[llvm] [AMDGPU][MC][True16] VOP3dot instruction update for true16/fake16 (PR #113474)

Thu Dec 5 09:51:01 PST 2024

================
@@ -4700,44 +4700,132 @@ v_pack_b32_f16_e64_dpp v5, -v1, |v2| op_sel:[1,0,0] row_xmask:0 row_mask:0x1 ban
 v_pack_b32_f16_e64_dpp v255, -|v255|, -|v255| op_sel:[0,1,0] row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:1 fi:1
 // GFX11: [0xff,0x13,0x11,0xd7,0xfa,0xfe,0x03,0x60,0xff,0x6f,0x0d,0x30]
 
-v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+v_dot2_f16_f16_e64_dpp v0.l, v1, v2, v3.l quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
 // GFX11: encoding: [0x00,0x00,0x66,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
 
-v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 op_sel:[1,1,0,0] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+v_dot2_f16_f16_e64_dpp v0.l, v1, v2, v3.l op_sel:[1,1,0,0] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
 // GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid op_sel operand
 
-v_dot2_f16_f16_e64_dpp v0, s1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+v_dot2_f16_f16_e64_dpp v0.h, v1, v2, v3.h op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+// GFX11: encoding: [0x00,0x60,0x66,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
+
+v_dot2_f16_f16_e64_dpp v0.h, |v1|, -v2, -|s3| op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+// GFX11: encoding: [0x00,0x65,0x66,0xd6,0xfa,0x04,0x0e,0xc0,0x01,0xe4,0x04,0x00]
+
+v_dot2_f16_f16_e64_dpp v0.l, s1, v2, v3.l quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
 // GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
 
-v_dot2_f16_f16_e64_dpp v0, v1, s2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+v_dot2_f16_f16_e64_dpp v0.l, v1, s2, v3.l quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
 // GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
 
-v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
-// GFX11: encoding: [0x00,0x60,0x66,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, v3.l quad_perm:[3,2,1,0]
+// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0x0e,0x04,0x01,0x1b,0x00,0xff]
 
-v_dot2_f16_f16_e64_dpp v0, |v1|, -v2, -|s3| op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
-// GFX11: encoding: [0x00,0x65,0x66,0xd6,0xfa,0x04,0x0e,0xc0,0x01,0xe4,0x04,0x00]
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, v255.h quad_perm:[0,1,2,3]
+// GFX11: encoding: [0x05,0x20,0x66,0xd6,0xfa,0x04,0xfe,0x07,0x01,0xe4,0x00,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, s3 row_mirror
+// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0x0e,0x00,0x01,0x40,0x01,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, s105 row_half_mirror
+// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0xa6,0x01,0x01,0x41,0x01,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, ttmp15 row_shl:1
+// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0xee,0x01,0x01,0x01,0x01,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, vcc_hi row_shl:15
+// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0xae,0x01,0x01,0x0f,0x01,0xff]
 
-v_dot2_f16_f16_e64_dpp v5, v1, v2, 0.5 quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf
-// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0xc2,0x03,0x01,0x1b,0x00,0xff]
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, vcc_lo row_shr:1
+// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0xaa,0x01,0x01,0x11,0x01,0xff]
 
-v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
-// GFX11: encoding: [0x00,0x00,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, -|m0| row_shr:15
+// GFX11: encoding: [0x05,0x04,0x66,0xd6,0xfa,0x04,0xf6,0x81,0x01,0x1f,0x01,0xff]
 
-v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 op_sel:[1,1,0,0] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+v_dot2_f16_f16_e64_dpp v5.l, -|v1|, v2, -|exec_hi| row_ror:1
+// GFX11: encoding: [0x05,0x05,0x66,0xd6,0xfa,0x04,0xfe,0xa1,0x01,0x21,0x01,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, v1, -|v2|, -|exec_lo| row_ror:15
+// GFX11: encoding: [0x05,0x06,0x66,0xd6,0xfa,0x04,0xfa,0xc1,0x01,0x2f,0x01,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, |v1|, -v2, null row_share:0 row_mask:0xf bank_mask:0xf
+// GFX11: encoding: [0x05,0x01,0x66,0xd6,0xfa,0x04,0xf2,0x41,0x01,0x50,0x01,0xff]
+
+v_dot2_f16_f16_e64_dpp v5.l, -v1, |v2|, -1 row_share:15 row_mask:0x0 bank_mask:0x1
+// GFX11: encoding: [0x05,0x02,0x66,0xd6,0xfa,0x04,0x06,0x23,0x01,0x5f,0x01,0x01]
+
+v_dot2_f16_f16_e64_dpp v5.l, -|v1|, -|v2|, 0.5 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 fi:0
+// GFX11: encoding: [0x05,0x03,0x66,0xd6,0xfa,0x04,0xc2,0x63,0x01,0x60,0x09,0x13]
+
+v_dot2_f16_f16_e64_dpp v255.h, -|v255|, -|v255|, -|src_scc| row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
+// GFX11: encoding: [0xff,0x47,0x66,0xd6,0xfa,0xfe,0xf7,0xe3,0xff,0x6f,0x05,0x30]
+
+v_dot2_f16_f16_e64_dpp v5.l, v1, v2, v255.h quad_perm:[0,1,2,3]
+// GFX11: encoding: [0x05,0x20,0x66,0xd6,0xfa,0x04,0xfe,0x07,0x01,0xe4,0x00,0xff]
+
+v_dot2_f16_f16_e64_dpp v255.h, -|v255|, -|v255|, -|src_scc| row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
+// GFX11: encoding: [0xff,0x47,0x66,0xd6,0xfa,0xfe,0xf7,0xe3,0xff,0x6f,0x05,0x30]
+
+v_dot2_bf16_bf16_e64_dpp v0.l, v1, v2, v3.l op_sel:[1,1,0,0] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
 // GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid op_sel operand
 
-v_dot2_bf16_bf16_e64_dpp v0, s1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+v_dot2_bf16_bf16_e64_dpp v0.h, v1, v2, v3.h op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+// GFX11: encoding: [0x00,0x60,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
+
+v_dot2_bf16_bf16_e64_dpp v0.h, |v1|, -v2, -|s3| op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
+// GFX11: encoding: [0x00,0x65,0x67,0xd6,0xfa,0x04,0x0e,0xc0,0x01,0xe4,0x04,0x00]
+
+v_dot2_bf16_bf16_e64_dpp v0.l, s1, v2, v3.l quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
 // GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
 
-v_dot2_bf16_bf16_e64_dpp v0, v1, s2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
+v_dot2_bf16_bf16_e64_dpp v0.l, v1, s2, v3.l quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
 // GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
 
-v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
-// GFX11: encoding: [0x00,0x60,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, v3.l quad_perm:[3,2,1,0]
+// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0x1b,0x00,0xff]
 
-v_dot2_bf16_bf16_e64_dpp v0, |v1|, -v2, -|s3| op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
-// GFX11: encoding: [0x00,0x65,0x67,0xd6,0xfa,0x04,0x0e,0xc0,0x01,0xe4,0x04,0x00]
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, v3.l quad_perm:[0,1,2,3]
+// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x00,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, v255.h row_mirror
+// GFX11: encoding: [0x05,0x20,0x67,0xd6,0xfa,0x04,0xfe,0x07,0x01,0x40,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, s3 row_half_mirror
+// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0x0e,0x00,0x01,0x41,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, s105 row_shl:1
+// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0xa6,0x01,0x01,0x01,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, ttmp15 row_shl:15
+// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0xee,0x01,0x01,0x0f,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, vcc_hi row_shr:1
+// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0xae,0x01,0x01,0x11,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, |v1|, v2, -vcc_lo row_shr:15
+// GFX11: encoding: [0x05,0x01,0x67,0xd6,0xfa,0x04,0xaa,0x81,0x01,0x1f,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, -v2, |m0| row_ror:1
+// GFX11: encoding: [0x05,0x04,0x67,0xd6,0xfa,0x04,0xf6,0x41,0x01,0x21,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, -|v1|, v2, -|exec_hi| row_ror:15
+// GFX11: encoding: [0x05,0x05,0x67,0xd6,0xfa,0x04,0xfe,0xa1,0x01,0x2f,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, -|v2|, -|exec_lo| row_share:0 row_mask:0xf bank_mask:0xf
+// GFX11: encoding: [0x05,0x06,0x67,0xd6,0xfa,0x04,0xfa,0xc1,0x01,0x50,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, -v1, |v2|, null row_share:15 row_mask:0x0 bank_mask:0x1
+// GFX11: encoding: [0x05,0x02,0x67,0xd6,0xfa,0x04,0xf2,0x21,0x01,0x5f,0x01,0x01]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, -|v1|, -|v2|, -1 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 fi:0
+// GFX11: encoding: [0x05,0x03,0x67,0xd6,0xfa,0x04,0x06,0x63,0x01,0x60,0x09,0x13]
+
+v_dot2_bf16_bf16_e64_dpp v255.h, -|v255|, -|v255|, -|src_scc| row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
+// GFX11: encoding: [0xff,0x47,0x67,0xd6,0xfa,0xfe,0xf7,0xe3,0xff,0x6f,0x05,0x30]
+
+v_dot2_bf16_bf16_e64_dpp v5.l, v1, v2, v255.h row_mirror
+// GFX11: encoding: [0x05,0x20,0x67,0xd6,0xfa,0x04,0xfe,0x07,0x01,0x40,0x01,0xff]
+
+v_dot2_bf16_bf16_e64_dpp v255.h, -|v255|, -|v255|, -|src_scc| row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
+// GFX11: encoding: [0xff,0x47,0x67,0xd6,0xfa,0xfe,0xf7,0xe3,0xff,0x6f,0x05,0x30]
 
-v_dot2_bf16_bf16_e64_dpp v5, v1, v2, 0 quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf
-// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0x02,0x02,0x01,0x1b,0x00,0xff]
----------------
broxigarchen wrote:

resolve test divergenced

https://github.com/llvm/llvm-project/pull/113474
