[llvm] [AMDGPU] Add clamp support to v_add_{max|min}_{i|u}32 (PR #164489)
    Stanislav Mekhanoshin via llvm-commits 
    llvm-commits at lists.llvm.org
       
    Tue Oct 21 14:04:44 PDT 2025
    
    
  
================
@@ -237,64 +237,76 @@
 # GFX1250: v_mad_nc_i64_i32 v[2:3], s4, v7, v[8:9] clamp ; encoding: [0x02,0x80,0xfb,0xd6,0x04,0x0e,0x22,0x04]
 
 0x02,0x00,0x60,0xd6,0x04,0x08,0x09,0x04
-# GFX1250: v_add_min_i32_e64 v2, s4, 4, v2         ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x08,0x09,0x04]
+# GFX1250: v_add_min_i32 v2, s4, 4, v2             ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x08,0x09,0x04]
 
 0x02,0x00,0x60,0xd6,0x04,0x0e,0x22,0x04
-# GFX1250: v_add_min_i32_e64 v2, s4, v7, v8        ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x0e,0x22,0x04]
+# GFX1250: v_add_min_i32 v2, s4, v7, v8            ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x0e,0x22,0x04]
 
 0x02,0x00,0x60,0xd6,0x04,0x01,0x05,0x02
-# GFX1250: v_add_min_i32_e64 v2, v4, 0, 1          ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x01,0x05,0x02]
+# GFX1250: v_add_min_i32 v2, v4, 0, 1              ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x01,0x05,0x02]
 
 0x02,0x00,0x60,0xd6,0x04,0x07,0x09,0x00
-# GFX1250: v_add_min_i32_e64 v2, v4, 3, s2         ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x07,0x09,0x00]
+# GFX1250: v_add_min_i32 v2, v4, 3, s2             ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x07,0x09,0x00]
 
 0x02,0x00,0x60,0xd6,0x04,0x0f,0xfe,0x03,0x39,0x30,0x00,0x00
-# GFX1250: v_add_min_i32_e64 v2, v4, v7, 0x3039    ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x0f,0xfe,0x03,0x39,0x30,0x00,0x00]
+# GFX1250: v_add_min_i32 v2, v4, v7, 0x3039        ; encoding: [0x02,0x00,0x60,0xd6,0x04,0x0f,0xfe,0x03,0x39,0x30,0x00,0x00]
+
+0x00,0x80,0x5f,0xd6,0x01,0x05,0x0e,0x04
+# GFX1250: v_add_max_u32 v0, v1, v2, v3 clamp      ; encoding: [0x00,0x80,0x5f,0xd6,0x01,0x05,0x0e,0x04]
+
+0x00,0x80,0x60,0xd6,0x01,0x05,0x0e,0x04
+# GFX1250: v_add_min_i32 v0, v1, v2, v3 clamp      ; encoding: [0x00,0x80,0x60,0xd6,0x01,0x05,0x0e,0x04]
 
 0x02,0x00,0x5e,0xd6,0x04,0x08,0x09,0x04
-# GFX1250: v_add_max_i32_e64 v2, s4, 4, v2         ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x08,0x09,0x04]
+# GFX1250: v_add_max_i32 v2, s4, 4, v2             ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x08,0x09,0x04]
 
 0x02,0x00,0x5e,0xd6,0x04,0x0e,0x22,0x04
-# GFX1250: v_add_max_i32_e64 v2, s4, v7, v8        ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x0e,0x22,0x04]
+# GFX1250: v_add_max_i32 v2, s4, v7, v8            ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x0e,0x22,0x04]
 
 0x02,0x00,0x5e,0xd6,0x04,0x01,0x05,0x02
-# GFX1250: v_add_max_i32_e64 v2, v4, 0, 1          ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x01,0x05,0x02]
+# GFX1250: v_add_max_i32 v2, v4, 0, 1              ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x01,0x05,0x02]
 
 0x02,0x00,0x5e,0xd6,0x04,0x07,0x09,0x00
-# GFX1250: v_add_max_i32_e64 v2, v4, 3, s2         ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x07,0x09,0x00]
+# GFX1250: v_add_max_i32 v2, v4, 3, s2             ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x07,0x09,0x00]
 
 0x02,0x00,0x5e,0xd6,0x04,0x0f,0xfe,0x03,0x39,0x30,0x00,0x00
-# GFX1250: v_add_max_i32_e64 v2, v4, v7, 0x3039    ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x0f,0xfe,0x03,0x39,0x30,0x00,0x00]
+# GFX1250: v_add_max_i32 v2, v4, v7, 0x3039        ; encoding: [0x02,0x00,0x5e,0xd6,0x04,0x0f,0xfe,0x03,0x39,0x30,0x00,0x00]
 
 0x02,0x00,0x61,0xd6,0x04,0x08,0x09,0x04
-# GFX1250: v_add_min_u32_e64 v2, s4, 4, v2         ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x08,0x09,0x04]
+# GFX1250: v_add_min_u32 v2, s4, 4, v2             ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x08,0x09,0x04]
 
 0x02,0x00,0x61,0xd6,0x04,0x0e,0x22,0x04
-# GFX1250: v_add_min_u32_e64 v2, s4, v7, v8        ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x0e,0x22,0x04]
+# GFX1250: v_add_min_u32 v2, s4, v7, v8            ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x0e,0x22,0x04]
 
 0x02,0x00,0x61,0xd6,0x04,0x01,0x05,0x02
-# GFX1250: v_add_min_u32_e64 v2, v4, 0, 1          ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x01,0x05,0x02]
+# GFX1250: v_add_min_u32 v2, v4, 0, 1              ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x01,0x05,0x02]
 
 0x02,0x00,0x61,0xd6,0x04,0x07,0x09,0x00
-# GFX1250: v_add_min_u32_e64 v2, v4, 3, s2         ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x07,0x09,0x00]
+# GFX1250: v_add_min_u32 v2, v4, 3, s2             ; encoding: [0x02,0x00,0x61,0xd6,0x04,0x07,0x09,0x00]
----------------
rampitec wrote:
This is correct though, as this is VOP3 only.
https://github.com/llvm/llvm-project/pull/164489
    
    
More information about the llvm-commits
mailing list