[llvm] [AMDGPU] Prefer v_madak_f32 over v_madmk_f32 to reduce vgpr pressure (PR #72506)

[Jay Foad via llvm-commits]

================
@@ -7,7 +7,6 @@ define amdgpu_ps float @_amdgpu_ps_main() #0 {
 ; GFX10:       ; %bb.0: ; %.entry
 ; GFX10-NEXT:    image_sample v[0:1], v[0:1], s[0:7], s[0:3] dmask:0x3 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    v_mov_b32_e32 v4, 0
-; GFX10-NEXT:    v_mov_b32_e32 v7, 0x3ca3d70a
----------------
jayfoad wrote:

Previously: literal materialized in vgpr

https://github.com/llvm/llvm-project/pull/72506