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

Improve naming of masked load/store instructions in builtins.

Browse Source 
Now, use _i32 suffixes, rather than _32, etc.  Also cleaned up the m4
macro to generate these functions, using WIDTH to get the target width,
etc.
This commit is contained in:
Matt Pharr
2012-06-07 13:51:08 -07:00
parent 91d22d150f
commit b86d40091a
13 changed files with 299 additions and 308 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
builtins/target-sse2.ll
View File

@@ -350,16 +350,16 @@ reduce_equal(4)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; masked store
define void @__masked_store_blend_32(<4 x i32>* nocapture, <4 x i32>,
<4 x i32> %mask) nounwind alwaysinline {
define void @__masked_store_blend_i32(<4 x i32>* nocapture, <4 x i32>,
<4 x i32> %mask) nounwind alwaysinline {
%val = load <4 x i32> * %0, align 4
%newval = call <4 x i32> @__vselect_i32(<4 x i32> %val, <4 x i32> %1, <4 x i32> %mask)
store <4 x i32> %newval, <4 x i32> * %0, align 4
ret void
}
define void @__masked_store_blend_64(<4 x i64>* nocapture %ptr, <4 x i64> %new,
<4 x i32> %mask) nounwind alwaysinline {
define void @__masked_store_blend_i64(<4 x i64>* nocapture %ptr, <4 x i64> %new,
<4 x i32> %mask) nounwind alwaysinline {
%oldValue = load <4 x i64>* %ptr, align 8
; Do 4x64-bit blends by doing two <4 x i32> blends, where the <4 x i32> values
@@ -552,10 +552,10 @@ define <4 x double> @__max_varying_double(<4 x double>, <4 x double>) nounwind r
masked_store_blend_8_16_by_4()
gen_masked_store(4, i8, 8)
gen_masked_store(4, i16, 16)
gen_masked_store(4, i32, 32)
gen_masked_store(4, i64, 64)
gen_masked_store(i8)
gen_masked_store(i16)
gen_masked_store(i32)
gen_masked_store(i64)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; unaligned loads/loads+broadcasts
@@ -565,10 +565,10 @@ load_and_broadcast(i16)
load_and_broadcast(i32)
load_and_broadcast(i64)
masked_load(4, i8, 8, 1)
masked_load(4, i16, 16, 2)
masked_load(4, i32, 32, 4)
masked_load(4, i64, 64, 8)
masked_load(i8, 1)
masked_load(i16, 2)
masked_load(i32, 4)
masked_load(i64, 8)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; gather/scatter