added half2 math operations

1. They use SDWA + LLVM IR
2. Added these functions to test
3. Need to do exp, exp10, log, log10, rint

Change-Id: I06176acc6cb8bb054495310531777406a41b54e4

hipamd/src/hip_ir.ll

@@ -86,4 +86,64 @@ define i32 @__hip_hc_ir_hsub2_int(i32 %a, i32 %b) #1 {
   ret i32 %1
 }
 
+define i32 @__hip_hc_ir_h2ceil_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_ceil_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_ceil_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2cos_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_cos_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_cos_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2exp2_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_exp_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_exp_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2floor_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_floor_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_floor_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2log2_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_log_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_log_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2rcp_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_rcp_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_rcp_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2rsqrt_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_rsq_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_rsq_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2sin_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_sin_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_sin_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2sqrt_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_sqrt_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_sqrt_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
+define i32 @__hip_hc_ir_h2trunc_int(i32 %a) #1 {
+  %1 = tail call i32 asm sideeffect "v_trunc_f16 $0, $1","=v,v"(i32 %a)
+  tail call void asm sideeffect "v_trunc_f16_sdwa $0, $1 dst_sel:WORD_1 dst_unused:UNUSED_PRESERVE src0_sel:WORD_1","v,v"(i32 %1, i32 %a)
+  ret i32 %1
+}
+
 attributes #1 = { alwaysinline nounwind }