From 11057ee0226ddffd82a9b7afab5fc738a9d901fb Mon Sep 17 00:00:00 2001
From: Marek Olsak <marek.olsak@amd.com>
Date: Tue, 3 Feb 2015 17:38:01 +0000
Subject: [PATCH] R600/SI: Trivial instruction definition corrections for VI
 (v2)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf8
Content-Transfer-Encoding: 8bit

- V_MAC_LEGACY_F32 exists on VI, but it's VOP3-only.

- Define CVT_PK opcodes which are different between SI and VI. These are
  unused. The idea is to define all chip differences.

v2: keep V_MUL_LO_U32

Tested-by: Michel Dänzer <michel.daenzer@amd.com>
llvm-svn: 227988
---
 llvm/lib/Target/R600/SIInstrInfo.td    |  1 +
 llvm/lib/Target/R600/SIInstructions.td | 35 ++++++++++++++++++++++------------
 2 files changed, 24 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Target/R600/SIInstrInfo.td b/llvm/lib/Target/R600/SIInstrInfo.td
index e2e5bde..b793b5c 100644
--- a/llvm/lib/Target/R600/SIInstrInfo.td
+++ b/llvm/lib/Target/R600/SIInstrInfo.td
@@ -785,6 +785,7 @@ def VOP_F32_F32_I32 : VOPProfile <[f32, f32, i32, untyped]>;
 def VOP_F64_F64_F64 : VOPProfile <[f64, f64, f64, untyped]>;
 def VOP_F64_F64_I32 : VOPProfile <[f64, f64, i32, untyped]>;
 def VOP_I32_F32_F32 : VOPProfile <[i32, f32, f32, untyped]>;
+def VOP_I32_F32_I32 : VOPProfile <[i32, f32, i32, untyped]>;
 def VOP_I32_I32_I32 : VOPProfile <[i32, i32, i32, untyped]>;
 def VOP_I32_I32_I32_VCC : VOPProfile <[i32, i32, i32, untyped]> {
   let Src0RC32 = VCSrc_32;
diff --git a/llvm/lib/Target/R600/SIInstructions.td b/llvm/lib/Target/R600/SIInstructions.td
index 7b203d6..1f589dd 100644
--- a/llvm/lib/Target/R600/SIInstructions.td
+++ b/llvm/lib/Target/R600/SIInstructions.td
@@ -1543,12 +1543,6 @@ defm V_WRITELANE_B32 : VOP2SI_3VI_m <
 // These instructions only exist on SI and CI
 let SubtargetPredicate = isSICI in {
 
-let isCommutable = 1 in {
-defm V_MAC_LEGACY_F32 : VOP2Inst <vop2<0x6>, "v_mac_legacy_f32",
-  VOP_F32_F32_F32
->;
-} // End isCommutable = 1
-
 defm V_MIN_LEGACY_F32 : VOP2Inst <vop2<0xd>, "v_min_legacy_f32",
   VOP_F32_F32_F32, AMDGPUfmin_legacy
 >;
@@ -1569,6 +1563,12 @@ defm V_LSHL_B32 : VOP2Inst <vop2<0x19>, "v_lshl_b32", VOP_I32_I32_I32, shl>;
 } // End isCommutable = 1
 } // End let SubtargetPredicate = SICI
 
+let isCommutable = 1 in {
+defm V_MAC_LEGACY_F32 : VOP2_VI3_Inst <vop23<0x6, 0x28e>, "v_mac_legacy_f32",
+  VOP_F32_F32_F32
+>;
+} // End isCommutable = 1
+
 defm V_BFM_B32 : VOP2_VI3_Inst <vop23<0x1e, 0x293>, "v_bfm_b32", VOP_I32_I32_I32,
   AMDGPUbfm
 >;
@@ -1585,14 +1585,25 @@ defm V_LDEXP_F32 : VOP2_VI3_Inst <vop23<0x2b, 0x288>, "v_ldexp_f32",
   VOP_F32_F32_I32, AMDGPUldexp
 >;
 
-////def V_CVT_PKACCUM_U8_F32 : VOP2_U8 <0x0000002c, "v_cvt_pkaccum_u8_f32", []>;
-////def V_CVT_PKNORM_I16_F32 : VOP2_I16 <0x0000002d, "v_cvt_pknorm_i16_f32", []>;
-////def V_CVT_PKNORM_U16_F32 : VOP2_U16 <0x0000002e, "v_cvt_pknorm_u16_f32", []>;
+
+defm V_CVT_PKACCUM_U8_F32 : VOP2_VI3_Inst <vop23<0x2c, 0x1f0>, "v_cvt_pkaccum_u8_f32",
+  VOP_I32_F32_I32>; // TODO: set "Uses = dst"
+
+defm V_CVT_PKNORM_I16_F32 : VOP2_VI3_Inst <vop23<0x2d, 0x294>, "v_cvt_pknorm_i16_f32",
+  VOP_I32_F32_F32
+>;
+defm V_CVT_PKNORM_U16_F32 : VOP2_VI3_Inst <vop23<0x2e, 0x295>, "v_cvt_pknorm_u16_f32",
+  VOP_I32_F32_F32
+>;
 defm V_CVT_PKRTZ_F16_F32 : VOP2_VI3_Inst <vop23<0x2f, 0x296>, "v_cvt_pkrtz_f16_f32",
- VOP_I32_F32_F32, int_SI_packf16
+  VOP_I32_F32_F32, int_SI_packf16
+>;
+defm V_CVT_PK_U16_U32 : VOP2_VI3_Inst <vop23<0x30, 0x297>, "v_cvt_pk_u16_u32",
+  VOP_I32_I32_I32
+>;
+defm V_CVT_PK_I16_I32 : VOP2_VI3_Inst <vop23<0x31, 0x298>, "v_cvt_pk_i16_i32",
+  VOP_I32_I32_I32
 >;
-////def V_CVT_PK_U16_U32 : VOP2_U16 <0x00000030, "v_cvt_pk_u16_u32", []>;
-////def V_CVT_PK_I16_I32 : VOP2_I16 <0x00000031, "v_cvt_pk_i16_i32", []>;
 
 //===----------------------------------------------------------------------===//
 // VOP3 Instructions
-- 
2.7.4