, exec_lo ; encoding: [0x05,0x00,0xb8,0xd5,0x7e,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, exec_hi
+// GFX1250: v_bfrev_b32_e64 v5, exec_hi ; encoding: [0x05,0x00,0xb8,0xd5,0x7f,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, null
+// GFX1250: v_bfrev_b32_e64 v5, null ; encoding: [0x05,0x00,0xb8,0xd5,0x7c,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, -1
+// GFX1250: v_bfrev_b32_e64 v5, -1 ; encoding: [0x05,0x00,0xb8,0xd5,0xc1,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, 0.5
+// GFX1250: v_bfrev_b32_e64 v5, 0.5 ; encoding: [0x05,0x00,0xb8,0xd5,0xf0,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v5, src_scc
+// GFX1250: v_bfrev_b32_e64 v5, src_scc ; encoding: [0x05,0x00,0xb8,0xd5,0xfd,0x00,0x00,0x00]
+
+v_bfrev_b32_e64 v255, 0xaf123456
+// GFX1250: v_bfrev_b32_e64 v255, 0xaf123456 ; encoding: [0xff,0x00,0xb8,0xd5,0xff,0x00,0x00,0x00,0x56,0x34,0x12,0xaf]
+
+v_ceil_f16_e64 v5, v1
+// GFX1250: v_ceil_f16_e64 v5, v1 ; encoding: [0x05,0x00,0xdc,0xd5,0x01,0x01,0x00,0x00]
+
+v_ceil_f16_e64 v5, v255
+// GFX1250: v_ceil_f16_e64 v5, v255 ; encoding: [0x05,0x00,0xdc,0xd5,0xff,0x01,0x00,0x00]
+
+v_ceil_f16_e64 v5, s1
+// GFX1250: v_ceil_f16_e64 v5, s1 ; encoding: [0x05,0x00,0xdc,0xd5,0x01,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, s105
+// GFX1250: v_ceil_f16_e64 v5, s105 ; encoding: [0x05,0x00,0xdc,0xd5,0x69,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, vcc_lo
+// GFX1250: v_ceil_f16_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xdc,0xd5,0x6a,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, vcc_hi
+// GFX1250: v_ceil_f16_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xdc,0xd5,0x6b,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, ttmp15
+// GFX1250: v_ceil_f16_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xdc,0xd5,0x7b,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, m0
+// GFX1250: v_ceil_f16_e64 v5, m0 ; encoding: [0x05,0x00,0xdc,0xd5,0x7d,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, exec_lo
+// GFX1250: v_ceil_f16_e64 v5, exec_lo ; encoding: [0x05,0x00,0xdc,0xd5,0x7e,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, exec_hi
+// GFX1250: v_ceil_f16_e64 v5, exec_hi ; encoding: [0x05,0x00,0xdc,0xd5,0x7f,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, null
+// GFX1250: v_ceil_f16_e64 v5, null ; encoding: [0x05,0x00,0xdc,0xd5,0x7c,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, -1
+// GFX1250: v_ceil_f16_e64 v5, -1 ; encoding: [0x05,0x00,0xdc,0xd5,0xc1,0x00,0x00,0x00]
+
+v_ceil_f16_e64 v5, 0.5 mul:2
+// GFX1250: v_ceil_f16_e64 v5, 0.5 mul:2 ; encoding: [0x05,0x00,0xdc,0xd5,0xf0,0x00,0x00,0x08]
+
+v_ceil_f16_e64 v5, src_scc mul:4
+// GFX1250: v_ceil_f16_e64 v5, src_scc mul:4 ; encoding: [0x05,0x00,0xdc,0xd5,0xfd,0x00,0x00,0x10]
+
+v_ceil_f16_e64 v255, -|0xfe0b| clamp div:2
+// GFX1250: v_ceil_f16_e64 v255, -|0xfe0b| clamp div:2 ; encoding: [0xff,0x81,0xdc,0xd5,0xff,0x00,0x00,0x38,0x0b,0xfe,0x00,0x00]
+
+v_ceil_f32_e64 v5, v1
+// GFX1250: v_ceil_f32_e64 v5, v1 ; encoding: [0x05,0x00,0xa2,0xd5,0x01,0x01,0x00,0x00]
+
+v_ceil_f32_e64 v5, v255
+// GFX1250: v_ceil_f32_e64 v5, v255 ; encoding: [0x05,0x00,0xa2,0xd5,0xff,0x01,0x00,0x00]
+
+v_ceil_f32_e64 v5, s1
+// GFX1250: v_ceil_f32_e64 v5, s1 ; encoding: [0x05,0x00,0xa2,0xd5,0x01,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, s105
+// GFX1250: v_ceil_f32_e64 v5, s105 ; encoding: [0x05,0x00,0xa2,0xd5,0x69,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, vcc_lo
+// GFX1250: v_ceil_f32_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xa2,0xd5,0x6a,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, vcc_hi
+// GFX1250: v_ceil_f32_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xa2,0xd5,0x6b,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, ttmp15
+// GFX1250: v_ceil_f32_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xa2,0xd5,0x7b,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, m0
+// GFX1250: v_ceil_f32_e64 v5, m0 ; encoding: [0x05,0x00,0xa2,0xd5,0x7d,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, exec_lo
+// GFX1250: v_ceil_f32_e64 v5, exec_lo ; encoding: [0x05,0x00,0xa2,0xd5,0x7e,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, exec_hi
+// GFX1250: v_ceil_f32_e64 v5, exec_hi ; encoding: [0x05,0x00,0xa2,0xd5,0x7f,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, null
+// GFX1250: v_ceil_f32_e64 v5, null ; encoding: [0x05,0x00,0xa2,0xd5,0x7c,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, -1
+// GFX1250: v_ceil_f32_e64 v5, -1 ; encoding: [0x05,0x00,0xa2,0xd5,0xc1,0x00,0x00,0x00]
+
+v_ceil_f32_e64 v5, 0.5 mul:2
+// GFX1250: v_ceil_f32_e64 v5, 0.5 mul:2 ; encoding: [0x05,0x00,0xa2,0xd5,0xf0,0x00,0x00,0x08]
+
+v_ceil_f32_e64 v5, src_scc mul:4
+// GFX1250: v_ceil_f32_e64 v5, src_scc mul:4 ; encoding: [0x05,0x00,0xa2,0xd5,0xfd,0x00,0x00,0x10]
+
+v_ceil_f32_e64 v255, -|0xaf123456| clamp div:2
+// GFX1250: v_ceil_f32_e64 v255, -|0xaf123456| clamp div:2 ; encoding: [0xff,0x81,0xa2,0xd5,0xff,0x00,0x00,0x38,0x56,0x34,0x12,0xaf]
+
+v_ceil_f64_e64 v[6:7], v[2:3]
+// GFX1250: v_ceil_f64_e64 v[6:7], v[2:3] ; encoding: [0x06,0x00,0x98,0xd5,0x02,0x01,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], v[254:255]
+// GFX1250: v_ceil_f64_e64 v[6:7], v[254:255] ; encoding: [0x06,0x00,0x98,0xd5,0xfe,0x01,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], s[2:3]
+// GFX1250: v_ceil_f64_e64 v[6:7], s[2:3] ; encoding: [0x06,0x00,0x98,0xd5,0x02,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], s[104:105]
+// GFX1250: v_ceil_f64_e64 v[6:7], s[104:105] ; encoding: [0x06,0x00,0x98,0xd5,0x68,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], vcc
+// GFX1250: v_ceil_f64_e64 v[6:7], vcc ; encoding: [0x06,0x00,0x98,0xd5,0x6a,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], ttmp[14:15]
+// GFX1250: v_ceil_f64_e64 v[6:7], ttmp[14:15] ; encoding: [0x06,0x00,0x98,0xd5,0x7a,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], exec
+// GFX1250: v_ceil_f64_e64 v[6:7], exec ; encoding: [0x06,0x00,0x98,0xd5,0x7e,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], null
+// GFX1250: v_ceil_f64_e64 v[6:7], null ; encoding: [0x06,0x00,0x98,0xd5,0x7c,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], -1
+// GFX1250: v_ceil_f64_e64 v[6:7], -1 ; encoding: [0x06,0x00,0x98,0xd5,0xc1,0x00,0x00,0x00]
+
+v_ceil_f64_e64 v[6:7], 0.5 mul:2
+// GFX1250: v_ceil_f64_e64 v[6:7], 0.5 mul:2 ; encoding: [0x06,0x00,0x98,0xd5,0xf0,0x00,0x00,0x08]
+
+v_ceil_f64_e64 v[6:7], -|src_scc| mul:4
+// GFX1250: v_ceil_f64_e64 v[6:7], -|src_scc| mul:4 ; encoding: [0x06,0x01,0x98,0xd5,0xfd,0x00,0x00,0x30]
+
+v_ceil_f64_e64 v[254:255], 0xaf123456 clamp div:2
+// GFX1250: v_ceil_f64_e64 v[254:255], 0xaf123456 clamp div:2 ; encoding: [0xfe,0x80,0x98,0xd5,0xff,0x00,0x00,0x18,0x56,0x34,0x12,0xaf]
+
+v_cls_i32_e64 v5, v1
+// GFX1250: v_cls_i32_e64 v5, v1 ; encoding: [0x05,0x00,0xbb,0xd5,0x01,0x01,0x00,0x00]
+
+v_cls_i32_e64 v5, v255
+// GFX1250: v_cls_i32_e64 v5, v255 ; encoding: [0x05,0x00,0xbb,0xd5,0xff,0x01,0x00,0x00]
+
+v_cls_i32_e64 v5, s1
+// GFX1250: v_cls_i32_e64 v5, s1 ; encoding: [0x05,0x00,0xbb,0xd5,0x01,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, s105
+// GFX1250: v_cls_i32_e64 v5, s105 ; encoding: [0x05,0x00,0xbb,0xd5,0x69,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, vcc_lo
+// GFX1250: v_cls_i32_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xbb,0xd5,0x6a,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, vcc_hi
+// GFX1250: v_cls_i32_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xbb,0xd5,0x6b,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, ttmp15
+// GFX1250: v_cls_i32_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xbb,0xd5,0x7b,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, m0
+// GFX1250: v_cls_i32_e64 v5, m0 ; encoding: [0x05,0x00,0xbb,0xd5,0x7d,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, exec_lo
+// GFX1250: v_cls_i32_e64 v5, exec_lo ; encoding: [0x05,0x00,0xbb,0xd5,0x7e,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, exec_hi
+// GFX1250: v_cls_i32_e64 v5, exec_hi ; encoding: [0x05,0x00,0xbb,0xd5,0x7f,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, null
+// GFX1250: v_cls_i32_e64 v5, null ; encoding: [0x05,0x00,0xbb,0xd5,0x7c,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, -1
+// GFX1250: v_cls_i32_e64 v5, -1 ; encoding: [0x05,0x00,0xbb,0xd5,0xc1,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, 0.5
+// GFX1250: v_cls_i32_e64 v5, 0.5 ; encoding: [0x05,0x00,0xbb,0xd5,0xf0,0x00,0x00,0x00]
+
+v_cls_i32_e64 v5, src_scc
+// GFX1250: v_cls_i32_e64 v5, src_scc ; encoding: [0x05,0x00,0xbb,0xd5,0xfd,0x00,0x00,0x00]
+
+v_cls_i32_e64 v255, 0xaf123456
+// GFX1250: v_cls_i32_e64 v255, 0xaf123456 ; encoding: [0xff,0x00,0xbb,0xd5,0xff,0x00,0x00,0x00,0x56,0x34,0x12,0xaf]
+
+v_clz_i32_u32_e64 v5, v1
+// GFX1250: v_clz_i32_u32_e64 v5, v1 ; encoding: [0x05,0x00,0xb9,0xd5,0x01,0x01,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, v255
+// GFX1250: v_clz_i32_u32_e64 v5, v255 ; encoding: [0x05,0x00,0xb9,0xd5,0xff,0x01,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, s1
+// GFX1250: v_clz_i32_u32_e64 v5, s1 ; encoding: [0x05,0x00,0xb9,0xd5,0x01,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, s105
+// GFX1250: v_clz_i32_u32_e64 v5, s105 ; encoding: [0x05,0x00,0xb9,0xd5,0x69,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, vcc_lo
+// GFX1250: v_clz_i32_u32_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xb9,0xd5,0x6a,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, vcc_hi
+// GFX1250: v_clz_i32_u32_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xb9,0xd5,0x6b,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, ttmp15
+// GFX1250: v_clz_i32_u32_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xb9,0xd5,0x7b,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, m0
+// GFX1250: v_clz_i32_u32_e64 v5, m0 ; encoding: [0x05,0x00,0xb9,0xd5,0x7d,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, exec_lo
+// GFX1250: v_clz_i32_u32_e64 v5, exec_lo ; encoding: [0x05,0x00,0xb9,0xd5,0x7e,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, exec_hi
+// GFX1250: v_clz_i32_u32_e64 v5, exec_hi ; encoding: [0x05,0x00,0xb9,0xd5,0x7f,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, null
+// GFX1250: v_clz_i32_u32_e64 v5, null ; encoding: [0x05,0x00,0xb9,0xd5,0x7c,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, -1
+// GFX1250: v_clz_i32_u32_e64 v5, -1 ; encoding: [0x05,0x00,0xb9,0xd5,0xc1,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, 0.5
+// GFX1250: v_clz_i32_u32_e64 v5, 0.5 ; encoding: [0x05,0x00,0xb9,0xd5,0xf0,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v5, src_scc
+// GFX1250: v_clz_i32_u32_e64 v5, src_scc ; encoding: [0x05,0x00,0xb9,0xd5,0xfd,0x00,0x00,0x00]
+
+v_clz_i32_u32_e64 v255, 0xaf123456
+// GFX1250: v_clz_i32_u32_e64 v255, 0xaf123456 ; encoding: [0xff,0x00,0xb9,0xd5,0xff,0x00,0x00,0x00,0x56,0x34,0x12,0xaf]
+
+v_cos_f16_e64 v5, v1
+// GFX1250: v_cos_f16_e64 v5, v1 ; encoding: [0x05,0x00,0xe1,0xd5,0x01,0x01,0x00,0x00]
+
+v_cos_f16_e64 v5, v255
+// GFX1250: v_cos_f16_e64 v5, v255 ; encoding: [0x05,0x00,0xe1,0xd5,0xff,0x01,0x00,0x00]
+
+v_cos_f16_e64 v5, s1
+// GFX1250: v_cos_f16_e64 v5, s1 ; encoding: [0x05,0x00,0xe1,0xd5,0x01,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, s105
+// GFX1250: v_cos_f16_e64 v5, s105 ; encoding: [0x05,0x00,0xe1,0xd5,0x69,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, vcc_lo
+// GFX1250: v_cos_f16_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xe1,0xd5,0x6a,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, vcc_hi
+// GFX1250: v_cos_f16_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xe1,0xd5,0x6b,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, ttmp15
+// GFX1250: v_cos_f16_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xe1,0xd5,0x7b,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, m0
+// GFX1250: v_cos_f16_e64 v5, m0 ; encoding: [0x05,0x00,0xe1,0xd5,0x7d,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, exec_lo
+// GFX1250: v_cos_f16_e64 v5, exec_lo ; encoding: [0x05,0x00,0xe1,0xd5,0x7e,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, exec_hi
+// GFX1250: v_cos_f16_e64 v5, exec_hi ; encoding: [0x05,0x00,0xe1,0xd5,0x7f,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, null
+// GFX1250: v_cos_f16_e64 v5, null ; encoding: [0x05,0x00,0xe1,0xd5,0x7c,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, -1
+// GFX1250: v_cos_f16_e64 v5, -1 ; encoding: [0x05,0x00,0xe1,0xd5,0xc1,0x00,0x00,0x00]
+
+v_cos_f16_e64 v5, 0.5 mul:2
+// GFX1250: v_cos_f16_e64 v5, 0.5 mul:2 ; encoding: [0x05,0x00,0xe1,0xd5,0xf0,0x00,0x00,0x08]
+
+v_cos_f16_e64 v5, src_scc mul:4
+// GFX1250: v_cos_f16_e64 v5, src_scc mul:4 ; encoding: [0x05,0x00,0xe1,0xd5,0xfd,0x00,0x00,0x10]
+
+v_cos_f16_e64 v255, -|0xfe0b| clamp div:2
+// GFX1250: v_cos_f16_e64 v255, -|0xfe0b| clamp div:2 ; encoding: [0xff,0x81,0xe1,0xd5,0xff,0x00,0x00,0x38,0x0b,0xfe,0x00,0x00]
+
+v_cos_f32_e64 v5, v1
+// GFX1250: v_cos_f32_e64 v5, v1 ; encoding: [0x05,0x00,0xb6,0xd5,0x01,0x01,0x00,0x00]
+
+v_cos_f32_e64 v5, v255
+// GFX1250: v_cos_f32_e64 v5, v255 ; encoding: [0x05,0x00,0xb6,0xd5,0xff,0x01,0x00,0x00]
+
+v_cos_f32_e64 v5, s1
+// GFX1250: v_cos_f32_e64 v5, s1 ; encoding: [0x05,0x00,0xb6,0xd5,0x01,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, s105
+// GFX1250: v_cos_f32_e64 v5, s105 ; encoding: [0x05,0x00,0xb6,0xd5,0x69,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, vcc_lo
+// GFX1250: v_cos_f32_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xb6,0xd5,0x6a,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, vcc_hi
+// GFX1250: v_cos_f32_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xb6,0xd5,0x6b,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, ttmp15
+// GFX1250: v_cos_f32_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xb6,0xd5,0x7b,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, m0
+// GFX1250: v_cos_f32_e64 v5, m0 ; encoding: [0x05,0x00,0xb6,0xd5,0x7d,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, exec_lo
+// GFX1250: v_cos_f32_e64 v5, exec_lo ; encoding: [0x05,0x00,0xb6,0xd5,0x7e,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, exec_hi
+// GFX1250: v_cos_f32_e64 v5, exec_hi ; encoding: [0x05,0x00,0xb6,0xd5,0x7f,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, null
+// GFX1250: v_cos_f32_e64 v5, null ; encoding: [0x05,0x00,0xb6,0xd5,0x7c,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, -1
+// GFX1250: v_cos_f32_e64 v5, -1 ; encoding: [0x05,0x00,0xb6,0xd5,0xc1,0x00,0x00,0x00]
+
+v_cos_f32_e64 v5, 0.5 mul:2
+// GFX1250: v_cos_f32_e64 v5, 0.5 mul:2 ; encoding: [0x05,0x00,0xb6,0xd5,0xf0,0x00,0x00,0x08]
+
+v_cos_f32_e64 v5, src_scc mul:4
+// GFX1250: v_cos_f32_e64 v5, src_scc mul:4 ; encoding: [0x05,0x00,0xb6,0xd5,0xfd,0x00,0x00,0x10]
+
+v_cos_f32_e64 v255, -|0xaf123456| clamp div:2
+// GFX1250: v_cos_f32_e64 v255, -|0xaf123456| clamp div:2 ; encoding: [0xff,0x81,0xb6,0xd5,0xff,0x00,0x00,0x38,0x56,0x34,0x12,0xaf]
+
+v_ctz_i32_b32_e64 v5, v1
+// GFX1250: v_ctz_i32_b32_e64 v5, v1 ; encoding: [0x05,0x00,0xba,0xd5,0x01,0x01,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, v255
+// GFX1250: v_ctz_i32_b32_e64 v5, v255 ; encoding: [0x05,0x00,0xba,0xd5,0xff,0x01,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, s1
+// GFX1250: v_ctz_i32_b32_e64 v5, s1 ; encoding: [0x05,0x00,0xba,0xd5,0x01,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, s105
+// GFX1250: v_ctz_i32_b32_e64 v5, s105 ; encoding: [0x05,0x00,0xba,0xd5,0x69,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, vcc_lo
+// GFX1250: v_ctz_i32_b32_e64 v5, vcc_lo ; encoding: [0x05,0x00,0xba,0xd5,0x6a,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, vcc_hi
+// GFX1250: v_ctz_i32_b32_e64 v5, vcc_hi ; encoding: [0x05,0x00,0xba,0xd5,0x6b,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, ttmp15
+// GFX1250: v_ctz_i32_b32_e64 v5, ttmp15 ; encoding: [0x05,0x00,0xba,0xd5,0x7b,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, m0
+// GFX1250: v_ctz_i32_b32_e64 v5, m0 ; encoding: [0x05,0x00,0xba,0xd5,0x7d,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, exec_lo
+// GFX1250: v_ctz_i32_b32_e64 v5, exec_lo ; encoding: [0x05,0x00,0xba,0xd5,0x7e,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, exec_hi
+// GFX1250: v_ctz_i32_b32_e64 v5, exec_hi ; encoding: [0x05,0x00,0xba,0xd5,0x7f,0x00,0x00,0x00]
+
+v_ctz_i32_b32_e64 v5, null
+// GFX1250: v_ctz_i32_b32_e64 v5, null ; encoding: [0x05,0x00,0xba,0xd5,0x7c,0x00,0x00,0x00]
+
+v_ctz_i32_b32...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/149567
More information about the llvm-commits
mailing list