test/Transforms/SLPVectorizer/AMDGPU/bitcast-int-fp.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=slp-vectorizer < %s | FileCheck -check-prefixes=NOPACKED %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a -passes=slp-vectorizer < %s | FileCheck -check-prefixes=PACKED,GFX90A %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -passes=slp-vectorizer < %s | FileCheck -check-prefixes=PACKED,GFX950 %s

; A bitcast from i32 to float is a no-op, so it gives no saving when it is
; vectorized.

; Only the bitcasts feed the stored pair. A vector bitcast saves nothing over
; two scalar bitcasts.
define void @bitcast_only(i32 %a, i32 %b, ptr addrspace(3) %out) {
; NOPACKED-LABEL: define void @bitcast_only(
; NOPACKED-SAME: i32 [[A:%.*]], i32 [[B:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0:[0-9]+]] {
; NOPACKED-NEXT:    [[AF:%.*]] = bitcast i32 [[A]] to float
; NOPACKED-NEXT:    [[BF:%.*]] = bitcast i32 [[B]] to float
; NOPACKED-NEXT:    [[R0:%.*]] = insertelement <2 x float> poison, float [[AF]], i64 0
; NOPACKED-NEXT:    [[R1:%.*]] = insertelement <2 x float> [[R0]], float [[BF]], i64 1
; NOPACKED-NEXT:    store <2 x float> [[R1]], ptr addrspace(3) [[OUT]], align 8
; NOPACKED-NEXT:    ret void
;
; PACKED-LABEL: define void @bitcast_only(
; PACKED-SAME: i32 [[A:%.*]], i32 [[B:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0:[0-9]+]] {
; PACKED-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> poison, i32 [[A]], i64 0
; PACKED-NEXT:    [[TMP2:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[B]], i64 1
; PACKED-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[TMP2]] to <2 x float>
; PACKED-NEXT:    store <2 x float> [[TMP3]], ptr addrspace(3) [[OUT]], align 8
; PACKED-NEXT:    ret void
;
  %af = bitcast i32 %a to float
  %bf = bitcast i32 %b to float
  %r0 = insertelement <2 x float> poison, float %af, i64 0
  %r1 = insertelement <2 x float> %r0, float %bf, i64 1
  store <2 x float> %r1, ptr addrspace(3) %out, align 8
  ret void
}

; The fmul pair is a packed fmul, so the pair is still vectorized.
define void @bitcast_fmul(i32 %a, i32 %b, float %x, float %y, ptr addrspace(3) %out) {
; NOPACKED-LABEL: define void @bitcast_fmul(
; NOPACKED-SAME: i32 [[A:%.*]], i32 [[B:%.*]], float [[X:%.*]], float [[Y:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0]] {
; NOPACKED-NEXT:    [[AF:%.*]] = bitcast i32 [[A]] to float
; NOPACKED-NEXT:    [[BF:%.*]] = bitcast i32 [[B]] to float
; NOPACKED-NEXT:    [[M0:%.*]] = fmul float [[AF]], [[X]]
; NOPACKED-NEXT:    [[M1:%.*]] = fmul float [[BF]], [[Y]]
; NOPACKED-NEXT:    [[R0:%.*]] = insertelement <2 x float> poison, float [[M0]], i64 0
; NOPACKED-NEXT:    [[R1:%.*]] = insertelement <2 x float> [[R0]], float [[M1]], i64 1
; NOPACKED-NEXT:    store <2 x float> [[R1]], ptr addrspace(3) [[OUT]], align 8
; NOPACKED-NEXT:    ret void
;
; PACKED-LABEL: define void @bitcast_fmul(
; PACKED-SAME: i32 [[A:%.*]], i32 [[B:%.*]], float [[X:%.*]], float [[Y:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0]] {
; PACKED-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> poison, i32 [[A]], i64 0
; PACKED-NEXT:    [[TMP2:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[B]], i64 1
; PACKED-NEXT:    [[TMP3:%.*]] = bitcast <2 x i32> [[TMP2]] to <2 x float>
; PACKED-NEXT:    [[TMP4:%.*]] = insertelement <2 x float> poison, float [[X]], i64 0
; PACKED-NEXT:    [[TMP5:%.*]] = insertelement <2 x float> [[TMP4]], float [[Y]], i64 1
; PACKED-NEXT:    [[TMP6:%.*]] = fmul <2 x float> [[TMP3]], [[TMP5]]
; PACKED-NEXT:    store <2 x float> [[TMP6]], ptr addrspace(3) [[OUT]], align 8
; PACKED-NEXT:    ret void
;
  %af = bitcast i32 %a to float
  %bf = bitcast i32 %b to float
  %m0 = fmul float %af, %x
  %m1 = fmul float %bf, %y
  %r0 = insertelement <2 x float> poison, float %m0, i64 0
  %r1 = insertelement <2 x float> %r0, float %m1, i64 1
  store <2 x float> %r1, ptr addrspace(3) %out, align 8
  ret void
}
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX90A: {{.*}}
; GFX950: {{.*}}