Skip to content

AMDGPU: Support V_PK_MAXIMUM3_F16 and V_PK_MINIMUM3_F16 on gfx1250 #150307

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Jul 23, 2025
Merged

AMDGPU: Support V_PK_MAXIMUM3_F16 and V_PK_MINIMUM3_F16 on gfx1250 #150307

merged 1 commit into from
Jul 23, 2025

Conversation

@changpeng
Copy link
Contributor

@changpeng changpeng commented Jul 23, 2025

No description provided.

@llvmbot llvmbot added backend:AMDGPU llvm:mc Machine (object) code labels Jul 23, 2025
@llvmbot
Copy link
Member

llvmbot commented Jul 23, 2025

@llvm/pr-subscribers-backend-amdgpu

Author: Changpeng Fang (changpeng)

Changes

Patch is 24.44 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/150307.diff

5 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/VOP3PInstructions.td (+2)
  • (added) llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll (+99)
  • (added) llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll (+99)
  • (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s (+144)
  • (modified) llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt (+90)
diff --git a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
index 43f9bc946e6fe..23dfa4b07a0e4 100644
--- a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
@@ -2235,6 +2235,8 @@ defm V_PK_ADD_BF16     : VOP3P_Real_gfx1250<0x23>;
 defm V_PK_MUL_BF16     : VOP3P_Real_gfx1250<0x2a>;
 defm V_PK_MIN_NUM_BF16 : VOP3P_Real_gfx1250<0x2b>;
 defm V_PK_MAX_NUM_BF16 : VOP3P_Real_gfx1250<0x2c>;
+defm V_PK_MINIMUM3_F16 : VOP3P_Real_gfx1250<0x36>;
+defm V_PK_MAXIMUM3_F16 : VOP3P_Real_gfx1250<0x37>;
 
 defm V_PK_MINIMUM_F16 : VOP3P_Real_gfx12<0x1d>;
 defm V_PK_MAXIMUM_F16 : VOP3P_Real_gfx12<0x1e>;
diff --git a/llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll b/llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll
new file mode 100644
index 0000000000000..852c9cf8face8
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll
@@ -0,0 +1,99 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-SDAG %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-GISEL %s
+
+define <2 x half> @fmaximum3_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v2f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, v2, v0, v1
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v2f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, v1, v1
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v2, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.maximum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.maximum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <2 x half> @fmaximum3_v2f16_vss(<2 x half> %a, <2 x half> inreg %b, <2 x half> inreg %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v2f16_vss:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, s1, v0, s0
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v2f16_vss:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, s0, s0
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, s1, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.maximum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.maximum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <3 x half> @fmaximum3_v3f16(<3 x half> %a, <3 x half> %b, <3 x half> %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v3f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v3f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_maximum_f16 v1, v1, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_maximum_f16 v1, v5, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <3 x half> @llvm.maximum.v3f16(<3 x half> %a, <3 x half> %b)
+  %res = call <3 x half> @llvm.maximum.v3f16(<3 x half> %c, <3 x half> %min)
+  ret <3 x half> %res
+}
+
+define <4 x half> @fmaximum3_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v4f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v4f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v1, v1, v3, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v1, v5, v1, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <4 x half> @llvm.maximum.v4f16(<4 x half> %a, <4 x half> %b)
+  %res = call <4 x half> @llvm.maximum.v4f16(<4 x half> %c, <4 x half> %min)
+  ret <4 x half> %res
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX1250: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll b/llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll
new file mode 100644
index 0000000000000..df9fb10badb9a
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll
@@ -0,0 +1,99 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-SDAG %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-GISEL %s
+
+define <2 x half> @fminimum3_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v2f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, v2, v0, v1
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v2f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, v1, v1
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v2, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.minimum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.minimum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <2 x half> @fminimum3_v2f16_vss(<2 x half> %a, <2 x half> inreg %b, <2 x half> inreg %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v2f16_vss:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, s1, v0, s0
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v2f16_vss:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, s0, s0
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, s1, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.minimum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.minimum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <3 x half> @fminimum3_v3f16(<3 x half> %a, <3 x half> %b, <3 x half> %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v3f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v3f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_minimum_f16 v1, v1, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_minimum_f16 v1, v5, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <3 x half> @llvm.minimum.v3f16(<3 x half> %a, <3 x half> %b)
+  %res = call <3 x half> @llvm.minimum.v3f16(<3 x half> %c, <3 x half> %min)
+  ret <3 x half> %res
+}
+
+define <4 x half> @fminimum3_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v4f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v4f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v1, v1, v3, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v1, v5, v1, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <4 x half> @llvm.minimum.v4f16(<4 x half> %a, <4 x half> %b)
+  %res = call <4 x half> @llvm.minimum.v4f16(<4 x half> %c, <4 x half> %min)
+  ret <4 x half> %res
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX1250: {{.*}}
diff --git a/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s b/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s
index 4b1a6e655b4db..0710344e3f057 100644
--- a/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s
+++ b/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s
@@ -1025,3 +1025,147 @@ v_pk_fma_bf16 v5, src_scc, vcc_lo, -1 op_sel:[0,1,0] op_sel_hi:[1,0,1] neg_lo:[0
 v_pk_fma_bf16 v255, 0xfe0b, vcc_hi, null op_sel:[0,0,1] op_sel_hi:[1,1,0] neg_lo:[1,1,1] neg_hi:[1,1,1] clamp
 // GFX1250: v_pk_fma_bf16 v255, 0xfe0b, vcc_hi, null op_sel:[0,0,1] op_sel_hi:[1,1,0] neg_lo:[1,1,1] neg_hi:[1,1,1] clamp ; encoding: [0xff,0xa7,0x11,0xcc,0xff,0xd6,0xf0,0xf9,0x0b,0xfe,0x00,0x00]
 // GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0] ; encoding: [0x08,0x00,0x36,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1] ; encoding: [0x08,0x60,0x36,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x36,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x3c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x5c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x9c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0] ; encoding: [0x08,0x41,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0] ; encoding: [0x08,0x42,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1] ; encoding: [0x08,0x44,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 clamp
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 clamp  ; encoding: [0x08,0xc0,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v1, v4, v9, v16
+// GFX1250: v_pk_minimum3_f16 v1, v4, v9, v16       ; encoding: [0x01,0x40,0x36,0xcc,0x04,0x13,0x42,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v1, v2, v5, 1.0
+// GFX1250: v_pk_minimum3_f16 v1, v2, v5, 1.0       ; encoding: [0x01,0x40,0x36,0xcc,0x02,0x0b,0xca,0x1b]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0] ; encoding: [0x08,0x00,0x37,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1] ; encoding: [0x08,0x60,0x37,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x37,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x3c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x5c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x9c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0] ; encoding: [0x08,0x41,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0] ; encoding: [0x08,0x42,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1] ; encoding: [0x08,0x44,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 clamp
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 clamp  ; encoding: [0x08,0xc0,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v1, v4, v9, v16
+// GFX1250: v_pk_maximum3_f16 v1, v4, v9, v16       ; encoding: [0x01,0x40,0x37,0xcc,0x04,0x13,0x42,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v1, v2, v5, 1.0
+// GFX1250: v_pk_maximum3_f16 v1, v2, v5, 1.0       ; encoding: [0x01,0x40,0x37,0xcc,0x02,0x0b,0xca,0x1b]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
diff --git a/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt b/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt
index 921db7e22a598..64106de88260f 100644
--- a/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt
+++ b/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt
@@ -725,3 +725,93 @@
 
 # GFX1250: v_pk_fma_bf16 v5, vcc_lo, ttmp15, v3    ; encoding: [0x05,0x40,0x11,0xcc,0x6a,0xf6,0x0c,0x1c]
 0x05,0x40,0x11,0xcc,0x6a,0xf6,0x0c,0x1c
+
+# GFX1250: v_pk_minimum3_f16 v1, v2, v5, 1.0       ; encoding: [0x01,0x40,0x36,0xcc,0x02,0x0b,0xca,0x1b]
+0x01,0x40,0x36,0xcc,0x02,0x0b,0xca,0x1b
+
+# GFX1250: v_pk_min...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Jul 23, 2025

@llvm/pr-subscribers-mc

Author: Changpeng Fang (changpeng)

Changes

Patch is 24.44 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/150307.diff

5 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/VOP3PInstructions.td (+2)
  • (added) llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll (+99)
  • (added) llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll (+99)
  • (modified) llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s (+144)
  • (modified) llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt (+90)
diff --git a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
index 43f9bc946e6fe..23dfa4b07a0e4 100644
--- a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
@@ -2235,6 +2235,8 @@ defm V_PK_ADD_BF16     : VOP3P_Real_gfx1250<0x23>;
 defm V_PK_MUL_BF16     : VOP3P_Real_gfx1250<0x2a>;
 defm V_PK_MIN_NUM_BF16 : VOP3P_Real_gfx1250<0x2b>;
 defm V_PK_MAX_NUM_BF16 : VOP3P_Real_gfx1250<0x2c>;
+defm V_PK_MINIMUM3_F16 : VOP3P_Real_gfx1250<0x36>;
+defm V_PK_MAXIMUM3_F16 : VOP3P_Real_gfx1250<0x37>;
 
 defm V_PK_MINIMUM_F16 : VOP3P_Real_gfx12<0x1d>;
 defm V_PK_MAXIMUM_F16 : VOP3P_Real_gfx12<0x1e>;
diff --git a/llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll b/llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll
new file mode 100644
index 0000000000000..852c9cf8face8
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fmaximum3.v2f16.ll
@@ -0,0 +1,99 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-SDAG %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-GISEL %s
+
+define <2 x half> @fmaximum3_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v2f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, v2, v0, v1
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v2f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, v1, v1
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v2, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.maximum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.maximum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <2 x half> @fmaximum3_v2f16_vss(<2 x half> %a, <2 x half> inreg %b, <2 x half> inreg %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v2f16_vss:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, s1, v0, s0
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v2f16_vss:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, s0, s0
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, s1, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.maximum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.maximum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <3 x half> @fmaximum3_v3f16(<3 x half> %a, <3 x half> %b, <3 x half> %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v3f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v3f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_maximum_f16 v1, v1, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_maximum_f16 v1, v5, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <3 x half> @llvm.maximum.v3f16(<3 x half> %a, <3 x half> %b)
+  %res = call <3 x half> @llvm.maximum.v3f16(<3 x half> %c, <3 x half> %min)
+  ret <3 x half> %res
+}
+
+define <4 x half> @fmaximum3_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
+; GFX1250-SDAG-LABEL: fmaximum3_v4f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_maximum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fmaximum3_v4f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v1, v1, v3, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_pk_maximum3_f16 v1, v5, v1, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <4 x half> @llvm.maximum.v4f16(<4 x half> %a, <4 x half> %b)
+  %res = call <4 x half> @llvm.maximum.v4f16(<4 x half> %c, <4 x half> %min)
+  ret <4 x half> %res
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX1250: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll b/llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll
new file mode 100644
index 0000000000000..df9fb10badb9a
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fminimum3.v2f16.ll
@@ -0,0 +1,99 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-SDAG %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250,GFX1250-GISEL %s
+
+define <2 x half> @fminimum3_v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v2f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, v2, v0, v1
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v2f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, v1, v1
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v2, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.minimum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.minimum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <2 x half> @fminimum3_v2f16_vss(<2 x half> %a, <2 x half> inreg %b, <2 x half> inreg %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v2f16_vss:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, s1, v0, s0
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v2f16_vss:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, s0, s0
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, s1, v0, v0
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <2 x half> @llvm.minimum.v2f16(<2 x half> %a, <2 x half> %b)
+  %res = call <2 x half> @llvm.minimum.v2f16(<2 x half> %c, <2 x half> %min)
+  ret <2 x half> %res
+}
+
+define <3 x half> @fminimum3_v3f16(<3 x half> %a, <3 x half> %b, <3 x half> %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v3f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v3f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_minimum_f16 v1, v1, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_minimum_f16 v1, v5, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <3 x half> @llvm.minimum.v3f16(<3 x half> %a, <3 x half> %b)
+  %res = call <3 x half> @llvm.minimum.v3f16(<3 x half> %c, <3 x half> %min)
+  ret <3 x half> %res
+}
+
+define <4 x half> @fminimum3_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
+; GFX1250-SDAG-LABEL: fminimum3_v4f16:
+; GFX1250-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v2
+; GFX1250-SDAG-NEXT:    v_pk_minimum3_f16 v1, v5, v1, v3
+; GFX1250-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GFX1250-GISEL-LABEL: fminimum3_v4f16:
+; GFX1250-GISEL:       ; %bb.0: ; %entry
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v0, v2, v2
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v1, v1, v3, v3
+; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v0, v4, v0, v0
+; GFX1250-GISEL-NEXT:    v_pk_minimum3_f16 v1, v5, v1, v1
+; GFX1250-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+entry:
+  %min = call <4 x half> @llvm.minimum.v4f16(<4 x half> %a, <4 x half> %b)
+  %res = call <4 x half> @llvm.minimum.v4f16(<4 x half> %c, <4 x half> %min)
+  ret <4 x half> %res
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX1250: {{.*}}
diff --git a/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s b/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s
index 4b1a6e655b4db..0710344e3f057 100644
--- a/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s
+++ b/llvm/test/MC/AMDGPU/gfx1250_asm_vop3p.s
@@ -1025,3 +1025,147 @@ v_pk_fma_bf16 v5, src_scc, vcc_lo, -1 op_sel:[0,1,0] op_sel_hi:[1,0,1] neg_lo:[0
 v_pk_fma_bf16 v255, 0xfe0b, vcc_hi, null op_sel:[0,0,1] op_sel_hi:[1,1,0] neg_lo:[1,1,1] neg_hi:[1,1,1] clamp
 // GFX1250: v_pk_fma_bf16 v255, 0xfe0b, vcc_hi, null op_sel:[0,0,1] op_sel_hi:[1,1,0] neg_lo:[1,1,1] neg_hi:[1,1,1] clamp ; encoding: [0xff,0xa7,0x11,0xcc,0xff,0xd6,0xf0,0xf9,0x0b,0xfe,0x00,0x00]
 // GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0] ; encoding: [0x08,0x00,0x36,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1] ; encoding: [0x08,0x60,0x36,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x36,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x3c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x5c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1] ; encoding: [0x08,0x40,0x36,0xcc,0x01,0x03,0x10,0x9c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0] ; encoding: [0x08,0x41,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0] ; encoding: [0x08,0x42,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1]
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1] ; encoding: [0x08,0x44,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v8, v1, s1, v4 clamp
+// GFX1250: v_pk_minimum3_f16 v8, v1, s1, v4 clamp  ; encoding: [0x08,0xc0,0x36,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v1, v4, v9, v16
+// GFX1250: v_pk_minimum3_f16 v1, v4, v9, v16       ; encoding: [0x01,0x40,0x36,0xcc,0x04,0x13,0x42,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_minimum3_f16 v1, v2, v5, 1.0
+// GFX1250: v_pk_minimum3_f16 v1, v2, v5, 1.0       ; encoding: [0x01,0x40,0x36,0xcc,0x02,0x0b,0xca,0x1b]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4        ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 op_sel_hi:[0,0,0] ; encoding: [0x08,0x00,0x37,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 op_sel:[0,0,1] op_sel_hi:[0,0,1] ; encoding: [0x08,0x60,0x37,0xcc,0x01,0x03,0x10,0x04]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,1,1] neg_hi:[1,1,1] ; encoding: [0x08,0x47,0x37,0xcc,0x01,0x03,0x10,0xfc]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[1,0,0] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x3c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,1,0] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x5c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_lo:[0,0,1] ; encoding: [0x08,0x40,0x37,0xcc,0x01,0x03,0x10,0x9c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[1,0,0] ; encoding: [0x08,0x41,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,1,0] ; encoding: [0x08,0x42,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1]
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 neg_hi:[0,0,1] ; encoding: [0x08,0x44,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v8, v1, s1, v4 clamp
+// GFX1250: v_pk_maximum3_f16 v8, v1, s1, v4 clamp  ; encoding: [0x08,0xc0,0x37,0xcc,0x01,0x03,0x10,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v1, v4, v9, v16
+// GFX1250: v_pk_maximum3_f16 v1, v4, v9, v16       ; encoding: [0x01,0x40,0x37,0xcc,0x04,0x13,0x42,0x1c]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
+
+v_pk_maximum3_f16 v1, v2, v5, 1.0
+// GFX1250: v_pk_maximum3_f16 v1, v2, v5, 1.0       ; encoding: [0x01,0x40,0x37,0xcc,0x02,0x0b,0xca,0x1b]
+// GFX12-ERR: :[[@LINE-2]]:1: error: instruction not supported on this GPU
diff --git a/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt b/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt
index 921db7e22a598..64106de88260f 100644
--- a/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt
+++ b/llvm/test/MC/Disassembler/AMDGPU/gfx1250_dasm_vop3p.txt
@@ -725,3 +725,93 @@
 
 # GFX1250: v_pk_fma_bf16 v5, vcc_lo, ttmp15, v3    ; encoding: [0x05,0x40,0x11,0xcc,0x6a,0xf6,0x0c,0x1c]
 0x05,0x40,0x11,0xcc,0x6a,0xf6,0x0c,0x1c
+
+# GFX1250: v_pk_minimum3_f16 v1, v2, v5, 1.0       ; encoding: [0x01,0x40,0x36,0xcc,0x02,0x0b,0xca,0x1b]
+0x01,0x40,0x36,0xcc,0x02,0x0b,0xca,0x1b
+
+# GFX1250: v_pk_min...
[truncated]

@changpeng changpeng requested review from rampitec and shiltian July 23, 2025 20:12
@changpeng changpeng merged commit 203ea0a into llvm:main Jul 23, 2025
12 checks passed
@changpeng changpeng deleted the minimum branch July 23, 2025 20:48
mahesh-attarde pushed a commit to mahesh-attarde/llvm-project that referenced this pull request Jul 28, 2025
@changpeng @rampitec @mahesh-attarde
AMDGPU: Support V_PK_MAXIMUM3_F16 and V_PK_MINIMUM3_F16 on gfx1250 (l…
e3ae8ff 
…lvm#150307)

Co-authored-by: Stanislav Mekhanoshin <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@rampitec rampitec rampitec approved these changes

@shiltian shiltian Awaiting requested review from shiltian

Assignees

No one assigned

Labels

backend:AMDGPU llvm:mc Machine (object) code

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@changpeng @llvmbot @rampitec