Skip to content
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

AMDGPU: Add encodings for minimum3/maximum3 f32 for gfx950 #117600

Merged
merged 1 commit into from
Nov 26, 2024
Merged

AMDGPU: Add encodings for minimum3/maximum3 f32 for gfx950 #117600

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion llvm/lib/Target/AMDGPU/AMDGPU.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -431,7 +431,9 @@ def FeatureGFX950Insts : SubtargetFeature<"gfx950-insts",
FeatureBF8ConversionScaleInsts,
FeatureFP4ConversionScaleInsts,
FeatureFP6BF6ConversionScaleInsts,
FeatureF16BF16ToFP6BF6ConversionScaleInsts]
FeatureF16BF16ToFP6BF6ConversionScaleInsts,
FeatureMinimum3Maximum3F32
]
>;

def FeatureGFX10Insts : SubtargetFeature<"gfx10-insts",
Expand Down
3 changes: 3 additions & 0 deletions llvm/lib/Target/AMDGPU/VOP3Instructions.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1937,6 +1937,9 @@ defm V_CVT_PK_BF16_F32: VOP3OpSel_Real_gfx9 <0x268>;
defm V_CVT_SR_FP8_F32 : VOP3OpSel_Real_gfx9_forced_opsel2 <0x2a4>;
defm V_CVT_SR_BF8_F32 : VOP3OpSel_Real_gfx9_forced_opsel2 <0x2a5>;

defm V_MINIMUM3_F32 : VOP3_Real_vi <0x2a8>;
defm V_MAXIMUM3_F32 : VOP3_Real_vi <0x2a9>;

defm V_BITOP3_B16 : VOP3_Real_BITOP3_gfx9<0x233, "v_bitop3_b16">;
defm V_BITOP3_B32 : VOP3_Real_BITOP3_gfx9<0x234, "v_bitop3_b32">;
let OtherPredicates = [HasFP8ConversionScaleInsts] in {
Expand Down
33 changes: 33 additions & 0 deletions llvm/test/MC/AMDGPU/gfx950_asm_features.s
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1149,3 +1149,36 @@ buffer_atomic_pk_add_bf16 v5, off, s[8:11], 0.5 offset:4095
// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: buffer_atomic_pk_add_bf16 v5, off, s[8:11], -4.0 offset:4095 ; encoding: [0xff,0x0f,0x48,0xe1,0x00,0x05,0x02,0xf7]
buffer_atomic_pk_add_bf16 v5, off, s[8:11], -4.0 offset:4095


// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v1, v2, v3, v4 ; encoding: [0x01,0x00,0xa9,0xd2,0x02,0x07,0x12,0x04]
v_maximum3_f32 v1, v2, v3, v4

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v1, -v2, -v3, -v4 ; encoding: [0x01,0x00,0xa9,0xd2,0x02,0x07,0x12,0xe4]
v_maximum3_f32 v1, -v2, -v3, -v4

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v1, -|v2|, -|v3|, -|v4| ; encoding: [0x01,0x07,0xa9,0xd2,0x02,0x07,0x12,0xe4]
v_maximum3_f32 v1, -|v2|, -|v3|, -|v4|

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v1, 0, 1.0, v3 ; encoding: [0x01,0x00,0xa9,0xd2,0x80,0xe4,0x0d,0x04]
v_maximum3_f32 v1, 0.0, 1.0, v3

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v2, 0, v3, 1.0 ; encoding: [0x02,0x00,0xa9,0xd2,0x80,0x06,0xca,0x03]
v_maximum3_f32 v2, 0.0, v3, 1.0

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v1, s8, v3, 1.0 ; encoding: [0x01,0x00,0xa9,0xd2,0x08,0x06,0xca,0x03]
v_maximum3_f32 v1, s8, v3, 1.0

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_maximum3_f32 v1, v2, s8, v3 ; encoding: [0x01,0x00,0xa9,0xd2,0x02,0x11,0x0c,0x04]
v_maximum3_f32 v1, v2, s8, v3

// NOT-GFX950: :[[@LINE+2]]:{{[0-9]+}}: error:
// GFX950: v_minimum3_f32 v0, v1, v2, v3 ; encoding: [0x00,0x00,0xa8,0xd2,0x01,0x05,0x0e,0x04]
v_minimum3_f32 v0, v1, v2, v3
33 changes: 33 additions & 0 deletions llvm/test/MC/AMDGPU/gfx950_err.s
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -353,3 +353,36 @@ buffer_atomic_pk_add_bf16 v5, off, s[8:11], s3 offset:4095 dlc

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: invalid operand for instruction
buffer_atomic_pk_add_bf16 v5, off, s[8:11], s3 offset:4095 glc slc dlc

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: instruction not supported on this GPU
v_maximum3_f16 v0, v1, v2, v3

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: instruction not supported on this GPU
v_minimum3_f16 v0, v1, v2, v3

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: instruction not supported on this GPU
v_maximum_f16 v0, v1, v2

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: instruction not supported on this GPU
v_minimum_f16 v0, v1, v2

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: instruction not supported on this GPU
v_maximum_f32 v0, v1, v2

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: instruction not supported on this GPU
v_minimum_f32 v0, v1, v2

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: invalid operand (violates constant bus restrictions)
v_maximum3_f32 v0, s1, s2, v3

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: invalid operand (violates constant bus restrictions)
v_maximum3_f32 v0, v3, s1, s2

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: invalid operand (violates constant bus restrictions)
v_maximum3_f32 v0, s1, v3, s2

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: invalid operand (violates constant bus restrictions)
v_minimum3_f32 v0, s1, s2, v3

// GFX950: :[[@LINE+1]]:{{[0-9]+}}: error: literal operands are not supported
v_minimum3_f32 v0, v1, v2, 0xdeadbeef
24 changes: 24 additions & 0 deletions llvm/test/MC/Disassembler/AMDGPU/gfx950_dasm_vop3.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -857,3 +857,27 @@

# GFX950: v_cvt_scalef32_2xpk16_bf6_f32 v[20:25], v[10:25], v[10:25], 11 ; encoding: [0x14,0x00,0x53,0xd2,0x0a,0x15,0x2e,0x02]
0x14,0x00,0x53,0xd2,0x0a,0x15,0x2e,0x02

# GFX950: v_maximum3_f32 v1, -v2, -v3, -v4 ; encoding: [0x01,0x00,0xa9,0xd2,0x02,0x07,0x12,0xe4]
0x01,0x00,0xa9,0xd2,0x02,0x07,0x12,0xe4

# GFX950: v_maximum3_f32 v1, -|v2|, -|v3|, -|v4| ; encoding: [0x01,0x07,0xa9,0xd2,0x02,0x07,0x12,0xe4]
0x01,0x07,0xa9,0xd2,0x02,0x07,0x12,0xe4

# GFX950: v_maximum3_f32 v1, 0, 1.0, v3 ; encoding: [0x01,0x00,0xa9,0xd2,0x80,0xe4,0x0d,0x04]
0x01,0x00,0xa9,0xd2,0x80,0xe4,0x0d,0x04

# GFX950: v_maximum3_f32 v1, s8, v3, 1.0 ; encoding: [0x01,0x00,0xa9,0xd2,0x08,0x06,0xca,0x03]
0x01,0x00,0xa9,0xd2,0x08,0x06,0xca,0x03

# GFX950: v_maximum3_f32 v1, v2, s8, v3 ; encoding: [0x01,0x00,0xa9,0xd2,0x02,0x11,0x0c,0x04]
0x01,0x00,0xa9,0xd2,0x02,0x11,0x0c,0x04

# GFX950: v_maximum3_f32 v1, v2, v3, v4 ; encoding: [0x01,0x00,0xa9,0xd2,0x02,0x07,0x12,0x04]
0x01,0x00,0xa9,0xd2,0x02,0x07,0x12,0x04

# GFX950: v_maximum3_f32 v2, 0, v3, 1.0 ; encoding: [0x02,0x00,0xa9,0xd2,0x80,0x06,0xca,0x03]
0x02,0x00,0xa9,0xd2,0x80,0x06,0xca,0x03

# GFX950: v_minimum3_f32 v0, v1, v2, v3 ; encoding: [0x00,0x00,0xa8,0xd2,0x01,0x05,0x0e,0x04]
0x00,0x00,0xa8,0xd2,0x01,0x05,0x0e,0x04
Loading