6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-08-05 13:26:55 +00:00
Code Issues Projects Releases Wiki Activity

AVX-512: Added more intrinsics for convert and min/max.

Browse Source 
Removed vzeroupper from AVX-512 mode - our optimization gude does not recommend to insert vzeroupper at all.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198557 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Elena Demikhovsky
2014-01-05 10:46:09 +00:00
parent 5c1c0bd714
commit 3a880de6e6
5 changed files with 132 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
include/llvm/IR/IntrinsicsX86.td
View File

@@ -2747,7 +2747,13 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_cvtdq2pd_512 : GCCBuiltin<"__builtin_ia32_cvtdq2pd512_mask">,
Intrinsic<[llvm_v8f64_ty], [llvm_v8i32_ty, llvm_v8f64_ty,
llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>;
llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_cvtudq2ps_512 : GCCBuiltin<"__builtin_ia32_cvtudq2ps512_mask">,
Intrinsic<[llvm_v16f32_ty], [llvm_v16i32_ty, llvm_v16f32_ty,
llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_cvtudq2pd_512 : GCCBuiltin<"__builtin_ia32_cvtudq2pd512_mask">,
Intrinsic<[llvm_v8f64_ty], [llvm_v8i32_ty, llvm_v8f64_ty,
llvm_i8_ty], [IntrNoMem]>;
}
// Vector load with broadcast
@@ -2800,18 +2806,18 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
// Arithmetic ops
let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_avx512_min_ps_512 : GCCBuiltin<"__builtin_ia32_minps512">,
Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty,
llvm_v16f32_ty], [IntrNoMem]>;
def int_x86_avx512_min_pd_512 : GCCBuiltin<"__builtin_ia32_minpd512">,
Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty,
llvm_v8f64_ty], [IntrNoMem]>;
def int_x86_avx512_max_ps_512 : GCCBuiltin<"__builtin_ia32_maxps512">,
Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty,
llvm_v16f32_ty], [IntrNoMem]>;
def int_x86_avx512_max_pd_512 : GCCBuiltin<"__builtin_ia32_maxpd512">,
Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty,
llvm_v8f64_ty], [IntrNoMem]>;
def int_x86_avx512_mask_max_ps_512 : GCCBuiltin<"__builtin_ia32_maxps512_mask">,
Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_v16f32_ty,
llvm_v16f32_ty, llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_max_pd_512 : GCCBuiltin<"__builtin_ia32_maxpd512_mask">,
Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_v8f64_ty,
llvm_v8f64_ty, llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_min_ps_512 : GCCBuiltin<"__builtin_ia32_minps512_mask">,
Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_v16f32_ty,
llvm_v16f32_ty, llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_min_pd_512 : GCCBuiltin<"__builtin_ia32_minpd512_mask">,
Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_v8f64_ty,
llvm_v8f64_ty, llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_pmaxu_d : GCCBuiltin<"__builtin_ia32_pmaxud512">,
Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty,