aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

New LLVM IR for getelementptr instruction

Browse Source
This commit is contained in:
Vsevolod Livinskiy
2015-03-05 12:51:28 +03:00
parent 0b1323e070
commit 29859e81ba
4 changed files with 141 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
builtins/target-avx-x2.ll
View File

@@ -427,7 +427,7 @@ define <16 x i32> @__masked_load_i32(i8 *, <16 x i32> %mask) nounwind alwaysinli
%val0 = call <8 x float> @llvm.x86.avx.maskload.ps.256(i8 * %0, <8 x float> %mask0)
%mask1 = shufflevector <16 x float> %floatmask, <16 x float> undef,
<8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%ptr1 = getelementptr i8 * %0, i32 32 ;; 8x4 bytes = 32
%ptr1 = getelementptr PTR_OP_ARGS(`i8', `%0, i32 32 ') ;; 8x4 bytes = 32
%val1 = call <8 x float> @llvm.x86.avx.maskload.ps.256(i8 * %ptr1, <8 x float> %mask1)
%retval = shufflevector <8 x float> %val0, <8 x float> %val1,
@@ -454,11 +454,11 @@ define <16 x i64> @__masked_load_i64(i8 *, <16 x i32> %mask) nounwind alwaysinli
%mask3d = bitcast <8 x i32> %mask3 to <4 x double>
%val0d = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8 * %0, <4 x double> %mask0d)
%ptr1 = getelementptr i8 * %0, i32 32
%ptr1 = getelementptr PTR_OP_ARGS(`i8', `%0, i32 32')
%val1d = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8 * %ptr1, <4 x double> %mask1d)
%ptr2 = getelementptr i8 * %0, i32 64
%ptr2 = getelementptr PTR_OP_ARGS(`i8', `%0, i32 64')
%val2d = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8 * %ptr2, <4 x double> %mask2d)
%ptr3 = getelementptr i8 * %0, i32 96
%ptr3 = getelementptr PTR_OP_ARGS(`i8', `%0, i32 96')
%val3d = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8 * %ptr3, <4 x double> %mask3d)
%val01 = shufflevector <4 x double> %val0d, <4 x double> %val1d,
@@ -504,7 +504,7 @@ define void @__masked_store_i32(<16 x i32>* nocapture, <16 x i32>,
<8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
call void @llvm.x86.avx.maskstore.ps.256(i8 * %ptr, <8 x float> %mask0, <8 x float> %val0)
%ptr1 = getelementptr i8 * %ptr, i32 32
%ptr1 = getelementptr PTR_OP_ARGS(`i8', `%ptr, i32 32')
call void @llvm.x86.avx.maskstore.ps.256(i8 * %ptr1, <8 x float> %mask1, <8 x float> %val1)
ret void
@@ -539,11 +539,11 @@ define void @__masked_store_i64(<16 x i64>* nocapture, <16 x i64>,
<4 x i32> <i32 12, i32 13, i32 14, i32 15>
call void @llvm.x86.avx.maskstore.pd.256(i8 * %ptr, <4 x double> %mask0d, <4 x double> %val0)
%ptr1 = getelementptr i8 * %ptr, i32 32
%ptr1 = getelementptr PTR_OP_ARGS(`i8', `%ptr, i32 32')
call void @llvm.x86.avx.maskstore.pd.256(i8 * %ptr1, <4 x double> %mask1d, <4 x double> %val1)
%ptr2 = getelementptr i8 * %ptr, i32 64
%ptr2 = getelementptr PTR_OP_ARGS(`i8', `%ptr, i32 64')
call void @llvm.x86.avx.maskstore.pd.256(i8 * %ptr2, <4 x double> %mask2d, <4 x double> %val2)
%ptr3 = getelementptr i8 * %ptr, i32 96
%ptr3 = getelementptr PTR_OP_ARGS(`i8', `%ptr, i32 96')
call void @llvm.x86.avx.maskstore.pd.256(i8 * %ptr3, <4 x double> %mask3d, <4 x double> %val3)
ret void