[llvm] [AMDGPU] Add new GFX12 image atomic float instructions (PR #76946)
Mirko BrkuĊĦanin via llvm-commits
llvm-commits at lists.llvm.org
Thu Jan 11 04:30:42 PST 2024
https://github.com/mbrkusanin updated https://github.com/llvm/llvm-project/pull/76946
>From b3442926bd31ca7ff59c4a24edccaa26960f2274 Mon Sep 17 00:00:00 2001
From: Mirko Brkusanin <Mirko.Brkusanin at amd.com>
Date: Thu, 4 Jan 2024 15:43:30 +0100
Subject: [PATCH 1/2] [AMDGPU] Add new GFX12 image atomic float instructions
---
llvm/include/llvm/IR/IntrinsicsAMDGPU.td | 3 +
llvm/lib/Target/AMDGPU/MIMGInstructions.td | 3 +
.../AMDGPU/llvm.amdgcn.image.atomic.flt.ll | 40 +++++++++
llvm/test/MC/AMDGPU/gfx12_asm_vimage.s | 81 +++++++++++++++++++
llvm/test/MC/AMDGPU/gfx12_asm_vimage_alias.s | 6 ++
.../Disassembler/AMDGPU/gfx12_dasm_vimage.txt | 81 +++++++++++++++++++
6 files changed, 214 insertions(+)
create mode 100644 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index e5596258847f9f..255f5106e543fb 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -1025,6 +1025,9 @@ defset list<AMDGPUImageDimIntrinsic> AMDGPUImageDimAtomicIntrinsics = {
defm int_amdgcn_image_atomic_xor : AMDGPUImageDimAtomic<"ATOMIC_XOR">;
defm int_amdgcn_image_atomic_inc : AMDGPUImageDimAtomic<"ATOMIC_INC">;
defm int_amdgcn_image_atomic_dec : AMDGPUImageDimAtomic<"ATOMIC_DEC">;
+ defm int_amdgcn_image_atomic_add_flt : AMDGPUImageDimFloatAtomic<"ATOMIC_ADD_FLT">;
+ defm int_amdgcn_image_atomic_min_flt : AMDGPUImageDimFloatAtomic<"ATOMIC_MIN_FLT">;
+ defm int_amdgcn_image_atomic_max_flt : AMDGPUImageDimFloatAtomic<"ATOMIC_MAX_FLT">;
defm int_amdgcn_image_atomic_cmpswap :
AMDGPUImageDimAtomicX<"ATOMIC_CMPSWAP", [AMDGPUArg<LLVMMatchType<0>, "src">,
diff --git a/llvm/lib/Target/AMDGPU/MIMGInstructions.td b/llvm/lib/Target/AMDGPU/MIMGInstructions.td
index 240366c8e7daae..bb263146fdec32 100644
--- a/llvm/lib/Target/AMDGPU/MIMGInstructions.td
+++ b/llvm/lib/Target/AMDGPU/MIMGInstructions.td
@@ -1553,6 +1553,9 @@ defm IMAGE_ATOMIC_DEC : MIMG_Atomic_Renamed <mimgopc<0x16, 0x16, 0x1c>
defm IMAGE_ATOMIC_FCMPSWAP : MIMG_Atomic <mimgopc<MIMG.NOP, MIMG.NOP, 0x1d, MIMG.NOP>, "image_atomic_fcmpswap", 1, 1>;
defm IMAGE_ATOMIC_FMIN : MIMG_Atomic <mimgopc<MIMG.NOP, MIMG.NOP, 0x1e, MIMG.NOP>, "image_atomic_fmin", 0, 1>;
defm IMAGE_ATOMIC_FMAX : MIMG_Atomic <mimgopc<MIMG.NOP, MIMG.NOP, 0x1f, MIMG.NOP>, "image_atomic_fmax", 0, 1>;
+defm IMAGE_ATOMIC_ADD_FLT : MIMG_Atomic <mimgopc<0x83, MIMG.NOP, MIMG.NOP, MIMG.NOP>, "image_atomic_add_flt", 0, 1>;
+defm IMAGE_ATOMIC_MIN_FLT : MIMG_Atomic <mimgopc<0x84, MIMG.NOP, MIMG.NOP, MIMG.NOP>, "image_atomic_min_num_flt", 0, 1, "image_atomic_min_flt">;
+defm IMAGE_ATOMIC_MAX_FLT : MIMG_Atomic <mimgopc<0x85, MIMG.NOP, MIMG.NOP, MIMG.NOP>, "image_atomic_max_num_flt", 0, 1, "image_atomic_max_flt">;
defm IMAGE_SAMPLE : MIMG_Sampler_WQM <mimgopc<0x1b, 0x1b, 0x20>, AMDGPUSample>;
let OtherPredicates = [HasExtendedImageInsts] in {
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll
new file mode 100644
index 00000000000000..0667beceee1744
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll
@@ -0,0 +1,40 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX12 %s
+; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX12 %s
+
+define amdgpu_ps float @atomic_min_flt_1d(<8 x i32> inreg %rsrc, float %data, i32 %s) {
+; GFX12-LABEL: atomic_min_flt_1d:
+; GFX12: ; %bb.0: ; %main_body
+; GFX12-NEXT: image_atomic_min_flt v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_waitcnt vmcnt(0)
+; GFX12-NEXT: ; return to shader part epilog
+main_body:
+ %v = call float @llvm.amdgcn.image.atomic.min.flt.1d.f32.f32(float %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
+ ret float %v
+}
+
+define amdgpu_ps float @atomic_max_flt_1d(<8 x i32> inreg %rsrc, float %data, i32 %s) {
+; GFX12-LABEL: atomic_max_flt_1d:
+; GFX12: ; %bb.0: ; %main_body
+; GFX12-NEXT: image_atomic_max_flt v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_waitcnt vmcnt(0)
+; GFX12-NEXT: ; return to shader part epilog
+main_body:
+ %v = call float @llvm.amdgcn.image.atomic.max.flt.1d.f32.f32(float %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
+ ret float %v
+}
+
+define amdgpu_ps float @atomic_add_flt_1d(<8 x i32> inreg %rsrc, float %data, i32 %s) {
+; GFX12-LABEL: atomic_add_flt_1d:
+; GFX12: ; %bb.0: ; %main_body
+; GFX12-NEXT: image_atomic_add_flt v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_waitcnt vmcnt(0)
+; GFX12-NEXT: ; return to shader part epilog
+main_body:
+ %v = call float @llvm.amdgcn.image.atomic.add.flt.1d.f32.f32(float %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
+ ret float %v
+}
+
+declare float @llvm.amdgcn.image.atomic.add.flt.1d.f32.f32(float, i32, <8 x i32>, i32, i32)
+declare float @llvm.amdgcn.image.atomic.min.flt.1d.f32.f32(float, i32, <8 x i32>, i32, i32)
+declare float @llvm.amdgcn.image.atomic.max.flt.1d.f32.f32(float, i32, <8 x i32>, i32, i32)
diff --git a/llvm/test/MC/AMDGPU/gfx12_asm_vimage.s b/llvm/test/MC/AMDGPU/gfx12_asm_vimage.s
index 72bc164c5e9bdc..2d935b9523592e 100644
--- a/llvm/test/MC/AMDGPU/gfx12_asm_vimage.s
+++ b/llvm/test/MC/AMDGPU/gfx12_asm_vimage.s
@@ -883,6 +883,87 @@ image_atomic_dec_uint v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_R
image_atomic_dec_uint v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT
// GFX12: encoding: [0x00,0x80,0x45,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+image_atomic_add_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
+// GFX12: encoding: [0x00,0xc0,0x60,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_add_flt v1, [v2, v3], s[4:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+// GFX12: encoding: [0x01,0xc0,0x60,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00]
+
+image_atomic_add_flt v4, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_3D
+// GFX12: encoding: [0x02,0xc0,0x60,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_add_flt v255, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_CUBE
+// GFX12: encoding: [0x03,0xc0,0x60,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_add_flt v[0:1], [v4, v5], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY
+// GFX12: encoding: [0x04,0xc0,0xe0,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00]
+
+image_atomic_add_flt v[1:2], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY
+// GFX12: encoding: [0x05,0xc0,0xe0,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_add_flt v[3:4], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA
+// GFX12: encoding: [0x06,0xc0,0xe0,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_add_flt v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY
+// GFX12: encoding: [0x07,0xc0,0xe0,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07]
+
+image_atomic_add_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT
+// GFX12: encoding: [0x00,0xc0,0x60,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_min_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
+// GFX12: encoding: [0x00,0x00,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_min_flt v1, [v2, v3], s[4:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+// GFX12: encoding: [0x01,0x00,0x61,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00]
+
+image_atomic_min_flt v4, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_3D
+// GFX12: encoding: [0x02,0x00,0x61,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_min_flt v255, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_CUBE
+// GFX12: encoding: [0x03,0x00,0x61,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_min_flt v[0:1], [v4, v5], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY
+// GFX12: encoding: [0x04,0x00,0xe1,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00]
+
+image_atomic_min_flt v[1:2], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY
+// GFX12: encoding: [0x05,0x00,0xe1,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_min_flt v[3:4], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA
+// GFX12: encoding: [0x06,0x00,0xe1,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_min_flt v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY
+// GFX12: encoding: [0x07,0x00,0xe1,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07]
+
+image_atomic_min_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT
+// GFX12: encoding: [0x00,0x00,0x61,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_max_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
+// GFX12: encoding: [0x00,0x40,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_max_flt v1, [v2, v3], s[4:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+// GFX12: encoding: [0x01,0x40,0x61,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00]
+
+image_atomic_max_flt v4, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_3D
+// GFX12: encoding: [0x02,0x40,0x61,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_max_flt v255, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_CUBE
+// GFX12: encoding: [0x03,0x40,0x61,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_max_flt v[0:1], [v4, v5], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY
+// GFX12: encoding: [0x04,0x40,0xe1,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00]
+
+image_atomic_max_flt v[1:2], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY
+// GFX12: encoding: [0x05,0x40,0xe1,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_max_flt v[3:4], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA
+// GFX12: encoding: [0x06,0x40,0xe1,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+
+image_atomic_max_flt v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY
+// GFX12: encoding: [0x07,0x40,0xe1,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07]
+
+image_atomic_max_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT
+// GFX12: encoding: [0x00,0x40,0x61,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+
image_bvh_intersect_ray v[4:7], [v9, v10, v[11:13], v[14:16], v[17:19]], s[4:7]
// GFX12: encoding: [0x10,0x40,0xc6,0xd3,0x04,0x08,0x00,0x11,0x09,0x0a,0x0b,0x0e]
diff --git a/llvm/test/MC/AMDGPU/gfx12_asm_vimage_alias.s b/llvm/test/MC/AMDGPU/gfx12_asm_vimage_alias.s
index bf70545ff23c6c..43c462c9b6d4f6 100644
--- a/llvm/test/MC/AMDGPU/gfx12_asm_vimage_alias.s
+++ b/llvm/test/MC/AMDGPU/gfx12_asm_vimage_alias.s
@@ -23,3 +23,9 @@ image_atomic_inc v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
image_atomic_dec v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
// GFX12: image_atomic_dec_uint v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x80,0x45,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_min_num_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
+// GFX12: image_atomic_min_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x00,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+
+image_atomic_max_num_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D
+// GFX12: image_atomic_max_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x40,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
diff --git a/llvm/test/MC/Disassembler/AMDGPU/gfx12_dasm_vimage.txt b/llvm/test/MC/Disassembler/AMDGPU/gfx12_dasm_vimage.txt
index eff2e09e31b1fe..e509e5d2168e92 100644
--- a/llvm/test/MC/Disassembler/AMDGPU/gfx12_dasm_vimage.txt
+++ b/llvm/test/MC/Disassembler/AMDGPU/gfx12_dasm_vimage.txt
@@ -883,6 +883,87 @@
# GFX12: image_atomic_dec_uint v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT ; encoding: [0x00,0x80,0x45,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
0x00,0x80,0x45,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00
+# GFX12: image_atomic_add_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0xc0,0x60,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+0x00,0xc0,0x60,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
+
+# GFX12: image_atomic_add_flt v1, [v2, v3], s[4:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ; encoding: [0x01,0xc0,0x60,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00]
+0x01,0xc0,0x60,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00
+
+# GFX12: image_atomic_add_flt v4, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_3D ; encoding: [0x02,0xc0,0x60,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x02,0xc0,0x60,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_add_flt v255, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_CUBE ; encoding: [0x03,0xc0,0x60,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x03,0xc0,0x60,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_add_flt v[0:1], [v4, v5], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY ; encoding: [0x04,0xc0,0xe0,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00]
+0x04,0xc0,0xe0,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00
+
+# GFX12: image_atomic_add_flt v[1:2], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x05,0xc0,0xe0,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x05,0xc0,0xe0,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_add_flt v[3:4], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA ; encoding: [0x06,0xc0,0xe0,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x06,0xc0,0xe0,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_add_flt v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY ; encoding: [0x07,0xc0,0xe0,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07]
+0x07,0xc0,0xe0,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07
+
+# GFX12: image_atomic_add_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT ; encoding: [0x00,0xc0,0x60,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+0x00,0xc0,0x60,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00
+
+# GFX12: image_atomic_min_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x00,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+0x00,0x00,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
+
+# GFX12: image_atomic_min_flt v1, [v2, v3], s[4:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ; encoding: [0x01,0x00,0x61,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00]
+0x01,0x00,0x61,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00
+
+# GFX12: image_atomic_min_flt v4, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_3D ; encoding: [0x02,0x00,0x61,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x02,0x00,0x61,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_min_flt v255, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_CUBE ; encoding: [0x03,0x00,0x61,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x03,0x00,0x61,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_min_flt v[0:1], [v4, v5], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY ; encoding: [0x04,0x00,0xe1,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00]
+0x04,0x00,0xe1,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00
+
+# GFX12: image_atomic_min_flt v[1:2], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x05,0x00,0xe1,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x05,0x00,0xe1,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_min_flt v[3:4], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA ; encoding: [0x06,0x00,0xe1,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x06,0x00,0xe1,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_min_flt v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY ; encoding: [0x07,0x00,0xe1,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07]
+0x07,0x00,0xe1,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07
+
+# GFX12: image_atomic_min_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT ; encoding: [0x00,0x00,0x61,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+0x00,0x00,0x61,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00
+
+# GFX12: image_atomic_max_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x40,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00]
+0x00,0x40,0x61,0xd0,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
+
+# GFX12: image_atomic_max_flt v1, [v2, v3], s[4:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ; encoding: [0x01,0x40,0x61,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00]
+0x01,0x40,0x61,0xd0,0x01,0x08,0x00,0x00,0x02,0x03,0x00,0x00
+
+# GFX12: image_atomic_max_flt v4, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_3D ; encoding: [0x02,0x40,0x61,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x02,0x40,0x61,0xd0,0x04,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_max_flt v255, [v4, v5, v6], s[8:15] dmask:0x1 dim:SQ_RSRC_IMG_CUBE ; encoding: [0x03,0x40,0x61,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x03,0x40,0x61,0xd0,0xff,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_max_flt v[0:1], [v4, v5], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY ; encoding: [0x04,0x40,0xe1,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00]
+0x04,0x40,0xe1,0xd0,0x00,0x10,0x00,0x00,0x04,0x05,0x00,0x00
+
+# GFX12: image_atomic_max_flt v[1:2], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x05,0x40,0xe1,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x05,0x40,0xe1,0xd0,0x01,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_max_flt v[3:4], [v4, v5, v6], s[8:15] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA ; encoding: [0x06,0x40,0xe1,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00]
+0x06,0x40,0xe1,0xd0,0x03,0x10,0x00,0x00,0x04,0x05,0x06,0x00
+
+# GFX12: image_atomic_max_flt v[254:255], [v4, v5, v6, v7], s[96:103] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY ; encoding: [0x07,0x40,0xe1,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07]
+0x07,0x40,0xe1,0xd0,0xfe,0xc0,0x00,0x00,0x04,0x05,0x06,0x07
+
+# GFX12: image_atomic_max_flt v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_NT ; encoding: [0x00,0x40,0x61,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00]
+0x00,0x40,0x61,0xd0,0x00,0x00,0x20,0x00,0x00,0x00,0x00,0x00
+
# GFX12: image_bvh_intersect_ray v[4:7], [v9, v10, v[11:13], v[14:16], v[17:19]], s[4:7] ; encoding: [0x10,0x40,0xc6,0xd3,0x04,0x08,0x00,0x11,0x09,0x0a,0x0b,0x0e]
0x10,0x40,0xc6,0xd3,0x04,0x08,0x00,0x11,0x09,0x0a,0x0b,0x0e
>From e221d7344684b4b481f5ef8c21af09b28c5addbb Mon Sep 17 00:00:00 2001
From: Mirko Brkusanin <Mirko.Brkusanin at amd.com>
Date: Thu, 11 Jan 2024 13:20:45 +0100
Subject: [PATCH 2/2] test updates
---
llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll
index 0667beceee1744..31f3fb5a66a331 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.flt.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -march=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX12 %s
-; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX12 %s
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefix=GFX12 %s
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefix=GFX12 %s
define amdgpu_ps float @atomic_min_flt_1d(<8 x i32> inreg %rsrc, float %data, i32 %s) {
; GFX12-LABEL: atomic_min_flt_1d:
More information about the llvm-commits
mailing list